Original vegetation and area of habitat

A restoration plan needs to know two things about every converted cell: what would grow back if it were restored, and which species would gain from it. This part answers both for Serra Verde, the imaginary region introduced in the previous part. You find the original vegetation of its pasture and cropland from a map of the past, then measure the area of habitat of eight species today and before conversion, and what each converted cell would give each species if it were restored. At the end you have the inputs of the restoration plans of the next two parts, and a first look at which species those plans will favour.

The idea#

Fractions of land cover#

Maps of land cover usually give each cell one class. At 1 km, though, a cell is rarely all forest or all pasture. The global restoration plans of Strassburg et al. (2019, 2020), which Terraflux follows, work instead with the share of every class in every cell: a cell may be 0.6 forest, 0.3 pasture and 0.1 cropland. Classes are either natural (forest, savanna, mangrove: what restoration brings back) or anthropic (pasture, cropland: what restoration replaces); the others (urban, water) are neither and are left alone.

What was there before#

Restoring a cell's pasture turns it back into the natural vegetation that was there, and that is not always the vegetation around it today. Strassburg et al. decide the original vegetation of each cell's anthropic share in three steps:

1. The past. If a map of the natural cover at an earlier date says what the cell held, its anthropic share is given back to the natural classes in the proportions they had then.
2. The ecoregion. If the past map says nothing (no natural cover left), or the cell is almost entirely converted (more than threshold, 95%, of its land anthropic), the proportions of the natural classes of its ecoregion are used instead.
3. A failsafe. If neither tells, the mean proportions over the whole map.

The area of habitat#

A species' range map (from IUCN or a distribution model) draws the outline of where it occurs, but much of the land inside it is not habitat: a forest bird's range includes the pastures and towns within it. The area of habitat (AOH) of a species is the part of its range covered by the land cover it lives in, within its known elevations (Brooks et al., 2019). With shares of land cover, each cell of the range counts for its area times the share of the classes the species lives in.

From habitat lost to extinction risk#

Why measure habitat before and after? Because the species-area relationship links the habitat a species has left to its risk of extinction. If a species keeps a fraction \(r\) of its original habitat, its risk is taken as \(1 - r^z\) with \(z \approx 0.25\), the exponent of the species-area curve (Thomas et al., 2004; Strassburg et al., 2019). The curve is steep near the left: a species that keeps half its habitat is far more at risk than one that keeps 90%, and each hectare restored for it avoids far more risk.

When to use it#

• Restoration planning: the original vegetation and the habitat gains are the inputs of restoration.plan in the next part.
• Red List assessments: AOH is a recommended measure for the IUCN Red List, a tighter estimate of the area a species occupies than its range (Brooks et al., 2019).
• Habitat loss by species: the potential and the current habitat give each species' loss, the basis of indicators of the impact of land conversion.

What it is not for: AOH counts habitat, not its quality, its fragmentation or whether it is occupied. For the habitat suitability of one species from its records, see species distributions; for patches and connectivity, habitat patterns and habitat networks.

What you need#

All maps on one grid, in a projected CRS (or with cell_area given in km²). For Serra Verde (60 × 90 cells of 1 km, EPSG:31983):

• land_cover.tif: the share (0 to 1) of six classes in every cell, as six bands named forest, savanna, mangrove, pasture, cropland and urban. Band names matter: the operators find the classes by them.
• past_cover.tif: the share of the three natural classes in the past, before most of the conversion (bands forest, savanna, mangrove).
• ecoregions.tif: four ecoregions, 1 coast, 2 lowland forest, 3 montane forest, 4 savanna.
• features.tif: eight bands, one per species or habitat type (jaguar, lowland_tapir, harpy_eagle, maned_wolf, golden_lion_tamarin, neotropical_otter, araucaria_forest, mangrove), each a suitability from 0 to 1. A cell above 0 is in the range.
• A table of the classes each species lives in (the suitability setting below).

Enlarge

Build the model#

The ecoregions' natural shares#

restoration.original_vegetation takes the ecoregion's natural proportions as a stack (a set of maps on one grid, each with a name): for every cell, the shares of forest, savanna and mangrove of its ecoregion. Here they are the mean proportions of the natural classes in each ecoregion's past cover, computed once outside the model:

Three raster.calc nodes write them on the map from the ecoregion codes, for example for forest:

0.31 if e == 1 else 0.98 if e == 2 else 0.91 if e == 3 else 0.2

and stack.of joins them into one stack, naming each layer as its input (forest, savanna, mangrove).

The original vegetation#

Add restoration.original_vegetation. Connect cover.raster (the land cover read by raster.read, every band as a layer) to land_cover, past.raster to past and eco_shares.stack to ecoregions. Set natural = ["forest", "savanna", "mangrove"] and anthropic = ["pasture", "cropland"]. Its outputs:

• original: the original share of each natural class (today's natural land plus the anthropic share given back).
• restore: the proportions restored land returns to in each cell (restore_to = "current" makes it the cell's current natural cover where it has some, instead of the original).
• source: where each cell's proportions came from, 0 the past, 1 the ecoregion, 2 the failsafe.
• anthropic: the share of anthropic land, the land a plan can restore.
• failsafe: the failsafe proportions of each natural class.

The area of habitat#

Add sdm.area_of_habitat. Its inputs: ranges (a stack, one layer per species: here ranges.raster, the features read whole), land_cover, and the vegetation's original, restore and anthropic. Its settings:

• natural: the natural classes, as above.
• suitability: the classes each species lives in, {species: [classes]}; a species not listed lives in every natural class (unless include_unlisted = false).
• elevation_limits with an elevation map: [[species, lowest, highest], ...] removes from each range the cells outside the species' elevations, as Brooks et al. (2019) recommend.
• cell_area: a cell's area in km², read from a projected grid by default.

Its outputs are table (per species, in the order of the ranges' layers: current and potential habitat in km², range_cells, and valid, 1 when the species counts in a plan) and gain (a stack: the habitat each species gains per unit of each cell restored).

format = "terraflux/1"

[[nodes]]
id = "cover"
op = "raster.read"
params = { path = "land_cover.tif" }
note = "Six bands: the share of each class."

[[nodes]]
id = "past"
op = "raster.read"
params = { path = "past_cover.tif" }

[[nodes]]
id = "ecoregions"
op = "raster.read"
params = { path = "ecoregions.tif" }

[[nodes]]
id = "eco_forest"
op = "raster.calc"
inputs = { e = "ecoregions.raster" }
params = { expression = "0.31 if e == 1 else 0.98 if e == 2 else 0.91 if e == 3 else 0.2" }

[[nodes]]
id = "eco_savanna"
op = "raster.calc"
inputs = { e = "ecoregions.raster" }
params = { expression = "0 if e == 1 else 0.01 if e == 2 else 0.09 if e == 3 else 0.8" }

[[nodes]]
id = "eco_mangrove"
op = "raster.calc"
inputs = { e = "ecoregions.raster" }
params = { expression = "0.69 if e == 1 else 0.01 if e == 2 else 0" }

[[nodes]]
id = "eco_shares"
op = "stack.of"
inputs = { forest = "eco_forest.map", savanna = "eco_savanna.map", mangrove = "eco_mangrove.map" }

[[nodes]]
id = "vegetation"
op = "restoration.original_vegetation"
inputs = { land_cover = "cover.raster", past = "past.raster", ecoregions = "eco_shares.stack" }
params = { natural = ["forest", "savanna", "mangrove"], anthropic = ["pasture", "cropland"] }

[[nodes]]
id = "ranges"
op = "raster.read"
params = { path = "features.tif" }
note = "Eight species: a cell above 0 is in the range."

[[nodes]]
id = "habitat"
op = "sdm.area_of_habitat"
inputs = { ranges = "ranges.raster", land_cover = "cover.raster", original = "vegetation.original", restore = "vegetation.restore", anthropic = "vegetation.anthropic" }
params = { natural = ["forest", "savanna", "mangrove"], suitability = { jaguar = ["forest", "savanna"], lowland_tapir = ["forest"], harpy_eagle = ["forest"], maned_wolf = ["savanna"], golden_lion_tamarin = ["forest"], neotropical_otter = ["forest", "mangrove"], araucaria_forest = ["forest"], mangrove = ["mangrove"] } }

Read the results#

Where the original vegetation came from#

Why give the ecoregions at all? Without them, the rule gives the nearly converted cells equal shares of every natural class: a third forest, a third savanna, a third mangrove, even inland. Drag the handle to compare:

With the ecoregions

Without

Here it changes little for the species (the inland mangrove is outside the mangrove's range), but the golden lion tamarin's potential habitat drops from 101.8 to 97.9 km² and the lowland tapir's from 781 to 761 km² without the ecoregions, because a third of those cells' forest became savanna and mangrove. On real data, where whole districts are converted, the difference can be large. In the source map, code 1 marks these cells whether or not ecoregions were given.

Habitat today and before conversion#

The absolute areas say little about need. What matters is the share left: the golden lion tamarin, in two small lowland fragments, keeps 56% of its habitat, and the mangrove 65%; the harpy eagle, in remote mountain forest, has lost almost nothing.

Enlarge

On the curve, the tamarin and the mangrove carry nearly two thirds of the total risk (0.136 + 0.104 of 0.365). Restoring their habitat moves them down the steep part of the curve; restoring the jaguar's moves it along the flat part. A plan that minimises extinction risk will go to them first, as the next part shows.

What restoring each cell gives each species#

Enlarge

Species

The gain maps are where the plan will act: the tamarin's two fragments, the coastal strip of the mangrove, the river valleys of the otter and the tapir.

How to report it#

Report the land cover data and classes (which are natural, which anthropic), the past map and how the ecoregions' proportions were made, the share of converted land each source explained, and per species the range source, the suitable classes, the elevation limits, and the current and potential habitat with the share left.

Settings that change the answer#

• Suitability. Which classes a species lives in decides its habitat. If the jaguar lived only in forest, its habitat would shrink to the forest of its range. Take the classes from the IUCN habitat classification of each species, and check them with experts.
• Elevation limits. With the relative elevation of elevation.tif connected to elevation and elevation_limits = [["araucaria_forest", 0.8, 1.0]], the araucaria forest keeps only the cells of its range at 0.8 or above: its range falls from 1,050 to 635 cells, its current habitat from 976 to 620 km² and its potential from 1,020 to 635 km². The share left rises from 95.8% to 97.7%, because the conversion happened at lower elevations. Limits remove habitat that the land cover alone would count, which is why Brooks et al. (2019) include them in the definition of AOH.
• The threshold of restoration.original_vegetation (0.95): below it the past map is trusted even for heavily converted cells; above it, the ecoregion's proportions are used for fewer of them.
• The ranges. A suitability map above 0 makes a generous range. A higher threshold (each band read by its name with raster.read, a raster.calc of f > 0.5 on it, and the results joined with stack.of), or ranges from a distribution model, make a tighter one, and all the areas fall.

Common pitfalls#

• Band names. The operators find classes and species by the names of the layers. A GeoTIFF without band descriptions gives layers named by number, and the settings must use those names.
• Shares, not classes. The land cover is a stack of shares from 0 to 1 that sum to at most 1 in each cell. A map of classes must be turned into shares first (one raster.calc per class, lc == 3, then stack.of).
• The order of the species. The table's rows follow the order of the ranges' layers, with no names; keep that order when you read it.
• No original vegetation. Give the original input: without it sdm.area_of_habitat cannot compute the gains a plan needs.
• Units. Areas are in km² from a projected grid (or cell_area); on a grid in degrees, each cell's true area on the ellipsoid is used.

Going further#

• The plan that restores this habitat where it avoids most extinction risk: A restoration plan, the next part.
• Where human pressure is highest: Ecological intactness, the previous part.
• Habitat suitability from species records: species distributions.
• The operators: restoration.original_vegetation, sdm.area_of_habitat.

References#

• Brooks, T. M., Pimm, S. L., Akçakaya, H. R., Buchanan, G. M., Butchart, S. H. M., Foden, W., et al. (2019). Measuring terrestrial area of habitat (AOH) and its utility for the IUCN Red List. Trends in Ecology & Evolution, 34(11), 977–986. https://doi.org/10.1016/j.tree.2019.06.009
• Strassburg, B. B. N., Beyer, H. L., Crouzeilles, R., Iribarrem, A., Barros, F., de Siqueira, M. F., et al. (2019). Strategic approaches to restoring ecosystems can triple conservation gains and halve costs. Nature Ecology & Evolution, 3(1), 62–70. https://doi.org/10.1038/s41559-018-0743-8
• Strassburg, B. B. N., Iribarrem, A., Beyer, H. L., Cordeiro, C. L., Crouzeilles, R., Jakovac, C. C., et al. (2020). Global priority areas for ecosystem restoration. Nature, 586(7831), 724–729. https://doi.org/10.1038/s41586-020-2784-9
• Thomas, C. D., Cameron, A., Green, R. E., Bakkenes, M., Beaumont, L. J., Collingham, Y. C., et al. (2004). Extinction risk from climate change. Nature, 427(6970), 145–148. https://doi.org/10.1038/nature02121