FragScape v2.

03

User Guide

17/4/2020

Mathieu Chailloux (INRAE / UMR TETIS) - [email protected]

Jennifer Amsallem (INRAE / UMR TETIS) - [email protected]
Jean-Pierre Chéry (AgroParisTech / UMR TETIS) - [email protected]
Table of contents

1 Overview 3
1.1 Landscape fragmentation metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Landscape division . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Splitting index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.3 Effective mesh size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.4 Cross-Boundary Connection (CBC) method . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Computation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Vector mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Raster mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Graphical User Interface (GUI) overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Steps 6
2.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Land cover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Additional data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3 Example 10

4 To go further... 11
4.1 Execution time and memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.1 Vector mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.2 Raster mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.3 Results comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.4 Configuration file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

5 FAQ 13
5.1 Error messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Bibliography 13

17/4/2020 FragScape v2.03 - User manual 2/ 14

1 OVERVIEW
FragScape is a QGIS plugin that computes landscape fragmentation metrics defined in paper ”Landscape
division, splitting index, and effective mesh size: new measures of landscape fragmentation” (Jaeger
2000). Among these metrics, effective mesh size has been widely used to quantify landscape fragmen-
tation. FragScape defines a 4 steps process from raw data to computed metrics and allow user to save
configuration so that results can be reproduced with same context.

1.1 LANDSCAPE FRAGMENTATION METRICS

Jaeger defined in his paper (Jaeger 2000) three new measures of landscape fragmentation:
• landscape division
• splitting index
• effective mesh size
To compute these measures, landscape elements assessed as fragmenting are removed. Remaining ar-
eas are called patches. Landscape is then composed of n patches. A patch area is denoted by Ai with
1 ≤ i ≤ n. The total area of the region is denoted by At .

1.1.1 Landscape division

The degree of coherence (C or COH), an auxiliary measure, is defined as the probability that two points
chosen randomly in a region are connected (e.g. not separated by fragmentation elements such as roads
or urban areas):
The degree of landscape division (D or DIV I) is defined as the probability that two points chosen ran-
domly in a region are not connected:
∑n
Ai
C= ( )² D =1−C
A t
i=1

1.1.2 Splitting index

The splitting index (S or SP LI) is defined as the number of patches one gets when dividing the total re-
gion into parts of equal size (meshes) in such a way that this new configuration leads to the same degree
of fragmentation of initial configuration:
A2
S = ∑n t 2
i=1 Ai
It can be interpreted as the effective mesh number of a grid with a constant mesh size dividing the region
into S patches which all have the size At /S.

1.1.3 Effective mesh size

The effective mesh size (m or M SIZ) denotes the size of the areas when the region is divided into S
areas (each of the same size At/S) with the same degree of landscape division as for the initial configu-
ration:
1 ∑ 2
n
At
m= = Ai
S At
i=1

Splitting density (s or SDEN ) is defined as the number of meshes per unit area. Net product (N or
N P RO) is defined as the product of the effective mesh size and the total area of the region:
S At 1 ∑n
s= = ∑n 2 = N = m.A t = A2i
At A
i=1 i m
i=1
17/4/2020 FragScape v2.03 - User manual 3/ 14
1.1.4 Cross-Boundary Connection (CBC) method

As other patch-based landscape metrics, above metrics can be biased by the boundaries and the extent
of a reporting unit if the boundaries fragment patches. This issue is called the ”boundary problem” and
has been addressed in paper ”Modification of the effective mesh size for measuring landscape fragmen-
tation to solve the boundary problem” (Moser et al. 2007).
New method called cross-boundary connections (CBC) includes area outside boundaries. The complete
area of a patch, regardless of boundaries, is denoted by Acmpl
i . The formula of effective mesh size ac-
cording to CBC method (mCBC or CBC_M SIZ) is:

1 ∑
n
mCBC = Ai .Acmpl
i
At
i=1

Encart 1 : Métriques CBC

In CBC mode, only 2 metrics are defined : effective mesh size and net product

1.2 COMPUTATION METHODS

Metrics are computed from a layer of natural areas patches.
FragScape is designed to include patch layer creation from raw data (land cover, roads, ...) following a
step-by-step procedure (cf section 2): natural areas selection from land cover and integration of addi-
tional data.
There are 2 computation methods depending on input data format, extent, precision needs and available
computing resources. Vector mode is appropriate for vector land cover in case the amount of data is
reasonable (cf section 4.1). Raster mode is appropriate otherwise (raster land cover, large extent, high
geometric precision, ...).

1.2.1 Vector mode

In vector mode, features are selected from land cover layer and dissolved (one feature of type Multi-
Polygon).
It is then possible to integrate additional data such as roads network, hydrographic network, or any miss-
ing data in initial layer. For each data source, features can be selected (paved roads for instance) and a
buffer can be applied to modelize footprint for linear data. These selections are then merged with land
cover, by union or difference depending on their contribution to fragmentation or to natural areas. Re-
sulting layer is then dissolved and casted to single geometry (Polygon) to finally get a correct patch layer
for metrics computation.

1.2.2 Raster mode

In raster mode, input data can be vector or raster but output layers are in raster format anycase (raster-
ization and reprojection according to extent and resolution parameters).
Resolution value is very important because it defines computation precision but also random-access
memory (RAM) needed. If land cover is already in raster format, resolution shall be the same than land
cover layer.
Land cover layer is reclassified: 1 for natural areas (selected classes), 0 for fragmentation data
(unchecked classes). Additional data is reprojected and classified same way. Resulting layers are then
merged according to specified ranking order in graphical user interface.

17/4/2020 FragScape v2.03 - User manual 4/ 14

1.3 INSTALLATION
FragScape is a QGIS plugin. FragScape is cross-platform: tests have been performed on Ubuntu bionic,
Windows 10 and macOS Sierra.

Encart 2 : Prerequisites

QGIS version must be superior to 3.4.0.

• QGIS version must be superior to 3.4.0.

• Python libraries scipy et numpy must be already installed to use raster mode (cf section
5.1).

To install FragScape, open QGIS:

• go to Extension menu
• open Install/Manage extensions dialog
• go to Parameters tab and check that Show experimental plugins option is checked
• go back to All tab, search for FragScape, select it and click on Install plugin button

Once installed, FragScape icon shall appear in tools panel.

If not, go to Extension menu and a FragScape entry shall be present.
If not, installation failed. Please check error message or contact support team.

1.4 GRAPHICAL USER INTERFACE (GUI) OVERVIEW

Figure 1 show an overview of FragScape GUI. It contains 4 main components :

• top icons bar : action icons (configuration management, language switch)

• right panel : description of current step
• bottom progress bar : shows progress of current process
• main frame : current step content

In main frame, current step can be composed of :

• parameters that must be set (such as Workspace)

• visualisation table that displays current configuration/results
• action buttons (such as Launch selection)

17/4/2020 FragScape v2.03 - User manual 5/ 14

 FIGURE 1: FragScape v2.0 Graphical User Interface

2 STEPS
FragScape defines a 4 step procedure from raw data to computed metrics.

2.1 PARAMETERS
First step is to define global parameters used for current FragScape project.
Workspace must be set before any processing as it defines FragScape outputs path. Output file of each
step is stored in Workspace/outputs. Be careful when setting workspace as existing file can be overriden.
Mode defines processing chain executed to compute metrics (cf section 1.2). In raster mode, Resolution
must be set to define pixel size in georeferenced unit (meter for a metric projection).
Extent layer defines data extent: data are clipped at layer limits. Optional in vector mode. In CBC
mode, data extent must be larger than study area.
Projection is a projected coordinate reference system that must be set according to data geographic
extent. It defines entities shape and area.
If option Save intermediate layers is checked, intermediate layers are stored in Workspace/tmp.
Otherwise, they are stored in QGIS processing temporary directory (path is displayed in log when a layer
is created).

17/4/2020 FragScape v2.03 - User manual 6/ 14

2.2 LAND COVER
Second step is to select target features from land cover layer. To do so:

• Select land cover layer (Input layer parameter).

• Choose Mode and Selection field in input layer is vector
• Press button Show field values
• Check land cover classes matching natural areas in toSelect column of table
• Press button Launch selection. Output layer is loaded in QGIS and saved in W orkspace/outputs
(landuseSelectionDissolve.gpkg in vector mode, landuseSelectionW arp.tif in raster mode).

Encart 3 : Selection mode

There are 2 selections modes in In Vector mode:
• By field values to extract unique values from Selection field
• By expression to extract features verifying specified expression (all features if expression
is empty)
In Raster mode, unique values are extracted from first band.

Figure 2 shows an example of land cover selection interface.

FIGURE 2: FragScape v2.0 Land cover tab

17/4/2020 FragScape v2.03 - User manual 7/ 14

2.3 ADDITIONAL DATA
Third step is to integrate additional data that would be missing in land cover. For instance: roads, river
courses, wildlife crossing, ...
For each data source, user should:

• Select Input layer

• (optional) Filter input features according to specified Expression (all features if expression is
empty). Expression can be built with widget.
• Specify Buffer expression for line and point data in vector mode to modelize footprint. Expression
must be a number and can be built with widget.
• Specify an Identifier (unique in project) for current selection
• Press Save selection button. Specified selection appears as a new line in visualisation table.

Once all data selections saved, user should rank lines (for instance wildlife corridors on top of roads) and
then press Integrate additional data button.
For each line, data is selected, buffer is applied (if defined) and layer is rasterized in raster mode. Output
layers are then merged and integrated to result of previous step.
Final layer is loaded in QGIS and stored and savec in output directory (landuseF ragmSingleGeom.gpkg
in vector mode, landuseF ragm.tif in raster mode).

FIGURE 3: FragScape v2.0 Additional data tab

17/4/2020 FragScape v2.03 - User manual 8/ 14

2.4 RESULTS
Fourth step is to compute fragmentation metrics. To do so:

• Specify Input layer (result of setp 3 by default).

• Specify Reporting layer. Metrics are computed for each feature of reporting layer. To compute
metrics for an entire region, specify a layer with a single feature.
• Check Include CBC metrics if needed (see section 1.1.4)
• Select Unit of area (from square meter to square kilometer)
• Specify Output layer. If not specified, a memory layer is created.
• Press Compute metrics button.

FIGURE 4: FragScape v2.0 Results tab

Figure 4 shows results step interface. Once metrics are computed, output layer is loaded in QGIS and
global effective mesh size (on the whole territory) is displayed. Output layer contains an attribute for
each metric defined in section 1.1 and new fields:

• patches: number of patches

• At: area of the reporting unit
• sum_Ai: intersection area of patches and reporting unit
• layer/path: temporary layer containing initial reporting unit
• divisor : divisor matching unit of area (for instance 100 for are unit)

17/4/2020 FragScape v2.03 - User manual 9/ 14

3 EXAMPLE
This section illustrates FragScape use case with provided sample data (subdirectory sample_data in
FragScape plugin directory).

(a) Input data (b) Step2

(c) Step 3 (d) Step 4

FIGURE 5: FragScape use case : from raw data to effective mesh size

Figure 5 shows input data and each step result.

To reproduce results:

• Copy sample_data to a local directory

• Open FragScape
• Set workspace to sample_data/CUT
• Open configuration file EPCI_Clermontais_2012_CUT.xml from button
• Check that configuration has been correcty loaded
• Run steps 2 to 4

17/4/2020 FragScape v2.03 - User manual 10/ 14

4 TO GO FURTHER...

4.1 EXECUTION TIME AND MEMORY

Use of FragScape depends on available computing resource when applied to large territory with high
level of geometric precision.

4.1.1 Vector mode

In vector mode, execution time can be very long depending on study area extent and geometric preci-
sion. Given execution times are indicative values.
Figure 6 show the evolution of execution time according to region extent (small region, big region, coun-
try) from Corine Land Cover (vector data):

Study area Step 2 Step 3 Step 4

Hérault <1mn 1mn 1mn
Occitanie 5mn 11mn 2mn
France 122h 19h 5h

FIGURE 6: Execution time by extent

Figure 7 show the execution time according to data source geometric precision (Corine Land Cover vs
OCcupation du Sol Grande Échelle) on a same territory (Hérault):

Cas de test Étape 2 Étape 3 Étape 4

CLC <1mn 1mn 1mn
OCSGE 6h 35h 3mn

FIGURE 7: Temps d’exécution - CLC vs OCSGE

If execution time is too long, user can switch to raster mode which is much faster but leads to a loss of
geometric precision depending on resolution.

4.1.2 Raster mode

In raster mode, critical resource is the available live memory (RAM). RAM needs depends on the amount
of data (number of pixels) that is directly linked to tuple (extent, resolution). If a memory error
occurs, user can change resolution and try to relaunch computation.

4.2 ALGORITHMS
Algorithms (available in QGIS processing toolbox) implement specific treatments developped for
FragScape. Figure 8 shows available algorithms. Groups Raster et Vector gather steps described in
section 2.

17/4/2020 FragScape v2.03 - User manual 11/ 14

 • Compare results layer : computes dif-
ference between 2 FragScape results layer
for each field (cf section 4.3)
• Raster Effective Mesh Size : com-
putes fragmentation metrics in raster
mode without reporting layer
• Raster Effective Mesh Size
(Cross-Boundary Connection) : com-
putes fragmentation metrics in raster
and CBC modes
• Raster Effective Mesh Size per
feature : computes fragmentation
metrics in raster and CUT modes
FIGURE 8: FragScapeAlgorithms

• Integrates vector data to land cover : applies geometric difference/union between natu-
ral areas layer and additional data in vector mode
• Prepare land cover data : selection of natural areas from land cover layer in vector mode
• Vector Effective Mesh Size (Global) : computes fragmentation metrics in vector mode on
the whole territory (features are dissolved if needed)
• Vector Effective Mesh Size per feature : computes fragmentation metrics in vector mode
for each feature of reporting layer

4.3 RESULTS COMPARISON

FragScape finality is to study fragmentation evolution and so to compare results on a same territory at
different times. Algorithm Compare results layer computes difference between 2 output layers of
FragScape on each field.
Difference on effective_mesh_size and net_product fields is performed using CBC value if available.
Field variation contains effective mesh size evolution in percentage: (Bval − Aval )/(Bval + Aval ).

4.4 CONFIGURATION FILE

Configuration is saved as an XML file and thus can be opened in a text editor. Figure 9 shows the begining
of configuration file sample_data/ECPI_Clermontais_2012/CBC/ECPI_Clermontais_2012_CBC.xml

FIGURE 9: Example of a configuration file

In Landuse tag, one can see attributes such as in_layer (input layer), select_mode (0 meaning selection
mode By field values) and select_field (selection field of input layer is CODE_12). For each loaded field
value, a LanduseFieldItem tag exists and contains same attributes as in FragScape (value, description, toS-
elect).

17/4/2020 FragScape v2.03 - User manual 12/ 14

Such file can be manually edited if needs be. For instance if relative paths must be changed for a new
project (../Source becoming ../../Source), updating it in FragScape tables or creating a new project
can be avoided by editing new paths in configuration file and then reloading it.

5 FAQ

• Fields are not loaded in field/expression widget , why ? If they don’t appear, it is because asso-
ciated layer is not loaded even if its path is displayed in combo box. Select another layer and then
re-select initial layer.
• Which method should I use, CUT or CBC ? CBC method has been designed to address bound-
ary problem and then should be used. CUT method is available to allow comparison with already
computed results, or in case boundaries are not a problem.
• Elements of fragmentation are already included in my land cover layer, should I run step 3 ? In
FragScape 2.0, it is possible to specify step 4 input layer so taht step 3 is optional.
• Can I apply FragScape processing to layer not produced by FragScape ? To apply FragScape specific
processing to specific data, one can use FragScape algorithms described in section 4.2.

Encart 4 : Good practices

• Do not use spaces and special characters in file names.

• Do not use special characters in field values.
• Save FragScape configuration at each step.
• Check each step result.
• If a problem occurs, save configuration, exit FragScape relaunch FragScape and re-open
saved configuration. If problem still occurs, exit and relaunch QGIS. If problem still occurs,
contact support team.

5.1 ERROR MESSAGES

• Layer XXX is already loaded in QGIS, please remove it. FragScape cannot delete file if it is already
loaded. Just remove it from QGIS project and relaunch FragScape processing.
• The process cannot access the file because it is being used by another process: XXX. Check that
XXX file is not used by another process. If not, save configuration, save QGIS project, exit QGIS,
relaunch QGIS, relaunch FragScape, re-open configuration and relaunch FragScape processing.
• Algorithm XXX not found This error occurs if FragScape installation failed. Try to uninstall and
reinstall FragScape. If error remains, please contact support team.
• NameError: name ’np’|’scipy’ is not defined Library numpy|scipy is not installed. Install it and re-
launch QGIS. On Linux, install package python-numpy|python-scipy. On Windows, use OsGeo4W in-
staller.

If an unknown error occurs, please report it at https://github.com/MathieuChailloux/FragScape/

issues.

17/4/2020 FragScape v2.03 - User manual 13/ 14

References
[1] Jochen Jaeger. “Landscape division, splitting index, and effective mesh size: New measures of
landscape fragmentation”. In: Landscape Ecology 15 (Feb. 2000), pp. 115–130. DOI: 10 . 1023 / A :
1008129329289.
[2] Brigitte Moser et al. “Modification of the effective mesh size for measuring landscape fragmen-
tation to solve the boundary problem”. In: Landscape Ecology 22 (Mar. 2007), pp. 447–459. DOI:
10.1007/s10980-006-9023-0.

17/4/2020 FragScape v2.03 - User manual 14/ 14