The network of a forest's structure (MSPA)

The graphs of parts 1 to 4 joined patches by the distance a species can cross. There is another way to see the links of a forest network: look for the forest itself joining the forest, the hedgerows, riparian strips and narrow woods that physically run from one block to another. Morphological spatial pattern analysis (MSPA, the subject of the habitat patterns series) finds them, as bridges, and GuidosToolbox turns its cores and bridges into a network. In this part you run MSPA on the forests around Hannover, build that structural network, measure how coherent it is, and rank its bridges and cores by how much of it would be cut without them. At the end you will know what a structural network says that a dispersal graph does not, and the other way round.

The idea#

Cores as nodes, bridges as links#

MSPA classifies every cell of a habitat map by its shape (Vogt et al., 2007; Soille & Vogt, 2009): cores are the interior of the habitat, farther than the edge width from its boundary; bridges are narrow strips of habitat that join two different cores; loops join a core to itself, branches lead nowhere, islets are too small to hold any core. Vogt et al. (2009) proposed to read this as a network: the cores are the nodes, the bridges the links. Two cores joined by a bridge belong to the same component, and a component with at least one bridge is a network; a core without any bridge is isolated.

The importance of a bridge is the share of that connectivity lost if it were cut (the component may split in two); the importance of a stepping-stone core, a core with two bridges or more, is the share lost without it. Saura et al. (2011) combined MSPA with Conefor's dPC in this way to find the key structural connectors of a forest landscape.

Structure and function, again#

A structural network has no dispersal distance: a bridge exists or it does not. It answers where the forest itself provides a corridor, which matters for species that never leave the woods (many forest plants, ground-dwelling invertebrates, some small mammals) and for planning green infrastructure, where the bridges are the things to protect. It ignores everything a species might cross; the dispersal graphs of parts 1 to 4 ignore the shape of the forest. Used together they tell the corridors that exist from the gaps that are crossed.

When to use it#

• Mapping green infrastructure: the cores as hubs and the bridges as links of a regional network, and the bridges ranked by importance to protect (Vogt et al., 2009; Saura et al., 2011).
• Species bound to the habitat, for which a gap of open land is a barrier at any distance.
• Following a network over time: the number of networks, the coherence and the most important bridges at two dates.

Not for species that cross open land: for them use the graphs of part 1 to part 4. And remember that MSPA can only find the bridges the map draws (the habitat patterns series shows how a coarse map hides them).

What you need#

• A habitat map that MSPA can read: GuidosToolbox's coding (0 missing, 1 background, 2 habitat), or any map with foreground naming the habitat classes. Here the land cover of the series, clc_100m.tif, with the forests (23, 24, 25) as foreground.
• An edge width, the depth of the edge effect in cells. With cells of 100 m, edge_width = 1 makes every forest strip two cells (200 m) wide or less a bridge or a branch, never a core. Choose it as the MSPA tutorial explains.

The application's template Network of a forest (MSPA) (Templates > Connectivity) has the model of this part on a landscape drawn by chance; replace its first node with a raster.read of your map.

Build the model#

format = "terraflux/1"

[[nodes]]
id = "land"
op = "raster.read"
params = { path = "clc_100m.tif" }
note = "CORINE Land Cover, cells of 100 m."

[[nodes]]
id = "mspa"
op = "landscape.mspa"
inputs = { map = "land.raster" }
params = { foreground = [23, 24, 25], edge_width = 1 }
note = "Cores, bridges, branches... of the forests, with an edge of 100 m."

[[nodes]]
id = "components"
op = "networks.gtn_components"
inputs = { map = "mspa.map" }
note = "Networks of cores joined by bridges; ECA and coherence in summary."

[[nodes]]
id = "importance"
op = "networks.gtn_importance"
inputs = { map = "mspa.map" }
note = "dPC of every bridge and stepping-stone core."

[[nodes]]
id = "conefor"
op = "networks.conefor"
inputs = { map = "mspa.map" }
note = "The same network as Conefor's input files."

[[plots]]
id = "plot"
title = "The most important links and nodes"
layers = [{ from = "importance.map" }]

Both network operators read the MSPA map only: they need its codes, so connect mspa.map directly, not a reclassified copy.

Read the results#

MSPA of the forest#

Enlarge

The forest at 100 m is mostly core: the map is coarse, and only the strips of forest that survived the resampling show as bridges. 0.3% of the map, about 750 cells, is bridge.

The networks#

Enlarge

networks.gtn_components's summary:

A coherence of 32% means the cores offer as much "structurally reachable" core as a single block of a third of their area. The comparison with part 2 is instructive. There, for a species that crosses 250 m of open land, the forest's ECA was 29% of the forest; here, for a species that never leaves it, the cores' coherence is 32% of the core. At 100 m, most of the links between forest blocks are gaps of farmland, not strips of forest, so the structural network is about as fragmented as the functional one of a poor disperser. A finer map (25 m or 10 m) would show the hedgerows and riparian strips that CORINE merges into farmland.

The most important bridges and cores#

Enlarge

GuidosToolbox's colours mark the links and nodes whose importance is within 1, 5 and 10% of the most important, not the top 1, 5 and 10% of their number: when one bridge is far more important than every other, as here, it is the only one coloured. The importance output has the values themselves; drawn on a zoom of the north-east:

Enlarge

A single bridge, link 36, seven cells of forest (7 ha, a strip a few hundred metres long), carries 7.3% of the network's connectivity: it ties the largest core of the north-east to a chain of cores to the south. Cut it, and the network falls in two. The importance drops fast after it: the 112 bridges are mostly redundant or join small cores. Among the 56 stepping-stone cores, core 80 (762 ha, seven bridges) is the most important; core 217, a single cell of core with two bridges, ranks second: structure, not size, makes a stepping stone.

Settings that change the answer#

• The edge width. A wider edge shrinks the cores and turns more strips into bridges: there are fewer, smaller cores and more links. Coherence can rise or fall; report the width with every number.
• The resolution of the map. It decides which strips exist. At 25 m the same forests have many more bridges; at 100 m most hedgerows are gone. Never compare networks made from maps of different resolution.
• 8 or 4 neighbours (connectivity of landscape.mspa): with 4, bridges that hold by a corner break, and cores they joined fall apart.
• What counts as habitat: adding transitional woodland (class 29) to the foreground makes new bridges across clearings.

Common pitfalls#

• Expecting functional links. Two woods 50 m apart across a field are not joined: there is no bridge. For species that cross fields, use the dispersal graphs.
• Reading "top 1%" as the top 1% of links. GuidosToolbox's classes are relative to the largest importance. Sort the links table by dpc for a ranking.
• Huge single-cell cores. At a coarse resolution and edge width 1, one interior cell is already a core (core 217). A wider edge or a finer map avoids such artefacts.
• The border. Cores cut by the map's edge lose bridges outside it.

Going further#

• Habitat patterns: MSPA itself, its classes and settings.
• Part 3: importance with dispersal, for comparison.
• Part 6: ProtConn: the connectivity of a network of reserves.
• Costs and corridors: where restoring a corridor would join two forest objects.
• The operators: landscape.mspa, networks.gtn_components, networks.gtn_importance, networks.conefor.

References#

• Saura, S., Vogt, P., Velázquez, J., Hernando, A., & Tejera, R. (2011). Key structural forest connectors can be identified by combining landscape spatial pattern and network analyses. Forest Ecology and Management, 262(2), 150–160. https://doi.org/10.1016/j.foreco.2011.03.017
• Soille, P., & Vogt, P. (2009). Morphological segmentation of binary patterns. Pattern Recognition Letters, 30(4), 456–459. https://doi.org/10.1016/j.patrec.2008.10.015
• Vogt, P., Ferrari, J. R., Lookingbill, T. R., Gardner, R. H., Riitters, K. H., & Ostapowicz, K. (2009). Mapping functional connectivity. Ecological Indicators, 9(1), 64–71. https://doi.org/10.1016/j.ecolind.2008.01.011
• Vogt, P., Riitters, K. H., Iwanowski, M., Estreguil, C., Kozak, J., & Soille, P. (2007). Mapping landscape corridors. Ecological Indicators, 7(2), 481–488. https://doi.org/10.1016/j.ecolind.2006.11.001
• Vogt, P., & Riitters, K. (2017). GuidosToolbox: universal digital image object analysis. European Journal of Remote Sensing, 50(1), 352–361. https://doi.org/10.1080/22797254.2017.1330650