Water Systems Modelling: Model

Classification/Types of models

By Dr. Ilyas Masih, Senior Lecturer in Water

Resources Planning, IHE Delft Institute for Water
Education. Email ([email protected])
 Analogy from other
Model Classification sciences: e.g. Biology
Why classify water systems models?
• It is a scientific method (e.g. classification in
biology and chemistry)
• It is required when there are many things
• Are there many models around ?? (Yes )
• So we need to classify them to understand:
• Process representation
• Type / level of mathematics involved
• How the simulation is accomplished
• Which part of the water system is
modelled
• What is the potential use of the model
• What spatial and temporal
scales/dimensions are represented
• etc.??
Part I- Physical process and spatial scale
driven classification
 Physically-based distributed models

Physically based models:

• Model describes the natural system using the basic
mathematical representations (partial differential equations) of
the flows of mass, continuity, momentum and various forms of
energy.
• Based on the understanding of the physics of the processes
involved.

Distributed Models:
• Model takes into account the spatial variations in all variables
and parameters (climate, soils, vegetation, demand/supply, ...).
Explains processes more realistically.
• Mainly found for groundwater models, improved rainfall-
runoff models, water quality models
 Physically-based distributed models
Example: Mike-She (e.g. Abbott et al., 1986 a, 1986b; Graham and Butts, 2005)

Source: Graham and Butts, 2005

 Physically-based distributed models
Example: Mike-She (e.g. Abbott et al., 1986 a, 1986b; Graham and Butts, 2005)

Overland flow: St. Venant equations

Unsaturated flow: 1-D Richard’s equation

Saturated zone flow: Boussinesq equation

Note: Details on equations can be found at Abbott et al., 1986

Source: Graham and Butts, 2005
or any standard text book on hydrology
 Physically-based distributed models
Example: Mike-She (e.g. Abbott et al., 1986 a, 1986b; Graham and Butts, 2005)

Applications
•River basin management and planning
• Water supply design, management and optimization
• Irrigation and drainage
• Soil and water management
• Surface water impact from groundwater withdrawal
•Conjunctive use of groundwater and surface water
• Wetland management and restoration
• Ecological evaluations
• Groundwater management
• Environmental impact assessments
•Aquifer vulnerability mapping
•Contamination from waste disposal
• Surface water and groundwater quality remediation
• Floodplain studies
• Impact of land use and climate change
• Impact of agriculture (irrigation, drainage, nutrients and pesticides, etc.)
 Physically-based distributed models
Advantages
• Simulation of actual processes according to physical laws (close to reality)
• Universal applicability
• Can account for spatial variations in time and space (climate, soils, land cover, ...)
• Hydrological information generated at many points (not only at catchment outlet)
• Can isolate problem areas
• Useful for ungauged catchments, real time flow forecasting
• Higher reliability of results, if successfully applied

Limitations:
• Require huge amount of input data
• Computational intensive (require more computer power and simulation time)
• Over-parameterized description in case of simple applications
• Scale issues (simulation at grid scale, mathematical descriptions are based on
experimental conditions)
• Requires in-depth knowledge of processes
• High cost
Hydrological model classification (adopted from Freeze and
Harlan, 1969): An example of conceptual hydrologic model of
systems hydrology
 Conceptual and Lumped Models

Conceptual Model
• Model is constructed on the basis of the physical processes that we
‘read’ into our observations of the catchment. Physical processes
are represented in simple conceptual way and with simple
equations (partial understanding)
• The physical significance is not usually so clear that the parameters
can be assessed from direct measurements. Instead, it is necessary
to estimate the parameters from calibrations, but some physical
relevance is given.

Lumped Model
• The catchment is regarded as one single unit; the variables and
parameters are thus representing average values for the entire
catchment. It ignores spatial variability.
 Conceptual and Lumped Models

Advantages
• Easy for use
• Ease for obtaining input data
• Require less time and resources
• Low cost
• Applicable to many cases, though do not offer wide possibilities

Disadvantages
• No or limited spatial variability
• Output is single catchment representative values
• Driving processes are not adequately represented
• Limited applicability in terms of regional applications, system
understanding and support in decision making.
 Conceptual and Lumped Models

Example: HBV model:

http://www.smhi.se/sgn0106/if/hy
drologi/hbv.htm#table2
(e.g. Bergström 1996;
Uhlenbrook et al., 2010;
Masih et al., 2010)

Main Water Balance Equation

where:
P = precipitation
E = evapotranspiration
Q = runoff
SP = snow pack
SM = soil moisture
UZ = upper groundwater zone
LZ =lower groundwater zone
lakes = lake volume
 Soil module-HBV
1
act. evaporation / pot. evaporation

 SM (t ) 
E act  E pot min  ,1
 FC  LP 

0
FC * LP FC
Soil moisture (mm)
FC is a model parameters, and not necessarily equal to measured values of field capacity
Runoff Generation Routine-HBV

recharge

UZL Q0=K0·(SUZ-UZL)
SUZ
E P Q1=K1·SUZ
PERC
LAKE

SLZ
Q2=K2·SLZ
runoff
 Conceptual and Lumped Models
Example: HBV model:
http://www.smhi.se/sgn0106/if/hydrologi/hbv.htm#table2
(e.g. Bergström 1996; Uhlenbrook et al., 2010; Masih et al., 2010)

Applications of the HBV model

• to extend runoff data series (or filling gaps)
• for data quality control
• for water balance studies
• for runoff forecasting (flood warning and reservoir operation)
• to compute design floods for dam safety
• to investigate the effects of changes within the catchment
• to simulate discharge from ungauged/poorly catchments
• to simulate climate change and land use change effects
 What are the similarities and differences between
conceptual diagram of the HBV model and the one
given by Freez and Harlaan, 1969?

HBV model conceptual diagram

 What are the similarities and differences between
conceptual diagram of the HBV model and the one
given for Mike-she model by Graham and Butts,
2005?

HBV model conceptual diagram

 (semi) Physically-based (semi) distributed models
(e.g. SWAT model, Arnold et al., 1998)

Schematic of water balance concept in SWAT

Evaporation and Precipitation

Transpiration

Surface Runoff
Root Zone Infiltration/plant uptake/
Soil moisture redistribution
Vadose Lateral Flow
(unsaturated)
Zone
Shallow Revap from Percolation to
(unconfined) shallow aquifer shallow aquifer Return Flow
Aquifer
Confining Layer
Deep Flow out of watershed Recharge to
(confined)
Aquifer deep aquifer
 Model structure

• Basin (watershed)
• Sub-basin (sub-watershed)
• Hydrological response unit
(HRU)
 Model structure: HRU definition
A unique combination of land use,
soil and slope

Land use map

+ HRU
Soil map
definition

Slope
 Model structure: HRU definition
A unique combination of land use, soil and slope
 HRU outputs are aggregated at sub-basin level-HRU’s are
virtual in a sub-basin(not geo- referenced)
A
R GIS subbasin HRU # of pixels
1 6
C 2 5
Pixels : 20
V Pixel area = A 3 8
I GEO - REFERENCED
4 1

E
W SWAT DATABASES
HRU parameters

Output 1 x A x 6 input Output 1

Output 2 x A x 5
S
Total input Output 2
+
output W
Output 3 x A x 8 input Output 3
A
Output 4 x A x 1 input Output 4
NOT GEO - REFERENCED
T
1 unit area
 Hydrological model classification (adopted from Freeze and
Harlan, 1969)
Hydrologic Simulation

Physical Hydrology System Investigation

Physical Physically-based Parametric/

Deterministic Methods Stochastic Methods
Models mathematical
Methods
Result in conceptual
(empirical) models
Distributed Mathematical Models Lumped (e.g. Lumped), also
termed as black/grey
box

Digital Computer Analog Computer

Also called Grey /White Solution solution
Box (e.g. distributed)

Physically-based digitally simulated Also called white box

response model models
Part II application domain based classification
 Model Classification based on the application
domain

• Watershed/Rainfall-runoff models (HEC-1, HBV, Mike She, SWAT)

• Water quality models (WASP, QUAL2E, SOBEK, DUFLOW)
• Water quantity-quality models (Mike She, SWAT, RIBASIM, WEAP)
• Groundwater model (MODFLOW; SWAP)
• Stream hydraulics models (HEC-RAS, Mike 11)
• Morphological models (SOBEK-1D, Delft 3D, Mike 21)
• Soil-Vegetation-Atmosphere system models (SWAP, DAISY,
Oryza2000).
• River/Reservoir systems models (WAFLEX, WEAP, RBASIM,
AQUARIUS, HYDROSIM, MODSIM, SWAT)
A brief on River/Reservoir System Models
 River/Reservoir system analysis models
(also termed as Arc-Node type of models)

• Widely used in planning studies to aid in the formulation and

evaluation of alternative plans for responding to water related
problems and needs
• Evaluate operating policies for reservoir systems
• allocating water between multiple uses and users;
• minimizing the risks and consequences of water shortages and flooding;
• optimizing the beneficial use of water, energy, and land resources;
• managing environmental resources.
• Evaluate water allocation policies for existing and future water uses
across a river system/basin
• Prior appropriation vs social equity or economic productivity
• Integrated River Basin planning and management
Schematic of flow network, generally used in river/reservoir models:
example from Juizo and Liden, 2010. WEAP, WAFLEX and WRYM was
applied in this basin)
 River/Reservoir system analysis models

• These models are generally categorized as

• Simulation models
• Optimization models
• Models with combinations of simulation and optimization.
 River/Reservoir system analysis models
Simulation VS Optimization

Simulation addresses “what if questions” optimization can address “what should

be” questions.
Simulation predict system response to given design and operating policy
whereas optimization identify optimal design and operating policy
 River/Reservoir system analysis models

Simulation VS Optimization
• Optimization models automatically search for an "optimum“ set of
decision variable values, whereas simulation models are limited to
predicting system behavior for a user-specified set of variable
values.

• Simulation and optimization models can also be used in

combination. For example, a study could involve preliminary
screening of numerous alternatives using an optimization model,
followed by a more detailed evaluation of selected plans using a
simulation model.
Part III- Discovering other factors used in model typology
(Home Work)
 More options to classify models

• Deterministic models
• Stochastic models
• Probabilistic models
• Steady/unsteady state
models
• Other types? (specify)
Concluding Remarks
Model Classification: Concluding Remarks

• A wide variety of water resources systems models have been

developed over time, which provide profound opportunity for
water resources modellers to use them in water resources
management.

• The models could be classified based on several criteria:

• Most commonly used criteria in hydrologic simulation is the

physical processes and spatial scale considered in a model:
(e.g. Physically-based distributed models, conceptual-
lumped models).
• The application domain is another main classification
criterion used in model classification (rainfall-runoff models,
hydraulic models, river/reservoir systems models, water
quality models).
 Model Classification: Concluding Remarks
• The model classification is helpful to systematically study and
use the water resources system models.

• The water resources modeller should asks several questions

before selecting a model to study/address a water management
issue at hand.

• suitability of model to address the problem at hand,

Processes simulated in the model, spatial and temporal
scale of application, input data requirement, output of the
model, parameters, calibration and uncertainty analysis
options, reliability of the model code, possibility to change
the model code, available support/documentation, availability
(free).
• Model classification contribute in model screening/selection
Exercise on model classification
(Class and Home Work)
 Exercise: Using knowledge gained so far in the
module, classify the given models using guidelines
in the table below
Classification Criteria HBV WEAP MIKE-SHE
From physical process point
of view (physically based, semi-
physical based, conceptual,
empirical).
From spatial scale
representation (distributed,
semi-distributed, lumped)

From application domain

point of view

Other criteria/Remarks
Thanks