Several benefits and trade-offs

A restoration programme is rarely about one thing. Funders want carbon, conservation agencies want species, others want whole ecosystems back or landscapes less pressed by people, and every region wants its share. In this last part of the series you will give restoration.plan four benefits (the species' extinction risk of the previous part, carbon, ecological intactness and the risk of collapse of ecosystems), see how differently each one alone would plan Serra Verde's 100 km², put them on one scale, trace the trade-off between species and carbon with a dashboard, rank sets of weights as Strassburg et al. (2020) do, and measure what it costs to give every region its part.

The idea#

Four benefits#

Strassburg et al. (2020) planned the world's restoration for three benefits at once: biodiversity (the extinction risk avoided), climate (the carbon restored vegetation would hold) and cost. restoration.plan takes:

Carbon is linear: each km² adds the same, so a plan for carbon alone restores the cells with the most carbon per km², in full. Species and ecosystems are concave: each gives less as it recovers. Intactness is measured by its gradient (part 3): how fast the total intactness of the landscape rises per unit restored, recomputed as the plan proceeds.

The risk of collapse of an ecosystem follows the logic of the IUCN Red List of Ecosystems, where the loss of an ecosystem's extent is one of the main criteria (Keith et al., 2013): an ecoregion that keeps a share \(r\) of its natural area gets the risk \(1 - r^z\), as a species does.

One scale#

The benefits come in different units: a risk avoided of a few tenths, about a million tonnes of carbon, a sum of intactness in the hundreds. Adding them with weights of 1 lets the largest number win. A simple, transparent scale divides each benefit by the most it can reach alone (the best plan for it alone), so each counts from 0 to 1.

Strassburg et al. (2020) also compare sets of weights: for each set, a plan; for each plan, every benefit as a share of its best; and a score, the mean of those shares over their standard deviation, high for plans that do well on all benefits and evenly. The weights setting of restoration.plan does the same and returns the plan of the best-ranked set.

When to use it#

• Programmes with several goals: carbon finance with biodiversity safeguards, landscape restoration for water, people and species.
• Negotiation: the trade-off curve shows each party what a little more of their benefit costs the others.
• Targets per region (states, watersheds), and by ecosystem, so that no ecosystem is left out.

Remember what the benefits leave out: the costs of restoration and of the land (add a cost map and a budget, part 5), the chance that restoration succeeds (Crouzeilles et al., 2016, found it depends strongly on the landscape around it), and the people who live there.

What you need#

The model of part 5, plus, on the same grid:

• carbon_potential.tif: the carbon of restored vegetation in every cell, in tonnes per hectare;
• footprint.tif: the human footprint (0 to 50, part 3);
• the land cover's pasture and cropland bands, for the reduction of the footprint when they are restored;
• ecoregions.tif: four ecoregions (1 coast, 2 lowland forest, 3 montane forest, 4 savanna);
• regions.tif: three administrative regions.

Build the model#

Carbon and intactness#

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

[[nodes]]
id = "carbon"
op = "raster.calc"
inputs = { c = "carbon_t_ha.raster" }
params = { expression = "c * 100 / 1000000" }

[[nodes]]
id = "linear"
op = "stack.of"
inputs = { carbon = "carbon.map" }

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

[[nodes]]
id = "pasture"
op = "raster.read"
params = { path = "land_cover.tif", band = "pasture" }

[[nodes]]
id = "cropland"
op = "raster.read"
params = { path = "land_cover.tif", band = "cropland" }

[[nodes]]
id = "reduction"
op = "raster.calc"
inputs = { p = "pasture.raster", c = "cropland.raster" }
params = { expression = "(4 * p + 7 * c) / (p + c) if p + c > 0 else 0" }

The plan with three benefits#

[[nodes]]
id = "plan"
op = "restoration.plan"
inputs = { restorable = "vegetation.anthropic", table = "habitat.table", gain = "habitat.gain", linear = "linear.stack", hfi = "footprint.map", reduction = "reduction.map" }
params = { target = 100, weight = 3.89, linear_weights = [0.817], intactness_weight = 0.00672 }

weight multiplies the risk benefit, linear_weights the linear ones in the order of their layers, and intactness_weight the intactness. The values are the inverses of the best of each alone (0.257, 1.224 Mt and 148.9, found by three plans with one benefit each): each benefit counts from 0 to 1. The benefits output gives, per benefit, its value, its weight and the weighted value. Intactness takes its own settings (radius, gamma, beta, intactness_z, as in part 3) and intactness_form: linear (the gradient, recomputed as the plan proceeds, as Strassburg et al. do) or exact (the total intactness itself, which is not concave: the plan is then a local optimum with a bound).

Ecosystems#

[[nodes]]
id = "ecosystems"
op = "restoration.ecosystem_areas"
inputs = { land_cover = "cover.stack", ecoregions = "ecoregions.map" }
params = { natural = ["forest", "savanna", "mangrove"], anthropic = ["pasture", "cropland"] }

[[nodes]]
id = "plan_eco"
op = "restoration.plan"
inputs = { restorable = "vegetation.anthropic", table = "ecosystems.table", gain = "ecosystems.gain" }
params = { target = 100 }

[[nodes]]
id = "eco_targets"
op = "restoration.ecosystem_targets"
inputs = { table = "ecosystems.table" }
params = { target = 100 }

restoration.ecosystem_areas gives, per ecoregion, its natural area now (current) and before conversion (potential, the natural plus the converted land), in the form restoration.plan takes for its risk benefit (table, gain). restoration.ecosystem_targets answers a different question: how to share a target among the ecosystems for the least total risk of collapse (method = "exact", or the stepwise steps), each ecosystem between a least share (minimum) and its converted area. Its targets table (an ecosystem's code and its share) can be the regional targets of a plan, with the ecoregions as regions.

Sets of weights and regions#

[[nodes]]
id = "plan"
op = "restoration.plan"
inputs = { restorable = "vegetation.anthropic", table = "habitat.table", gain = "habitat.gain", linear = "linear.stack", hfi = "footprint.raster", reduction = "reduction.map" }
params = { target = 100, weights = [
  { extinction_risk = 3.89, carbon = 0, intactness = 0 },
  { extinction_risk = 0, carbon = 0.817, intactness = 0 },
  { extinction_risk = 0, carbon = 0, intactness = 0.00672 },
  { extinction_risk = 3.89, carbon = 0.817, intactness = 0.00672 },
  { extinction_risk = 7.78, carbon = 0.817, intactness = 0.00672 },
  { extinction_risk = 3.89, carbon = 0.817, intactness = 0 } ] }

With weights, a list of sets (the benefits named extinction_risk, the linear layers' names and intactness), the node solves a plan per set and returns the weights table (per set its weights, its benefits, Strassburg et al.'s mean, sd, score and rank) and the plan of the best-ranked set.

For targets per region, connect a map of region codes to regions and give target as {code: area}, for instance target = { 1 = 30, 2 = 40, 3 = 30 }; sense applies to each.

Read the results#

Each benefit alone asks for other places#

Enlarge

How each single-benefit plan does on the others, as a share of the best for each:

The plan for the species and the plan for carbon overlap more than either does with intactness: the coast is rich in carbon and is where the mangrove was lost. Intactness pulls the other way, towards the fringes of the large converted areas, where restoring a strip reconnects forest that is already there. A plan with all three keeps most of every benefit, which is the usual finding of such analyses: optimising jointly costs each benefit little compared with optimising one and hoping for the others (Strassburg et al., 2020).

Enlarge

The trade-off between species and carbon#

Enlarge

PlayWeight of the species (against carbon)

The first points of the curve are the cheapest: a small weight on the species buys a lot of risk avoided for almost no carbon. That is the place to start a negotiation. Planners and the people involved choose a point on the curve; the model makes the price of each point explicit.

Ranking sets of weights#

The weights table of the six sets above, each benefit as a share of its best:

The equal weights rank first by far: their plan is the most balanced (a mean of 79 % with little spread). The score rewards balance; a programme with a clear priority (carbon finance) would rather choose on the trade-off curve.

Ecosystems#

restoration.ecosystem_targets shares a target of 100 km² among the ecoregions for the least total risk of collapse: the coast and the lowland forest, which keep two thirds of their area, get 91 km²; the montane forest none. The savanna gets 9 km² although it has lost little: its whole converted area is only 9 km², and a small ecoregion's last km² still lowers its risk. Giving these as regional targets ensures every ecosystem gets its part of a plan driven by species or carbon.

Targets per region#

Sharing the effort costs 9 % of the benefit (0.234 against 0.257): region 1 must restore 30 km² where they help species that have lost little. That number, the price of the rule, is what a planner can put before the people who decide whether the rule is worth it.

Reporting#

Report every benefit with its data and units, how they were scaled (here by the best alone) and weighted, the trade-off curve or the sets of weights compared, and the plan's value for each benefit as a share of its best, so that readers see what each benefit gave up. With regional or ecosystem targets, report the targets and their price in benefit.

Settings that change the answer#

• The weights, as the slider shows, and the scaling. Scaling by the best alone is one choice; scaling by a reference (the benefit of restoring everything) gives other numbers. Say which.
• The reduction of the footprint per class (4 for pasture, 7 for cropland) and the parameters of intactness.
• The carbon map: the carbon of mature vegetation overstates what young restoration holds for decades.
• Regional and ecosystem targets, and their sense (at most, at least, exactly).

Common pitfalls#

• Weights of 1 on benefits in different units. The benefit with the largest numbers wins. Scale first.
• Weights read as importance. They are exchange rates between scaled benefits; their consequences, not their values, should be discussed.
• Linear benefits on a concave model. A plan for carbon alone restores whole cells with the most carbon; with species too, the mix changes in ways the weights alone do not tell. Look at the trade-off curve.
• Intactness on its own scale. Its gradient is in units of total intactness per unit restored; always scale it.
• Targets per region larger than a region's restorable land. Check the restorable area of each region first.

Going further#

• The series from the start: Distance to habitat, Costs and corridors, Ecological intactness, Original vegetation and habitat, A restoration plan.
• Choosing which places to protect rather than restore: Systematic conservation planning.
• The operators: restoration.plan, restoration.ecosystem_areas, restoration.ecosystem_targets.

References#

• Crouzeilles, R., Curran, M., Ferreira, M. S., Lindenmayer, D. B., Grelle, C. E. V., & Rey Benayas, J. M. (2016). A global meta-analysis on the ecological drivers of forest restoration success. Nature Communications, 7, 11666. https://doi.org/10.1038/ncomms11666
• Keith, D. A., Rodríguez, J. P., Rodríguez-Clark, K. M., Nicholson, E., Aapala, K., Alonso, A., et al. (2013). Scientific foundations for an IUCN Red List of Ecosystems. PLoS ONE, 8(5), e62111. https://doi.org/10.1371/journal.pone.0062111
• 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
• Venter, O., Sanderson, E. W., Magrach, A., Allan, J. R., Beher, J., Jones, K. R., et al. (2016). Sixteen years of change in the global terrestrial human footprint and implications for biodiversity conservation. Nature Communications, 7, 12558. https://doi.org/10.1038/ncomms12558