Systematic conservation planning: the ideas

Money, land and time for conservation are always short, so someone has to decide where. This series is about making that decision with methods that can be explained, repeated and defended: systematic conservation planning and the spatial prioritization at its heart. In this first part you will meet the ideas every method shares, using a small example you can follow by eye and a synthetic region, Serra Verde, that the whole series uses. At the end you will have made your first plan in Terraflux and read the table that says whether it does its job.

The idea#

Reserves chosen by chance#

For most of the twentieth century, protected areas were chosen ad hoc: for their scenery, for recreation, or because the land was too steep, remote or poor to be worth anything else. Pressey (1994) showed what this left behind in New South Wales: reserves piled up on rugged, unproductive land while lowland habitats, and the species that live only there, were left out. A decade later a global gap analysis, which asks what existing reserves already hold, found hundreds of threatened species occurring in no protected area at all, most of them with small ranges (Rodrigues et al., 2004).

Systematic conservation planning (Margules & Pressey, 2000) replaced these choices with an explicit, repeatable procedure: state clear goals for what should be conserved, measure what the existing reserves already achieve, choose new areas that fill the gaps as efficiently as possible, then implement, manage and monitor, and revise the plan as things change. The step in the middle, the choice of areas, is spatial prioritization (Moilanen, Wilson & Possingham, 2009), and it is what the prioritize operators of Terraflux do.

Complementarity#

The idea that most separates systematic planning from intuition is complementarity: the value of a site depends on what the other sites in the plan already hold (Kirkpatrick, 1983; Pressey et al., 1993). A second site with the same species as the first adds little; a poor site with a species found nowhere else adds a lot.

Take four sites, A to D, and four species. A holds three species, B two, C and D one each. If only two sites can be protected, the obvious choice is the two richest, A and B. But B holds nothing that A does not already have, so the pair protects three species. The pair A and C protects all four:

Enlarge

Both families of methods in this series use complementarity. Zonation (part 2) removes the sites one by one, always the one whose loss matters least given what remains; Marxan and integer linear programming (parts 3 to 5) search for the best set directly.

The core concepts#

Kukkala & Moilanen (2013) reviewed the concepts behind spatial prioritization. Five do most of the work:

Representation

Every feature (species, habitat, ecosystem service) should be in the plan, not only the common or charismatic ones.

Complementarity

Places are chosen for what they add to the others, as above.

Irreplaceability

How much the goals depend on a place. A site with the only population of a species is irreplaceable; a site with features found in many others is not. Part 3 and part 6 measure it.

Efficiency

Reaching the goals at the least cost, which leaves the rest of the landscape, and the rest of the budget, for other uses.

Adequacy and persistence

Being in a reserve is not the same as persisting there. Features need enough area, in large and connected pieces, to last (Cabeza & Moilanen, 2001). Representation is the first step; part 2 (connectivity in Zonation) and part 5 (compact and connected plans) take the next ones.

The ingredients of a planning problem#

Every prioritization, from a watershed to a continent, is the answer to five questions.

A raster is a map stored as a grid of square cells, each with a value: a habitat suitability, a land price, a class of land cover. A stack is several rasters on the same grid, one per feature. A planning unit is the smallest area a plan can select or leave out; on a raster it is usually a cell.

Costs can be the price of land, the income lost by not farming or logging it (the opportunity cost), the cost of managing it, or simply its area when nothing better is known (Naidoo et al., 2006). Without costs the best plan is to protect everything; with them the question becomes which places give the most for what they cost.

Targets say how much is enough: 30 % of the habitat of every species, 500 km² of mangrove, 2 million tonnes of carbon. They come from policy (Target 3 of the Kunming-Montreal Global Biodiversity Framework asks for 30 % of land and sea by 2030), from ecology (the area a viable population needs), or from the people involved.

Serra Verde#

The series uses Serra Verde, a synthetic coastal region that comes with landschaft, the library behind Terraflux's prioritization. It is imaginary but built to behave like real planning data: 60 × 90 cells of 1 km (4825 planning units on land; the sea has no value and is not part of the problem), mountains in the west where land is cheap, a coastal plain with three cities, a highway and two rivers where land is dear, three protected areas, and eight features from widespread to very rare.

Enlarge

Enlarge

Download the rasters to follow along: serra-verde.zip (the features, the cost, the protected areas, the cities, the habitat condition and two threats, with a short README). They are GeoTIFFs in a projected coordinate reference system (CRS: how the numbers of a map relate to places on Earth), UTM zone 23 South, with cells of 1000 m.

The richest cells against a plan#

Here is the experiment that shows why methods are needed. The goal is to protect 30 % of the habitat of every feature. Two strategies spend exactly the same money:

A. The richest cells. Add up the eight suitability maps (a "hotspot" score), then select cells from the highest score down until the money runs out.

B. An optimized plan. Ask Terraflux for the cheapest set of cells that meets all eight targets (a minimum set problem, part 4). Its cost, 37.1 million USD, is the budget of strategy A.

Enlarge

The richest cells are in the mountains, where the jaguar, the harpy eagle, the maned wolf and the araucaria forest overlap. Every mountain cell adds a little more of the same features. The golden lion tamarin, the mangrove and the tapir live in places with few other features, so those cells never reach the top of a richness list. The optimized plan values sets: once the mountain cells hold 30 % of the araucaria forest, more of them add nothing to the targets, and the next dollar goes to a tamarin fragment or a stretch of mangrove. That is complementarity at the scale of a region. The plan is also efficient: where two places hold the same features it takes the cheaper one, and it stays away from the dear land around the cities whenever there is an alternative.

Two families of methods#

Scoring ignores complementarity, as the experiment showed. The other three use it, and they belong to two families:

• Ranking (Zonation; Moilanen et al., 2005, 2022) orders every cell from the least to the most important. It needs no target and no budget: the best 10 % lies inside the best 30 %, so one map answers every budget. This suits the early stages of planning, while targets are still discussed, and very large maps.
• Optimization finds the best set of areas for a target or a budget. The two classic formulations are the minimum set problem (meet every target at the least cost) and the maximal coverage problem (do the most within a budget; Church et al., 1996). Marxan (Ball et al., 2009) solves them with simulated annealing, a heuristic that improves a plan by random changes and gives good answers without knowing how good. Written as integer linear programs (Hanson et al., 2025), they are solved exactly, with a proof of how close the plan is to the best possible (Underhill, 1994; Rodrigues & Gaston, 2002; Beyer et al., 2016).

Enlarge

The two families are complementary, not rivals: a ranking to explore, and an optimized plan to decide. Part 6 compares them on Serra Verde.

When to use it#

Spatial prioritization answers questions such as:

• Where should new protected areas go to represent every ecosystem and species, at the least cost? This is the classic use, from national reserve systems to the global analyses of where protected areas should grow (Montesino Pouzols et al., 2014).
• What does a target cost? The cost of 17 %, 30 % or 50 % of every feature, and what a given budget achieves.
• Which places are irreplaceable, so that development should avoid them (impact avoidance, offsets)?
• Where should restoration go, for the most biodiversity or carbon per dollar (Strassburg et al., 2020)? See Distance and restoration.
• Which actions against which threats, where? (multi-action planning, part 7).
• How good is an existing or proposed reserve network? A gap analysis is an evaluation (part 6).

What it is not:

• It is a decision-support tool, not a decision. The answer depends on the goals, data and costs you give it. Different, equally reasonable goals give different maps, which is why stakeholders should take part in choosing them.
• It is only as good as its data. Species distribution models have errors, costs are estimates, and both pass into the plan.
• High threat is not high priority. A very threatened place may be a poor investment if action there is unlikely to work (Game et al., 2013).
• It does not replace field knowledge. A selected area may be impractical for reasons no map holds.

What you need#

• Features: a raster per feature, all on the same grid: species distribution models or ranges, habitat maps, carbon. Values are amounts: presence (0/1), probability or suitability (0 to 1), hectares of habitat, tonnes. The Species distributions tutorials make such maps.
• A cost raster on the same grid. Cells with no value are not planning units (the sea in Serra Verde). A cost of zero is a free planning unit, which the solver will gladly select.
• Optionally: existing protected areas (to lock in), places that cannot be protected (cities, to lock out), threats, habitat condition.
• A projected coordinate system with cells of equal area (metres, not degrees), so that a cell means the same area everywhere and boundaries are measured in metres.

In the application, the templates of the Templates menu, group Conservation and restoration, start the main methods with data drawn by chance: Zonation, Marxan and irreplaceability, Conservation planning (linear programming), Irreplaceability of a plan and Multi-action planning. Replace their nlm nodes with raster.read nodes of your own maps.

Build the model#

The first plan of the series: the cheapest set of cells that holds 30 % of each of the eight features, and its evaluation.

format = "terraflux/1"

[[nodes]]
id = "features"
op = "raster.read"
params = { path = "features.tif" }
note = "Serra Verde's eight features, a band each."

[[nodes]]
id = "cost"
op = "raster.read"
params = { path = "cost.tif" }
note = "Thousand USD per km²; the sea has no value."

[[nodes]]
id = "plan"
op = "prioritize.optimize"
inputs = { features = "features.raster", cost = "cost.raster" }
params = { objective = "min_set", relative_targets = 0.3, gap = 0.001 }

[[nodes]]
id = "evaluation"
op = "prioritize.evaluate"
inputs = { features = "features.raster", cost = "cost.raster", solution = "plan.solution" }
params = { relative_targets = 0.3 }

[outputs.coverage]
from = "evaluation.coverage"

[[plots]]
id = "show"
title = "Plan for 30 % of every feature"
layers = [{ from = "plan.solution" }]

From the command line, save the model as first-plan.toml next to the rasters and run it:

terraflux run first-plan.toml --out output

The outputs of the model ([outputs.coverage] above) are written to the output folder, tables as CSV files.

Read the results#

The plan selects 812 cells (17 % of the land) for 37.1 million USD, and the solver proves it is within 0.01 % of the cheapest plan possible. The coverage table is the first thing to read after any solve:

Most targets are met exactly: a minimum set buys no more than it needs. Two features are well above their target, the harpy eagle and the araucaria forest. They come "for free": they live in the cheap western mountains where the plan must go anyway for the jaguar and the maned wolf. A feature far above its target is a sign of where the plan's cheap land is, not a sign of a problem.

What to report in a paper, at a minimum: the planning units (their size and number), the features and their data, the cost layer and its units, the targets, the objective and the solver with its optimality gap, then the plan's cost, area and the share of every feature it holds, and a map. Part 6 gives a full checklist and shows how to compare plans.

Settings that change the answer#

The target sets the price. The slider shows the cheapest plan for targets from 10 % to 50 % of every feature:

Enlarge

PlayTarget (% of every feature)

The other settings that change a plan, each the subject of a part of this series:

• the method: a ranking (part 2) or an optimized set (parts 3 and 4);
• the cost layer: area, land price or opportunity cost give different plans (part 2 shows the effect in Zonation);
• the objective: meet every target, or do the most with a budget (part 4);
• the shape: compact, connected plans cost more but last better (part 5);
• what is already protected, and what cannot be (part 4).

Common pitfalls#

• Zero instead of no value. Cells that are not part of the problem (sea, outside the study area) must have no value in the cost raster. A cost of 0 makes them free, and every method will select them.
• Features on different grids. All features and the cost must share one grid (extent, cell size, CRS). Align them first with raster.match (see Getting started).
• Geographic coordinates. Cells of 0.01° shrink towards the poles, so their areas differ and boundaries are in degrees. Reproject to an equal-area or a local projected CRS before planning.
• Features of very different totals. A widespread species has a far larger total than a rare one. Relative targets handle this, but budget objectives and rankings without weights favour the widespread (parts 2 and 4).
• The edge of the study area. A species rare inside your borders may be common next door; a plan within administrative borders gives priority to the edges of ranges. Plan within the ecological context when you can.
• Threat is not priority. Do not use a threat map as a feature to protect. Threats belong in the cost, in locked-out areas, in the condition of habitat (part 2), or as threats to act against (part 7).

Going further#

The series, in order:

1. The ideas (this page).
2. Ranking a landscape with Zonation: rank maps, performance curves, weights, costs, connectivity, condition and Zonation 5 projects.
3. Reserves with Marxan, and irreplaceability: simulated annealing, the boundary length modifier and its calibration, selection frequency, ProxIRR.
4. Optimal plans with linear programming: objectives, targets, budgets, locked areas, gaps and solvers, portfolios.
5. Compact and connected plans, and zones: penalties, constraints, connectivity, hexagons and management zones.
6. Evaluating, comparing and reporting plans.
7. Multi-action planning: threats and actions.

Restoration planning is in Distance and restoration. The operators are documented in the operator reference.

References#

• Ball, I. R., Possingham, H. P., & Watts, M. E. (2009). Marxan and relatives: software for spatial conservation prioritization. In A. Moilanen, K. A. Wilson & H. P. Possingham (Eds.), Spatial Conservation Prioritization (pp. 185–195). Oxford University Press. https://doi.org/10.1093/oso/9780199547760.003.0014
• Beyer, H. L., Dujardin, Y., Watts, M. E., & Possingham, H. P. (2016). Solving conservation planning problems with integer linear programming. Ecological Modelling, 328, 14–22. https://doi.org/10.1016/j.ecolmodel.2016.02.005
• Cabeza, M., & Moilanen, A. (2001). Design of reserve networks and the persistence of biodiversity. Trends in Ecology & Evolution, 16(5), 242–248. https://doi.org/10.1016/S0169-5347(01)02125-5
• Church, R. L., Stoms, D. M., & Davis, F. W. (1996). Reserve selection as a maximal covering location problem. Biological Conservation, 76(2), 105–112. https://doi.org/10.1016/0006-3207(95)00102-6
• Game, E. T., Kareiva, P., & Possingham, H. P. (2013). Six common mistakes in conservation priority setting. Conservation Biology, 27(3), 480–485. https://doi.org/10.1111/cobi.12051
• Hanson, J. O., Schuster, R., Strimas-Mackey, M., Morrell, N., Edwards, B. P. M., Arcese, P., Bennett, J. R., & Possingham, H. P. (2025). Systematic conservation prioritization with the prioritizr R package. Conservation Biology, 39(1), e14376. https://doi.org/10.1111/cobi.14376
• Kirkpatrick, J. B. (1983). An iterative method for establishing priorities for the selection of nature reserves: an example from Tasmania. Biological Conservation, 25(2), 127–134. https://doi.org/10.1016/0006-3207(83)90056-3
• Kukkala, A. S., & Moilanen, A. (2013). Core concepts of spatial prioritisation in systematic conservation planning. Biological Reviews, 88(2), 443–464. https://doi.org/10.1111/brv.12008
• Margules, C. R., & Pressey, R. L. (2000). Systematic conservation planning. Nature, 405(6783), 243–253. https://doi.org/10.1038/35012251
• Moilanen, A., Franco, A. M. A., Early, R. I., Fox, R., Wintle, B., & Thomas, C. D. (2005). Prioritizing multiple-use landscapes for conservation: methods for large multi-species planning problems. Proceedings of the Royal Society B, 272(1575), 1885–1891. https://doi.org/10.1098/rspb.2005.3164
• Moilanen, A., Lehtinen, P., Kohonen, I., Jalkanen, J., Virtanen, E. A., & Kujala, H. (2022). Novel methods for spatial prioritization with applications in conservation, land use planning and ecological impact avoidance. Methods in Ecology and Evolution, 13(5), 1062–1072. https://doi.org/10.1111/2041-210X.13819
• Moilanen, A., Wilson, K. A., & Possingham, H. P. (Eds.) (2009). Spatial Conservation Prioritization: Quantitative Methods and Computational Tools. Oxford University Press. https://doi.org/10.1093/oso/9780199547760.001.0001
• Montesino Pouzols, F., Toivonen, T., Di Minin, E., Kukkala, A. S., Kullberg, P., Kuusterä, J., Lehtomäki, J., Tenkanen, H., Verburg, P. H., & Moilanen, A. (2014). Global protected area expansion is compromised by projected land-use and parochialism. Nature, 516(7531), 383–386. https://doi.org/10.1038/nature14032
• Naidoo, R., Balmford, A., Ferraro, P. J., Polasky, S., Ricketts, T. H., & Rouget, M. (2006). Integrating economic costs into conservation planning. Trends in Ecology & Evolution, 21(12), 681–687. https://doi.org/10.1016/j.tree.2006.10.003
• Pressey, R. L. (1994). Ad hoc reservations: forward or backward steps in developing representative reserve systems? Conservation Biology, 8(3), 662–668. https://doi.org/10.1046/j.1523-1739.1994.08030662.x
• Pressey, R. L., Humphries, C. J., Margules, C. R., Vane-Wright, R. I., & Williams, P. H. (1993). Beyond opportunism: key principles for systematic reserve selection. Trends in Ecology & Evolution, 8(4), 124–128. https://doi.org/10.1016/0169-5347(93)90023-I
• Rodrigues, A. S. L., Andelman, S. J., Bakarr, M. I., Boitani, L., Brooks, T. M., Cowling, R. M., et al. (2004). Effectiveness of the global protected area network in representing species diversity. Nature, 428(6983), 640–643. https://doi.org/10.1038/nature02422
• Rodrigues, A. S. L., & Gaston, K. J. (2002). Optimisation in reserve selection procedures: why not? Biological Conservation, 107(1), 123–129. https://doi.org/10.1016/S0006-3207(02)00042-3
• 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
• Underhill, L. G. (1994). Optimal and suboptimal reserve selection algorithms. Biological Conservation, 70(1), 85–87. https://doi.org/10.1016/0006-3207(94)90302-6