Research and Development at DWD

DWD Database Reference for the

Global and Regional ICON and ICON-EPS
Forecasting System
Version 2.1.7

D. Reinert, F. Prill, H. Frank, M. Denhard, M. Baldauf, C. Schraff,

C. Gebhardt, C. Marsigli and G. Zängl

Offenbach am Main, 2021

Version: 2.1.7

DOI: 10.5676/DWD_pub/nwv/icon_2.1.7

The CC license “BY-NC-ND” allows others only to download the publication and share it with others
as long as they credit the publication, but they can’t change it in any way or use it commercially.

Publisher Editors
Deutscher Wetterdienst Daniel Reinert, FE13,
Business Area “Research and Development” Tel. +49 (69) 8062-2060, [email protected]
Frankfurter Straße 135 Helmut Frank, FE13,
63067 Offenbach Tel. +49 (69) 8062-2742, [email protected]
www.dwd.de Florian Prill, FE13,
Tel. +49 (69) 8062-2727, [email protected]

For data requests please contact: [email protected]

Revision History

Revision Date Author(s) Description

0.1.0 10.01.13 DR, FP Generated preliminary list of available GRIB2 output fields
0.2.0 12.07.13 DR, FP Added a short section describing the horizontal ICON grid.
AUMFL_S, AVMFL_S added to the list of available output fields
0.2.1 15.07.13 DR Provide newly available output fields in tabulated form. Change
levelType of 3D atmospheric fields from 105 (Hybrid) to 150 (
Generalized vertical height coordinate)
0.2.2 16.07.13 FP Short description of ICON’s vertical grid.
0.2.3 25.09.13 DR Added description of available First Guess and analysis fields
0.2.4 17.12.13 DR Added description of external paramater fields
0.3.0 24.01.14 DR Added information about horizontal output grids
0.3.1 24.01.14 DR Added information about newly available output field OMEGA
0.4.0 22.05.14 HF Added SKY-database documentation
0.4.1 15.07.14 DR Some documentation on statistical processing and minor up-
dates. New output fields ASWDIR_S, ASWDIFD_S, ASWDIFU_S,
DTKE_CON
0.4.2 10.09.14 DR New output fields CLCT_MOD, CLDEPTH
0.5.0 01.10.14 DR Description of IAU initialization method
0.5.1 15.10.14 DR Updated description of necessary input fields
0.5.2 31.10.14 DR Add full table with model half level heights
0.6.0 05.12.14 DR Add short introduction and fix some minor bugs
0.6.1 10.12.14 DR New output field APAB_S
0.7.0 16.12.14 DR Revised documentation of time invariant fields and a couple of
bug fixes
0.7.2 09.01.15 DR General GRIB2 description
0.8.0 15.01.15 FP, DR Couple of bug fixes regarding the available fields on triangular
and regular grids
0.8.1 16.01.15 FP, DR List of pressure-level variables available on triangular grids
0.8.2 16.01.15 FP List of height-level variables available on regular grids
0.8.3 16.01.15 DR List of variables exclusively available for V V = 0
0.8.4 06.02.15 FP, DR Details of internal interpolation onto lon-lat grids. Details re-
garding output frequency.
0.8.5 18.02.15 FP Additional pressure levels for regular grid output.
0.8.6 23.02.15 FP Formula for computing non-zero topography level height.
1.0.0 23.02.15 FP Additional table of model full levels.
1.0.1 24.02.15 DR Update on available forecast runs and time span.

iii
Revision History

1.0.2 27.02.15 FP Added tables for grid point with maximum topo height.
1.0.3 13.03.15 DR, FP Section on statistically processed fields.
1.1.0 15.04.15 FP, DR Section on ICON EU nest (preliminary).
1.1.1 07.07.15 HF Added SMA list, list of half levels for EU nest, modified output
lists to automatically write model level variables in the namelist
templates.
1.1.1 17.07.15 HF Preliminary add T_S because it is already written in operations.
Some other minor modifications.
1.1.2 14.08.15 FP Added note on ICON’s earth radius and a table summarizing
regular grids.
1.1.3 04.12.15 FP Added WW code table 7.1.
1.1.4 11.01.16 HF Updated examples how to retrieve ICON data from SKY.
1.1.5 22.01.16 AR Description of En-Var.
1.1.6 28.01.16 DR Extend tables by field specific lat-lon interpolation method.
1.1.7 11.04.16 DR, FP Add timeline of model changes.
1.1.8 06.07.16 HF Add DTKE_HSH and other minor corrections.
1.1.9 27.09.16 DR Update intro and timeline.
1.1.10 03.02.17 DR update lat-lon interpolation methods and timeline.
1.1.11 08.05.17 DR update version history.
1.1.12 13.07.17 DR Update description for output variable SOILTYP.
1.1.13 25.10.17 FP, DR Remove references to COSMO-EU
1.1.14 10.01.18 DR Bug fix regarding availability of CLCT_MOD on global domain
1.2.0 26.01.18 MD Documentation for ICON-EPS products added
1.2.1 31.01.18 FP Updated height tables (appendix)
1.2.2 12.03.18 DR Added new output fields EVAP_PL, and SMI. Further adaptions
to the list of available fields and updated timeline; info on down-
load of grids.
1.2.3 07.07.18 FP Added output field CAPE_ML, EVAP_PL, SMI for global domain,
native grid. Added output field ALB_SEAICE.
1.2.4 31.08.18 MD Updated table of probability products.
1.2.5 09.10.18 DR Updated history of model changes; updated description of out-
put products; added output fields SNOAG and HSNOW_MAX
1.2.6 01.11.18 HF Updated section on Sky database examples.
1.2.7 27.02.19 DR Updated history of model changes.
1.2.8 01.04.19 HF Updated list of pressure levels of EPS output.
1.2.9 30.04.19 HF Updated list of pressure levels of deterministic global ICON on
triangular grid; hourly output of 5 fields of ICON-EU on regular
grid until the end of the forecast.
1.2.10 08.08.19 DR Updated history of model changes and description for HZEROCL
1.2.11 30.09.19 HF, DR Updated output list of ICON-EPS. Updated description of T_G
and T_SO
1.2.12 20.04.20 DR Updated history of model changes and added new diagnostic
output field CEILING
1.2.13 10.06.20 DR Updated history of model changes

iv
 Revision History

2.0.0 21.01.20 MB, FP, DR, MD, Common database reference for all models ICON global/-EU/-
CS, CG D2; partly restructured chapters about output variables
2.0.1 17.02.20 MD, CG, FP, DR, Ensemble model output is now contained in chapters 10-12
MB
2.1.0 15.05.20 MD, CM, MB, FP New tables for all EPS products
2.1.1 05.06.20 MB, DR New native grid for ICON-D2
2.1.2 19.01.21 DR, CG, MB Update a few product descriptions; hints about the rotated lat-
lon grid output
2.1.3 16.02.21 DR Add description for DBZ_850, DBZ_CMAX
2.1.4 30.06.21 FP,DR Updated forecast range for ICON-D2(-EPS)
2.1.5 09.09.21 HF New output field LPI_CON_CI_MAX for ICON-EU(-EPS)
2.1.6 20.10.21 HF Output of ICON-EU runs to 51 h
2.1.7 17.11.21 HF Output of LPI_CON_CI_MAX for ICON-EU also on regular grid

v
Simulations are believed by no one except those who
conducted them.

Experimental results are believed by everyone except

those who conducted them.
ANONYMOUS
Contents
1. Introduction 1

2. History of model changes 3

3. Grid geometry 9
3.1. Horizontal grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2. Vertical grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3. Refined subregion over Europe (“local nest”) . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.4. ICON-D2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4. Mandatory input fields 17

4.1. Grid Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.1.1. Download of Predefined Grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2. External Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

5. Analysis fields 21
5.1. Ensemble Data Assimilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.2. Incremental analysis update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.3. Initial state for the convection-resolving ICON-D2 . . . . . . . . . . . . . . . . . . . . . 26
5.3.1. Interpolated initial conditions and initialisation . . . . . . . . . . . . . . . . . . . 26
5.3.2. Data assimilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

6. Output fields of the ICON model: General description 29

6.1. Available output fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.1.1. Time-constant (external parameter) fields . . . . . . . . . . . . . . . . . . . . . . 30
6.1.2. Multi-level fields on native hybrid vertical levels . . . . . . . . . . . . . . . . . . 33
6.1.3. Multi-level fields interpolated to pressure levels . . . . . . . . . . . . . . . . . . . 34
6.1.4. Single-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.1.5. Lake-specific single-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6.1.6. Soil-specific multi-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.1.7. Output fields for soil moisture analysis SMA . . . . . . . . . . . . . . . . . . . . 41

7. Extended description of available output fields 43

7.1. Cloud products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7.2. Atmospheric products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7.3. Near surface products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.3.1. General comment on statistically processed fields . . . . . . . . . . . . . . . . . . 44
7.4. Surface products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.5. Soil products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.6. Vertical Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

8. Remarks on statistical processing and horizontal interpolation 51

8.1. Statistically processed output fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.1.1. Time-averaged fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.1.2. Accumulated fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
8.1.3. Extreme value fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
8.2. Technical Details of the Horizontal Interpolation . . . . . . . . . . . . . . . . . . . . . . 54

vii
Contents

9. Global output fields 55

9.1. Time-constant (external parameter) fields . . . . . . . . . . . . . . . . . . . . . . . . . . 56
9.2. Multi-level fields on native hybrid vertical levels . . . . . . . . . . . . . . . . . . . . . . . 59
9.3. Multi-level fields interpolated to pressure levels . . . . . . . . . . . . . . . . . . . . . . . 61
9.4. Single-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
9.5. Soil-specific multi-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

10.EU Nest output fields 73

10.1. Time-constant (external parameter) fields . . . . . . . . . . . . . . . . . . . . . . . . . . 75
10.2. Multi-level fields on native hybrid vertical levels . . . . . . . . . . . . . . . . . . . . . . . 76
10.3. Multi-level fields interpolated to pressure levels . . . . . . . . . . . . . . . . . . . . . . . 78
10.4. Single-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
10.5. Soil-specific multi-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

11.ICON-D2 output fields 89

11.1. Time-constant fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
11.2. Multi-level fields on native hybrid vertical levels . . . . . . . . . . . . . . . . . . . . . . . 91
11.3. Multi-level fields interpolated to pressure levels . . . . . . . . . . . . . . . . . . . . . . . 91
11.4. Single-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
11.5. Soil-specific multi-level fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

12.Ensemble forecasts with ICON (global, nested, and limited area mode) 99
12.1. Initial Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
12.2. Ensemble Physics Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
12.3. Lateral boundary perturbations (limited area EPS) . . . . . . . . . . . . . . . . . . . . . 102
12.4. ICON Ensemble output fields in DWD databases . . . . . . . . . . . . . . . . . . . . . . 102
12.4.1. Model Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
12.4.2. Ensemble Products for the ICON-EPS (global) and ICON-EU-EPS . . . . . . . . 103
12.4.3. Ensemble Products for the ICON-D2-EPS . . . . . . . . . . . . . . . . . . . . . . 108

13.ICON data in the SKY data bases of DWD 113

13.1. SKY categories for ICON . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
13.2. Retrieving ICON data from SKY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
13.2.1. Deterministic products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
13.2.2. Ensemble products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

Appendix A. ICON standard level heights 117

A.1. Level heights for zero topography height . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
A.2. Non-zero topography heights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

Appendix B. Output on rotated latitude-longitude grids in ICON-D2 125

Bibliography 127

Glossary 129

viii
1. Introduction
The ICOsahedral Nonhydrostatic model ICON is the global and regional numerical weather prediction
model at DWD. It became operational at 2015-01-20, replacing the former operational global model
GME. In June 2015 a refined sub-region (nest) over Europe was activated (ICON-EU), in order
to replace the regional model COSMO-EU. On 2021-02-10 the convection-permitting model setup
ICON-D2 (i.e. using the limited-area mode of ICON) has replaced COSMO-D2. Since that date, the
entire NWP system at DWD is based on ICON.
The ICON modelling system as a whole is developed jointly by DWD and the Max-Planck Institute for
Meteorology in Hamburg (MPI-M). While ICON is the new working horse for short and medium range
weather forecast at DWD, it embodies the core of a new climate modelling system at MPI-M.
ICON analysis and forecast fields serve as initial and boundary data for a couple of different limited
area models: Since 2015-01-20, analysis and forecast fields of the deterministic forecast run at 13 km
horizontal resolution serve as initial and boundary data for
• RLMs (Relocatable Local Model) of the German armed forces,
• DWD’s wave models.
ICON-D2 (-EPS) is driven by the deterministic (ensemble) forecasts of the ICON-EU nest.
This document provides some basic information about ICON’s grid structure, data assimilation system,
numerical algorithms (see also Zängl et al. (2015)) and physical parameterizations (the latter two are
planned but not yet available). Furthermore, it provides an overview about the ICON analysis and
forecast fields stored in the database SKY at DWD. Some examples on how to read these data from
the database are given as well.

If you encounter bugs or inconsistencies, or if you have suggestions for improving this document, please
contact one of the following colleagues:

Daniel Reinert, FE13 Helmut Frank, FE13 Florian Prill, FE13

Tel: +49 (69) 8062-2060 Tel: +49 (69) 8062-2742 Tel: +49 (69) 8062-2727
Mail: [email protected] Mail: [email protected] Mail: [email protected]

1
2. History of model changes

The forecasting environment, which is composed of the ICON model and the data assimilation system,
is subject to continuous improvements and modifications. The most important ones in terms of forecast
quality and output products are depicted below. For additional information, the reader is referred to
the official change notifications which are available from

http://www.dwd.de/DE/fachnutzer/forschung_lehre/numerische_wettervorhersage/nwv_
aenderungen/nwv_aenderungen_node.html

Alternatively you can click on the timeline-date to see the corresponding change notification.

2015-01-20 First operational ICON forecast run at a horizontal resolution of 13 km and 90

vertical levels. Model top at 75 km

2015-02-25 4 additional forecast runs 03, 09, 15, 21 UTC up to 30 h lead time. Maximum
forecast lead times for 06 and 18 UTC runs extended to 120 h

2015-03-04 Improved wind gust diagnostic for mountainous regions

2015-07-07 Time interval over which max/min temperatures TMAX_2M, TMIN_2M are col-
lected changed to 6 h (formerly 3 h). Time interval over which maximum wind
gusts VMAX_10M are collected changed to 1 h (formerly 3 h).

2015-07-21 Launch of the ICON-EU nest with a horizontal resolution of 6.5 km and 50 ver-
tical levels. Model top at 50 km.

2015-09-02 Assimilation of selected microwave satellite radiance channels over land

(AMSU-A and ATMS), which have so far been assimilated only over the
oceans.

2015-12-01 Surface tile approach for land, sea-ice and lake points, including snow-tiles.
Each grid point can have up to 5 tiles consisting of 3 land tiles (dominant land-
use types), a lake tile or a sea-water plus sea-ice tile. Land tiles may have addi-
tional snow tiles.

2016-01-20 Launch of the ensemble data assimilation system (LETKF, Localized Ensem-
ble Transform Kalman Filter), providing a 40 member analysis ensemble at
R2B6N7. For deterministic forecasts the 3D-Var assimilation system is re-
placed by En-Var (Ensemble Variational analysis system) which makes use of
the ensemble-based model error covariances.

3
Chapter 2. History of model changes

2016-04-13 Improved version of the ICON model which

• accounts for oceanic salt content in the saturation vapour pressure
computation.
• contains modifications to the detrainment tendencies from the convection
parameterization in order to reduce spurious drizzle.
• makes use of the aerosol climatology for the computation of cloud droplet
number concentrations in the radiation parameterization.

2016-04-20 The pressure levels for regular grid output have been revised. Newly available
levels: 650, 550, 450, 350, 275, 225, 175, 125 hPa. Deprecated levels: 725, 20, 7,
3, 0.3 hPa.

2016-09-28 New version of the ICON model in which

• the convection parameterization has been improved with respect to
mixed-phase cloud processes
• the process of rain water evaporation below cloud base has been updated.

2016-11-30 New version of the ICON model in which

• the incremental analysis update (IAU) scheme has been improved
• the diagnostic relative humidity at 2 m above ground is used in the
assimilation system instead of the prognostic one at 10 m above ground.

2017-01-31 Modifications to the ICON-internal lat-lon interpolation for temperature and

specific humidity. Changes to the data assimilation system, i.e.
• bias correction for scatterometer data
• improvements regarding the usage of wind measurements from buoys

2017-03-15 New version of the ICON model:

• updated set of external parameters w.r.t. model orography
• reduced orography filtering for global domain
• new surface evaporation scheme and activation of interception storage
• regular update of the sea surface temperature (SST) by climatological
increments
• a limiter is applied to SSO-scheme wind tendencies from the middle
stratosphere onwards
Updates to the data assimilation scheme:
• Usage of Meteosat-8 Atmospheric motion vector (AMV) winds (instead of
Meteosat-7 AMVs)

2017-05-10 New version of the ICON model, which includes

• prognostic sea-ice albedo
• new ozone climatology over antarctica
• SSO scheme retuning
• usage of ice-coverage observations for the great lakes
• improved snow-analysis w.r.t. to small-scale structures

4
2017-09-29 Usage of high-resolution (0.083◦ ) NCEP sea surface temperature SST analysis

2017-10-11 Updated weather interpretation (WW) code tables

2017-10-25 Modifications w.r.t. the usage of 2 m relative humidity and 10 m winds from
SYNOP stations over land in the global data assimilation system.

2017-11-29 Usage of Dual-Metop Atmospheric Motion Vector (AMV) winds in the global
data assimilation system.

2018-01-17 Launch of the global ICON-EPS (Ensemble Prediction System) with 40 mem-
bers and 40(20) km horizontal resolution globally (over Europe).

2018-03-14 New version of the ICON model and improvements of the assimilation system,
which include
• assimilation of satellite moisture channels (IASI and MHS) over water
• updated RTTOV library (RTTOV12)
• migration from GRIB Master Tables Version 11 to Version 19
• updated model topography
• several measures for reducing the model bias for 2 m humidity and
temperature.

2018-04-25 Usage of FY-3C GPS radio occultation data in the global data assimilation
system.

2018-06-06 New version of the ICON model and improvements of the assimilation system,
which include
• tuning of the grid-scale microphysics and convection scheme in order to reduce
systematic temperature biases in the tropics
• improved version of the preprocessing tool for radiances (sat_pp)

2018-07-11 New version of the ICON model and improvements of the assimilation system,
which include
• tuning of the convection scheme in order to reduce systematic temperature
biases in the tropics
• OSTIA SST-analysis replacing the NCEP SST-analysis
• new output fields on global and nested triangular grid for vv = 0 h:
ALB_SEAICE, SMI, EVAP_PL
• new output field on global triangular grid for all output dates: CAPE_ML

2018-09-13 Additional ensemble products

• ensemble products for the global ICON-EPS are now available from database
category icrgle_main_fcprod_rout

5
Chapter 2. History of model changes

2018-09-19 New version of the ICON model and improvements of the assimilation system,
which include
• correction for misclassified glacier points
• fixed implementation of stratospheric background aerosol
• reduced error in the atmospheric water budget
• improved representation of (partial) snow coverage
• new output fields on global and nested triangular grid for vv = 0 h: SNOWAG,
HSNOW_MAX
• usage of high-resolution radiosonde data in BUFR-format

2018-12-05 Advances to the assimilation of humidity sensitive microwave radiances of

ATMS and MHS.
• activated assimilation over land for selected channels

2019-01-22 New version of the ICON model.

• Interpolation method for lat-lon output changed from RBF-interpolation to
barycentric interpolation. This applies to all time-averaged and accumulated
output fields.
• Bug fix in aerosol climatology.
• Efforts to reduce differences in atmospheric temperature biases between
deterministic and ensemble forecasts.

2019-02-27 Updates to the data assimilation scheme and ICON:

• taking into account error correlations during assimilation of IASI data
• assimilation of ATMS microwave radiances from NOAA-20
• ICON: tuning of molecular roughness length over water

2019-04-10 Updates to the data assimilation scheme:

• Assimilation of observations from Metop-C (i.e. AMSU-A and MHS radiances,
GRAS radio occultations, ASCAT scatterometer winds and AVHHR
atmospheric motion vectors)
• Assimilation of observations from ScatSat (i.e. scatterometer winds)

2019-06-04 Updates to the data assimilation scheme:

• Assimilation of humidity sensitive channels from SSMI/S (Special Sensor
Microwave - Image/Sounder) and GMI (Microwave Imager)
• Assimilation of humidity sensitive channels from AHI (Advanced Himawari
Imager), ABI (Advanced Baseline Imager), GOEA-Imager

2019-07-30 New ICON version (2.5.0-nwp0) and improvements to the assimilation scheme:
• improved cloud cover parameterization
• modified diagnostic regarding the height of the 0 degree Celsius isotherm
(HZEROCL)
• blacklist for SST observations from ships and buoys
• usage of high-res dropsonde data in BUFR-format

6
2019-10-22 New ICON version (2.5.0-nwp1) and improvements to the assimilation scheme:
• Limiter for overshooting convective updrafts in the convection scheme
• Tuning for revised cloud cover scheme
• Assimilation of SSMI/S radiances over land
• Inflation of assumed observation errors for IASI
• RTTOV smoothing for new humidity sensitive channels
• Update of data selection for mapped radiances

2019-11-25 First pre-operational model run of the convection-permitting ICON-D2 (in

limited-area mode) at a horizontal resolution of 2.1 km and 65 vertical levels

2019-11-27 Improvements to the assimilation scheme:

• Use of additional coastal surface observation in the U.S.
• First use of radiosonde descent data (from the German research vessel
’Polarstern’)
• Use of radiosonde humidity observations up to 200hPa

2020-02-19 New ICON version (2.5.0-nwp2) and improvements to the assimilation scheme:
• Correction of surface-level air temperature passed to the radiation scheme
• Assimilation of SAPHIR data

2020-04-21 New ICON version (2.6.0-nwp0) and improvements to the assimilation scheme:
• Corrected calculations of evaporation over snow and CAPE/CIN in ICON
• new diagnostic output field in deterministic runs of ICON-EU: CEILING
• Use of GNSS Zenith Total Delay (ZTD) data

2020-05-19 Improvements to the global assimilation system and ICON:

• Use of global Aeolus horizontal line-of-sight (HLOS) winds
• Use of radiosonde descent data over Germany
• Introduction of a skin temperature formulation in TERRA
• Use of analysed 2 m-temperature bias to vary uncertain model parameters
affecting evapotranspiration

2021-06-23 Extension of ICON-D2 forecasts up to 48h.

7
3. Grid geometry

3.1. Horizontal grid

The horizontal ICON grid consists of a set of spherical triangles that seamlessly span the entire sphere.
The grid is constructed from an icosahedron (see Figure 3.1a) which is projected onto a sphere. The
spherical icosahedron (Figure 3.1b) consists of 20 equilateral spherical triangles. The edges of each
triangle are bisected into equal halves or more generally into n equal sections. Connecting the new edge
points by great circle arcs yields 4 or more generally n2 spherical triangles within the original triangle
(Figure 3.2a, 3.2b).

(a) (b)

Figure 3.1.: Icosahedron before (a) and after (b) projection onto a sphere

(a) (b)

Figure 3.2.: (a) Bisection of the original triangle edges (b) More general division into n equal sections

ICON grids are constructed by an initial root division into n sections (Rn) followed by k bisection steps
(Bk), resulting in a RnBk grid. Figures 3.3a and 3.3b show R2B00 and R2B02 ICON grids. Such
grids avoid polar singularities of latitude-longitude grids (Figure 3.3c) and allow a high uniformity in
resolution over the whole sphere.

Throughout this document, the grid is referred to as the “RnBk grid” or “RnBk resolution”. For a

9
Chapter 3. Grid geometry

(a) (b) (c)

Figure 3.3.: (a) R2B00 grid. (b) R2B02 grid. (c) traditional regular latitude-longitude grid with polar
singularities

given resolution RnBk, the total number of cells, edges, and vertices can be computed from

nc = 20 n2 4k
ne = 30 n2 4k
nv = 10 n2 4k + 2

The average cell area ∆A can be computed from

4π re2
∆A = ,
nc

with the earth radius re , and nc the total number of cells. ICON uses an earth radius of

re = 6.371229 · 106 m.

Based on ∆A one can derive an estimate of the average grid resolution ∆x:

5050 · 103
r
p π re
∆x = ∆A = k
≈ [m]
5 n2 n 2k

Visually speaking, ∆x is the edge length of a square which has the same area as our triangular cell.

In Table 3.1, some characteristics of frequently used global ICON grids are given. The table contains
information about the total number of triangles (nc ), the average resolution ∆x, and the maximum/min-
imum cell area. The latter may be interpreted as the area for which the prognosed meteorological quan-
tities (like temperature, pressure, . . . ) are representative. Some additional information about ICON’s
horizontal grid can be found in Wan et al. (2013).

10
 3.2. Vertical grid

Table 3.1.: Characteristics of frequently used global ICON grids. ∆Amax and ∆Amin refer to the maximum
and minimum area of the grid cells, respectively.

Grid number of cells avg. resolution [km] ∆Amax [km2 ] ∆Amin [km2 ]
(nc )

R2B04 20480 157.8 25974.2 18777.3

R2B05 81920 78.9 6480.8 4507.5
R2B06 327680 39.5 1618.4 1089.6
R2B07 1310720 19.7 404.4 265.1
R3B07 2949120 13.2 179.7 116.3

The operational deterministic version of ICON is based on the R3B07 grid (∆x ≈
13 km), while the ensemble version (ICON-EPS) is based on the R2B6 grid (∆x ≈ 40 km)

3.2. Vertical grid

The vertical grid consists of a set of vertical layers with height-based vertical coordinates. Each of these
layers carries the horizontal 2D grid structure, thus forming the 3D structure of the grid. Close to
the surface the vertical layers are terrain following, while with increasing distance from the surface the
terrain signal is smoothed out and they gradually evolve into layers of constant height. The ICON grid
employs a Lorenz-type staggering with the vertical velocity defined at the boundaries of layers (half
levels) and the other prognostic variables in the center of the layer (full levels) (see Figure 3.4).
To improve simulations of flow past complex topography, the ICON model employs a smooth level
vertical (SLEVE) coordinate (Leuenberger et al., 2010). It allows for a faster transition to smooth
levels in the upper troposphere and lower stratosphere, as compared to the classical height-based Gal-
Chen coordinate. In the operational setup (deterministic and ensemble), the transition from terrain
following levels in the lower atmosphere to constant height levels is completed at z = 16 km. Model
levels above are flat. The required smooth large-scale contribution of the model topography is generated
by digital filtering with a ∇2 -diffusion operator.
Figure 3.5 shows the (half) level heights and layer thicknesses of the operational ICON setup with 90
vertical levels. This figure applies to the deterministic and the ensemble system, as both share the same
vertical grid.The table to the right shows the height above ground of selected half levels (for zero height
topography) and the corresponding pressure, assuming the US standard atmosphere. Standard heights
for all 91 half levels are given in Table A.1.
Please note that for grid cells with non-zero topography these values only represent rough
estimates of the true level height. Actual heights and layer thicknesses may vary consid-
erably from location to location, due to grid level stretching/compression over non-zero
topography (see e.g. the layer compression which is visible in Figure 3.4).

3.3. Refined subregion over Europe (“local nest”)

ICON has the capability for running global simulations with refined domains (so called nests). The
triangular mesh of the refined area is generated by bisection of triangles in the global “parent” grid,

11
Chapter 3. Grid geometry

half level 1
full level 1
half level 2
w
full level 2
vn , ρ, θv
half level 3
full level 3 w
half level 4

full levels half levels

Figure 3.4.: Illustration of ICON’s vertical levels. With num_lev layers, there are num_lev + 1 so-called
half levels. The half levels k − 1/2, k + 1/2 enclose layer k at the centers of which are the
corresponding full levels k, for k = 1, . . . , num_lev. Layer 1 is at the top of the atmosphere
and layer n at the bottom of the atmosphere. Half level num_lev + 1 coincides with the Earth’s
surface.

see Fig. 3.6. In the vertical the global grid extends into the mesosphere (which greatly facilitates the
assimilation of satellite data) whereas the nested domains extend only into the lower stratosphere in
order to save computing time. For the same orography the heights of levels 1–60 of the Europe nest are
the same as those of levels 31–90 of the global grid. In practice, however, near surface level heights of
nests and the global domain differ due to the fact that the underlying orography differs, with deeper
slopes and higher summits in the high resolution nests.
For each nesting level, the time step is automatically divided by a factor of two. Note that the grid
nests are computed in a concurrent fashion:
• Points that are covered by the refined subdomain additionally contain data for the global grid
state.
• The data points on the triangular grid are the cell circumcenters. Therefore the global grid data
points are closely located to nest data sites, but they do not coincide exactly (see Fig. 3.6).
Simulation on the global grid and the nested domain(s) are tightly coupled (two-way nesting): Bound-
ary data for the nest area is updated every time step (120 s/360 s in case of the operational determinis-
tic/ensemble system). Feedback of atmospheric prognostic variables (except precipitation) is computed
via relaxation on a 3 h time scale.
The operational ICON has one refined subregion over Europe (ICON-EU). Key figures like edge coor-
dinates and mesh size of the ICON-EU nest are given in Table 3.2. The geographical location of the
nest is visualised in Fig. 3.7.

Model simulations including the nested region over Europe are running regularly,
starting from

2015-07-21, 06 UTC (roma) for ICON-EU

12
 3.4. ICON-D2

2,600 level [m] [P a]

70,000
2,400 1 75 000 2.1
5 64 946 10.0
2,200
60,000
10 53 878 46.3
z [m] ∆z [m] 2,000 15 44 198 158.8
20 35 958 487.2
1,800
50,000 25 29 039 1 355.0
1,600 30 23 409 3 211.8
35 19 202 6 209.4
40,000 1,400
40 16 108 10 113.6
1,200 45 13 822 14 504.3
30,000 50 11 822 19 882.1
1,000
55 9 822 27 166.6
800
60 7 822 36 528.6
20,000
600 65 5 822 48 347.1
70 3 954 62 009.2
400
10,000 75 2 432 75 325.6
200 80 1 255 87 126.2
85 436 96 190.0
0 0
20 40 60 80 90 20 101 085.0
level

Figure 3.5.: Vertical half levels (blue) and layer thickness (red) of the ICON operational setup (deterministic
and ensemble). The table of selected pressure values (for zero height) is based on the 1976 US
standard atmosphere.

2018-01-17, 06 UTC (roma) for ICON-EU-EPS

Main forecasts starting at 00, 06, 12, 18 UTC reach out to 120 h, while additional short-
range forecasts starting at 03, 09, 15, 21 UTC provide data until +30 h.

3.4. ICON-D2
The horizontal domain of the regional model system ICON-D2 is depicted in Fig. 3.8. It is a regional
R19B07 grid with 542040 cells and a horizontal resolution of ∆x ≈ 2 km. The model domain completely
covers the areas of Germany, Switzerland, Austria, Denmark, Belgium and the Netherlands and parts of
the neighbouring countries. The same horizontal grid is used for the ensemble system ICON-D2-EPS.
The grid frame which is highlighted in Fig. 3.8 depicts the lateral boundary zone where the model is
forced by externally specified data. The lateral boundary zone has a total width of 14 cell rows. It
consists of two sub-zones which are named interpolation zone (rows 1-4) and nudging zone (rows 5-14).
Rows are counted positive towards the model interior. The interpolation zone provides the necessary
boundary conditions and contains interpolated forcing data of the driving model. In other words, no
prognostic computations are performed in the outermost 4 cell rows. In the nudging zone, the interior
flow (i.e. the prognostic solution) is nudged towards the data of the driving model (ICON-EU in our
case).
Figure 3.9 shows the (half) level heights and layer thicknesses of the pre-operational ICON-D2 setup.

13
Chapter 3. Grid geometry

Figure 3.6.: ICON grid refinement (zoom view). Blue and red dots indicate the cell circumcenters for the
global (“parent”) and the refined (“child”) domain, respectively.

Table 3.2.: Key figures of the ICON-EU and ICON-EU-EPS nest.

ICON-EU nest ICON-EU-EPS nest

geogr. coordinates 23.5◦ W – 62.5◦ E 23.5◦ W – 62.5◦ E

29.5◦ N – 70.5◦ N 29.5◦ N – 70.5◦ N

mesh size ≈ 6.5 km (R3B08) ≈ 20 km (R2B07)

659156 triangles 75948 triangles
vertical levels 60 levels 60 levels
upper boundary 22.5 km 22.5 km

14
 3.4. ICON-D2

Figure 3.7.: Horizontal extent of the ICON-EU/ICON-EU-EPS nest (greenish blue area) in a cylindrical
equidistant projection.

Figure 3.8.: ICON-D2 domain in cylindrical equidistant projection. It is comprised of a regional R19B07
grid with 542040 cells and a horizontal resolution of ∆x ≈ 2 km. The highlighted frame depicts
the lateral boundary zone where the model is forced by externally specified data sets. It has a
total width of 14 cell rows and consists of two sub-zones named interpolation zone (rows 1-4)
and nudging zone (rows 5-14). The prognostic region starts at cell row 5. In the nudging zone
the prognostic solution is nudged towards the data of the driving model.

15
Chapter 3. Grid geometry

22,000

2,400
20,000

2,200
18,000
2,000
z [m] ∆z [m]
16,000
1,800

14,000
1,600

12,000 1,400

10,000 1,200

1,000
8,000

800
6,000
600
4,000
400

2,000
200

0 0
10 20 30 40 50 60
level

Figure 3.9.: Vertical half levels (blue) and layer thickness (red) of the ICON-D2 operational setup (deter-
ministic and ensemble).

16
4. Mandatory input fields
Several input files are needed to perform runs of the ICON model. These can be divided into three
classes: Grid files, external parameters, and initialisation (analysis) files. The latter will be described
in Chapter 5.

4.1. Grid Files

In order to run ICON, it is necessary to load the horizontal grid information as an input parameter.
This information is stored within so-called grid files. For an ICON run, at least one global grid file
is required. For model runs with nested grids, additional files of the nested domains are necessary.
Optionally, a reduced radiation grid for the global domain may be used.
The unstructured triangular ICON grid resulting from the grid generation process is represented in
NetCDF format. This file stores coordinates and topological index relations between cells, edges and
vertices.
The most important data entries of the main grid file are

• cell (INTEGER dimension)

number of (triangular) cells
• vertex (INTEGER dimension)
number of triangle vertices
• edge (INTEGER dimension)
number of triangle edges
• clon, clat (double array, dimension: #triangles, given in radians)
longitude/latitude of the midpoints of triangle circumcenters
• vlon, vlat (double array, dimension: #triangle vertices, given in radians)
longitude/latitude of the triangle vertices
• elon, elat (double array, dimension: #triangle edges, given in radians)
longitude/latitude of the edge midpoints
• cell_area (double array, dimension: #triangles)
triangle areas
• vertex_of_cell (INTEGER array, dimensions: [3, #triangles])
The indices vertex_of_cell(:,i) denote the vertices that belong to the triangle i. The vertex_of_cell
index array is ordered counter-clockwise for each cell.
• edge_of_cell (INTEGER array, dimensions: [3, #triangles])
The indices edge_of_cell(:,i) denote the edges that belong to the triangle i.
• clon/clat_vertices (double array, dimensions: [#triangles, 3], given in radians)
clon/clat_vertices(i,:) contains the longitudes/latitudes of the vertices that belong to the
triangle i.
• neighbor_cell_index (INTEGER array, dimensions: [3, #triangles])
The indices neighbor_cell_index(:,i) denote the cells that are adjacent to the triangle i.

17
Chapter 4. Mandatory input fields

Figure 4.1.: Screenshots of the ICON download server hosted by the Max Planck Institute for Meteorology
in Hamburg.

• zonal/meridional_normal_primal_edge: (INTEGER array, #triangle edges)

components of the normal vector at the triangle edge midpoints.
• zonal/meridional_normal_dual_edge: (INTEGER array, #triangle edges)
These arrays contain the components of the normal vector at the facets of the dual control volume.
Note that each facet corresponds to a triangle edge and that the dual normal matches the direction
of the primal tangent vector but signs can be different.

4.1.1. Download of Predefined Grids

For fixed domain sizes and resolutions a list of grid files has been pre-built for the ICON model together
with the corresponding reduced radiation grids and the external parameters.
The contents of the primary storage directory are regularly mirrored to a public web site for download,
see Figure 4.1 for a screenshot of the ICON grid file server. The download server can be accessed via

http://icon-downloads.mpimet.mpg.de

The pre-defined grids are identified by a centre number, a subcentre number and a numberOfGridUsed,
the latter being simply an integer number, increased by one with every new grid that is registered in the
download list. Also contained in the download list is a tree-like illustration which provides information
on parent-child relationships between global and local grids, and global and radiation grids, respectively.
Note that the grid information of some of the older grids (no. 23 – 40) is split over two files: The users
need to download the main grid file itself and a grid connectivity file (suffix -grfinfo.nc).

4.2. External Parameters

External parameters are used to describe the properties of the earth’s surface. These data include e.g.
the orography, the land-sea-mask as well as parameters describing soil and surface properties, like the
soiltype or the plant cover fraction.
The ExtPar software (ExtPar – External parameter for Numerical Weather Prediction and Climate
Application) is able to generate external parameters for the ICON model. The generation is based on a
set of raw datafields which are listed in Table 4.1. For a more detailed overview of ExtPar, the reader
is referred to the User and Implementation Guide of Extpar.

18
 4.2. External Parameters

Table 4.1.: Raw datasets from which the ICON external parameter fields are derived.

Dataset Source Resolution

Global Land One-km Base Elevation Project NGDC 30”

(GLOBE)
Global Land Cover Map for 2009 (GlobCover 2009) ESA 10”
Global Land Cover Characteristics (GLCC) USGS 30”
Digital Soil Map of the World (DSMW) FAO 5’
Sea-viewing Wide Field-of-view Sensor (SeaWIFS) NASA/GSFC 2.5’
NDVI Climatotology
Climate Research Unit - Gridded climatology of CRU-UEA 0.5◦
1961-1990 monthly means (CRU-CL)
Global Aerosol Climatology Project (GACP) Aerosol NASA/GISS 4x5◦
Optical thickness
Global Lake Database (GLDB) DWD/RSHU/MeteoFrance 30”
Moderate Resolution Imaging Spectroradiometer NASA 5’
(MODIS) albedo

GlobCover 2009 is a land cover database covering the whole globe, except for Antarctica. Therefore, we
make use of GlobCover 2009 for 90◦ > φ > −56◦ (with φ denoting latitude) and switch to the coarser,
however globally available dataset GLCC for −56◦ ≥ ψ > −90◦ .
The products generated by the ExtPar software package are listed in Table 4.2 together with the
underlying raw dataset. These are mandatory input fields for assimilation- and forecast runs.

Table 4.2.: External parameter fields for ICON, produced by the ExtPar software package (in alphabetical
order)

ShortName Description Raw dataset

AER_SS12 Sea salt aerosol climatology (monthly fields) GACP

AER_DUST12 Total soil dust aerosol climatology (monthly fields) GACP
AER_ORG12 Organic aerosol climatology (monthly fields) GACP
AER_SO412 Total sulfate aerosol climatology (monthly fields) GACP
AER_BC12 Black carbon aerosol climatology (monthly fields) GACP
ALB_DIF12 Shortwave (0.3 − 5.0 µm) albedo for diffuse radiation MODIS
(monthly fields)
ALB_UV12 UV-visible (0.3 − 0.7 µm) albedo for diffuse radiation MODIS
(monthly fields)
ALB_NI12 Near infrared (0.7 − 5.0 µm) albedo for diffuse radiation MODIS
(monthly fields)
DEPTH_LK Lake depth GLDB
Continued on next page

19
Chapter 4. Mandatory input fields

Table 4.2.: continued

EMIS_RAD Surface longwave (thermal) emissivity GlobCover 2009

FOR_D (*) Fraction of deciduous forest GlobCover 2009
FOR_E (*) Fraction of evergreen forest GlobCover 2009
FR_LAKE Lake fraction (fresh water) GLDB
FR_LAND Land fraction (excluding lake fraction but including GlobCover 2009
glacier fraction)
FR_LUC Landuse class fraction
HSURF Orography height at cell centres GLOBE
LAI_MX (*) Leaf area index in the vegetation phase GlobCover 2009
NDVI_MAX Normalized differential vegetation index SeaWIFS
NDVI_MRAT proportion of monthly mean NDVI to yearly maximum SeaWIFS
(monthly fields)
PLCOV_MX Plant covering degree in the vegetation phase GlobCover 2009
(*)
ROOTDP (*) Root depth GlobCover 2009
RSMIN (*) Minimum stomatal resistance GlobCover 2009
SOILTYP Soil type DSMW
SSO_STDH Standard deviation of sub-grid scale orographic height GLOBE
SSO_THETA Principal axis-angle of sub-grid scale orography GLOBE
SSO_GAMMA Horizontal anisotropy of sub-grid scale orography GLOBE
SSO_SIGMA Average slope of sub-grid scale orography GLOBE
T_2M_CL Climatological 2m temperature (serves as lower bound- CRU-CL
ary condition for soil model)
Z0 (*) Surface roughness length (over land), containing a con- GlobCover 2009
tribution from subgrid-scale orography

Note that fields marked with (*) are not required in operational model runs. I.e. the surface roughness
Z0 is only required, if the additional contribution from sub-grid scale orography shall be taken into
account (i.e. for itype_z0=1). In operational runs this is not the case. Instead, land-cover class specific
roughness lengths are taken from a GlobCover-based lookup table. FOR_D, FOR_E, LAI_MX, PLCOV_MX,
RSMIN, and ROOTDP became obsolete with the activation of the surface tile approach (2015-03-04). The
latter 4 fields are replaced by land-cover class specific values taken from lookup tables.

Remarks on post-processing

Some of the external parameter fields are further modified by ICON. The following fields are affected:

DEPTH_LK HSURF FR_LAND FR_LAKE Z0

Thus, for consistency, the modified fields should be used for post-processing tasks rather than the
original external parameter fields. See Section 6.1.1 for more details.

20
5. Analysis fields
Numerical weather prediction (NWP) is an initial value problem. The ability to make a skillful forecast
relies heavily on an accurate estimate of the present atmospheric state, known as the analysis. In
general, an analysis is generated by optimally combining all available observations with a short-range
model forecast, known as first guess (FG) or background. Currently an atmospheric analysis is created
every 3 h. The 3-hourly first guess output provided by ICON comprises the following fields:

Table 5.1.: Available 3h first guess output fields from the forecast database
CAT_NAME=$model_ass_fc_$suite

Type GRIB shortName

Atmosphere VN, U, V, W, DEN, THETA_V, T, QV,

QC, QI, QR, QS, TKE, P
Surface (general) T_G, T_SO(0), QV_S, T_2M, TD_2M,
U_10M, V_10M, PS, Z0
Land specific W_SNOW, T_SNOW, RHO_SNOW,
H_SNOW, FRESHSNW, W_I,
T_SO(1:nlev_soil), W_SO, W_SO_ICE
Lake/sea ice specific T_MNW_LK, T_WML_LK, H_ML_LK,
T_BOT_LK, C_T_LK, T_B1_LK,
H_B1_LK, T_ICE, H_ICE, FR_ICE
Time invariant FR_LAND, HHL, CLON, CLAT, ELON,
ELAT, VLON, VLAT

Atmospheric analysis fields are computed every 3 hours (00, 03, 06,. . . 21 UTC) by the 3DVar data
assimilation system, which has recently been upgraded to an En-Var system (see Section 5.1). Sea
surface temperature T_SO(0) and sea ice cover FR_ICE are provided once per day (00 UTC) by the
SST-Analysis. A snow analysis is conducted every 3 hours, providing updated information on the snow
height H_SNOW and snow age FRESHSNW. In addition a soil moisture analysis (SMA) is conducted once
per day (00 UTC). It basically modifies the soil moisture content W_SO, in order to improve the 2 m
temperature forecast.
For the 3-hourly assimilation cycle and forecast runs, ICON must be provided with 2 input files: One
containing the First Guess (FG) and the other containing analysis (AN) fields, only. Variables for which
no analysis is available are always read from the first guess file (e.g. TKE). Other variables may be read
either from the first guess or the analysis file, depending on the starting time. E.g. for T_SO(0) the
first guess is read at 03, 06, 09, 12, 15, 18, 21 UTC, however, the analysis is read at 00 UTC when a
new SST analysis is available. In Table 5.2 the available and employed first guess and analysis fields
are listed as a function of starting time.

21
Chapter 5. Analysis fields

Table 5.2.: The leftmost column shows variables that are mandatory for the assimilation cycle and forecast
runs. Column 2 indicates, whether or not an analysis is performed for these variables. Columns
3 to 10 show the origin of these variables (analysis or first guess), depending on the starting
time.

ShortName Analysis 00 03 06 09 12 15 18 21

Atmosphere
VN – FG FG FG FG FG FG FG FG
THETA_V – FG FG FG FG FG FG FG FG
DEN – FG FG FG FG FG FG FG FG
W – FG FG FG FG FG FG FG FG
TKE – FG FG FG FG FG FG FG FG
QC, QI, QR, QS – FG FG FG FG FG FG FG FG
QV 3DVar AN AN AN AN AN AN AN AN
T 3DVar AN AN AN AN AN AN AN AN
P 3DVar AN AN AN AN AN AN AN AN
U, V 3DVar AN AN AN AN AN AN AN AN
Surface
Z0 – FG FG FG FG FG FG FG FG
T_G – FG FG FG FG FG FG FG FG
QV_S – FG FG FG FG FG FG FG FG
T_SO(0) (SST only) Ana_SST AN FG FG FG FG FG FG FG
T_SO(0:nlevsoil) – FG FG FG FG FG FG FG FG
W_SO_ICE – FG FG FG FG FG FG FG FG
W_SO SMA AN FG FG FG FG FG FG FG
W_I – FG FG FG FG FG FG FG FG
W_SNOW1 Ana_SNOW AN AN AN AN AN AN AN AN
T_SNOW – FG FG FG FG FG FG FG FG
RHO_SNOW1 Ana_SNOW AN AN AN AN AN AN AN AN
H_SNOW Ana_SNOW AN AN AN AN AN AN AN AN
FRESHSNW Ana_SNOW AN AN AN AN AN AN AN AN
SNOWC – FG FG FG FG FG FG FG FG
Sea ice/Lake
T_ICE – FG FG FG FG FG FG FG FG
H_ICE – FG FG FG FG FG FG FG FG
FR_ICE Ana_SST AN FG FG FG FG FG FG FG
Continued on next page

22
 5.1. Ensemble Data Assimilation

Table 5.2.: The leftmost column shows variables that are mandatory for the assimilation cycle and forecast
runs. Column 2 indicates, whether or not an analysis is performed for these variables. Columns
3 to 10 show the origin of these variables (analysis or first guess), depending on the starting
time.

ShortName Analysis 00 03 06 09 12 15 18 21

T_MNW_LK – FG FG FG FG FG FG FG FG
T_WML_LK – FG FG FG FG FG FG FG FG
H_ML_LK – FG FG FG FG FG FG FG FG
T_BOT_LK – FG FG FG FG FG FG FG FG
C_T_LK – FG FG FG FG FG FG FG FG
T_B1_LK – FG FG FG FG FG FG FG FG
H_B1_LK – FG FG FG FG FG FG FG FG

5.1. Ensemble Data Assimilation

Until 2016-01-20 the analyses were derived by a 3-hourly cycled 3-dimensional data assimilation system
(3D-Var).
From 2016-01-20 on the analysis system consists of the 3-hourly cycled Ensemble Variational Data
assimilation system (En-Var) providing initial fields for the deterministic ICON forecasts at 13 km reso-
lution, based on the 3-hour short range forecast (first guess) and the observations at the actual analysis
time. In the En-Var a part of the background error covariance matrix is derived from the statistics of a
3-hour short range ensemble forecast at lower resolution (currently 40 members at 40 km R2B06 reso-
lution with a 20 km nest over Europe). The En-Var deterministic analysis system is complemented by
an Ensemble Data Assimilation system (EDA), in the specific implementation of a Localized Ensemble
Transform Kalman Filter (LETKF). The EDA provides the initial fields for the 3-hourly cycled ICON
short range ensemble forecasts.
Both the deterministic and the ensemble data assimilation provide atmospheric analyses and analysis
increments as described in Table 5.2 and Section 5.2. However, The Ensemble Data assimilation cur-
rently does not run separate analyses for sea surface temperature, snow, and soil moisture. Instead these
fields are derived from the deterministic forecast and provided 3-hourly by the EDA in the following
way:
Sea Surface Temperature The sea surface temperature at ensemble resolution is interpolated (taking
the nearest neighbour) from the deterministic sea surface temperature. Ice fraction, ice height,
and ice temperature are taken from the deterministic first guess as well. As a SST analysis is
run once a day in the deterministic forecast system this mechanism ensures that the ensemble sea
surface temperature stays close to the observed one.
In addition the interpolated sea surface temperature is perturbed individually for each ensem-
ble member with prescribed spacial and temporal correlation length scales to account for the
uncertainties in the SST analysis.

1 Note that RHO_SNOW is read from the analysis, however it does not contain any new/independent information compared
to the model first guess, except for an initialisation of newly generated snow points and a limitation over glacier points.
W_SNOW is read from the analysis, too, however it is re-diagnosed within the ICON-code based on the analysed snow
height H_SNOW and the former mentioned snow density RHO_SNOW.

23
Chapter 5. Analysis fields

Soil Moisture
The ensemble mean of soil moisture is adjusted to its value in the deterministic run. This procedure
ensures that the mean ensemble soil moisture stays close to the analysed one, as a soil moisture
analysis is run once a day in the deterministic forecast system. By adjusting only the ensemble
mean the ensemble spread is preserved.
Snow
For each ensemble member the mean ensemble snow cover is adjusted to its deterministic value.
The data assimilation system also provides a couple of fields, which are not modified with respect to
their guess values, so that a full set of nominal analysis fields is available.

Table 5.3.: Fields provided by the ensemble analysis system. The column Increment indicates if an analysis
increment is provided. Analysis indicates if the field is analysed by the LETKF (letkf), taken from
the first guess (fg), interpolated (det) from, or (mean) adjusted to the respective deterministic
quantity, or additionally perturbed (per).

ShortName Type Increment Analysis

T Atmosphere yes letkf Temperature

U " yes letkf U-Component of Wind
V " yes letkf V-Component of Wind
QV " yes letkf Specific Humidity
P " yes letkf Pressure
QC " letkf Cloud Mixing Ratio
QI " letkf Cloud Ice Mixing Ratio
QR " fg Rain Mixing Ratio
QS " fg Snow Mixing Ratio
H_SNOW Snow yes mean Snow Depth
FRESHSNW " yes mean Fresh snow factor
QV_S Surface fg Surface Specific Humidity
W_I " fg Plant Canopy Surface Water
Z0 " fg Surface Roughness length
T_SO(0) Sea surface det+per Sea Surface Temperature
H_ICE " det Sea Ice Thickness
FR_ICE " det Sea Ice Cover
W_SO Soil yes mean Soil moisture
W_SO_ICE " fg Soil ice content
T_SO " fg Soil temperature

24
 5.2. Incremental analysis update

5.2. Incremental analysis update

Analysis fields provided by the data assimilation system are usually not perfectly balanced, leading
to e.g. the generation of spurious gravity waves. Thus, atmospheric models generally require some
initialisation procedure in order to minimize spin-up effects and to prevent the accumulation of noise.
In ICON, a method known as Incremental Analysis Update (IAU) (Bloom et al., 1996, Polavarapu
et al., 2004) is applied. The basic idea is quite simple: Rather than adding the analysis increments
∆xA = xA − xF G ( i.e. the difference between the analysis xA and the model first guess xF G ) in one
go, they are incorporated into the model in small drips over many timesteps (see Figure 5.1).

Figure 5.1.: Incremental Analysis Update. Analysis increments are added to the background state (FG) in
small drips over some time interval rather than in one go. Currently, increments for U, V, P, T,
QV are treated in this way.

Mathematically speaking, during forward integration the model is forced with appropriately weighted
analysis increments:
Z
dx A
= Ax + g(t)∆x , with g(t) dt = 1 (5.1)
dt

x is the discrete model state, A is a matrix representing the (non)-linear dynamics of the system and
g(t) is a weighting function, which is non-zero over some time-interval ∆t.
This drip by drip incorporation acts as a low pass filter in frequency domain on the analysis increments
such that small scale unbalanced modes are effectively filtered (see Bloom et al. (1996)). The filter
characteristic depends on the weighting function g(t). It should be noted that IAU only filters the
increments and not the background state, such that regions where analysis increments are zero remain
unaffected. This method is currently applied to the prognostic atmospheric fields π, ρ, vn , qv , based on
analysis increments provided for u, v, p, t and qv . π denotes the Exner pressure.
The method sounds incredibly simple, however there are a few technical aspects to be taken care of when
implementing this into an operational system: Figure 5.2 shows how the IAU-method is implemented
in ICON for a 3 h assimilation run starting at midnight. Analysis increments are applied over a 3 h time
window, centered at the actual model start time. As indicated by the blue line, constant weights are
used:
∆t
g(t) = , for − T /2 < t < T /2 (5.2)
T

25
Chapter 5. Analysis fields

T is the window width and ∆t is the fast physics time step. The key point in terms of technical
implementation is that the model must be started 90 minutes prior to the actual starting time of the
assimilation run. The model is started from the 22:30 UTC first guess. The analysis increments for
U, V, P, T, QV, whose validity time is 00:00 UTC are added over 3 hours until at 1:30 the free forecast
starts. Then, two first guess data sets are written into the database. One at 1:30 UTC, which will be
used for starting the next 3 h assimilation run, and a second one at 3:00 UTC, which serves as input
for the assimilation system itself. Thus in general, using the IAU method requires some care in terms
of reading and writing the right fields at the right times.

Figure 5.2.: Time line for an ICON assimilation run starting at 00:00 UTC.

This method is not restricted to atmospheric fields, but also applicated to assimilated soil and surface
fields, specifically soil moisture W_SO, and snow quantities H_SNOW and FRESHSNW.

5.3. Initial state for the convection-resolving ICON-D2

5.3.1. Interpolated initial conditions and initialisation

For experimental predictions and simulations with the ICON-D2 the initial state can be determined by
interpolation from the analysis of a driving model (normally ICON / ICON-EU, or also the IFS). With
interpolated initial conditions one should generally note that the calculated initial state is not very
well defined due to the difference in the horizontal and vertical resolution. Therefore, a settling period
occurs (spin-up, approx. 3–6 hours), during which the flow adjusts to the high-resolution topography1 .

5.3.2. Data assimilation

KENDA-LETKF

The initial conditions will be generated with the data assimilation system KENDA (’Kilometer-scale
Ensemble Data Assimilation’, Schraff et al. (2016)) which is based on the method of the “Local Ensemble

1 The digital filter initialisation (DFI) by Lynch (1997) used earlier, is difficult to adapt for the 2-time-level based
integration used in ICON-D2 according to previous experience.

26
 5.3. Initial state for the convection-resolving ICON-D2

Transform Kalman Filter” (LETKF, Hunt et al. (2007)). With this method, one can simultaneously and
consistently provide initial conditions both for the deterministic ICON-D2 in the form of a deterministic
analysis as well as for the COSMO-D2-EPS using a whole ensemble of suitably disturbed analyses.
For the calculation of the analysis, the information is combined from the current observations and the
previous short-term forecast, in the case of the current configuration of KENDA this is a 1-hour forecast.
The weighting of these components is based on the estimation of the respective uncertainties, where the
errors of the predictions in particular on the convective scale depends to a high degree on the situation
and weather. In KENDA, these forecasting uncertainties can be estimated with the help of an ensemble
from (currently) 40 appropriately slightly different 1-hour forecasts. The analysis procedure allows
not only to estimate the most likely actual state of the atmosphere, but also the analysis error. This
estimate is used in the generation of an entire ensemble of different analyses (with the same number
as the incoming prediction Ensemble) in such a way that the analysis ensemble mean corresponds to
the most likely current state and that the spread of the analysis ensemble corresponds to the estimated
analysis error. As a result, the estimation of the analysis error influences directly the spread of the
subsequent ensemble forecasts (in the data assimilation cycle or as an actual short-term forecast ICON-
D2-EPS), which in turn serves as a measure of the forecasting uncertainty. However, it should be
mentioned that the ensemble spread generally only describes random errors, but not the systematic
analysis errors or forecast errors.
To sufficiently take into account the uncertainty in the heat flux from the surface of the earth into
the atmosphere, additional explicit random errors in the sea surface temperature and soil moisture
are applied, so that ensemble members have a spread of 1 K or approx. 15 % relative soil moisture
(between wilting point and field capacity). Without these disturbances, the ensemble spread and thus
the estimate of the uncertainty of analysis and prediction in the planetary boundary layer would become
underestimated.
Because the analysis ensemble mean in the atmosphere is not a very well balanced model state, and,
as used as an initial state for a deterministic forecast, would lead to a slightly increased spin-up in
the first forecast hours, an additional undistorted model run (’Control Run’ or ’deterministic run’) is
determined. Based on this 1 hour forecast (’deterministic first guess’), the deviations of the observations
from this run, and the estimation of the forecast errors from the LETKF (in the form of the ’Kalman
Gain’ for the ensemble mean) a ’deterministic’ analysis is calculated. This serves as the initial condition
for the subsequent ’Control Run’ or for the actual deterministic short-term prediction.
The so-called ’control variables’ of the LETKF, i.e. the variables which are changed (’analysed’) by
LETKF are currently: 3-D wind components, temperature, specific humidity, cloud water, and cloud
ice on all model levels, as well as pressure at the bottom model surface. In the areas above it, the pressure
is adjusted in such a way that the entire analysis corrections (analysis increments) are hydrostatically
balanced. For all other model variables, the analysis is just the 1 hour forecast (’First Guess’).
Only conventional observations are currently assimilated in the LETKF (data from radio sondes, aircraft,
wind profilers, ground stations). To assimilate precipitation rates from radar data, the LETKF is
combined with ’Latent Heat Nudging’ (LHN). This is done by application of LHN between the LETKF
analysis steps and at the beginning of the predictions in the deterministic model run.

Latent Heat Nudging

In order to have a sufficiently good forecast quality of precipitation, especially in the short range, the
use of radar-based precipitation information is essential for the determination of a reasonable initial
condition. Currently, quality proven products of near-ground precipitation rates are used in a temporal
resolution of 5 minutes and a horizontal resolution of 1 km × 1 km from the DWD radar network and
foreign radar stations. These data are aggregated to the ICON-D2 model grid and are brought in GRIB
format as precipitation analyses into the database. With the help of the “latent heat nudging” method
these radar precipitation data are assimilated during the forward integration of the (ICON) model into
the model state (Stephan et al., 2008). To do this, one determines temperature increments from the
ratio between observed and modelled precipitation as well as from model based latent heating rates. The

27
Chapter 5. Analysis fields

temperature changes take place while maintaining the relative humidity, whereby the specific humidity
is adjusted accordingly. The increments introduced influence the dynamics of the model in that the
model precipitation adjusts to the observation.

Soil moisture analysis

The soil moisture is adapted by relaxing the soil moisture index (SMI) of ICON-D2 towards the SMI
of ICON-EU (which uses a soil moisture analysis, see above).

Further external analysis

Once a day there is an analysis of the sea surface temperature carried out. Based on the previous
analysis as ’first guess’, the new analysis is produced by using all observations from ships and buoys
of the previous 2 days with the aid of a correction procedure. In low data areas this is complemented
through the global analysis, based on the analysis by NCEP, which is also based on satellite data.
Furthermore, a snow depth analysis is carried out every 6 hours. It is based on a simple weighted
averaging of SYNOP snow depth observations. The weighting depends on the horizontal and vertical
distances the target grid points. In areas with low data density, an attempt is made to derive the snow
depth increments from SYNOP precipitation and temperatures.

28
6. Output fields of the ICON model:
General description
ICON output fields are exclusively available in the Gridded Binary Format, 2nd edition (GRIB2), with
the exception of meteogram data (NetCDF). GRIB is a bit-oriented data storage format which was
developed by World Meteorological Organization (WMO) to facilitate the exchange of large volumes of
gridded data between weather prediction centres.
In GRIB2, a product (i.e. a variable/field) is identified by a set of three parameters
• Discipline (see GRIB2 code table 0.0)
• ParameterCategory (see GRIB2 code table 4.1)
• ParameterNumber (see GRIB2 code table 4.2),
augmented by a large number of additional metadata in order to uniquely describe the nature of the
data. Noteworthy examples of additional metadata are
• typeOfFirstfixedSurface and typeOfSecondFixedSurface (see GRIB2 code table 4.5)
• typeOfStatisticalProcessing, former known as stepType (instant, accum, avg, max, min, diff, rms,
sd, cov, . . . ), describing the statistical process used to calculate the field
just to name a few.
A documentation of the official WMO GRIB2 code tables can be found here:
http://www.wmo.int/pages/prog/www/WMOCodes.html
or
http://www.wmo.int/pages/prog/www/WMOCodes/WMO306_vI2/LatestVERSION/WMO306_vI2_GRIB2_
CodeFlag_en.pdf
In the following tables typeOfFirstFixedSurface and typeOfSecondFixedSurface will be abbreviated by
Lev-Typ 1/2.
For decoding and encoding GRIB2 messages, the DWD in general and ICON in particular makes use
of the ecCodes package developed by ECMWF (it replaces the older GRIB_API). ecCodes includes
both programming interfaces for reading and writing GRIB2- (and the older GRIB1-) data in Fortran-,
C-, and Python-programs and command line tools for analysing and further processing of GRIB-fields.
Examples for the latter are
> grib_ls gribfile
for a listing of the repository of a gribfile, or
> grib_dump gribfile
for extensive information about the single grib fields. To see the meta data in ’pure’ form (i.e. only
so-called ’coded keys’ are displayed) and ordered by GRIB sections then
> grib_dump -O gribfile
should be used.
Further information can be found at
https://software.ecmwf.int/wiki/display/ECC/ecCodes+Home

29
Chapter 6. Output fields of the ICON model: General description

6.1. Available output fields

In the following tables, all output variables of the ICON model are listed together with the most
important GRIB keys for their identification. These tables are of interest in particular for those users,
who don’t use the ecCodes (or GRIB_API) software together with the national DWD GRIB tables.
Please note, that the individual models (ICON global, ICON-EU nest, ICON-D2) deliver only a subset
of these GRIB fields. The concrete output for each model is described in sections 9, 10, and 11.
For details regarding the algorithm for interpolation onto the lat-lon grid, see Section 8.2. In the tables
below, the specific algorithm used for lat-lon interpolation is indicated in the column LL IntpType. If
nothing is specified, then an RBF-based interpolation method is used.

6.1.1. Time-constant (external parameter) fields

Table 6.1 provides an overview of the available time-constant fields. As mentioned in Section 4.2,
there are two types of such variables. The one type is delivered as an external (invariant) field; such
fields are available from the database category CAT_NAME=$model_const_an_$suite. In the later tables
9.1, 10.2, and 11.2 they are denoted by ’invar’. The other type of variables (in particular DEPTH_LK,
HSURF, FR_LAND, FR_LAKE and Z0) is modified by ICON. Thus, the latter should not be taken from
the const_an database category, unless you definitely know what you are doing. For convenience, the
modified invariant fields (and some more) are stored in the forecast database categories for step s[h] = 0
(CAT_NAME=$model_$run_fc_$suite) (such variables are denoted by ’t=0’ in the above mentioned
tables).
See Section 13.1 for more details on the database categories and Section 13 for sample retrievals.

Table 6.1.: Time-constant fields or variables exclusively available for V V = 0 from the forecast databases

LL IntpType
Lev-Typ 1/2
ShortName

Discipline

stepType
Category
Number

Description

Unit

Date/Time (YYYY-MM-DDThh) D=0001-01-01T00

ALB_SEAICE Sea ice albedo 0/19/234 1/– inst – %

CLAT Geographical latitude of native grid 0/191/1 1/– inst – Deg. N
triangle cell center

CLON Geographical longitude of native 0/191/2 1/– inst – Deg. E

grid triangle cell center

DEPTH_LK Lake depth 1/2/0 1/162 inst m

ELAT Geographical latitude of native grid 0/191/1 1/– inst – Deg. N
triangle edge midpoint

ELON Geographical longitude of native 0/191/2 1/– inst – Deg. E

grid triangle edge midpoint

EMIS_RAD Longwave surface emissivity 2/3/199 1/– inst – 1

Continued on next page

30
 6.1. Available output fields

Table 6.1.: continued

EVAP_PL Evaporation of plants (integrated 2/0/198 1/– acc kg m−2

since "nightly reset")

FOR_D Fraction of deciduous forest 2/0/30 1/– inst – 1

(possible range [0, 1])

FOR_E Fraction of evergreen forest 2/0/29 1/– inst – 1

(possible range [0, 1])

FR_LAKE Fresh water lake fraction (possible 1/2/2 1/– inst 1

range [0, 1])

FR_LAND Land fraction (possible range [0, 1]) 2/0/0 1/– inst 1
FR_LUC Land use class fraction (possible 2/0/36 1/– inst – 1
range [0, 1])
HHL Geometric height of model half 0/3/6 150/101 inst m
levels above msl

H_SNOW Snow depth 0/1/11 1/– inst m

HSNOW_MAX Maximum snow height during 0/1/235 1/– inst m
contiguous accumulating snow
period

HSURF Geometric height of the earths 0/3/6 1/101 inst m

surface above msl

LAI Leaf area index 2/0/28 1/– inst 1

LAI_MX Leaf area index in the vegetation 2/0/28 1/– max – 1
phase

NDVIRATIO ratio of current NDVI (normalized 2/0/192 1/– inst – 1

differential vegetation index) to
annual max

NDVI_MAX Normalized differential vegetation 2/0/31 1/– max – 1

index

PLCOV Plant cover 2/0/4 1/– inst %

PLCOV_MX Plant covering degree in the 2/0/4 1/– max – 1
vegetation phase
RLAT Geographical latitude of rotated 0/191/1 1/– inst Deg. N
lat-lon grid cell centers
RLON Geographical longitude of rotated 0/191/2 1/– inst Deg. E
lat-lon grid cell centers

ROOTDP Root depth of vegetation 2/0/32 1/– inst m

RSMIN Minimum stomatal resistance 2/0/16 1/– inst – s m−1
SMI Soil moisture index 2/3/200 106/106 inst 1
SNOAG Snow age 0/1/17 1/– inst d
SOILTYP Soil type of land fraction (9 types 2/3/196 1/– inst NNB 1
[1, . . . , 9])

Continued on next page

31
Chapter 6. Output fields of the ICON model: General description

Table 6.1.: continued

SSO_GAMMA Anisotropy of sub-gridscale 0/3/24 1/– inst – 1

orography

SSO_SIGMA Slope of sub-gridscale orography 0/3/22 1/– inst – 1

SSO_STDH Standard deviation of sub-grid 0/3/20 1/– inst – m
scale orography

SSO_THETA Angle of sub-gridscale orography 0/3/21 1/– inst – rad

T_2M_CL Climatological 2 m temperature 0/0/0 103/– inst – K
(used as lower bc. for soil model)

VLAT Geographical latitude of native grid 0/191/1 1/– inst Deg. N

triangle vertex

VLON Geographical longitude of native 0/191/2 1/– inst Deg. E

grid triangle vertex

Z0 Surface roughness (above land and 2/0/1 1/– inst m

water)

Date/Time (YYYY-MM-DDThh) D=1111-01-11T11

AER_SS12 Sea salt aerosol climatology 0/20/102 1/– avg – 1

(monthly fields)

AER_DUST12 Total soil dust aerosol climatology 0/20/102 1/– avg – 1

(monthly fields)

AER_ORG12 Organic aerosol climatology 0/20/102 1/– avg – 1

(monthly fields)

AER_SO412 Total sulfate aerosol climatology 0/20/102 1/– avg – 1

(monthly fields)
AER_BC12 Black carbon aerosol climatology 0/20/102 1/– avg – 1
(monthly fields)
ALB_DIF12 Shortwave (0.3 − 5.0 µm) albedo for 0/19/18 1/– avg – 1
diffuse radiation (monthly fields)
ALB_UV12 UV-visible (0.3 − 0.7 µm) albedo for 0/19/222 1/– avg – 1
diffuse radiation (monthly fields)

ALB_NI12 Near infrared (0.7 − 5.0 µm) albedo 0/19/223 1/– avg – 1
for diffuse radiation (monthly
fields)
NDVI_MRAT ratio of monthly mean NDVI 0/0/192 1/– avg – 1
(normalized differential vegetation
index) to annual max

32
 6.1. Available output fields

6.1.2. Multi-level fields on native hybrid vertical levels

Table 6.2.: Hybrid multi-level forecast (V V > 0) and initialised analysis (V V = 0) products

LL IntpType
Lev-Typ 1/2
ShortName

Discipline

stepType
Category
Number
Description

Unit
CLC Cloud cover 0/6/22 150/150 inst %
DEN Density of moist air 0/3/10 150/150 inst – kg m−3
DTKE_CON Buoyancy-production of TKE due 0/19/219 150/– inst – m2 s−3
to sub grid scale convection

DTKE_HSH Production of TKE due to 0/19/220 150/– inst – m2 s−3

horizontal shear

P Pressure 0/3/0 150/150 inst Pa

QC Cloud mixing ratio 3
0/1/22 150/150 inst kg kg−1
QG Graupel mixing ratio3 0/1/32 150/150 inst kg kg−1
QI Cloud ice mixing ratio3 0/1/82 150/150 inst kg kg−1
QR Rain mixing ratio3 0/1/24 150/150 inst – kg kg−1
QS Snow mixing ratio3 0/1/25 150/150 inst – kg kg−1
QV Specific humidity 0/1/0 150/150 inst BCT kg kg−1
Q_SEDIM Specific mass of sedimenting 0/1/196 150/150 inst BCT kg kg−1
particles

T Temperature 0/0/0 150/150 inst BCT K

TKE Turbulent kinetic energy 0/19/11 150/– inst m2 s−2
U Zonal wind 0/2/2 150/150 inst m s−1
V Meridional wind 0/2/3 150/150 inst m s−1
W Vertical wind 0/2/9 150/– inst m s−1

3 for the time being, erroneously encoded as mixing ratios instead of specific quantities

33
Chapter 6. Output fields of the ICON model: General description

6.1.3. Multi-level fields interpolated to pressure levels

Table 6.3.: Multi-level forecast (V V > 0) and initialised analysis (V V = 0) products interpolated to
pressure levels

LL IntpType
Lev-Typ 1/2
ShortName

Discipline

stepType
Category
Number
Description

Unit
CLC Cloud cover 0/6/22 100/– inst %
FI Geopotential 0/3/4 100/– inst m2 s−2
OMEGA Vertical velocity in pressure 0/2/8 100/– inst Pa s−1
coordinates (ω = dp/dt)

RELHUM Relative humidity (with respect to 0/1/1 100/– inst %

water)

T Temperature 0/0/0 100/– inst BCT K

U Zonal wind 0/2/2 100/– inst m s−1
V Meridional wind 0/2/3 100/– inst m s−1

6.1.4. Single-level fields

Table 6.4.: Single-level forecast (V V > 0) and initialised analysis (V V = 0) products

LL IntpType
Lev-Typ 1/2
ShortName

Discipline

stepType
Category
Number

Description
Unit

ALB_RAD Shortwave broadband albedo for 0/19/1 1/– inst %

diffuse radiation

ALHFL_S Latent heat net flux at surface 0/0/10 1/– avg BCT W m−2
(average since model start)

APAB_S Photosynthetically active radiation 0/4/10 1/– avg BCT W m−2

flux at surface (average since model
start)

ASHFL_S Sensible heat net flux at surface 0/0/11 1/– avg BCT W m−2
(average since model start)

ASOB_S Net short-wave radiation flux at 0/4/9 1/– avg BCT W m−2
surface (average since model start)

Continued on next page

34
 6.1. Available output fields

Table 6.4.: continued

ASOB_T Net short-wave radiation flux at 0/4/9 8/– avg BCT W m−2
top of atmosphere (TOA) (average
since model start)

ASWDIFD_S Surface down solar diffuse radiation 0/4/199 1/– avg BCT W m−2
(average since model start)

ASWDIFU_S Surface up solar diffuse radiation 0/4/8 1/– avg BCT W m−2
(average since model start)

ASWDIR_S Surface down solar direct radiation 0/4/198 1/– avg BCT W m−2
(average since model start)

ATHB_S Net long-wave radiation flux at 0/5/5 1/– avg BCT W m−2
surface (average since model start)

ATHB_T Net long-wave radiation flux at 0/5/5 8/– avg BCT W m−2
TOA (average since model start)

AUMFL_S U-momentum flux at surface 0/2/17 1/– avg BCT m

1/2
u0 w 0 (average since model start)

AVMFL_S V-momentum flux at surface 0/2/18 1/– avg BCT m

1/2
v 0 w0 (average since model start)
CAPE_CON Convective available potential 0/7/6 1/– inst NNB J kg−1
energy
CAPE_ML Mixed layer CAPE 0/7/6 192/– inst NNB J kg−1
CEILING Ceiling height (above MSL) 0/6/13 2/101 inst m
CIN_ML Mixed layer convective inhibition 0/7/7 192/– inst NNB J kg−1
CLCH High level clouds 0/6/22 100/100 inst %
CLCL Low level clouds 0/6/22 100/1 inst %
CLCM Mid level clouds 0/6/22 100/100 inst %
CLCT Total cloud cover 0/6/1 1/– inst %
CLCT_MOD Modified total cloud cover for 0/6/199 1/– inst 1
media

CLDEPTH Modified cloud depth for media 0/6/198 1/– inst 1

DBZ_850 Radar Reflectivity in approx. 0/15/1 1/– inst – dBZ
850 hPa

DBZ_CMAX Column Maximum Radar 0/15/1 1/8 inst – dBZ

Reflectivity

FRESHSNW Fresh snow factor (weighting 0/1/203 1/– inst – 1

function for albedo indicating
freshness of snow)

FR_ICE Sea/lake ice cover (possible range: 10/2/0 1/– inst – 1

[0, 1])

GRAU_GSP4 Large scale graupel (accumulated 0/1/75 1/– accu BCT kg m−2
since model start)

Continued on next page

35
Chapter 6. Output fields of the ICON model: General description

Table 6.4.: continued

HBAS_CON Height of convective cloud base 0/6/26 2/101 inst NNB m

above MSL

HBAS_SC Height of shallow convective cloud 0/6/192 2/101 inst NNB m

base above MSL

H_ICE Sea/Lake ice thickness (Max: 3 m) 10/2/1 1/– inst m

H_SNOW Snow depth 0/1/11 1/– inst m
HTOP_CON Height of convective cloud top 0/6/27 3/101 inst NNB m
above MSL

HTOP_DC Height of top of dry convection 0/6/196 3/101 inst NNB m

above MSL

HTOP_SC Height of shallow convective cloud 0/6/193 3/101 inst NNB m

top above MSL

HZEROCL Height of 0 degree Celsius isotherm 0/3/6 4/101 inst NNB m

above MSL

LPI Lightning Potential Index 0/17/192 1/– inst NNB J kg−1

LPI_MAX Maximum Lightning Potential 0/17/192 1/– max NNB J kg−1
Index

LPI_CON_CI_MAX Maximum Lightning Potential 0/17/193 1/– max NNB J kg−1

Index from convection scheme,
modified by convection index

PMSL Surface pressure reduced to MSL 0/3/1 101/– inst Pa

PRG_GSP Precipitation rate of large scale 0/1/75 1/– inst BCT kg m−2 s
graupel
PRR_GSP Precipitation rate of large scale rain 0/1/77 1/– inst BCT kg m−2 s
PRS_GSP Precipitation rate of large scale 0/1/56 1/– inst BCT kg m−2 s
snow

PS Surface pressure (not reduced) 0/3/0 1/– inst Pa

QV_2M Specific humidity at 2m above 0/1/0 103/– inst kg kg−1
ground

QV_S Surface specific humidity 0/1/0 1/– inst kg kg−1

RAIN_CON4 Convective rain (accumulated since 0/1/76 1/– accu BCT kg m−2
model start)
RAIN_GSP4 Large scale rain (accumulated since 0/1/77 1/– accu BCT kg m−2
model start)

RELHUM_2M Relative humidity at 2m above 0/1/1 103/– inst %

ground

RHO_SNOW Snow density 0/1/61 1/– inst kg m−3

RUNOFF_G Soil water runoff (accumulated 2/0/5 106/– accu BCT kg m−2
since model start)

Continued on next page

36
 6.1. Available output fields

Table 6.4.: continued

RUNOFF_S Surface water runoff (accumulated 2/0/5 106/– accu BCT kg m−2
since model start)

SDI2 Supercell Detection Index 2 0/7/193 1/– inst NNB s−1

SNOWC Snow cover 0/1/42 1/– inst %
SNOW_CON 4
Convective snowfall water 0/1/55 1/– accu BCT kg m−2
equivalent (accumulated since
model start)
SNOW_GSP4 Large scale snowfall water 0/1/56 1/– accu BCT kg m−2
equivalent (accumulated since
model start)

SNOWLMT Height of snowfall limit above MSL 0/1/204 4/101 inst NNB m
SOBS_RAD Net short-wave radiation flux at 0/4/9 1/– inst W m−2
surface (instantaneous)

SYNMSG_BT_CL_IR10.8 Synthetic MSG SEVIRI 3/1/14 –/– inst K

image brightness temp. at 10.8µm

SYNMSG_BT_CL_WV6.2 Synthetic MSG SEVIRI 3/1/14 –/– inst K

image brightness temp. at 6.2µm

T_2M Temperature at 2m above ground 0/0/0 103/– inst K

TCH Turbulent transfer coefficient for 0/0/19 1/– inst 1
heat and moisture (surface)

TCM Turbulent transfer coefficient for 0/2/29 1/– inst 1

momentum (surface)

TCOND_MAX column integrated condensate, 0/1/81 1/8 max kg m−2

maximum over the last hour

TCOND10_MAX as TCOND_MAX, but integration 0/1/81 20/8 max kg m−2

only above z(T = −10◦ C)
TD_2M Dew point temperature at 2m 0/0/6 103/– inst K
above ground
T_G Ground temperature (temperature 0/0/0 1/– inst K
at sfc-atm interface)
THBS_RAD Net long-wave radiation flux at 0/5/5 1/– inst W m−2
surface (instantaneous)

T_ICE Sea/Lake ice temperature (at 10/2/8 1/– inst K

ice-atm interface)

TMAX_2M Maximum temperature at 2m 0/0/0 103/– max K

above ground

TMIN_2M Minimum temperature at 2m above 0/0/0 103/– min K

ground

TOT_PREC4 Total precipitation (accumulated 0/1/52 1/– accu BCT kg m−2

since model start)

Continued on next page

37
Chapter 6. Output fields of the ICON model: General description

Table 6.4.: continued

TQC Column integrated cloud water 0/1/69 1/– inst kg m−2

(grid scale)

TQC_DIA Total column integrated cloud 0/1/215 1/– inst kg m−2

water (including sub-grid-scale
contribution)

TQG Column integrated graupel (grid 0/1/74 1/– inst kg m−2

scale)

TQI Column integrated cloud ice (grid 0/1/70 1/– inst kg m−2
scale)

TQI_DIA Total column integrated cloud ice 0/1/216 1/– inst kg m−2
(including sub-grid-scale
contribution)

TQR Column integrated rain (grid scale) 0/1/45 1/– inst kg m−2
TQS Column integrated snow (grid 0/1/46 1/– inst kg m−2
scale)

TQV Column integrated water vapour 0/1/64 1/– inst kg m−2

(grid scale)
TQV_DIA Total column integrated water 0/1/214 1/– inst kg m−2
vapour (including sub-grid-scale
contribution)

TWATER Column integrated water (grid 0/1/78 1/– inst kg m−2

scale)
T_S5 Temperature of the soil surface 2/3/18 1/– inst K
(equivalent to T_SO(0))
T_SNOW Temperature of the snow surface 0/0/18 1/– inst K
U_10M Zonal wind at 10m above ground 0/2/2 103/– inst m s−1
UH_MAX updraft helicity (vertically 0/7/15 102/102 max m2 s−2
averaged), maximum over last hour

V_10M Meridional wind at 10m above 0/2/3 103/– inst m s−1

ground

VMAX_10M Maximum wind at 10 m above 0/2/22 103/– max m s−1

ground

VORW_CTMAX vorticity (vertically averaged), 0/2/206 102/102 max s−1

maximum of absolute value over
last hour

W_CTMAX vertical velocity (maximum value), 0/2/207 102/102 max m s−1

maximum over last hour

W_I Plant canopy surface water 2/0/13 1/– inst – kg m−2

W_SNOW Snow depth water equivalent 0/1/60 1/– inst kg m−2
WW Weather interpretation (WMO), 0/19/25 1/– inst NNB 1
see Table 7.1 for details.

Continued on next page

38
 6.1. Available output fields

Table 6.4.: continued

Z0 Surface roughness (above land and 2/0/1 1/– inst m

water)

6.1.5. Lake-specific single-level fields

Table 6.5.: Single-level forecast (V V > 0) and initialised analysis (V V = 0) products of the lake model
model

LL IntpType
Lev-Typ 1/2
ShortName

Discipline

stepType
Category
Number
Description

Unit
C_T_LK Shape factor with respect to the 1/2/10 162/166 inst – 1
temperature profile in the
thermocline

H_ML_LK Mixed-layer depth 1/2/0 1/166 inst – m

T_BOT_LK Temperature at the water-bottom 1/2/1 162/– inst – K
sediment interface

T_MNW_LK Mean temperature of the water 1/2/1 1/162 inst – K

column

T_WML_LK Mixed-layer temperature 1/2/1 1/166 inst – K

4 Note that the unit which is displayed, when inspecting the GRIB2 message with grib_dump is kg m−2 s−1 rather
than kg m−2 . Mathematically this is wrong, however, it is in accordance with the GRIB2 standard. To get the
mathematically correct unit for accumulated fields (typeOfStatisticalProcessing=1), the unit displayed by grib_dump
must be multiplied by s.
5 T_S is identical to T_SO at level 0. It will no longer be available in the future. Use T_SO(0) instead of T_S.

39
Chapter 6. Output fields of the ICON model: General description

6.1.6. Soil-specific multi-level fields

Table 6.6.: Multi-level forecast (V V > 0) and initialised analysis (V V = 0) products of the soil model

LL IntpType
Lev-Typ 1/2
ShortName

Discipline

stepType
Category
Number
Description

Unit
T_SO Soil temperature 2/3/18 106/– inst K
W_SO Soil moisture integrated over 2/3/20 106/106 inst kg m−2
individual soil layers (ice + liquid)

W_SO_ICE Soil ice content integrated over 2/3/22 106/106 inst NNB kg m−2
individual soil layers

Soil temperature is defined at the soil depths given in Table 6.7 6.7 (column 2). Levels 1 to 8 define the
full levels of the soil model. A zero gradient condition is assumed between levels 0 and 1, meaning that
temperatures at the surface-atmosphere interface are set equal to the temperature at the first full level
depth (0.5 cm). Temperatures are prognosed for layers 1 to 7. At the lowermost layer (mid-level height
1458 cm) the temperature is fixed to the climatological average 2 m-temperature.
Soil moisture W_SO is prognosed for layers 1 to 6. In the two lowermost layers W_SO is filled with W_SO(6)
(zero gradient condition).

Table 6.7.: Soil model: vertical distribution of levels and layers

level no. depth [cm] layer no. upper/lower bounds [cm]

0 0.0
1 0.5 1 0.0 — 1.0
2 2.0 2 1.0 — 3.0
3 6.0 3 3.0 — 9.0
4 18.0 4 9.0 — 27.0
5 54.0 5 27.0 — 81.0
6 162.0 6 81.0 — 243.0
7 486.0 7 243.0 — 729.0
8 1458.0 8 729.0 — 2187.0

40
 6.1. Available output fields

6.1.7. Output fields for soil moisture analysis SMA

The soil moisture analysis (SMA) requires the following fields from the main run at 00 UTC. They are
written only by this run and from forecast hour 2 to 24. As a soil moisture analysis is made for the
global and the nest domain, these fields are available for both domains, but only on the native grid.

Table 6.8.: Fields for SMA from 00 UTC run for forecast hours 2 to 24.

LL IntpType
Lev-Typ 1/2
ShortName

Discipline

stepType
Category
Number
Description

Unit
ALHFL_BS Latent heat flux from bare soil 2/0/193 1/– avg – W m−2
ALHFL_PL Latent heat flux from plants 2/0/194 106/106 avg – W m−2
RSTOM Stomatal resistance 2/0/195 1/– inst – s m−1

The latent heat flux from plants is defined at the same soil layers as the soil moisture W_SO.

41
7. Extended description of available output
fields
In order to facilitate the selection and interpretation of fields and to guard against possible misinterpre-
tation or misusage, the following section provides a more thorough description of the available output
fields.

7.1. Cloud products

CEILING Ceiling is that height above MSL (in m), where the large scale cloud coverage (more
precise: scale and sub-scale, but without the convective contribution) first exceeds
50% when starting from ground.

CLCT_MOD Modified total cloud cover (0 ≤ CLCT_MOD ≤ 1). Used for visualisation purpose (i.e.
gray-scale figures) in the media. It is derived from CLC, neglecting cirrus clouds
if there are only high clouds present at a given grid point. The reason for this
treatment is that the general public does not regard transparent cirrus clouds as
‘real’ clouds.

CLDEPTH Modified cloud depth (0 ≤ CLDEPTH ≤ 1). Used for visualisation purpose (i.e. gray-
scale figures) in the media. A cloud reaching a vertical extent of 700 hPa or more,
has CLDEPTH= 1.

DBZ_850 Synthetic Rayleigh-type approximation of radar reflectivity in dBZ, as a function of

DBZ_CMAX the model variables rain water QR, snow water content QR, graupel content QG and
temperature T. DBZ_850 is the radar reflectivity at approximately 850 hPa, while
DBZ_CMAX is the maximum within the entire grid column.
HBAS_CON Height of the convective cloud base in m above MSL. HBAS_CON is initialised with
−500 m at points where no convection is diagnosed.

HBAS_SC Height of the convective cloud base in m above MSL, but only the shallow convection
part is active.

HTOP_CON Same as HBAS_CON, but for cloud top.

HTOP_SC Same as HBAS_SC, but for cloud top.

7.2. Atmospheric products

HZEROCL Height of the 0◦ C isotherm above MSL. In case of multiple 0◦ C isotherms, HZEROCL
contains the uppermost one. If the temperature is below 0◦ C throughout the entire
atmospheric column, HZEROCL is set equal to the topography height (fill value).
Note that prior to 2019-07-30, HZEROCL contains the height of the lowermost 0◦ C
isotherm. At grid points where no 0◦ C isotherm could be diagnosed, a fill value of
0 is used.

43
Chapter 7. Extended description of available output fields

SNOWLMT Height of snow fall limit above MSL. It is defined as the height where the wet
bulb temperature Tw first exceeds 1.3◦ C (scanning mode from top to bottom). If
this threshold is never reached within the entire atmospheric column, SNOWLMT is
undefined (GRIB2 bitmap).

7.3. Near surface products

TD_2M Dew point temperature at 2 m above ground, i.e. the temperature to which the
air must be cooled, keeping its vapour pressure e constant, such that e equals the
saturation (or equilibrium) vapour pressure es .

es (Td ) = e

TMIN_2M Minimum temperature at 2 m above ground. Minima are collected over 6-hourly
intervals on all domains. (Prior to 2015-07-07 minima were collected over 3-hourly
intervals on the global grid.)

TMAX_2M Same, but for maximum 2 m temperature.

VMAX_10M Maximum wind gust at 10 m above ground. It is diagnosed from the turbulence
state in the atmospheric boundary layer, including a potential enhancement by the
SSO parameterization over mountainous terrain. In the presence of deep convection,
it contains an additional contribution due to convective gusts.
Maxima are collected over hourly intervals on all domains. (Prior to 2015-07-07
maxima were collected over 3-hourly intervals on the global grid.)

7.3.1. General comment on statistically processed fields

In GRIB2, the overall time interval over which a statistical process (like averaging, computation of
maximum/minimum) has taken place is encoded as follows:
The beginning of the overall time interval is defined by referenceTime + forecastTime, whereas the
end of the overall time interval is given by referenceTime + forecastTime + lengthOfTimeRange.
See Section 8.1 for more details on statistically processed fields.

7.4. Surface products

ALB_RAD =
b α Ratio of upwelling to downwelling diffuse radiative flux for wavelength interval
[0.3 µm, 5.0 µm]. Values over snow-free land points are based on a monthly
mean MODIS climatology. MODIS values have been limited to a minimum
value of 2 %.
net
ASOB_S =
b F sw,s Shortwave net radiation flux at the surface, averaged over forecast time. See
Section 8.1 for more information on time averaging.
↓ dif
ASWDIFD_S =
b F sw,s Downward solar diffuse radiation flux at the surface, averaged over forecast
time.
↑ dif
ASWDIFU_S =
b F sw,s Upward solar diffuse radiation flux at the surface, averaged over forecast time.

44
 7.4. Surface products

↓ dir
ASWDIR_S =
b F sw,s Downward solar direct radiation flux at the surface, averaged over forecast
time. This quantity is not directly provided by the radiation scheme. It is a-
posteriori diagnosed from the definition of the surface net shortwave radiation
net
flux Fsw,s
net ↓ dir ↓ dif ↑ dif
Fsw,s = Fsw,s + Fsw,s − Fsw,s .
↓ dir
Solving this equation for Fsw,s , one arrives at
↓ dir net ↓ dif ↑ dif
F sw,s = F sw,s − F sw,s + F sw,s .

The overbar denotes a time average over the forecast time.

↓ dif ↓ dir ↓ tot
From F sw,s and F sw,s the time averaged global radiation at the surface F sw,s can easily be computed
as follows:
↓ tot ↓ dif ↓ dir
F sw,s = F sw,s + F sw,s
↓ tot net
An estimate of F sw,s can also be derived from the surface net solar radiation flux F sw,s and albedo α:
net
↓ tot F sw,s
F sw,s =
1 − 0.01 α
However be aware that this is only approximately true, as α (ALB_RAD) is an instantaneous field. In
addition α only constitutes the albedo for the diffuse component of the incoming solar radiation (“white
net
sky” albedo). F sw,s , however, contains both diffuse and direct components. As a consequence the
reflection of the incoming direct radiation, which is dependent on the solar zenith angle (and described
by the so called “black sky” albedo), is not correctly taken into account.
FR_ICE Sea and lake ice cover. This is the fraction of water covered by ice. I.e. if a grid
cell contains land and water FR_ICE = 1 if the whole fraction of water of this grid
cell is covered by ice. At lake points no fractional ice cover is allowed, meaning that
FR_ICE is either 1 or 0.

H_ICE Ice thickness over sea and frozen fresh water lakes. The maximum allowable ice
thickness is limited to 3 m. New sea-ice points generated by the analysis are ini-
tialised with H_ICE = 0.5 m.

H_SNOW Snow depth in m. It is diagnosed from RHO_SNOW and W_SNOW according to

W_SNOW
H_SNOW =
RHO_SNOW
and is limited to H_SNOW ≤ 40 m.

LPI The Lightning Potential Index after Lynn and Yair (2010). It is calculated as a
vertical integral of the squared updraft velocity weighted by a function that es-
sentially contains the graupel concentration. Therefore, the graupel scheme must
be necessarily switched on and consequently the LPI can be calculated only in a
convection-permitting model setup.

LPI_MAX as LPI, but the maximum value over the last hour is delivered.

LPI_CON_CI_MAX The Modified Lightning Potential Index based on Lynn and Yair (2010) and
Lopez (2016). It is calculated in a similar way as the LPI, only that the updraft
velocity and hydrometeors are taken from the Bechtold-Tiedke convection scheme.
A further correction using the vertical equivalent potential temperature gradient is
applied. The variable contains the maximum since the last output.

45
Chapter 7. Extended description of available output fields

RHO_SNOW Snow density in kg/m3 . It can vary between 50 kg/m3 for fresh snow and 400 kg/m3
for compacted old snow. At snow-free points over land and over water RHO_SNOW is
set to 0 kg/m3 . Note that prior to 2019-07-30 RHO_SNOW was set to 50 kg/m3 over
snow-free land points.

SDI2 The supercell detection index detects the mesocyclone of a supercell. It is based
on the product of a correlation between vertical velocity and vorticity and the local
vorticity Wicker et al. (2005).

T_ICE Ice temperature over sea-ice and frozen lake points. Melting ice has a temperature
of 273.15 K. Ice-free points over land, sea, and lakes are set to T_SO(0).

T_G Temperature at the atmosphere-surface interface, i.e. the temperature of those parts
of the ground which are in direct contact with the atmosphere. E.g. at snow-free
land points it is the temperature of the soil surface, whereas at snow covered land
points it is the temperature of the snow surface.
At snow-free land points T_G is equal to T_SO(0). Likewise, at open water points
T_G is equal to T_SO(0), and represents the sea-surface temperature SST (for more
details on SST see description of T_SO(0) in Section 7.5). At other grid points one
has
• T_G = T_SNOW+(1−f_snow)∗(T_SO(0)−T_SNOW) over (partially) snow covered
grid points. f_snow is the grid point fraction that is snow covered.
• T_G = T_ICE over frozen sea and fresh water lakes

TOT_PREC Total precipitation accumulated since model start.

In simulations with deep convection parametrization it is TOT_PREC = RAIN_GSP +
SNOW_GSP + RAIN_CON + SNOW_CON, whereas in a convection-permitting setup it is
TOT_PREC = RAIN_GSP + SNOW_GSP + GRAU_GSP.

T_SNOW Temperature of snow surface. At snow-free points (H_SNOW = 0), T_SNOW contains
the temperature of the soil surface T_SO(0).

WW Significant weather of the last hour. The predicted weather will be diagnosed hourly
at each model grid point and coded as a key number. The latter is called ww-code
and represents weather phenomena within the last hour. The interpretation of such
weather phenomena from raw model output relies on an independent post-processing
method. This technique applies a number of thresholding processes based on WMO
criteria. Therefore, a couple of ww-codes may differ from the direct model output
(e.g. ww-category snow vs. SNOW_GSP/SNOW_CON). Due to limitations in temporal
and spatial resolution, not all ww-codes as defined by the WMO criteria can be
determined. However, the simulated ww-code is able to take the following values:
no significant weather/ cloud cover (0, 1, 2, 3), fog (45, 48), drizzle (51, 53, 55, 56,
57), rain (61, 63, 65, 66, 67), solid precip not in showers (71, 73, 75, 77), showery
precip (liquid & solid) (80, 81, 82, 85, 86), thunderstorm (95, 96, 99 (only ICON-
D2)) (see also Table 7.1).

W_I Water content of interception layer, i.e. the amount of precipitation intercepted by
vegetation canopies. Over water points, W_I is set to 0.

W_SNOW Snow depth water equivalent in kg/m2 . Set to 0 above water surfaces and snow-free
land points.

W_CTMAX updraft velocity; delivered is the maximum value between ground and 10 km above
ground and during the last hour.

46
 7.5. Soil products

Z0 Surface roughness length. Constant over land, where it depends only on the type of
land cover. I.e. it does not contain any contribution from subgrid-scale orography.
Over water, the roughness length usually varies with time. It is computed by the so
called Charnock-formula, which parameterizes the impact of waves on the roughness
length. Note that this field differs significantly from the external parameter field Z0
(see Table 4.2 or 9.1).

WW weather interpretation WW weather interpretation

45 Fog 48 Fog, depositing rime

51 Slight drizzle 53 Moderate drizzle
55 Heavy drizzle 56 Drizzle, freezing, slight
57 Drizzle, freezing, moderate or heavy 61 Slight rain, not freezing
63 Moderate rain, not freezing 65 Heavy rain, not freezing
66 Rain, freezing, slight 67 Rain, freezing, moderate or heavy
71 Slight fall of snowflakes 73 Moderate fall of snowflakes
75 Heavy fall of snowflakes 77 Snow grains
80 Rain shower(s), slight 81 Rain shower(s), moderate or heavy
82 Rain shower(s), violent 85 Snow shower(s), slight
86 Snow shower(s), moderate or heavy 95 Thunderstorm, slight or moderate
96 Thunderstorm with hail, or heavy thunderstorm

Table 7.1.: Weather interpretation (WW) code table for the ICON model. This table is a subset of the
WMO code table FM 94 BUFR/FM 95 CREX code table 0 20 003 – present weather. In the
case that none of the values provided in Table 7.1 is returned, the WW output contains the
total cloud cover, encoded in the following form: 0: clear sky 1: mainly clear 2: partly/generally
cloudy 3: cloudy/overcast.

7.5. Soil products

RUNOFF_G Water runoff from soil layers. Sum over forecast.

RUNOFF_S Surface water runoff from interception and snow reservoir and from limited infiltra-
tion rate. Sum over forecast.

SOILTYP Characterizes the dominant soiltype in a grid cell. The soiltype is assumed to be
the same for all soil levels. Currently 9 soiltypes are distinguished and encoded by
1-digit integers 1-9. The mapping between these integer numbers and soiltype short
names is given in Table 7.2, together with some soil-dependent hydraulic parameters.
For the full list of hydraulic and thermal parameters, the reader is referred to Doms
et al. (2011).

47
Chapter 7. Extended description of available output fields

Table 7.2.: Mapping between the the soiltype index stored in the field SOILTYP and
soiltype short names. The hydraulic parameters porosity and field capacity,
currently used by ICON, are given in terms of volume fractions.

index soiltype porosity field capacity

1 ice – –
2 rock – –
3 sand 0.364 0.196
4 sandyloam 0.445 0.260
5 loam 0.455 0.340
6 clayloam 0.475 0.370
7 clay 0.507 0.463
8 peat 0.863 0.763
9 sea water – –

T_SO Temperature of the soil and sea water. At land points T_SO(1:7) provides the prog-
nostic temperature of the soil. The full level depths at which the soil temperature
is defined are given in Table 6.7. The temperature at the uppermost level T_SO(0)
is not prognostic. It is rather set equal to the temperature at the first prognos-
tic level T_SO(1). The temperature at the lowermost level T_SO(8) is set to the
climatological 2 m temperature T_2M_CL.
At sea-points T_SO(0:7) provides the sea-surface temperature SST (same value at
all levels). So far, the SST in ICON is not prognostic. It is read from the analysis at
model start and is updated incrementally each day at 00 UTC based on its annual
climatological cycle.
Note that T_SO(0) does not necessarily represent the temperature at the interface
soil-atmosphere. I.e. over snow/ice covered surfaces, T_SO(0) represents the tem-
perature below snow/ice.

7.6. Vertical Integrals

TCOND_MAX The column integrated condensate (i.e. C, I, R, S, G), delivered is the maximum over
the last hour.
TCOND10_MAX As TCOND_MAX, but the vertical integration is restricted to heights above z(T =
−10◦ C).
TQx Column integrated water species x, derived from the 3D grid-scale prognostic quan-
tities Qx, with x ∈ {V, C, I, R, S, G}. TQx is based on the assumption that there would
be no sub-grid-scale variability. That assumption is particularly problematic for
precipitation generation, moist turbulence and radiation.
TQx_DIA Total column integrated water species x, with x ∈ {C, I, V}. Takes into account
the sub-grid-scale variability that includes simple treatments of turbulent motion
and convective detrainment. These cloud variables attempt to represent all model
included physical processes. They are also consistent with the cloud cover variables
CLC, CLCT, CLCH, CLCM and CLCL.

48
 7.6. Vertical Integrals

TWATER This is just the sum over all TQx (but can be independently calculated).

UH_MAX Updraft helicity (i.e. the product of vertical velocity and vorticity) that is vertically
averaged between 500 m and 6000 m above ground. Delivered is its maximum value
(either positive or negative) over the last hour.

VORW_CTMAX Vorticity, vertically averaged between the surface and 1500 m above ground. Deliv-
ered is its maximum value (either positive or negative) over the last hour.

49
8. Remarks on statistical processing and
horizontal interpolation

8.1. Statistically processed output fields

8.1.1. Time-averaged fields

The quantities

ALHFL_S ASHFL_S AUMFL_S AVMFL_S

APAB_S ASOB_S ASOB_T ATHB_S
ATHB_T ASWDIR_S ASWDIFD_S ASWDIFU_S

constitute time averages over the respective forecast time. The averaging process is performed from
forecast start (t0 = 0 s) till forecast end. Thus, time averaged fields which are written to the database
at t = ti contain averages for the elapsed time interval [t0 , ti ].
Let Ψ denote the instantaneous value of one of the above fields. The time average Ψ at time t stored
in the database is given as

1 t
Z
Ψ(t) = Ψ dt , for t > 0.
t 0

For t = 0, the average Ψ is equal to 0. If time averages are required for other time intervals [t1 , t2 ], with
t1 > 0, these can be computed as follows:
Z t2
1
Ψ(t2 − t1 ) = Ψ dt
t2 − t1 t1
Z t2 Z t1 
1
= Ψ dt − Ψ dt
t2 − t1 0 0
1  
= t2 Ψ(t2 ) − t1 Ψ(t1 )
t2 − t1

For this equation to work, it is of course necessary that the fields Ψ(t1 ) and Ψ(t2 ) are available from
the database.
The averaging process is fully reflected by the field’s GRIB2 metainfo. In order to check whether a
field contains the desired time average, it is advisable to check the content of the GRIB2 keys listed in
Table 8.1. I.e. productDefinitionTemplateNumber=8 indicates that the field in question is statistically
processed. The statistical process itself is specified by the key typeOfStatisticalProcessing. The
averaging interval (relative to the start of the forecast) is given by

[forecastTime, forecastTime+lengthOfTimeRange].

Since the averaging process starts at t = 0, the key forecastTime is set to 0.

51
Chapter 8. Remarks on statistical processing and horizontal interpolation

Table 8.1.: List of GRIB2 keys which provide information about the averaging process
Octet(s) Key Value Meaning

8-9 productDefinitionTemplateNumber 8 Average, accumulation, extreme values

or other statistically processed values at
a horizontal level or in a horizontal layer
in a continuous or non-continuous time
interval
19-22 forecastTime 0 Starting time of the averaging process
relative to the reference time.
47 typeOfStatisticalProcessing 0 Average
50-53 lengthOfTimeRange variable Time range over which statistical pro-
cessing is done

8.1.2. Accumulated fields

The quantities

RAIN_GSP SNOW_GSP GRAU_GSP RAIN_CON

SNOW_CON TOT_PREC RUNOFF_S RUNOFF_G

are accumulated over the respective forecast time. The accumulation process is performed from forecast
start (t0 = 0 s) till forecast end. Thus, fields which are written to the database at t = ti are accumulated
for the elapsed time interval [t0 , ti ].
Let Ψ denote the instantaneous value of one of the above fields. The accumulation Ψ̂ at time t stored
in the database is given as
Z t
Ψ̂(t) = Ψ dt , for t > 0.
0

For t = 0, the accumulation Ψ̂ is equal to 0. If accumulations are required for other time intervals
[t1 , t2 ], with t1 > 0, these can be computed as follows:
Z t2
Ψ̂(t2 − t1 ) = Ψ dt
t1
Zt2 Z t1
= Ψ dt − Ψ dt
0 0
= Ψ̂(t2 ) − Ψ̂(t1 )

For this equation to work, it is of course necessary that the fields Ψ̂(t1 ) and Ψ̂(t2 ) are available from
the database.
The accumulation process is fully reflected by the field’s GRIB2 metainfo. In order to check whether a
field contains the desired accumulation, it is advisable to check the content of the GRIB2 keys listed in
Table 8.2. I.e. productDefinitionTemplateNumber=8 indicates that the field in question is statistically
processed. The statistical process itself is specified by the key typeOfStatisticalProcessing. The
accumulation interval (relative to the start of the forecast) is given by

[forecastTime, forecastTime+lengthOfTimeRange].

Since the accumulation process starts at t = 0, the key forecastTime is set to 0.

52
 8.1. Statistically processed output fields

Table 8.2.: List of GRIB2 keys which provide information about the accumulation process
Octet(s) Key Value Meaning

8-9 productDefinitionTemplateNumber 8 Average, accumulation, extreme values

or other statistically processed values at
a horizontal level or in a horizontal layer
in a continuous or non-continuous time
interval
19-22 forecastTime 0 Starting time of the accumulation pro-
cess relative to the reference time.
47 typeOfStatisticalProcessing 1 Accumulation
50-53 lengthOfTimeRange variable Time range over which statistical pro-
cessing is done

8.1.3. Extreme value fields

The quantities

LPI_MAX LPI_CON_CI_MAX TCOND_MAX TCOND10_MAX

TMAX_2M TMIN_2M UH_MAX VMAX_10M
VORW_CTMAX W_CTMAX

represent extreme values, which are collected over certain time intervals χ, starting from the beginning
of the forecast. The interval χ is variable dependent:
• χ = 6 h for TMAX_2M, TMIN_2M
• χ = 1 h for LPI_MAX, LPI_CON_CI_MAX, TCOND_MAX, TCOND10_MAX, UH_MAX,
VMAX_10M, VORW_CTMAX, W_CTMAX
After χ hours of forecast the fields are re-initialized with 0 and the next χ-hourly collection phase is
started. This procedure is repeated till the end of the forecast.
Let Ψ denote the instantaneous value of one of the above fields. The maximum value Ψmax at time t
stored in the database is given as
Ψmax (t) = max(Ψ(t), Ψmax (t)) , for ti < t < ti + χ
Here, ti indicates the time when Ψmax was (re)-initialized the last time. For t = 0, the extreme value
Ψmax is equal to the instantaneous value Ψ.

Please note: Even though a 6 hour time window is used for temperatures, the database contains
hourly, 2-hourly, etc. extreme temperatures. This is because the extreme temperatures are written to
the database hourly, irrespective of the start/end of the 6-hourly time windows. Example: Extreme
temperatures which are written into the database after a forecast time of 8 hours, contain extreme values
collected over the last 2 hours. On the other hand, extreme temperatures written into the database
after 12 hours contain values collected over the last 6 hours. Thus, when dealing with those fields it is
very important to check the GRIB2 keys listed in Table 8.3.
productDefinitionTemplateNumber=8 indicates that the field in question is statistically processed.
The statistical process itself is specified by the key typeOfStatisticalProcessing. The time interval
(relative to the start of the forecast) over which the extreme value collection was performed is given by
[forecastTime, forecastTime+lengthOfTimeRange]. Since the collection process is restarted every χ
hours, the key forecastTime can differ from 0.

53
Chapter 8. Remarks on statistical processing and horizontal interpolation

Table 8.3.: List of GRIB2 keys which provide information about the extreme value process
Octet(s) Key Value Meaning

8-9 productDefinitionTemplateNumber 8 Average, accumulation, extreme values

or other statistically processed values at
a horizontal level or in a horizontal layer
in a continuous or non-continuous time
interval
19-22 forecastTime variable Starting time of the statistical process
relative to the reference time.
47 typeOfStatisticalProcessing 2,3 Maximum/Minimum
50-53 lengthOfTimeRange variable Time range over which statistical pro-
cessing is done

8.2. Technical Details of the Horizontal Interpolation

To facilitate the practical use of ICON output files, many fields are additionally delivered as interpo-
lated fields on a regular (i.e. geographical) lat-lon or a rotated lat-lon grid. Of course, this means a
minimal loss of information on the smallest scales. ICON currently supports three different methods
for interpolating data horizontally from the native triangular grid onto a lat-lon grid:
RBF Radial basis functions
BCT Barycentric interpolation
NNB Nearest-neighbor interpolation
The interpolation selected for a particular field is indicated in the previous tables which list all available
output fields.
Most of the output data on lat-lon grids is processed using an RBF-based interpolation method. The
algorithm approximates the input field with a linear combination of radial basis functions (RBF) located
at the data sites, see, for example, Ruppert (2007). RBF interpolation typically produces over- and un-
dershoots at position where the input field exhibits steep gradients. Therefore, the internal interpolation
algorithm performs a cut-off by default. Note that RBF-based interpolation is not conservative.
Barycentric interpolation (BCT) is a two-dimensional generalization of linear interpolation. This
method uses just three near-neighbors to interpolate and avoids over- and undershoots, since extremal
values are taken only in the data points. This interpolation makes sense for fields where the values
change in a roughly piecewise linear way.
A small number of output fields is treated differently, with a nearest-neighbor interpolation (NNB). The
nearest neighbor algorithm selects the value of the nearest point and does not consider the values of
neighboring points at all, yielding a piecewise-constant interpolant.

54
9. Global output fields
ICON forecasts are performed multiple times a day with varying forecast periods. An overview of the
forecast runs, including its forecast period and output intervals is provided in Figure 9.1.

Figure 9.1.: Time span covered by the various global ICON forecasts which are launched every three hours.
Output on the native (triangular) grid ( ) and the regular grid ( ) is generally available until
forecast end, as indicated by the lenght of the two bars shown for each forecast run. Output
fields are available hourly up to V V = 78 h and 3-hourly for larger forecast times (for exceptions
see the following tables).

Main forecasts are performed 4 times a day at 0, 6, 12, 18 UTC, covering a forecast time span of 180 h
for the 0 und 12 UTC runs and 120 h for the 6 und 18 UTC runs. Prior to 2015-02-25 the 6 and 18
UTC runs were restricted to 78 h. Additional short-range forecasts are performed at 3, 9, 15 and 21
UTC. The forecast time covered by these runs is limited to 30 h since one main purpose of these runs
is to provide boundary data for the high resolution ICON-D2 runs from the ICON nest. See Chapter
10 for more details on the ICON nest and the available output fields.

55
Chapter 9. Global output fields

In general, all time-dependent output fields are available hourly up to V V = 78 h and 3-hourly for larger
forecast times2 . Please note that for ICON fields the time unit is minutes rather than hours, and thus
differs from the previously used global model GME (hours).
Output is available on two distinct horizontal grids:
• The native triangular grid with an average resolution of 13 km that covers the earth with 2949120
triangles
• a regular latitude-longitude grid with a resolution of ∆λ = ∆Φ = 0.25◦ (see table 9.1)
On the native grid most output fields are defined on triangle cell (circum-)centers, except for VN, which
is defined on cell edges. On the lat-lon grid, all fields are defined on cell centers.

global lat-lon

geogr. coordinates 0.0◦ – 359.75◦

90.0◦ S – 90.0◦ N
mesh size 0.25◦
no. of grid points 1038240 (= 721 × 1440)

Table 9.1.: Summary of the latitude-longitude grid for ICON global output.

For details regarding the available fields, please see the tables below. A few remarks about the column
’Time range’: listed is the output time range in hours, followed by the output intervall (also in hours).
The time range is given for the longest runs (i.e. the 00 and 12 UTC runs); of course, for the shorter
runs at 03, 06, 09, 15, 18, 21 UTC, output is only available until the end of the forecast range.

9.1. Time-constant (external parameter) fields

One should distinguish between time-constant or invariant fields that can be extracted from the
database by means of the sky-category parameter (CAT_NAME=$model_const_an_$suite) and variables
exclusively available for V V = 0 that can be extracted via (CAT_NAME=$model_$run_fc_$suite,
s[h] = 0) .

Table 9.2.: Time-constant (external parameter) fields

native
latlon

Time range
EPS
Det.

ShortName level type

ALB_SEAICE t=0 X X
CLAT t=0 X X
t=0 X X
CLON t=0 X X
t=0 X X
Continued on next page

2 An exception here are the output fields VMAX_10M, U_10M and V_10M, which are available hourly throughout the forecast.
For the latter two this is because U_10M and V_10M are needed as input by the wave models.

56
 9.1. Time-constant (external parameter) fields

Table 9.2.: continued

C_T_LK t=0 X X
DEPTH_LK t=0 X X X
ELAT t=0 X X
t=0 X X
ELON t=0 X X
t=0 X X
EVAP_PL t=0 X X
FRESHSNW t=0 X X
FR_ICE t=0 X X
FR_LAKE t=0 X X X
FR_LAND t=0 X X X
t=0 X X
HHL t=0 X X X m
t=0 X X m
HSNOW_MAX t=0 X X
HSURF t=0 X X X
t=0 X X
H_ICE t=0 X X
H_ML_LK t=0 X X
H_SNOW t=0 X X
LAI t=0 X X X
P t=0 X X m
PLCOV t=0 X X X
QC t=0 X X m
QI t=0 X X m
QR t=0 X X m
QS t=0 X X m
QV t=0 X X m
QV_S t=0 X X
RHO_SNOW t=0 X X
RLAT t=0 X X
RLON t=0 X X
ROOTDP t=0 X X X
SMI t=0 X X soil
Continued on next page

57
Chapter 9. Global output fields

Table 9.2.: continued

SNOAG t=0 X X
SOILTYP t=0 X X X
T t=0 X X m
T_BOT_LK t=0 X X
T_G t=0 X X
T_ICE t=0 X X
T_MNW_LK t=0 X X
T_SNOW t=0 X X
T_SO t=0 X X soil
T_WML_LK t=0 X X
U t=0 X X m
V t=0 X X m
W t=0 X X m
W_I t=0 X X
W_SNOW t=0 X X
W_SO t=0 X X soil
W_SO_ICE t=0 X X soil
Z0 t=0 X X

Table 9.3.: Time-constant (external parameter) fields

native
latlon

Time range
EPS
Det.

ShortName level type

AER_BC12 invar X X X
AER_DIF12 invar X X X
AER_DUST12 invar X X X
AER_MRAT invar X X X
AER_NI12 invar X X X
AER_ORG12 invar X X X
AER_SO412 invar X X X
AER_SS12 invar X X X
AER_UV12 invar X X X
CLAT invar X X
CLON invar X X
Continued on next page

58
 9.2. Multi-level fields on native hybrid vertical levels

Table 9.3.: continued

DEPTH_LK invar X X X
EMIS_RAD invar X X X
FOR_D invar X X X
FOR_E invar X X X
FR_LAKE invar X X X
FR_LAND invar X X X
FR_LUC invar X X X
HSURF invar X X X
LAI_MX invar X X X
NDVI_MAX invar X X X
PLCOV_MX invar X X X
ROOTDP invar X X X
RSMIN invar X X X
SOILTYP invar X X X
SSO_GAMMA invar X X X
SSO_SIGMA invar X X X
SSO_STDH invar X X X
SSO_THETA invar X X X
T_2M_CL invar X X X
Z0 invar X X X

9.2. Multi-level fields on native hybrid vertical levels

In the following table 9.4 the denotations in the ’level types’ mean:
• ’m’: output on all model levels (either 90 on main levels or 91 on half (interface) levels)
• ’m25-ke1’: output on model levels 25 ... nlev+1 (=lowest level near the ground)
• ’m42-ke1’: on model levels, k=42... nlev+1
• ’m_3’: output on model levels 42 - 75

Table 9.4.: Multi-level fields on native hybrid vertical levels

native
latlon

Time range
EPS
Det.

ShortName level type

CLC 1–78, 1 h X X X m
81–180, 3 h X X X m
Continued on next page

59
Chapter 9. Global output fields

Table 9.4.: continued

DEN 1–78, 1 h X X m
81–180, 3 h X X m
DTKE_CON 1–180, 1 h X X m42-ke1
0–36, 1 h X X m_3
DTKE_HSH 1–180, 1 h X X m42-ke1
0–36, 1 h X X m_3
P 1–78, 1 h X X m
81–180, 3 h X X m
1–78, 1 h X X m25-ke1
81–180, 3 h X X m25-ke1
0–180, 6 h X X m
QC 1–78, 1 h X X m
81–180, 3 h X X m
1–78, 1 h X X m25-ke1
81–180, 3 h X X m25-ke1
0–180, 6 h X X m
QI 1–78, 1 h X X m
81–180, 3 h X X m
1–78, 1 h X X m25-ke1
81–180, 3 h X X m25-ke1
0–180, 6 h X X m
QR 1–78, 1 h X X m
81–180, 3 h X X m
QS 1–78, 1 h X X m
81–180, 3 h X X m
QV 1–78, 1 h X X m
81–180, 3 h X X m
1–78, 1 h X X m25-ke1
81–180, 3 h X X m25-ke1
0–180, 6 h X X m
T 1–78, 1 h X X m
81–180, 3 h X X m
1–78, 1 h X X m25-ke1
81–180, 3 h X X m25-ke1
Continued on next page

60
 9.3. Multi-level fields interpolated to pressure levels

Table 9.4.: continued

0–180, 6 h X X m
TKE 1–180, 1 h X X m42-ke1
0–36, 1 h X X m_3
U 1–78, 1 h X X m
81–180, 3 h X X m
1–78, 1 h X X m25-ke1
81–180, 3 h X X m25-ke1
0–180, 6 h X X m
V 1–78, 1 h X X m
81–180, 3 h X X m
1–78, 1 h X X m25-ke1
81–180, 3 h X X m25-ke1
0–180, 6 h X X m
W 1–78, 1 h X X m
81–180, 3 h X X m
1–78, 1 h X X m25-ke1
81–180, 3 h X X m25-ke1

9.3. Multi-level fields interpolated to pressure levels

There are several ’level types’ for output on pressure levels. In the following table 9.5 they are denoted
as
• p2: output on pressure levels 30, 50, 70, 100, 150, 200, 250, 300, 400, 500, 600, 700, 800, 850, 900,
925, 950, 1000 hPa
• p3: output on pressure levels 5, 10, 30, 50, 70, 100, 125, 150, 175, 200, 225, 250, 275, 300, 350,
400, 450, 500, 550, 600, 650, 700, 750, 775, 800, 825, 850, 875, 900, 925, 950, 975, 1000 hPa
• p4: output on pressure levels 0.1, 1, 2 hPa
• pe1: output on pressure levels 1, 2, 5, 10, 30, 50, 70, 100, 200 hPa
• pe2: output on pressure levels 250 hPa
• pe3: output on pressure levels 300, 400, 500, 700, 850, 900, 925, 950, 1000 hPa
• pe4: output on pressure levels 500 hPa

61
Chapter 9. Global output fields

Table 9.5.: Multi-level fields interpolated to pressure levels

native
latlon
Time range

EPS
Det.
ShortName level type

CLC 1–78, 1 h X X p3
81–180, 3 h X X p3
FI 1–78, 1 h X X p2
81–180, 3 h X X p2
1–78, 1 h X X p3, p4
81–180, 3 h X X p3, p4
0–75, 1 h X X pe3
75–180, 3 h X X pe3
0–180, 6 h X X pe2
0–72, 6 h X X pe1
72–180, 12 h X X pe1
OMEGA 1–78, 1 h X X p3
81–180, 3 h X X p3
0–75, 1 h X X pe4
75–180, 3 h X X pe4
RELHUM 1–78, 1 h X X p2
81–180, 3 h X X p2
1–78, 1 h X X p3
81–180, 3 h X X p3
0–75, 1 h X X pe3
75–180, 3 h X X pe3
0–180, 6 h X X pe2
0–72, 6 h X X pe1
72–180, 12 h X X pe1
T 1–78, 1 h X X p2
81–180, 3 h X X p2
1–78, 1 h X X p3, p4
81–180, 3 h X X p3, p4
0–75, 1 h X X pe3
75–180, 3 h X X pe3
0–180, 6 h X X pe2
Continued on next page

62
 9.4. Single-level fields

Table 9.5.: continued

0–72, 6 h X X pe1
72–180, 12 h X X pe1
U 1–78, 1 h X X p2
81–180, 3 h X X p2
1–78, 1 h X X p3, p4
81–180, 3 h X X p3, p4
0–75, 1 h X X pe3
75–180, 3 h X X pe3
0–180, 6 h X X pe2
0–72, 6 h X X pe1
72–180, 12 h X X pe1
V 1–78, 1 h X X p2
81–180, 3 h X X p2
1–78, 1 h X X p3, p4
81–180, 3 h X X p3, p4
0–75, 1 h X X pe3
75–180, 3 h X X pe3
0–180, 6 h X X pe2
0–72, 6 h X X pe1
72–180, 12 h X X pe1

9.4. Single-level fields

Table 9.6.: Single-level fields

native
latlon

Time range
EPS
Det.

ShortName

AER_BC12 invar, 0 h X X X
AER_DIF12 invar, 0 h X X X
AER_DUST12 invar, 0 h X X X
AER_MRAT invar, 0 h X X X
AER_NI12 invar, 0 h X X X
AER_ORG12 invar, 0 h X X X
AER_SO412 invar, 0 h X X X
Continued on next page

63
Chapter 9. Global output fields

Table 9.6.: continued

AER_SS12 invar, 0 h X X X
AER_UV12 invar, 0 h X X X
ALB_RAD 1–78, 1 h X X X
81–180, 3 h X X X
ALHFL_BS 2–24, 1 h X X
ALHFL_PL 2–24, 1 h X X
ALHFL_S 1–78, 1 h X X X
81–180, 3 h X X X
6–180, 6 h X X
APAB_S 1–78, 1 h X X
81–180, 3 h X X
ASHFL_S 1–78, 1 h X X X
81–180, 3 h X X X
6–180, 6 h X X
ASOB_S 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
ASOB_T 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
ASWDIFD_S 1–78, 1 h X X X
81–180, 3 h X X X
3–180, 3 h X X
ASWDIFU_S 1–78, 1 h X X X
81–180, 3 h X X X
6–180, 6 h X X
ASWDIR_S 1–78, 1 h X X X
81–180, 3 h X X X
3–180, 3 h X X
ATHB_S 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
Continued on next page

64
 9.4. Single-level fields

Table 9.6.: continued

75–180, 3 h X X
ATHB_T 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
AUMFL_S 1–78, 1 h X X X
81–180, 3 h X X X
AVMFL_S 1–78, 1 h X X X
81–180, 3 h X X X
CAPE_CON 1–78, 1 h X X
81–180, 3 h X X
CAPE_ML 1–78, 1 h X X X
81–180, 3 h X X X
CIN_ML 1–78, 1 h X X
81–180, 3 h X X
CLAT invar, 0 h X X
CLCH 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
CLCL 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
CLCM 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
CLCT 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
CLCT_MOD 1–78, 1 h X X X
81–180, 3 h X X X
Continued on next page

65
Chapter 9. Global output fields

Table 9.6.: continued

0–180, 6 h X X
CLDEPTH 1–78, 1 h X X X
81–180, 3 h X X X
CLON invar, 0 h X X
C_T_LK 1–78, 1 h X X
81–180, 3 h X X
DEPTH_LK invar, 0 h X X X
EMIS_RAD invar, 0 h X X X
FOR_D invar, 0 h X X X
FOR_E invar, 0 h X X X
FRESHSNW 1–78, 1 h X X
81–180, 3 h X X
FR_ICE 1–78, 1 h X X X
81–180, 3 h X X X
0–72, 6 h X X
72–180, 12 h X X
FR_LAKE invar, 0 h X X X
FR_LAND invar, 0 h X X X
FR_LUC invar, 0 h X X X
HBAS_CON 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
HSURF invar, 0 h X X X
HTOP_CON 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
HTOP_DC 1–78, 1 h X X X
81–180, 3 h X X X
HZEROCL 1–78, 1 h X X X
81–180, 3 h X X X
H_ICE 1–78, 1 h X X X
81–180, 3 h X X X
Continued on next page

66
 9.4. Single-level fields

Table 9.6.: continued

0–72, 6 h X X
72–180, 12 h X X
H_ML_LK 1–78, 1 h X X
81–180, 3 h X X
H_SNOW 1–78, 1 h X X X
81–180, 3 h X X X
0–72, 6 h X X
72–180, 12 h X X
LAI_MX invar, 0 h X X X
NDVI_MAX invar, 0 h X X X
PLCOV_MX invar, 0 h X X X
PMSL 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
PS 1–78, 1 h X X X
81–180, 3 h X X X
0–180, 6 h X X
QV_2M 1–78, 1 h X X
81–180, 3 h X X
QV_S 1–78, 1 h X X X
81–180, 3 h X X X
RAIN_CON 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
RAIN_GSP 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
RELHUM_2M 1–78, 1 h X X X
81–180, 3 h X X X
0–180, 6 h X X
RHO_SNOW 1–78, 1 h X X X
Continued on next page

67
Chapter 9. Global output fields

Table 9.6.: continued

81–180, 3 h X X X
ROOTDP invar, 0 h X X X
RSMIN invar, 0 h X X X
RSTOM 2–24, 1 h X X
RUNOFF_G 1–78, 1 h X X X
81–180, 3 h X X X
RUNOFF_S 1–78, 1 h X X X
81–180, 3 h X X X
SNOW_CON 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
SNOW_GSP 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
SOBS_RAD 1–78, 1 h X X
81–180, 3 h X X
0–75, 1 h X X
75–180, 3 h X X
SOILTYP invar, 0 h X X X
SSO_GAMMA invar, 0 h X X X
SSO_SIGMA invar, 0 h X X X
SSO_STDH invar, 0 h X X X
SSO_THETA invar, 0 h X X X
TCH 1–78, 1 h X X X
81–180, 3 h X X X
TCM 1–78, 1 h X X X
81–180, 3 h X X X
TD_2M 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
THBS_RAD 1–78, 1 h X X
Continued on next page

68
 9.4. Single-level fields

Table 9.6.: continued

81–180, 3 h X X
0–75, 1 h X X
75–180, 3 h X X
TKE 1–78, 1 h X X
81–180, 3 h X X
TMAX_2M 1–78, 1 h X X X
81–180, 3 h X X X
6–180, 6 h X X
TMIN_2M 1–78, 1 h X X X
81–180, 3 h X X X
6–180, 6 h X X
TOT_PREC 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
TQC 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
TQC_DIA 1–78, 1 h X X X
81–180, 3 h X X X
TQI 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
TQI_DIA 1–78, 1 h X X X
81–180, 3 h X X X
TQR 1–78, 1 h X X X
81–180, 3 h X X X
TQS 1–78, 1 h X X X
81–180, 3 h X X X
TQV 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
Continued on next page

69
Chapter 9. Global output fields

Table 9.6.: continued

75–180, 3 h X X
T_2M 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
T_2M_CL invar, 0 h X X X
T_BOT_LK 1–78, 1 h X X
81–180, 3 h X X
T_G 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
T_ICE 1–78, 1 h X X X
81–180, 3 h X X X
T_MNW_LK 1–78, 1 h X X
81–180, 3 h X X
T_S 1–78, 1 h X X
81–180, 3 h X X
T_SNOW 1–78, 1 h X X X
81–180, 3 h X X X
0–72, 6 h X X
72–180, 12 h X X
T_SO 0–180, 6 h X X
T_WML_LK 1–78, 1 h X X
81–180, 3 h X X
U_10M 1–78, 1 h X X
81–180, 3 h X X
0–180, 1 h X X
0–75, 1 h X X
75–180, 3 h X X
VMAX_10M 1–180, 1 h X X
0–180, 1 h X X
1–180, 1 h X X
V_10M 1–78, 1 h X X
Continued on next page

70
 9.5. Soil-specific multi-level fields

Table 9.6.: continued

81–180, 3 h X X
0–180, 1 h X X
0–75, 1 h X X
75–180, 3 h X X
WW 1–78, 1 h X X X
81–180, 3 h X X X
W_I 1–78, 1 h X X
81–180, 3 h X X
W_SNOW 1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X
W_SO 0–75, 1 h X X
75–180, 3 h X X
W_SO_ICE 81–180, 3 h X X X
Z0 invar, 0 h X X X
1–78, 1 h X X X
81–180, 3 h X X X
0–75, 1 h X X
75–180, 3 h X X

9.5. Soil-specific multi-level fields

• soil: soil levels = 0, 1, 3, 9, 27, 81, 243, 729 cm

Table 9.7.: Soil-specific multi-level fields

native
latlon

Time range
EPS
Det.

ShortName level type

T_SO 1–78, 1 h X X X soil

81–180, 3 h X X X soil
W_SO 1–78, 1 h X X X soil
81–180, 3 h X X X soil
W_SO_ICE 1–78, 1 h X X X soil

71
10. EU Nest output fields
This section contains a list of output fields that are available with the launch of the ICON-EU nest.
See Fig. 3.7 on page 15 for details regarding the nest location and extent. Forecasts on the EU-nest
are performed multiple times a day with varying forecast periods. Forecasts reaching out to 120 h are
performed at 00, 06, 12, and 18 UTC. Additional short-range forecasts reaching out to 51 h are performed
at 03, 09, 15 and 21 UTC. Its main purpose is to provide boundary data for the high resolution ICON-D2
(formerly: COSMO-D2 or COSMO-DE) runs. A schematic overview of the various forecasts, including
its forecast period and output intervals is provided in Figure 10.1.

Figure 10.1.: Time span covered by the various EU nest forecasts which are launched every three hours.
Output on the native (triangular) grid ( ) and the regular grid ( ) is generally available until
forecast end, as indicated by the lenght of the two bars shown for each forecast run. Output
on the native grid is available hourly to 51 h, and every 6 hours for later forecast times
(forecast time ≥ 54 h). Output on the regular grid is available hourly to 78 h, and every 3
hours for later forecast times.

Output is available on two distinct horizontal grids:

73
Chapter 10. EU Nest output fields

• a native triangular grid with an average resolution of 6.5 km, and

• a regular latitude-longitude grid with a resolution of ∆λ = ∆Φ = 0.0625◦ .
See Table 10.1 for a summary.
Output on the native (triangular) grid is hourly to 51 h, and every 6 hours for verification from
forecast time ≥ 54 h until the forecast end at 120 h. Output on the regular grid is hourly to 78 h,
and every 3 hours until forecast end. See also Figure 10.1. Output of the 10m wind U_10M, V_10M,
VMAX_10M and the solar radiation ASWDIR_S and ASWDIFD_S on the regular grid is hourly until
the end of the forecast.
In the subsequent tables the availability of specific fields on the native grid, on the lat-lon grid, or on
both grids is denoted. A few remarks about the column ’Time range’: listed is the output time range
in hours, followed by the output intervall (also in hours). The time range is given for the longest runs
(i.e. the 00 and 12 UTC runs); of course, for the shorter runs at 03, 06, 09, 15, 18, 21 UTC, output is
only available until the end of the forecast range.

EU nest lat-lon

geogr. coordinates 23.5◦ W – 62.5◦ E,

29.5◦ N – 70.5◦ N
mesh size 0.0625◦

Table 10.1.: Summary of the latitude-longitude grid for the ICON-EU nest output.

74
 10.1. Time-constant (external parameter) fields

10.1. Time-constant (external parameter) fields

Table 10.2.: Time-constant (external parameter) fields

native
latlon
Time range

EPS
Det.
ShortName level type

ALB_SEAICE t=0 X X
CLAT t=0 X X
t=0 X X
CLON t=0 X X
t=0 X X
DEPTH_LK t=0 X X X
ELAT t=0 X X
t=0 X X
ELON t=0 X X
t=0 X X
EVAP_PL t=0 X X
FR_LAKE t=0 X X X
FR_LAND t=0 X X X
t=0 X X
HHL t=0 X X X m
t=0 X X m_5
HSNOW_MAX t=0 X X
HSURF t=0 X X X
t=0 X X
H_SNOW t=0 X X
LAI t=0 X X X
PLCOV t=0 X X X
RLAT t=0 X X
RLON t=0 X X
ROOTDP t=0 X X X
SMI t=0 X X soil
SNOAG t=0 X X
SOILTYP t=0 X X X
Z0 t=0 X X

75
Chapter 10. EU Nest output fields

10.2. Multi-level fields on native hybrid vertical levels

In the following table 10.3 the denotations in the ’level types’ mean:
• ’m’: output on all model levels
• ’m12-ke1’: output on model levels 12 ... nlev+1 (=lowest level near the ground)
• ’m_4’: output on model levels 1 - 60
• ’m_5’: output on model levels 1 - 61
• ’m_6’: output on model levels 12 - 61

Table 10.3.: Multi-level fields on native hybrid vertical levels

native
latlon
Time range

EPS
Det.
ShortName level type

CLC 0–78, 1 h X X m
81–120, 3 h X X m
DTKE_CON 0–48, 1 h X X m12-ke1
0–36, 1 h X X m_6
DTKE_HSH 0–48, 1 h X X m12-ke1
0–36, 1 h X X m_6
P 0–51, 1 h X X m
54–120, 6 h X X m
0–78, 1 h X X m
81–120, 3 h X X m
0–51, 1 h X X m_4
54–78, 3 h X X m_4
78–120, 6 h X X m_4
QC 0–51, 1 h X X m
54–120, 6 h X X m
0–78, 1 h X X m
81–120, 3 h X X m
0–51, 1 h X X m_4
54–78, 3 h X X m_4
78–120, 6 h X X m_4
QI 0–51, 1 h X X m
54–120, 6 h X X m
0–78, 1 h X X m
81–120, 3 h X X m
Continued on next page

76
 10.2. Multi-level fields on native hybrid vertical levels

Table 10.3.: continued

0–51, 1 h X X m_4
51–120, 6 h X X m_4
QR 0–51, 1 h X X m
54–120, 6 h X X m
0–51, 1 h X X m_4
QS 0–51, 1 h X X m
54–120, 6 h X X m
0–51, 1 h X X m_4
QV 0–51, 1 h X X m
54–120, 6 h X X m
0–78, 1 h X X m
81–120, 3 h X X m
0–51, 1 h X X m_4
54–78, 3 h X X m_4
78–120, 6 h X X m_4
T 0–51, 1 h X X m
54–120, 6 h X X m
0–78, 1 h X X m
81–120, 3 h X X m
0–51, 1 h X X m_4
54–78, 3 h X X m_4
78–120, 6 h X X m_4
TKE 0–48, 1 h X X m12-ke1
0–78, 1 h X X m
81–120, 3 h X X m
0–36, 1 h X X m_6
U 0–51, 1 h X X m
54–120, 6 h X X m
0–78, 1 h X X m
81–120, 3 h X X m
0–51, 1 h X X m_4
54–120, 6 h X X m_4
V 0–51, 1 h X X m
54–120, 6 h X X m
Continued on next page

77
Chapter 10. EU Nest output fields

Table 10.3.: continued

0–78, 1 h X X m
81–120, 3 h X X m
0–51, 1 h X X m_4
54–120, 6 h X X m_4
W 0–51, 1 h X X m
54–120, 6 h X X m
0–78, 1 h X X m
81–120, 3 h X X m
0–3, 3 h X X m_5

10.3. Multi-level fields interpolated to pressure levels

In the following table 10.4, the ’level type’ means
• p5: output on pressure levels 50, 70, 100, 125, 150, 175, 200, 225, 250, 275, 300, 350, 400, 450,
500, 550, 600, 650, 700, 750, 775, 800, 825, 850, 875, 900, 925, 950, 975, 1000 hPa
• pe4: output on pressure level 500 hPa
• pe5: output on pressure levels 300, 400, 500, 700, 850, 900, 925, 950, 1000 hPa
On the native (triangular) grid no output is generated for pressure levels.

Table 10.4.: Multi-level fields interpolated to pressure levels

native
latlon

Time range
EPS
Det.

ShortName level type

CLC 0–78, 1 h X X p5
81–120, 3 h X X p5
FI 0–78, 1 h X X p5
81–120, 3 h X X p5
0–75, 1 h X X pe5
75–120, 3 h X X pe5
OMEGA 0–78, 1 h X X p5
81–120, 3 h X X p5
0–75, 1 h X X pe4
75–120, 3 h X X pe4
RELHUM 0–78, 1 h X X p5
81–120, 3 h X X p5
Continued on next page

78
 10.4. Single-level fields

Table 10.4.: continued

0–75, 1 h X X pe5
75–120, 3 h X X pe5
T 0–78, 1 h X X p5
81–120, 3 h X X p5
0–75, 1 h X X pe5
75–120, 3 h X X pe5
U 0–78, 1 h X X p5
81–120, 3 h X X p5
0–75, 1 h X X pe5
75–120, 3 h X X pe5
V 0–78, 1 h X X p5
81–120, 3 h X X p5
0–75, 1 h X X pe5
75–120, 3 h X X pe5

10.4. Single-level fields

Table 10.5.: Single-level fields native

latlon

Time range
EPS
Det.

ShortName

ALB_RAD 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
ALHFL_BS 2–24, 1 h X X
ALHFL_PL 2–24, 1 h X X
ALHFL_S 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
APAB_S 0–78, 1 h X X
81–120, 3 h X X
ASHFL_S 0–51, 1 h X X
Continued on next page

79
Chapter 10. EU Nest output fields

Table 10.5.: continued

54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
ASOB_S 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
ASOB_T 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
ASWDIFD_S 0–51, 1 h X X
54–120, 6 h X X
0–120, 1 h X X
0–75, 1 h X X
75–120, 3 h X X
ASWDIFU_S 0–78, 1 h X X
81–120, 3 h X X
6–120, 6 h X X
ASWDIR_S 0–51, 1 h X X
54–120, 6 h X X
0–120, 1 h X X
0–75, 1 h X X
75–120, 3 h X X
ATHB_S 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
ATHB_T 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
Continued on next page

80
 10.4. Single-level fields

Table 10.5.: continued

AUMFL_S 0–78, 1 h X X
81–120, 3 h X X
AVMFL_S 0–78, 1 h X X
81–120, 3 h X X
CAPE_CON 0–78, 1 h X X
81–120, 3 h X X
CAPE_ML 0–78, 1 h X X
81–120, 3 h X X
CIN_ML 0–78, 1 h X X
81–120, 3 h X X
CLCH 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
CLCL 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
CLCM 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
CLCT 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
Continued on next page

81
Chapter 10. EU Nest output fields

Table 10.5.: continued

CLCT_MOD 0–78, 1 h X X
81–120, 3 h X X
CLDEPTH 0–78, 1 h X X
81–120, 3 h X X
C_T_LK 0–51, 1 h X X
54–120, 6 h X X
FRESHSNW 0–51, 1 h X X
54–120, 6 h X X
0–3, 3 h X X
FR_ICE 0–51, 1 h X X
54–120, 6 h X X
0–120, 6 h X X
HBAS_CON 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
HTOP_CON 0–48, 1 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
HTOP_DC 0–78, 1 h X X
81–120, 3 h X X
HZEROCL 0–78, 1 h X X
81–120, 3 h X X
H_ICE 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–6, 3 h X X
6–120, 6 h X X
H_ML_LK 0–51, 1 h X X
54–120, 6 h X X
H_SNOW 0–78, 1 h X X
Continued on next page

82
 10.4. Single-level fields

Table 10.5.: continued

81–120, 3 h X X
0–120, 6 h X X
LPI_CON_CI_MAX 0–48, 1 h X X X
51–72, 3 h X X X
78–120, 6 h X X X
0–48, 1 h X X
51–72, 3 h X X
78–120, 6 h X X
PMSL 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
PS 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
QV_2M 0–78, 1 h X X
81–120, 3 h X X
QV_S 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–51, 1 h X X
RAIN_CON 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
RAIN_GSP 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
RELHUM_2M 0–78, 1 h X X
Continued on next page

83
Chapter 10. EU Nest output fields

Table 10.5.: continued

81–120, 3 h X X
0–120, 6 h X X
RHO_SNOW 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
RSTOM 2–24, 1 h X X
RUNOFF_G 0–78, 1 h X X
81–120, 3 h X X
RUNOFF_S 0–78, 1 h X X
81–120, 3 h X X
SNOWLMT 0–78, 1 h X X
81–120, 3 h X X
SNOW_CON 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
SNOW_GSP 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
SOBS_RAD 0–75, 1 h X X
75–120, 3 h X X
SYNMSG_BT_CL_IR10.8 0–78, 1 h X X
81–120, 3 h X X
SYNMSG_BT_CL_WV6.2 0–78, 1 h X X
81–120, 3 h X X
TCH 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
TCM 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
Continued on next page

84
 10.4. Single-level fields

Table 10.5.: continued

81–120, 3 h X X
TD_2M 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
THBS_RAD 0–75, 1 h X X
75–120, 3 h X X
TMAX_2M 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
6–120, 6 h X X
TMIN_2M 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
6–120, 6 h X X
TOT_PREC 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
TQC 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
TQI 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
TQR 0–78, 1 h X X
Continued on next page

85
Chapter 10. EU Nest output fields

Table 10.5.: continued

81–120, 3 h X X
TQS 0–78, 1 h X X
81–120, 3 h X X
TQV 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
T_2M 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
T_BOT_LK 0–51, 1 h X X
54–120, 6 h X X
T_G 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
T_ICE 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–3, 3 h X X
T_MNW_LK 0–51, 1 h X X
54–120, 6 h X X
T_SNOW 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–51, 1 h X X
T_SO 0–51, 1 h X X
Continued on next page

86
 10.4. Single-level fields

Table 10.5.: continued

54–120, 6 h X X
0–51, 1 h X X
T_WML_LK 0–51, 1 h X X
54–120, 6 h X X
U_10M 0–51, 1 h X X
54–120, 6 h X X
0–120, 1 h X X
0–75, 1 h X X
75–120, 3 h X X
VMAX_10M 0–51, 1 h X X
54–120, 6 h X X
0–120, 1 h X X
1–120, 1 h X X
V_10M 0–51, 1 h X X
54–120, 6 h X X
0–120, 1 h X X
0–75, 1 h X X
75–120, 3 h X X
WW 0–78, 1 h X X
81–120, 3 h X X
W_I 0–51, 1 h X X
54–120, 6 h X X
0–3, 3 h X X
W_SNOW 0–51, 1 h X X
54–120, 6 h X X
0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X
W_SO 0–51, 1 h X X
54–120, 6 h X X
0–75, 1 h X X
75–120, 3 h X X
W_SO_ICE 0–51, 1 h X X
Continued on next page

87
Chapter 10. EU Nest output fields

Table 10.5.: continued

54–120, 6 h X X
Z0 0–78, 1 h X X
81–120, 3 h X X
0–75, 1 h X X
75–120, 3 h X X

10.5. Soil-specific multi-level fields

• soil: soil levels = 0, 1, 3, 9, 27, 81, 243, 729 cm

Table 10.6.: Soil-specific multi-level fields

native
latlon
Time range

EPS
ShortName Det. level type

T_SO 0–78, 1 h X X soil

81–120, 3 h X X soil
W_SO 0–78, 1 h X X soil
81–120, 3 h X X soil
W_SO_ICE 0–78, 1 h X X soil
81–120, 3 h X X soil

88
11. ICON-D2 output fields
This section contains a list of output fields that are available with the launch of ICON-D2. See Fig. 3.8
for details regarding its location and extent. Forecasts of ICON-D2 are performed 8 times a day for the
forecast times 00, 03, 06, 09, 12, 15, 18, 21 UTC, with a forecast range of 48h. Prior to 2021-06-23, the
forecast range was limited to 27h with the exception of the 03 UTC run which reached 45h. During
the pre-operational phase (i.e. prior to 2021-02-10) forecasts of ICON-D2 have been performed 2 times
a day for the forecast times 00 and 12 UTC, with a forecast range of 27h.
Output is available on two distinct horizontal grids:
• a native triangular grid with an average resolution of about 2.1 km, and
• a rotated (!) latitude-longitude grid with a resolution of ∆λ = ∆Φ = 0.02◦ (remark: this horizontal
lat-lon grid is exactly the same as for the former COSMO-D2!). See Table 11.1 for a summary.
The geographical coordinates of every rotated grid point can be found in the fields RLON and RLAT.
This information should be sufficient for the most users (otherwise some more details can be found
in appendix B).
Note that there are a few differences to some of the former COSMO-D2 fields:
– although the velocity components u and v are given on the rotated lat-lon grid points,
too, their components now are the purely (i.e. unrotated) zonal and meridional components,
respectively (in the Grib-Metadata ResolutionAndComponentFlags the 5th bit is 0, whereas
in the former COSMO it was 1).
– Now every variable is interpolated to the same cell center point (whereas in the former
COSMO again the velocity components have been staggered by the half grid mesh size) (in
the Grib-Metadata scanningMode the last four bits are all zero, i.e. no staggering).
– The vertical model levels for 3D fields are slightly different to COSMO. In any case the height
values are given by the vertical averaging of the two neighbouring HHL-values (HHL-fields are
delivered on the rotated lat-lon grid, too).
The output of the most variables takes place hourly. A few variables, which are in partiular of interest
in the cases of deep convection, are delivered every 15 min.
The model area of ICON-D2 (Fig. 3.8) completely contains the areas of Germany, Switzerland and
Austria and also parts of the neighbouring coutries.
The rotated latitude-longitude output grid contains 651 × 716 = 466116 grid points with a grid mesh
size of 0.02◦ (∼ 2,2 km).

Left bottom (SW) corner: λ = 07.50◦ W ϕ = 06.30◦ S λg = 00.25◦ W ϕg = 43.19◦ N

Right bottom (SE) corner: λ = 05.50◦ E ϕ = 06.30◦ S λg = 17.54◦ E ϕg = 43.42◦ N
Left top (NW) corner: λ = 07.50◦ W ϕ = 08.00◦ N λg = 03.84◦ W ϕg = 57.31◦ N
Right top (NE) corner: λ = 05.50◦ E ϕ = 08.00◦ N λg = 20.21◦ E ϕg = 57.62◦ N

Table 11.1.: Rotated coordinates (λ, ϕ) and geographical coordinates (λg , ϕg ) of the four corner points of
the lat-lon grid.

In the subsequent tables the availability of specific fields on the native grid, on the lat-lon grid, or on
both grids is denoted.

89
Chapter 11. ICON-D2 output fields

11.1. Time-constant fields

Table 11.2.: Time-constant fields

native
latlon
Time range

EPS
Det.
ShortName level type

CLAT t=0 X X
t=0 X X
CLON t=0 X X
t=0 X X
DEPTH_LK t=0 X X X
t=0 X X
ELAT t=0 X X
t=0 X X
ELON t=0 X X
t=0 X X
FR_ICE t=0 X X X
t=0 X X
FR_LAKE t=0 X X X
t=0 X X
FR_LAND t=0 X X X
t=0 X X
HHL t=0 X X X m
t=0 X X m
HSURF t=0 X X X
t=0 X X
LAI t=0 X X X
t=0 X X
PLCOV t=0 X X X
t=0 X X
RLAT t=0 X X
RLON t=0 X X
ROOTDP t=0 X X X
t=0 X X
SOILTYP t=0 X X X
t=0 X X

90
 11.3. Multi-level fields interpolated to pressure levels

11.2. Multi-level fields on native hybrid vertical levels

In the following table 11.3, the denotation ’m’ in the ’level types’ means: output on all model levels.

Table 11.3.: Multi-level fields on native hybrid vertical levels

native
latlon
Time range

EPS
Det.
ShortName level type

CLC 0–48, 1 h X X X m
0–48, 1 h X X m
P 0–48, 1 h X X X m
0–48, 1 h X X m
QC 0–48, 1 h X X X m
0–48, 1 h X X m
QG 0–48, 1 h X X X m
0–48, 1 h X X m
QI 0–48, 1 h X X X m
0–48, 1 h X X m
QR 0–48, 1 h X X X m
QS 0–48, 1 h X X X m
QV 0–48, 1 h X X X m
0–48, 1 h X X m
Q_SEDIM 0–48, 1 h X X X m
0–48, 1 h X X m
T 0–48, 1 h X X X m
0–48, 1 h X X m
TKE 0–48, 1 h X X X m
U 0–48, 1 h X X X m
0–48, 1 h X X m
V 0–48, 1 h X X X m
0–48, 1 h X X m
W 0–48, 1 h X X X m
0–48, 1 h X X m

11.3. Multi-level fields interpolated to pressure levels

In the following table, the ’level type’ p1 means output on the pressure levels 200, 250, 300, 400, 500,
600, 700, 850, 950, 975, 1000 hPa.

91
Chapter 11. ICON-D2 output fields

Table 11.4.: Multi-level fields interpolated to pressure levels

native
latlon
Time range

EPS
Det.
ShortName level type

FI 0–48, 1 h X X X p1
0–48, 1 h X X p1
OMEGA 0–48, 1 h X X X p1
0–48, 1 h X X p1
RELHUM 0–48, 1 h X X X p1
0–48, 1 h X X p1
T 0–48, 1 h X X X p1
0–48, 1 h X X p1
U 0–48, 1 h X X X p1
0–48, 1 h X X p1
V 0–48, 1 h X X X p1
0–48, 1 h X X p1

11.4. Single-level fields

Table 11.5.: Single-level fields

native
latlon

Time range
EPS
Det.

ShortName

ALB_RAD 0–48, 1 h X X X
0–48, 1 h X X
ALHFL_S 0–48, 1 h X X X
0–48, 1 h X X
APAB_S 0–48, 1 h X X X
ASHFL_S 0–48, 1 h X X X
0–48, 1 h X X
ASOB_S 0–48, 1 h X X X
0–48, 1 h X X
ASOB_T 0–48, 1 h X X X
0–48, 1 h X X
ASWDIFD_S 0–48, 0.25 h X X X
Continued on next page

92
 11.4. Single-level fields

Table 11.5.: continued

0–48, 1 h X X
ASWDIFU_S 0–48, 1 h X X X
0–48, 1 h X X
ASWDIR_S 0–48, 0.25 h X X X
0–48, 1 h X X
ATHB_S 0–48, 1 h X X X
0–48, 1 h X X
ATHB_T 0–48, 1 h X X X
0–48, 1 h X X
AUMFL_S 0–48, 1 h X X X
AVMFL_S 0–48, 1 h X X X
CAPE_CON 0–48, 1 h X X
CAPE_ML 0–48, 0.25 h X X X
0–48, 1 h X X
CEILING 0–48, 1 h X X X
0–48, 1 h X X
CIN_ML 0–48, 0.25 h X X X
0–48, 1 h X X
CLCH 0–48, 1 h X X X
0–48, 1 h X X
CLCL 0–48, 1 h X X X
0–48, 1 h X X
CLCM 0–48, 1 h X X X
0–48, 1 h X X
CLCT 0–48, 1 h X X X
0–48, 1 h X X
CLCT_MOD 0–48, 1 h X X X
0–48, 1 h X X
CLDEPTH 0–48, 1 h X X X
C_T_LK 0–48, 1 h X X
DBZ_850 0–48, 0.25 h X X
0–48, 0.25 h X X
DBZ_CMAX 0–48, 0.25 h X X
0–48, 0.25 h X X
Continued on next page

93
Chapter 11. ICON-D2 output fields

Table 11.5.: continued

ECHOTOP 0–48, 0.25 h X X

0–48, 0.25 h X X
FRESHSNW 0–48, 1 h X X
0–48, 1 h X X
GRAU_GSP 0–48, 0.25 h X X X
0–48, 1 h X X
HBAS_SC 0–48, 0.25 h X X X
0–48, 1 h X X
HTOP_DC 0–48, 1 h X X X
HTOP_SC 0–48, 0.25 h X X X
0–48, 1 h X X
HZEROCL 0–48, 0.25 h X X X
H_ICE 0–48, 1 h X X X
H_ML_LK 0–48, 1 h X X
H_SNOW 0–48, 1 h X X X
0–48, 1 h X X
LPI 0–48, 0.25 h X X X
0–48, 0.25 h X X
LPI_MAX 0–48, 1 h X X X
0–48, 1 h X X
PMSL 0–48, 1 h X X X
0–48, 1 h X X
PRG_GSP 0–48, 0.25 h X X X
PRR_GSP 0–48, 0.25 h X X X
PRS_GSP 0–48, 0.25 h X X X
PS 0–48, 1 h X X X
0–48, 1 h X X
QV_S 0–48, 1 h X X X
0–48, 1 h X X
RAIN_CON 0–48, 1 h X X
RAIN_GSP 0–48, 0.25 h X X X
0–48, 1 h X X
RELHUM_2M 0–48, 1 h X X X
0–48, 1 h X X
Continued on next page

94
 11.4. Single-level fields

Table 11.5.: continued

RHO_SNOW 0–48, 1 h X X X
0–48, 1 h X X
RUNOFF_G 0–48, 1 h X X X
0–48, 1 h X X
RUNOFF_S 0–48, 1 h X X X
0–48, 1 h X X
SDI_2 0–48, 0.25 h X X X
0–48, 1 h X X
SNOWC 0–48, 1 h X X
0–48, 1 h X X
SNOWLMT 0–48, 0.25 h X X X
SNOW_CON 0–48, 1 h X X
SNOW_GSP 0–48, 0.25 h X X X
0–48, 1 h X X
SYNMSG_BT_CL_IR10.8 0–48, 0.25 h X X X
SYNMSG_BT_CL_WV6.2 0–48, 0.25 h X X X
TCH 0–48, 1 h X X
TCM 0–48, 1 h X X
TCOND10_MAX 0–48, 1 h X X X
0–48, 1 h X X
TCOND_MAX 0–48, 1 h X X X
0–48, 1 h X X
TD_2M 0–48, 1 h X X X
0–48, 1 h X X
TMAX_2M 0–48, 1 h X X X
0–48, 1 h X X
TMIN_2M 0–48, 1 h X X X
0–48, 1 h X X
TOT_PREC 0–48, 0.25 h X X X
0–48, 0.25 h X X
TQC 0–48, 0.25 h X X X
0–48, 1 h X X
TQC_DIA 0–48, 1 h X X
0–48, 1 h X X
Continued on next page

95
Chapter 11. ICON-D2 output fields

Table 11.5.: continued

TQG 0–48, 0.25 h X X X

0–48, 1 h X X
TQI 0–48, 0.25 h X X X
0–48, 1 h X X
TQI_DIA 0–48, 1 h X X
0–48, 1 h X X
TQR 0–48, 0.25 h X X X
TQS 0–48, 0.25 h X X X
TQV 0–48, 0.25 h X X X
0–48, 1 h X X
TQV_DIA 0–48, 1 h X X
0–48, 1 h X X
TWATER 0–48, 1 h X X X
0–48, 1 h X X
T_2M 0–48, 1 h X X X
0–48, 1 h X X
T_BOT_LK 0–48, 1 h X X
T_G 0–48, 1 h X X X
0–48, 1 h X X
T_ICE 0–48, 1 h X X X
T_MNW_LK 0–48, 1 h X X
T_SNOW 0–48, 1 h X X X
0–48, 1 h X X
T_WML_LK 0–48, 1 h X X
UH_MAX 0–48, 1 h X X X
0–48, 1 h X X
U_10M 0–48, 1 h X X X
0–48, 1 h X X
VMAX_10M 0–48, 1 h X X X
0–48, 1 h X X
VORW_CTMAX 0–48, 1 h X X X
0–48, 1 h X X
V_10M 0–48, 1 h X X X
0–48, 1 h X X
Continued on next page

96
 11.5. Soil-specific multi-level fields

Table 11.5.: continued

WW 0–48, 1 h X X X
W_CTMAX 0–48, 1 h X X X
0–48, 1 h X X
W_I 0–48, 1 h X X X
0–48, 1 h X X
W_SNOW 0–48, 1 h X X X
0–48, 1 h X X
Z0 0–48, 1 h X X X

11.5. Soil-specific multi-level fields

• soil: soil levels = 0, 1, 3, 9, 27, 81, 243, 729 cm

Table 11.6.: Soil-specific multi-level fields

native
Time range latlon
EPS
Det.

ShortName level type

SMI 0–48, 1 h X X X soil

0–48, 1 h X X soil
T_SO 0–48, 1 h X X X soil
0–48, 1 h X X soil
W_SO 0–48, 1 h X X X soil
0–48, 1 h X X soil
W_SO_ICE 0–48, 1 h X X X soil
0–48, 1 h X X soil

97
12. Ensemble forecasts with ICON (global,
nested, and limited area mode)
There are two ensemble systems based on ICON running under operational conditions at DWD. The
global ICON ensemble suite is operational since January 2018 and generates 40 short to medium range
forecasts at approx. 40 km (R2B06) horizontal resolution on the global scale with a 20 km (R2B07)
nesting area over Europe. It runs 8 times a day providing boundary conditions for the ICON-D2-EPS.
At 00/06/12/18 UTC, the EU-nest runs within the global model up to 120h. At 00/12 UTC, only the
global system is further integrated up to 180 h lead time. Whereas at 03/09/15/21 UTC the maximum
forecast lead time is limited to +30 h for both, the global domain and the EU nest.
With the operational start of ICON-D2 (2021-02-10), a convection-permitting ensemble ICON-D2-EPS
(analogous to the former COSMO-D2-EPS) with 20 members and the same resolution and grid as the
deterministic ICON-D2 (≈ 2 km, R19B07) is performed. It runs 8 times a day with 48 hours of forecasts
for the 00, 06, 09, 12, 15, 18, 21 UTC runs. Prior to 2021-06-23 the forecast range was limited to 27
hours, with the exception of the 03 UTC run (45 hours).
The main purpose of an ensemble system is to estimate forecast uncertainty by running a number
of possible physically consistent scenarios of future development. The different scenarios arise from
uncertainties in initial conditions and model error. For limited area ensembles, an additional source of
forecast uncertainty is the uncertainty in the boundary conditions. In the following sections we explain
the techniques used in the ICON-EPS to simulate the effects of those error sources on the forecast and
describe its output data.

12.1. Initial Perturbations

In the ICON ensembles the initial perturbations are set by the EDA system for the ICON EPS (global
domain and EU nest, Section 5.1) and the KENDA system for ICON-D2-EPS (Section 5.3.2). Both
systems are based on a Local Ensemble Transform Kalman Filter (LETKF). The implementation of the
filter follows the paper of Hunt et al. (2007). The algorithm establishes an assimilation cycle of 3 hours
(1 hour for KENDA) and solves the underlying equations in ensemble space spanned by a background
ensemble of 40 members.
Since the ICON-D2-EPS runs with 20 forecast members, only the first 20 of the 40 KENDA members
are used as initial conditions for ICON-D2-EPS. Over a large number of ICON-D2-EPS runs, this is
statistically equivalent to a random selection of 20 members as there are no statistically distinguishable
members in KENDA by construction. More details of the implementation can be found in Schraff et al.
(2016) and Section 5 of this document.

In the context of ensemble forecasting it is important to note that the LETKF establishes a square
root filter with multiple variance inflation techniques (see Anlauf et al. (2017), Freitag and Potthast
(2013), Schraff et al. (2016) The ”Kalman gain” from adding observations reduces the uncertainty in the
analysis and thus the variance in the analysis ensemble. By the time this would lead to underestimation
of the true background error compared to the observation error and the analysis ensemble must be
re-inflated in each analysis step to stabilise the ensemble variance. The analysis increments as well as
the partly random variance inflation techniques introduce imbalances in the initial states of the forecast
ensemble, which are damped using an incremental analysis update scheme (IAU; Section 5.2). All the
modifications from the analysis cycle lead to a new analysis ensemble. The new properties and relative

99
Chapter 12. Ensemble forecasts with ICON (global, nested, and limited area mode)

arrangements of the analysis members determine the spread growth and thus the quality of the forecast
uncertainty estimation.

12.2. Ensemble Physics Perturbations

To simulate the model error a simple methodology for perturbing various physics tuning parameters
has been implemented in ICON for the ensemble mode.
At the beginning of each forecast the actual values of a predefined set of tuning parameters are calculated
using a random number generator. The user can specify a range for the variation of each parameter in
the namelist ensemble_pert_nml. The randomised perturbation of physics is activated by setting the
parameter use_ensemble_pert=.true. in the same namelist.
For most parameters, the perturbation is applied in an additive symmetric way:

pert_param = ref _param + 2. ∗ (rand _num − 0.5) ∗ range ,

where rand _num ∈ [0, 1] is drawn from a uniform distribution.

For a few exceptions, only positive variations are retained (implying that no perturbations are applied for
random numbers below 0.5, e.g. capdcfac_et), or the perturbation is multiplicative (e.g. rlam_heat).
There are two options to control the randomisation using parameters in the same namelist:
timedep_pert: The randomisation depends on the member ID with timedep_pert=0. With timedep_pert=1,
it depends on the forecast start time (not taking into account time differences finer than hours)
and the member ID. A value of timedep_pert=2 results in time-dependent perturbations vary-
ing sinusoidally within their range. The randomisation is accomplished by a phase shift of the
sinusoidal wave depending on the member ID.
itype_pert_gen: The random number rand _num is used at face value (itype_pert_gen=1 ) or is set
to 0 for rand _num < 0.5 or 1 otherwise (itype_pert_gen=2 ).
In the current implementation, the set-up is timedep_pert=0 and itype_pert_gen=1 for the global
ICON-EPS and its EU nest whereas the set-up is timedep_pert=1 and itype_pert_gen=2 for ICON-
D2-EPS, i.e. only the boundary values of the specified range of a parameter are used besides its default
value.
Table 12.1 lists the parameters which are perturbed. Parameter ranges can be slightly different between
EPS set-ups:

Table 12.1.: List of parameters which are perturbed in the global/EU ICON-EPS (glo) and/or ICON-D2-
EPS (D2). The perturbation mode is either additive (a) or multiplicative (m), possibly limited
to positive (+) deviations from the default value.

Parameter Description glo D2 mode

gkwake Low level wake drag constant X a

gkdrag Gravity wake drag constant X a
gfrcrit Critical Froude number X a
entrorg Entrainment parameter in convection scheme X X a
valid for dx=20 km (depends on model resolu-
tion)
q_crit Critical value for normalised super-saturation X a+
zvz0i Terminal fall velocity of ice X X a
Continued on next page

100
 12.2. Ensemble Physics Perturbations

Table 12.1.: continued

Parameter Description glo D2 mode

rprcon Coefficient for conversion of cloud water into X a
precipitation. If perturbed, its perturbation is
forced to be anticorrelated to the perturbation
of zvzi0 to compensate for a temperature bias
in the upper tropical troposphere.
rdepths Maximum allowed depth of shallow convection X a
rain_n0fac Intercept parameter of raindrop size distribu- X X m
tion
gfluxlaun Variability range for non-orographic gravity X a
wave launch momentum flux
capdcfac_et Maximum fraction of CAPE diurnal cycle cor- X a+
rection applied in the extratropics
capdcfac_tr Maximum fraction of CAPE diurnal cycle cor- X a+
rection applied in the tropics
lowcapefac Tuning parameter for diurnal-cycle correction X a
in convection scheme: reduction factor for
low-cape situations.
negpblcape Tuning parameter for diurnal-cycle correc- X a
tion in convection scheme: maximum negative
PBL CAPE allowed in the modified CAPE
closure.
rhebc_land RH threshold for onset of evaporation below X a
cloud base over land
rhebc_ocean RH threshold for onset of evaporation below X a
cloud base over sea
rcucov Convective area fraction used for computing X a
evaporation below cloud base
texc Excess value for temperature used in test par- X X a
cel ascent
qexc Excess fraction of grid-scale QV used in test X X a
parcel ascent
box_liq Box width for liquid cloud diagnostic in cloud X X a
cover scheme
box_liq_asy Asymmetry factor for liquid cloud cover diag- X X a
nostic
tur_len Asymptotic maximal turbulent distance X X a
tkhmin Scaling factor for minimum vertical diffusion X X a
coefficient (proportional to Ri−2/3 ) for heat
and moisture
Continued on next page

101
Chapter 12. Ensemble forecasts with ICON (global, nested, and limited area mode)

Table 12.1.: continued

Parameter Description glo D2 mode

tkmmin Scaling factor for minimum vertical diffusion X X a
coefficient (proportional to Ri−2/3 ) for mo-
mentum
rlam_heat Scaling factor of the laminar boundary layer X X m
for heat (scalars). The change in rlam_heat is
accompanied by an inverse change of rat_sea
in order to keep the evaporation over water
(controlled by rlam_heat∗rat_sea) the same.
a_hshr Length scale factor for the separated horizon- X a
tal shear mode
a_stab Factor for stability correction of turbulent X a+
length scale
c_diff Length scale factor for vertical diffusion of X m
TKE
c_soil Evaporating fraction of soil X a
minsnowfrac Lower limit of snow cover fraction to which X a
melting snow is artificially reduced in the con-
text of the snow-tile approach
cwimax_ml Scaling parameter for maximum interception X m
storage
alpha0 Lower bound of velocity-dependent Charnock X a/m
parameter

12.3. Lateral boundary perturbations (limited area EPS)

For models running in limited area mode, the provision of lateral boundaries is an additional source of
forecast uncertainty.
This is accounted for in ICON-D2-EPS by using as hourly lateral boundary conditions forecast members
of the ICON-EU-EPS which have started 3 hours before the start of the respective ICON-D2-EPS. Only
the members 1–20 of the 40 members of ICON-EU-EPS are used for driving the 20 members of ICON-
D2-EPS. As for the reduction to KENDA members 1–20 for the perturbation of initial conditions
(Section 12.1), this is statistically equivalent to a random selection of 20 members of ICON-EU-EPS
over a large number of ICON-D2-EPS runs. However, there are promising approaches in the research
community to select a suitable subset of 20 out ouf 40 members at each forecast following certain
optimisation criteria. The application and suitability of such methods has not yet been tested for the
combination of ICON-EU-EPS and ICON-D2-EPS.

12.4. ICON Ensemble output fields in DWD databases

In SKY the data is stored in different categories and data base subsystems. These are identified by the
cat=CAT_NAME parameter. The ICON ensemble categories start with the string ico for ICON data on
its native grid. Next follows a two-letter string to identify the domain of ICON: gl for the global domain,
eu for the nest over Europe, la for the limited area domain. The ensemble data is further characterised

102
 12.4. ICON Ensemble output fields in DWD databases

by a final e. The category parameters run, type, and suite have the same meaning for all forecast
models of DWD. See section 13.1 for an explanation and available values. Hence, the full category
name for data from an operational ensemble forecast run of the ICON-EPS is icogle_main_fc_rout
for a global field and icoeue_main_fc_rout for the nesting area over Europe. For ICON-D2-EPS
the full category name is icolae_main_fc_vera. The ensemble output data is stored exclusively on
the native grid. For interpolation to other grid types, please use postprocessing software like CDO (or
fieldextra), which are able to read native ICON grids. The instructions manual for interpolating ICON
model fields with CDO can be found on the DWD web pages https://www.dwd.de by typing CDO in
the search tool of the web page.
Ensemble members or ranges of ensemble members are specified in the SKY language by the parameter
enum=N U M or enum=N U M 1 − N U M 2 where N U M is the member id. See Section 13 for SKY
retrieval examples.

12.4.1. Model Output

The model output fields are collected in the tables of chapter 9 (ICON global), 10 (ICON-EU nest),
and 11 (for ICON-D2).

12.4.2. Ensemble Products for the ICON-EPS (global) and ICON-EU-EPS

The EPS products are stored in the roma database and can be identified by the category type fcprod.
This leads to the category name icreue_main_fcprod_rout for ensemble products on the EU domain
and icrgle_main_fcprod_rout for global products. The products are generated with fieldextra
on regular latitude/longitude grids with resolutions of 0.25◦ and 0.5◦ , respectively. The ICON-EPS
provides ensemble products in three different categories, which can be accessed by using the SKY bank
parameters derivedForecast (deriv), percentile (perc) and exceedance probability (probt):

1. Mean and extreme values

Unweighted mean of all members (deriv = 0)
Spread of all members (deriv = 4)
Minimum of all ensemble members (deriv = 8)
Maximum of all ensemble members (deriv = 9)
2. Percentiles,
i.e. physical values of a forecast parameter (e.g. T_2M,. . . ), which define the
perc=10,25,50,75,90 [%] parts of the ensemble distribution.
3. Exceedance Probabilities
Probability of event above lower limit (probt=3)
Probability of event below upper limit (probt=4)

The thresholds for the exceedance probabilities are given by the DWD alert thresholds which are used
for issuing weather warnings1 and follow the WMO recommendations2 for the global fields. Ensemble
products are generated every 6 hours up to 120 h lead time on the EU domain and for the global fields
up to 180 h twice a day (00/12 UTC). This is done for different accumulation periods depending on
the forecast parameter. In the following tables, the accumulation time range is given in hours. The
meaning of the level types is:
• (no key): 2D field

1 https://www.dwd.de/DE/wetter/warnungen_aktuell/kriterien/warnkriterien.html
2 http://www.wmo.int/pages/prog/www/DPS/Publications/WMO_485_Vol_I.pdf

103
Chapter 12. Ensemble forecasts with ICON (global, nested, and limited area mode)

• pe4: pressure level 500 hPa

• pe7: pressure level 850 hPa
• pe8: pressure levels 500, 850 hPa
• pe9: pressure levels 250, 500, 850 hPa
All products are delivered only on the lat/lon grid (i.e. not on the ICON native grid).

Table 12.2.: EPS products from ICON global. See table 12.4 for a description of the various product types.

level type
Time range
ShortName Accum. Product
timerange type

CAPE_ML 6–180, 6 h perct_1

6–180, 6 h probt_12
CLCH 6–180, 6 h perct_1
6–180, 6 h probt_13
CLCL 6–180, 6 h perct_1
6–180, 6 h probt_13
CLCM 6–180, 6 h perct_1
6–180, 6 h probt_13
CLCT 6–180, 6 h perct_1
6–180, 6 h probt_13
FI 6–180, 6 h pe4 perct_1
PMSL 6–180, 6 h perct_1
SP 6–180, 6 h pe9 perct_1
SP_10M 6–180, 6 h perct_1
6–180, 6 h probt_8
T 6–180, 6 h pe8 perct_1
6–180, 6 h pe7 probt_14
TD_2M 6–180, 6 h perct_1
TMAX_2M 12–180, 6 h 720 perct_1
12–180, 6 h 720 probt_11
12–180, 6 h 1440 perct_1
12–180, 6 h 1440 probt_11
TMIN_2M 12–180, 6 h 720 perct_1
12–180, 6 h 720 probt_10
12–180, 6 h 1440 perct_1
Continued on next page

104
 12.4. ICON Ensemble output fields in DWD databases

Table 12.2.: continued

12–180, 6 h 1440 probt_10

TOT_PREC 6–180, 6 h 360 perct_1
6–180, 6 h 360 probt_4
6–180, 6 h 720 perct_1
6–180, 6 h 720 probt_5
6–180, 6 h 1440 perct_1
6–180, 6 h 1440 probt_3
6–180, 6 h 2880 perct_1
6–180, 6 h 2880 probt_6
6–180, 6 h 4320 perct_1
6–180, 6 h 4320 probt_7
TOT_SNOW 6–180, 6 h 360 perct_1
6–180, 6 h 360 probt_1
6–180, 6 h 720 perct_1
6–180, 6 h 720 probt_2
6–180, 6 h 1440 perct_1
6–180, 6 h 1440 probt_2
T_2M 6–180, 6 h perct_1
T_G 12–180, 6 h 720 perct_1
12–180, 6 h 720 probt_9
VMAX_10M 6–180, 6 h 360 perct_1
6–180, 6 h 360 probt_8
6–180, 6 h 720 perct_1
6–180, 6 h 720 probt_8
6–180, 6 h 1440 perct_1
6–180, 6 h 1440 probt_8

Table 12.3.: EPS products from ICON-EU. See table 12.4 for a description of the various product types.
level type

Time range
ShortName Accum. Product
timerange type

CAPE_ML 6–120, 6 h perct_1

Continued on next page

105
Chapter 12. Ensemble forecasts with ICON (global, nested, and limited area mode)

Table 12.3.: continued

6–120, 6 h probt_12
CLCH 6–120, 6 h perct_1
6–120, 6 h probt_13
CLCL 6–120, 6 h perct_1
6–120, 6 h probt_13
CLCM 6–120, 6 h perct_1
6–120, 6 h probt_13
CLCT 6–120, 6 h perct_1
6–120, 6 h probt_13
FI 6–120, 6 h pe4 perct_1
PMSL 6–120, 6 h perct_1
SP 6–120, 6 h pe9 perct_1
SP_10M 6–120, 6 h perct_1
6–120, 6 h probt_8
T 6–120, 6 h pe8 perct_1
TD_2M 6–120, 6 h perct_1
TMAX_2M 12–120, 6 h 720 perct_1
12–120, 6 h 720 probt_11
12–120, 6 h 1440 perct_1
12–120, 6 h 1440 probt_11
TMIN_2M 12–120, 6 h 720 perct_1
12–120, 6 h 720 probt_10
12–120, 6 h 1440 perct_1
12–120, 6 h 1440 probt_10
TOT_PREC 6–120, 6 h 360 perct_1
6–120, 6 h 360 probt_4
6–120, 6 h 720 perct_1
6–120, 6 h 720 probt_5
6–120, 6 h 1440 perct_1
6–120, 6 h 1440 probt_3
6–120, 6 h 2880 perct_1
6–120, 6 h 2880 probt_6
6–120, 6 h 4320 perct_1
6–120, 6 h 4320 probt_7
Continued on next page

106
 12.4. ICON Ensemble output fields in DWD databases

Table 12.3.: continued

TOT_SNOW 6–120, 6 h 360 perct_1

6–120, 6 h 360 probt_1
6–120, 6 h 720 perct_1
6–120, 6 h 720 probt_2
6–120, 6 h 1440 perct_1
6–120, 6 h 1440 probt_2
T_2M 6–120, 6 h perct_1
T_G 12–120, 6 h 720 perct_1
12–120, 6 h 720 probt_9
VMAX_10M 6–120, 6 h 360 perct_1
6–120, 6 h 360 probt_8
6–120, 6 h 720 perct_1
6–120, 6 h 720 probt_8
6–120, 6 h 1440 perct_1
6–120, 6 h 1440 probt_8

Table 12.4.: Meaning of the ’product type’ key for EPS-products from ICON (global) and ICON-EU EPS.

Product type Description

perct_1 perct; mean; spread; min; max; 10; 25; 50; 75; 90 (ensemble distribution)
probt_1 1.0; 5.0; 10.0 (precipitation thresholds (in mm))
probt_2 1.0; 5.0; 10.0; 15.0; 20.0; 25.0; 30.0; 50.0 (precipitation thresholds (in mm))
probt_3 1.0; 5.0; 10.0; 20.0; 25.0; 30.0; 50.0; 80.0; 100.0 (precipitation thresholds (in mm))
probt_4 20.0; 35.0; 60.0 (precipitation thresholds (in mm))
probt_5 25.0; 40.0; 70.0 (precipitation thresholds (in mm))
probt_6 40.0; 60.0; 90.0; 150.0 (precipitation thresholds (in mm))
probt_7 50.0; 100.0; 150.0; 250.0 (precipitation thresholds (in mm))
probt_8 11.0; 12.5; 14.0; 18.0; 21.0; 25.0; 29.0; 33.0; 39.0 (wind speed thresholds (in m/s))
probt_9 273.15 (temperature thresholds (in K))
probt_10 253.15; 263.15; 273.15; 293.15 (temperature thresholds (in K))
probt_11 263.15; 273.15; 298.15; 303.15; 308.15 (temperature thresholds (in K))
probt_12 750.0; 1000.0; 1500.0; 2000.0; 3000.0; 4000.0 (CAPE thresholds)
probt_13 50.0; 87.5 (cloud cover thresholds (in %))
probt_14 -1.0; -1.5; -2.0; 1.0; 1.5; 2.0 (temperature anomalie thresholds)

107
Chapter 12. Ensemble forecasts with ICON (global, nested, and limited area mode)

Most of the parameters are available on both domains, but there are exceptions: SP250, FI500,T_SO
and the temperature anomaly are available on the global domain only. The latter is calculated for thresh-
olds of ±1, ±1.5 and ±2 standard deviations with respect to the reanalysis climatology ERA_INTERIM3 .
The global products are available via the WMO WIS/WMS system or directly as grib files and charts
on the opendata server of DWD4 in /weather/wmc/icon-eps. A graphical user interface for direct access
to the charts is available on the DWD website5 . The dissemination of the EU-Nest ensemble product
grib files via the opendata server of DWD is planned to start in October 2018.

12.4.3. Ensemble Products for the ICON-D2-EPS

The ICON-D2-EPS products are stored in the vera database and can be identified by the category type
fcprod. This leads to the category name icrlae_main_fcprod_vera. The products are generated
with fieldextra on the COSMO-D2 grid (rotated latlon grid). The ICON-D2-EPS provides ensem-
ble products in three different categories, which can be accessed by using the SKY bank parameters
derivedForecast (deriv), percentile (perc) and exceedance probability (probt):
1. Mean and extreme values
Unweighted mean of all members (deriv = 0)
Spread of all members (Standard Deviation) (deriv = 4)
Spread of all members (Interquartile Range) (deriv = 7)
Minimum of all ensemble members (deriv = 8)
Maximum of all ensemble members (deriv = 9)
i.e. ’spread’ is computed as interquartile range for TOT_PREC and TOT_SNOW and as standard
deviation for all the other products.
2. Percentiles,
i.e. physical values of a forecast parameter (e.g. T_2M,. . . ), which define the
perc=10,25,30,50,75,90 [%] parts of the ensemble distribution.
3. Exceedance Probabilities
Probability of event above lower limit (probt=3)
Probability of event below upper limit (probt=4)
4. Grid specification
Products on the rotated latlon COSMO-D2 grid (localTypeOfEnsembleProductGeneration=1)
Products on the upscaled (10x10 grid point) grid (localTypeOfEnsembleProductGeneration=101)

The thresholds for the exceedance probabilities are given by the DWD alert thresholds which are used
for issuing weather warnings6 and follow the WMO recommendations7 .
In addition, exceedance probability products for TOT_PREC and LPI are generated on an upscaled grid,
10x10 grid points. In order to distinguish the products on the upscaled grid from the ones on the
COSMO-D2 grid, the grib parameter localTypeOfEnsembleProductGeneration is used.
Ensemble products are generated every 1 hour (not for all parameters, see the tables) up to 27 h lead
time eight times per day (00/03/06/09/12/15/18/21 UTC). This is done for different accumulation
periods, depending on the forecast parameter. In the following tables, the accumulation time range is
given in minutes.

3 https://www.ecmwf.int/en/forecasts/datasets/archive-datasets/reanalysis-datasets/era-interim
4 https://www.dwd.de/opendata
5 https://www.dwd.de/EN/ourservices/wmc/wmc.html
6 https://www.dwd.de/DE/wetter/warnungen_aktuell/kriterien/warnkriterien.html
7 http://www.wmo.int/pages/prog/www/DPS/Publications/WMO_485_Vol_I.pdf

108
 12.4. ICON Ensemble output fields in DWD databases

Table 12.5.: EPS products from ICON-D2. See table 12.6 for a description of the various product types.

level type
Time range
ShortName Accum. Product
timerange type

CAPE_ML 6–24, 6 h 360 perct_1d

6–24, 6 h 360 probt_7d
CLCL 1–48, 1 h perct_1d
1–48, 1 h probt_6d
CLCT 1–48, 1 h perct_1d
1–48, 1 h probt_6d
DBZ_CMAX 1–48, 1 h perct_1d
1–48, 1 h probt_13d
LPI 1–48, 1 h perct_1d
1–48, 1 h probt_14d
1–48, 1 h probt_14d_ups
TMAX_2M 6–24, 6 h 360 perct_1d
6–24, 6 h 360 probt_12d
12–24, 12 h 720 perct_1d
12–24, 12 h 720 probt_12d
TMIN_2M 6–24, 6 h 360 perct_1d
6–24, 6 h 360 probt_11d
12–24, 12 h 720 perct_1d
12–24, 12 h 720 probt_11d
TOT_PREC 6–48, 1 h 360 perct_1d
6–48, 1 h 360 probt_3d
1–48, 1 h 60 perct_1d
1–48, 1 h 60 probt_1d
12–48, 1 h 720 perct_1d
12–48, 1 h 720 probt_9d
12–48, 1 h 720 probt_9d_ups
1–48, 1 h 60 probt_1d_ups
6–48, 1 h 360 probt_3d_ups
TOT_SNOW 6–48, 1 h 360 perct_1d
6–48, 1 h 360 probt_5d
Continued on next page

109
Chapter 12. Ensemble forecasts with ICON (global, nested, and limited area mode)

Table 12.5.: continued

1–48, 1 h 60 perct_1d
1–48, 1 h 60 probt_4d
12–48, 1 h 720 perct_1d
12–48, 1 h 720 probt_2d
T_2M 1–48, 1 h perct_1d
1–48, 1 h probt_11d
T_G 1–48, 1 h perct_1d
1–48, 1 h probt_10d
VMAX_10M 6–24, 6 h 360 perct_1d
6–24, 6 h 360 probt_8d
1–48, 1 h 60 perct_1d
1–48, 1 h 60 probt_8d
12–24, 12 h 720 perct_1d
12–24, 12 h 720 probt_8d

110
 12.4. ICON Ensemble output fields in DWD databases

Table 12.6.: Meaning of the ’product type’ key for EPS-products from ICON-D2-EPS.

Product type Description

perct_1d perct; mean; spread; min; max; 10; 25; 30; 50; 75; 90 (ensemble distribution)
probt_1d_ups as probt_1d, but upscaled to 10x10 gridpoints
probt_1d > 0.1; > 1.0; > 2.0; > 5.0; > 10.0; > 15.0; > 25.0; > 40.0 (probability thresholds)
probt_2d > 0.1; > 1.0; > 5.0; > 10.0; > 15.0; > 20.0; > 25.0; > 30.0; > 50.0 (probability
thresholds)
probt_3d > 0.1; > 1.0; > 2.0; > 5.0; > 10.0; > 20.0; > 35.0; > 60.0 (probability thresholds)
probt_3d_ups as probt_3d, but upscaled to 10x10 gridpoints
probt_4d > 0.1; > 1.0; > 2.0; > 5.0 (probability thresholds)
probt_5d > 0.1; > 5.0; > 10.0 (probability thresholds)
probt_6d < 50.0; > 87.5 (probability thresholds)
probt_7d > 750.0; > 1000.0; > 1500.0; > 2000.0 (probability thresholds)
probt_8d > 11.0; > 12.5; > 14.0; > 18.0; > 21.0; > 25.0; > 29.0; > 33.0; > 39.0
(probability thresholds)
probt_9d > 25.0; > 40.0; > 70.0 (probability thresholds)
probt_9d_ups as probt_9d, but upscaled to 10x10 gridpoints
probt_10d < 273.15 (probability thresholds)
probt_11d < 253.15; < 263.15; < 273.15; >= 293.15; >= 298.15; >= 303.15; >= 308.15
(probability thresholds)
probt_12d < 263.15; < 273.15; >= 298.15; >= 303.15; >= 308.15 (probability thresholds)
probt_13d > 28.0; > 37.0; > 46.0; > 54.0 (probability thresholds)
probt_14d > 0.1; > 1.0; > 5.0; > 10.0; > 20.0 (probability thresholds)
probt_14d_ups as probt_14d, but upscaled to 10x10 gridpoints

111
13. ICON data in the
SKY data bases of DWD
GRIB data of the numerical weather prediction models are stored in the data base SKY at DWD.
Documentation on the SKY system is available in the intranet of DWD at IT/Messnetz/Technik →
Datenmanagement (technisch) → Management der DWD Fachdaten -Dokumentation → SKY. Here,
some remarks are given on the SKY categories for ICON data, and some examples are given how to
retrieve data from the data base.

13.1. SKY categories for ICON

In SKY the data is stored in different categories and data base subsystems. These are identified by the
cat=CAT_NAME parameter. The name of a category is made up of 4 parts:

$model_$run_$type_$suite

run, type, and suite are general for all forecast models of DWD. They can have the following values:
• run: main for main forecast runs, ass for assimilation runs, pre for pre-assimilation runs, const
for invariant data.
• type: an for analysis data, fc for forecast data, fcprod for EPS products.
• suite:
– rout for operational data in db=roma,
– para1 for pre-operational data in db=parma, or vera for pre-operational data in db=vera.
– exp or exp1 for data from experiments in db=numex. The category extension exp1 is used
for experiments of the NUMEX wizard, a special NUMEX user.
Data from experiments is additionally identified by the parameter exp=N U M where N U M
is the experiment number.
The model part of the sky-categories for ICON itself is constructed by several substrings. It starts
• with the string ico for ICON data on the native ICON grid, or
• with icr for data on a regular lat-lon or a rotated lat-lon grid.
Next follows a two-letter string to identify the domain of ICON;
• gl for the global domain,
• eu for the nest over Europe,
• la for the limited-area model ICON-D2.
Further particulars of the category name differ for the global and nested domain. In case of deterministic
runs of the global domain, gl is followed by the mesh width of the model in units of 100 m, and then
the number of levels preceded by the letter l. As an example icogl130l90 is on the native grid from a
global model with a mesh width 13 km (grid R3B07) and 90 levels. icrgl400l90 is data on a regular
grid from a global model with mesh width 40 km (R2B06) and 90 levels.

113
Chapter 13. ICON data in the SKY data bases of DWD

For the nested domain, the specification of the mesh width and number of levels is omitted. As an
example, icreu is the ICON nest over Europe (with a mesh width of 6.5 km and 60 levels), interpolated
to a regular lat-lon grid.
The same holds for the limited-area model ICON-D2: e.g., icrla is the ICON-D2 interpolated to a
rotated lat-lon grid.
For ensemble forecasts or ensemble analyses the first part of the category is extended by an e (for
instance icogle or icrgle). Ensemble members or ranges of ensemble members are specified by the
parameter enum=N U M or enum=N U M 1/to/N U M 2 where N U M is the member id.

Hence, the full category name for data from a global operational deterministic forecast run of ICON on
a regular grid will be icrgl130l90_main_fc_rout. The initial analysis for this run is in category
icogl130l90_main_an_rout.

Since 2014-08-12 12 UTC ICON is running pre-operationally at DWD. Hence, forecast data
was available in the sky database db=parma in categories icogl130l90_main_fc_para and
icrgl130l90_main_fc_para. Data of the present pre-operational ICON runs is in db=parma
in categories icogl130l90_main_fc_para1 and icrgl130l90_main_fc_para1.

Since 2015-01-20 06 UTC the global ICON model is running operationally at DWD. Forecast
data is available in the sky database db=roma in categories icogl130l90_main_fc_rout and
icrgl130l90_main_fc_rout.

Since 2016-01-20 06 UTC an ensemble data assimilation for ICON is running operationally
at DWD. Analysis data is available in the sky database db=roma in the ensemble cat-
egories icogle_main_an_rout and icoeue_main_an_rout. First guess data is in
icogle_pre_fc_rout and icogle_pre_fc_rout. The ensemble runs of ICON write data only
on the native ICON grid. Data on regular grids must be interpolated from the native grid.

Since 2018-01-17 06 UTC the global ICON-EPS is running operationally at DWD. Forecast
data is available in the sky database db=roma in categories icogle_main_fc_rout and
icoeue_main_fc_rout for the EU domain.

Since 2019-11-25 00 UTC the limited-area model ICON-D2 is running pre-operationally at DWD.
Forecast data is available in the sky database db=vera in categories icola_main_fc_vera,
icrla_main_fc_vera, icolae_main_fc_vera, icrlae_main_fc_vera.

Data from the operational forecast runs of ICON on the native ICON grid, cat =
ico*_main_fc_rout, is kept in the database roma only for 15 months! Analysis and first
guess data is kept forever.

13.2. Retrieving ICON data from SKY

Here we shall give several examples how to retrieve ICON data from SKY. The parameter d specifies
the reference or initial date, s is the forecast step, p the parameter or variable name, and f the name of
the GRIB data file.

114
 13.2. Retrieving ICON data from SKY

13.2.1. Deterministic products

ICON global

• Retrieve the 2 m temperature and dew point temperature for forecast hours 3 to 78 every 3 hours
of today’s run at 00 UTC on the global domain from an ICON run on a R3B07 grid with 90 levels
to file icon2mdat
read db = roma cat = i c o g l 1 3 0 l 9 0 _ m a i n _ f c _ r o u t d = t00 s [ h ]=3/ to /78/ by /3 p = t_2m , td_2m bin
f = icon2mdat

• Retrieve the analysis of T on the native grid and model levels (lv=genv) from yesterday 18 UTC:
read db = roma cat = i c o g l 1 3 0 l 9 0 _ m a i n _ a n _ r o u t d = t18 -1 d p = T lv = genv gptype =0 bin f =
t_icon_ana

• Get the 6, 12, 18, and 24 hour forecast of the 2 m temperature from a forecast in experiment 10096
on 2015-09-05 at 00 UTC from an ICON run on a R3B07 grid with 90 levels. Retrieve data on
the regular lat/lon grid:
read db = numex cat = i c r g l 1 3 0 l 9 0 _ m a i n _ f c _ e x p exp =10096 d =2015090500 s [ h ]=6 ,12 ,18 ,24 p =
t_2m bin f = t_2m_fc . grb

• Retrieve all available time-invariant (constant) fields on the native grid and store them in the file
const_icongl. Write reference date (d), forecast step (s), level type (lv), value of first level (lv1),
decoding date (dedat), store date (stdat), and validity date (valdat) in information file icr.info. It
is important to set invar=true.
read db = roma cat = i c o g l 1 3 0 l 9 0 _ c o n s t _ a n _ r o u t invar = true info = metaData metaArray =d ,s ,p
, lv , lv1 , dedat , stdat bin infof = icr . info f = const_icongl

ICON-EU

• Retrieve accumulated precipitation of the ICON-EU nest on the regular grid every 6 hours to 72
hours from the yesterday’s operational run at 12 UTC:
read db = roma cat = i c r e u _ m a i n _ f c _ r o u t d = t12 -1 d s [ h ]=6/ to /72/ by /6 p = tot_prec bin f =
tot_prec_ieu

• List the data on pressure levels of the 18 hours forecast from 06 UTC of ICON-EU nest on the
regular grid. Write reference date (d), forecast step (s), level type (lv), value of first level (lv1),
decoding date (dedat), and store date (stdat) in information file icr.info.
read db = roma cat = i c r e u _ m a i n _ f c _ r o u t d =06 step [ h ]=18 lv = P info = metaData metaArray =d ,
s ,p , lv , lv1 , dedat , stdat sort =p , lv1 infof = icr . info

ICON-D2

• Retrieve accumulated precipitation of the pre-operational ICON-D2 on the rotated lat-lon grid
every 3 hours to 27 hours from the first day of the pre-operational start at 12 UTC:
read db = vera cat = i c r l a _ m a i n _ f c _ v e r a d =2019112512 s [ h ]=3/ to /27/ by /3 p = tot_prec bin f
= tot_prec_id2

• Retrieve the surface net short-wave radiation flux (averaged since model start) of the operational
ICON-D2 on the rotated lat-lon grid every hour to 12 hours for today’s run at 00 UTC:
read db = roma cat = i c r l a _ m a i n _ f c _ r o u t d = t00 s [ h ]=0/ to /12/ by /1 p = asob_s bin f =
asob_s_id2

115
Chapter 13. ICON data in the SKY data bases of DWD

13.2.2. Ensemble products

ICON global

• ICON-EPS: Retrieve the 2 m temperature and dew point temperature for forecast hours 3 to 78
every 3 hours of today’s run at 00 UTC on the global domain from an ICON-EPS run on a R2B06
grid to file iconEPS2mdat
read db = roma cat = i c o g l e _ m a i n _ f c _ r o u t d = t00 s [ h ]=3/ to /78/ by /3 p = t_2m , td_2m bin f =
iconEPS2mdat

• Retrieve temperature in 850 hPa from the first guess of the 40 ensemble members of the EDA on
the 40 km grid in the parallel suite yesterday at 21 UTC. Sort the data by ensemble member.
read db = parma cat = i c o g l e _ a s s _ f c _ p a r a 1 enum =1/ to / d = t21 -1 d s =3 p = T lv = P lv1 =85000
bin f = T850_eps sort = enum

ICON-EU

• ICON-EPS: Retrieve 90% percentile (on regular lat/lon grid) of accumulated precipitation (avail-
able accumulation periods) at forecast hour 72 of today’s run at 06 UTC on the EU domain.
read db = roma cat = i c r e u e _ m a i n _ f c p r o d _ r o u t d = t06 s [ h ]=72 perc =90 p = TOT_PREC bin f =
i co nE P S_ RR 7 2_ 9 0

• Retrieve ensemble spread (deriv=4) of CAPE for forecast hours 6 to 120 every 6 hours of today’s
run at 00 UTC on the EU domain (on regular lat/lon grid).
read db = roma cat = i c r e u e _ m a i n _ f c p r o d _ r o u t d = t00 s [ h ]=6/ to /120/ by /6 deriv =4 p = CAPE_ML
bin f = i c o n E P S _ C A P E _ s p r e a d

• Retrieve probabilities of TMIN_2M of the last 12 h and 24 h for any available threshold, where the
probability of event is above lower limit (probt=3), for all available forecast hours of today’s run
at 00 UTC on the EU domain.
read db = roma cat = i c r e u e _ m a i n _ f c p r o d _ r o u t d = t00 probt =3 p = TMIN_2M bin f =
iconEPS_TMIN_2M_probt3

ICON-D2

• Retrieve ensemble spread (deriv=4) of T_2M for all forecast hours (from 0 to 27 every hour) of
a run starting at 00 UTC.
read db = vera cat = i c r l a e _ m a i n _ f c p r o d _ v e r a d =2020050100 s [ h ]=0/ to /27/ by /1 deriv =4 p =
T_2M bin f = i c o l a e _ T 2 M _ s p r e a d

• Retrieve probabilities of TOT_PREC for any available threshold, where the probability of event
is above lower limit (probt=3), for the forecast hours between 0 and 24 with 6-hourly step, all
available accumulations (1h, 6h, 12h), of one run at 00 UTC.
read db = vera cat = i c r l a e _ m a i n _ f c p r o d _ v e r a d =2020050100 s [ h ]=0/ to /24/ by /6 probt =3 p =
TOT_PREC bin f = icolae _TP_pr ob

• Retrieve 90% percentile of CAPE_ML (6h accumulation period) at forecast hour 12 of today’s
run at 00 UTC.
read db = vera cat = i c r l a e _ m a i n _ f c p r o d _ v e r a d = t00 s [ h ]=12 perc =90 p = CAPE_ML bin f =
icolae_CAPE_perc

116
 117

A. ICON standard level heights

A.1. Level heights for zero topography height

ICON standard half level heights z h0 are listed in Table A.1 for ICON global/ICON-EU and in Table A.3
for ICON-D2. Please note that these values correspond to the actual level heights only at grid points
with zero topography height, e. g. at ocean grid points.
If full level heights z f 0 are required, these can be deduced as follows: Let i denote the full level index
for which the height is wanted. Then the full level height zif 0 is given by

zih0 + zi+1
h0
zif 0 = .
2
See Table A.2 for a list of all full level heights of the operational setup for ICON global/ICON-EU and
Table A.4 for ICON-D2.

A.2. Non-zero topography heights

The prerequisite ”zero topography height” is seldom met in real applications. Instead the user has to
compute the model level height for each grid point separately. To this end the invariant fields HSURF
and HHL are provided where HHL is the geometric height of model half levels above sea level. The level
height above ground can therefore be computed by the following formula:

zih (x) = HHL(x) − HSURF(x)

zih (x) + zi+1

h
(x)
zif (x) =
2

As an example, Tables A.5 and A.6 show these model heights for a special grid point over India with a
quite high surface elevation.
118

Table A.1.: Standard heights zih0 (i.e. for zero topography height) for all 91 vertical half levels of the global
13 km domain and the 61 vertical half levels for the 6.5 km EU nest.

level index height level index height level index height

global EU nest [m] global EU nest [m] global EU nest [m]

1 - 75 000.000 32 2 21 569.375 63 33 6 621.524

2 - 72 363.546 33 3 20 731.107 64 34 6 221.524
3 - 69 842.381 34 4 19 942.837 65 35 5 821.524
4 - 67 357.797 35 5 19 201.585 66 36 5 421.524
5 - 64 946.444 36 6 18 504.545 67 37 5 033.731
6 - 62 606.299 37 7 17 849.081 68 38 4 659.952
7 - 60 335.466 38 8 17 232.713 69 39 4 300.121
8 - 58 132.167 39 9 16 653.108 70 40 3 954.183
9 - 55 976.216 40 10 16 108.074 71 41 3 622.092
10 - 53 877.930 41 11 15 595.549 72 42 3 303.815
11 - 51 824.685 42 12 15 113.594 73 43 2 999.329
12 - 49 826.951 43 13 14 660.386 74 44 2 708.624
13 - 47 890.748 44 14 14 234.210 75 45 2 431.707
14 - 46 014.776 45 15 13 821.524 76 46 2 168.596
15 - 44 197.795 46 16 13 421.524 77 47 1 919.330
16 - 42 438.627 47 17 13 021.524 78 48 1 683.966
17 - 40 736.151 48 18 12 621.524 79 49 1 462.584
18 - 39 089.298 49 19 12 221.524 80 50 1 255.291
19 - 37 497.048 50 20 11 821.524 81 51 1 062.224
20 - 35 958.428 51 21 11 421.524 82 52 883.557
21 - 34 472.507 52 22 11 021.524 83 53 719.514
22 - 33 038.397 53 23 10 621.524 84 54 570.373
23 - 31 655.249 54 24 10 221.524 85 55 436.493
24 - 30 322.249 55 25 9 821.524 86 56 318.336
25 - 29 038.622 56 26 9 421.524 87 57 216.516
26 - 27 803.623 57 27 9 021.524 88 58 131.880
27 - 26 617.350 58 28 8 621.524 89 59 65.677
28 - 25 488.963 59 29 8 221.524 90 60 20.000
29 - 24 416.908 60 30 7 821.524 91 61 0.000
30 - 23 408.796 61 31 7 421.524
31 1 22 460.814 62 32 7 021.524
 119

Table A.2.: Standard heights zif 0 (i.e. for zero topography height) for all 90 vertical full levels of the global
13 km domain.

level index height [m] level index height [m] level index height [m]
1 73 681.773 31 22 015.095 61 7 221.524
2 71 102.963 32 21 150.241 62 6 821.524
3 68 600.089 33 20 336.972 63 6 421.524
4 66 152.120 34 19 572.211 64 6 021.524
5 63 776.371 35 18 853.065 65 5 621.524
6 61 470.883 36 18 176.813 66 5 227.628
7 59 233.817 37 17 540.897 67 4 846.842
8 57 054.191 38 16 942.910 68 4 480.037
9 54 927.073 39 16 380.591 69 4 127.152
10 52 851.308 40 15 851.812 70 3 788.138
11 50 825.818 41 15 354.572 71 3 462.954
12 48 858.849 42 14 886.990 72 3 151.572
13 46 952.762 43 14 447.298 73 2 853.976
14 45 106.285 44 14 027.867 74 2 570.165
15 43 318.211 45 13 621.524 75 2 300.151
16 41 587.389 46 13 221.524 76 2 043.963
17 39 912.725 47 12 821.524 77 1 801.648
18 38 293.173 48 12 421.524 78 1 573.275
19 36 727.738 49 12 021.524 79 1 358.938
20 35 215.467 50 11 621.524 80 1 158.757
21 33 755.452 51 11 221.524 81 972.891
22 32 346.823 52 10 821.524 82 801.536
23 30 988.749 53 10 421.524 83 644.943
24 29 680.436 54 10 021.524 84 503.433
25 28 421.123 55 9 621.524 85 377.415
26 27 210.487 56 9 221.524 86 267.426
27 26 053.157 57 8 821.524 87 174.198
28 24 952.936 58 8 421.524 88 98.779
29 23 912.852 59 8 021.524 89 42.839
30 22 934.805 60 7 621.524 90 10.000
120

Table A.3.: Standard heights zih0 (i.e. for zero topography height) for all 66 vertical half levels of ICON-D2.

level idx. height [m] level idx. height [m] level idx. height [m] level idx. height [m]
1 22 000.000 19 8 256.329 37 3 333.549 55 702.132
2 19 401.852 20 7 890.952 38 3 138.402 56 606.827
3 18 013.409 21 7 539.748 39 2 949.656 57 516.885
4 16 906.264 22 7 201.825 40 2 767.143 58 432.419
5 15 958.169 23 6 876.388 41 2 590.708 59 353.586
6 15 118.009 24 6 562.725 42 2 420.213 60 280.598
7 14 358.139 25 6 260.200 43 2 255.527 61 213.746
8 13 661.439 26 5 968.239 44 2 096.537 62 153.438
9 13 016.363 27 5 686.321 45 1 943.136 63 100.277
10 12 414.654 28 5 413.976 46 1 795.234 64 55.212
11 11 850.143 29 5 150.773 47 1 652.748 65 20.000
12 11 318.068 30 4 896.323 48 1 515.610 66 0.000
13 10 814.653 31 4 650.265 49 1 383.761
14 10 336.841 32 4 412.272 50 1 257.155
15 9 882.112 33 4 182.043 51 1 135.760
16 9 448.359 34 3 959.301 52 1 019.556
17 9 033.796 35 3 743.791 53 908.539
18 8 636.893 36 3 535.279 54 802.721
 121

Table A.4.: Standard heights zif 0 (i.e. for zero topography height) for all 65 vertical full levels of ICON-D2.

level idx. height [m] level idx. height [m] level idx. height [m] level idx. height [m]
1 20 700.926 18 8 446.611 35 3 639.535 52 964.048
2 18 707.630 19 8 073.640 36 3 434.414 53 855.630
3 17 459.836 20 7 715.350 37 3 235.976 54 752.427
4 16 432.216 21 7 370.787 38 3 044.029 55 654.479
5 15 538.089 22 7 039.106 39 2 858.399 56 561.856
6 14 738.074 23 6 719.557 40 2 678.926 57 474.652
7 14 009.789 24 6 411.462 41 2 505.461 58 393.002
8 13 338.901 25 6 114.219 42 2 337.870 59 317.092
9 12 715.508 26 5 827.280 43 2 176.032 60 247.172
10 12 132.398 27 5 550.148 44 2 019.836 61 183.592
11 11 584.105 28 5 282.374 45 1 869.185 62 126.857
12 11 066.360 29 5 023.548 46 1 723.991 63 77.745
13 10 575.747 30 4 773.294 47 1 584.179 64 37.606
14 10 109.477 31 4 531.269 48 1 449.686 65 10.000
15 9 665.235 32 4 297.157 49 1 320.458
16 9 241.077 33 4 070.672 50 1 196.457
17 8 835.344 34 3 851.546 51 1 077.658
122

Table A.5.: Height above ground zih (x) (half levels) for the grid point with maximum topography height in
the operational setup R03B07, 13 km spatial resolution.

Example: Height above ground HHL - HSURF

Location with max. surface height

CLON/CLAT = 88.180 / 27.938

HSURF = 6521.735 m

level idx. height [m] level idx. height [m] level idx. height [m] level idx. height [m]
1 68 478.265 26 21 281.888 51 5 327.376 76 866.489
2 65 841.812 27 20 095.614 52 5 131.536 77 758.180
3 63 320.647 28 18 967.228 53 4 935.699 78 657.641
4 60 836.062 29 17 895.173 54 4 739.875 79 564.832
5 58 424.710 30 16 887.062 55 4 544.039 80 479.667
6 56 084.562 31 15 939.079 56 4 348.198 81 402.130
7 53 813.730 32 15 047.640 57 4 152.369 82 332.014
8 51 610.433 33 14 209.372 58 3 956.552 83 269.290
9 49 454.480 34 13 421.103 59 3 760.682 84 213.850
10 47 356.194 35 12 679.851 60 3 564.867 85 165.505
11 45 302.948 36 11 982.810 61 3 369.022 86 124.377
12 43 305.214 37 11 327.347 62 3 173.185 87 90.297
13 41 369.015 38 10 710.978 63 2 977.389 88 62.129
14 39 493.042 39 10 131.372 64 2 781.538 89 40.050
15 37 676.058 40 9 586.339 65 2 585.719 90 20.103
16 35 916.890 41 9 102.469 66 2 389.888 91 0.000
17 34 214.417 42 8 626.524 67 2 201.043
18 32 567.562 43 8 180.073 68 2 020.364
19 30 975.312 44 7 761.408 69 1 847.844
20 29 436.690 45 7 357.246 70 1 683.365
21 27 950.772 46 6 966.873 71 1 527.137
22 26 516.663 47 6 578.056 72 1 378.952
23 25 133.513 48 6 191.036 73 1 238.808
24 23 800.515 49 5 852.974 74 1 106.748
25 22 516.886 50 5 567.850 75 982.565
 123

Table A.6.: Height above ground zif (x) (full levels) for the grid point with maximum topography height in
the operational setup R03B07, 13 km spatial resolution.

Example: Height above ground, full levels

Location with max. surface height

CLON/CLAT = 88.180 / 27.938

HSURF = 6521.735 m

level idx. height [m] level idx. height [m] level idx. height [m] level idx. height [m]
1 67 160.038 25 21 899.387 49 5 710.412 73 1 172.778
2 64 581.230 26 20 688.751 50 5 447.613 74 1 044.657
3 62 078.355 27 19 531.421 51 5 229.456 75 924.527
4 59 630.386 28 18 431.200 52 5 033.618 76 812.334
5 57 254.636 29 17 391.117 53 4 837.787 77 707.911
6 54 949.146 30 16 413.070 54 4 641.957 78 611.237
7 52 712.081 31 15 493.359 55 4 446.119 79 522.250
8 50 532.456 32 14 628.506 56 4 250.284 80 440.899
9 48 405.337 33 13 815.237 57 4 054.460 81 367.072
10 46 329.571 34 13 050.477 58 3 858.617 82 300.652
11 44 304.081 35 12 331.330 59 3 662.774 83 241.570
12 42 337.114 36 11 655.078 60 3 466.944 84 189.678
13 40 431.028 37 11 019.162 61 3 271.103 85 144.941
14 38 584.550 38 10 421.175 62 3 075.287 86 107.337
15 36 796.474 39 9 858.855 63 2 879.463 87 76.213
16 35 065.653 40 9 344.404 64 2 683.628 88 51.090
17 33 390.989 41 8 864.497 65 2 487.803 89 30.077
18 31 771.437 42 8 403.299 66 2 295.465 90 10.052
19 30 206.001 43 7 970.741 67 2 110.703
20 28 693.731 44 7 559.327 68 1 934.104
21 27 233.718 45 7 162.060 69 1 765.604
22 25 825.088 46 6 772.464 70 1 605.251
23 24 467.014 47 6 384.546 71 1 453.044
24 23 158.700 48 6 022.005 72 1 308.880
 125

B. Output on rotated latitude-longitude grids

in ICON-D2
The output on the structured lat-lon grid of ICON-D2 takes place on a rotated latitude-longitude grid,
i.e. in comparison to the standard latitude-longitude output of ICON global or ICON-EU, this spherical
coordinate system has rotated poles. The reason for this is to achieve a relatively constant grid mesh
size on the sphere1 , which saves storage space by avoiding senseless high spatial sampling in the northern
(i.e. close to the pole) part of the domain.
For most of the users it may be sufficient just to read the two fields RLON and RLAT, which contain the
geographical coordinates (longite and latitude, respectively) of every rotated grid point.
For those who are interested in the underlying transformations or need them for some reason, we list
here the transformation formulas between the rotated and the geographical coordinates. First, the north
pole of the rotated output grid for ICON-D2 is shifted to the position λN = 170◦ W and ϕN = 40◦ N.
(i.e. somewhere into the pacific), Therefore, the equator of the rotated grid goes roughly through the
center of the output domain (i.e. through the center of Germany).
The ’rotated latitude/longitude grid’ is coded in the Grid Description Section (GDS=section 3) of the
ICON-D2 output GRIB-files by gridDefinitionTemplateNumber=1 and gridType=rotated_ll.
It should be noted, that ICON still assumes a perfect sphere of the earth with a radius of 6371,229 km.
The transformation relations between the geographical coordinates (λg , ϕg ) and the rotated coordinates
(λ, ϕ) can be derived from simple geometric relations of spherical geometry. They are:
• From rotated to geographical coordinates
 
cos ϕ sin λ
λg = λN − arctan ,
sin ϕ cos ϕN − sin ϕN cos ϕ cos λ

ϕg = arcsin {sin ϕ sin ϕN + cos ϕ cos λ cos ϕN } ,

• and for the backtransformation from geographical to rotated coordinates

 
− cos ϕg sin(λg − λN )
λ = arctan ,
− cos ϕg sin ϕN cos(λg − λN ) + sin ϕg cos ϕN

ϕ = arcsin {sin ϕg sin ϕN + cos ϕg cos ϕN cos(λg − λN )} .

Note, that all angles are given in arcs (not in degrees). To get the angle in degrees, one has to multiply
by 180/π ≈ 57, 2957795. Take care that the arctan is correctly evaluated in all 4 quadrants.2 .
In the dwdlib (in particular in the library libmisc.a) the four Fortran functions RLSTORL, PHSTOPH,
RLTORLS and PHTOPHS are contained, which calculate the transformtions. These programs give and
expect angles in degrees.

1 In former models that used a spherical coordinate system, not only for output but as the base for their numerical grid
(e.g. the COSMO model), the use of a rotated grid was necessary to avoid too narrow grid cells near the poles (the so
called ’pole problem’). Narrow grid cells induce strong time step restrictions and therefore would result in inefficient
code.
2 Most programming languages have an extension of the standard arctan-function, e.g. in Fortran the function
ATAN2(numerator, denominator), which takes into account the corrct quadrant.
Chapter B. Output on rotated latitude-longitude grids in ICON-D2

• RLSTORL calculates geographic longitude (RL) from rotated longitude and latitude.
• PHSTOPH calculates geographic latitude (PH) from rotated longitude and latitude.
• RLTORLS calculates rotated longitude (RLS) from geographic longitude and latitude.
• PHTOPHS calculates rotated latitude (PHS) from geographic longitude and latitude.
For the transformation of many points or even whole fields dwdlib also contains the better optimized rou-
tines PLSTOPL, PLTOPLS, APLSTPL, and APLTPLS. An online description can be get via disdwd PLSTOPL
or man libmisc.

126
Bibliography
Anlauf, H., A. Rhodin, A. Cress, R. Faulwetter, M. Lange, G. Paul, C. Kopken-Watts, O. O. Stiller,
and R. Potthast, 2017: Deterministic global data assimilation for the ICON model - system suite
and evaluation. Mon. Wea. Rev., in preparation.
Bloom, S. C., L. L. Takacs, A. M. D. Silva, and D. Ledvina, 1996: Data assimilation using incremental
analysis updates. Mon. Wea. Rev., 124, 1256–1270.
Doms, G., J. Förstner, E. Heise, H.-J. Herzog, D. Mironov, M. Raschendorfer, T. Reinhardt, B. Ritter,
R. Schrodin, J.-P. Schulz, and G. Vogel, 2011: A description of the nonhydrostatic regional COSMO
model. Part II: Physical parameterization. Technical report, Deutscher Wetterdienst. Available at
http://www.cosmo-model.org/.
Freitag, M. and R. Potthast, 2013: Large Scale Inverse Problems Computational Methods and Applica-
tions In the Earth Sciences. De Gruyter.
Hunt, B. R., E. J. Kostelich, and I. Szunyogh, 2007: Efficient data assimilation for spatiotemporal
chaos: A local ensemble transform Kalman filter. Physica D: Nonlinear Phenomena, 230, 112–126.
Leuenberger, D., M. Koller, and C. Schär, 2010: A generalization of the sleve vertical coordinate. Mon.
Wea. Rev., 138, 3683–3689.
Lopez, P., 2016: A Lightning Parameterization for the ECMWF Integrated Forecasting System. Mon.
Wea. Rev., 144(9), 3057–3075.
Lynch, P., 1997: The Dolph-Chebyshev window: A simple optimal filter. Mon. Wea. Rev., 125, 655–660.
Lynn, B. and Y. Yair, 2010: Prediction of lightning flash density with the WRF model. Adv. Geosci.,
23, 11–16.
Polavarapu, S., S. Ren, A. M. Clayton, D. Sankey, and Y. Rochon, 2004: On the relationship between
incremental analysis updating and incremental digital filtering. Mon. Wea. Rev., 132, 2495–2502.
Ruppert, T., 2007: Diplomarbeit: Vector field reconstruction by radial basis functions. Master’s thesis,
Technical University Darmstadt, Department of Mathematics.
Schraff, C., H. Reich, A. Rhodin, A. Schomburg, K. Stephan, A. Perianez, and R. Potthast, 2016:
Kilometre-scale ensemble data assimilation for the COSMO model (KENDA). Q.J.R. Meteorol. Soc.,
142, 1453–1472.
Stephan, K., S. Klink, and C. Schraff, 2008: Assimilation of radar-derived rain rates into the
convective-scale model COSMO-DE at DWD. Quart. J. Royal Met. Soc., 134(634), 1315–1326.
DOI: 10.1002/qj.269.
Wan, H., M. A. Giorgetta, G. Zängl, M. Restelli, D. Majewski, L. Bonaventura, K. Fröhlich, D. Reinert,
P. Ripodas, L. Kornblueh, and J. Förstner, 2013: The ICON-1.2 hydrostatic atmospheric dynamical
core on triangular grids – Part 1: Formulation and performance of the baseline version. Geosci. Model
Dev., 6, 735–763.
Wicker, L., J. Kain, S. Weiss, and D. Bright, 2005: A Brief Description of the Supercell Detection
Index. http://www.spc.noaa.gov/exper/Spring_2005/SDI-docs.pdf.
Zängl, G., D. Reinert, P. Ripodas, and M. Baldauf, 2015: The ICON (ICOsahedral Non-hydrostatic)
modelling framework of DWD and MPI-M: Description of the non-hydrostatic dynamical core. Q.J.R.
Meteorol. Soc., 141, 563–579.

127
Glossary
CRU-CL Climate Research Unit - Gridded climatology of 1961-1990 monthly means. 19, 20
CRU-UEA Climate Research Unit - University of east Anglia. 19

DSMW Digital Soil Map of the World. 19, 20

ESA European Space Agency. 19

FAO Food and Agricultural Organization. 19

GACP Global Aerosol Climatology Project. 19

GLCC Global Land Cover Characteristics. 19
GLDB Global Lake Database. 19, 20
GlobCover 2009 Global Land Cover Map for 2009. 19, 20
GLOBE Global Land One-km Base Elevation Project. 19, 20
GRIB2 Gridded Binary Format, 2nd edition. 29, 44
GSFC Goddard Space Flight Center. 19

MODIS Moderate Resolution Imaging Spectroradiometer. 19

NASA National Aeronautics and Space Administration. 19

NGDC NOAA National Geophysical Data Center. 19

SeaWIFS Sea-viewing Wide Field-of-view Sensor. 19, 20

TOA top of atmosphere. 35

USGS U.S. Geological Service. 19

WMO World Meteorological Organization. 29

129
Deutscher Wetterdienst
Business Area “Research and Development”
Frankfurter Straße 135
63067 Offenbach
Germany