Glossary

The terms of the course, with the chapter where each is introduced.

Accessible area (M)

The area a species has been able to reach by dispersal over the relevant period; in a distribution model, where its background is drawn. Ch. 13

Additive benefit function (ABF)

A Zonation rule: the marginal loss of a cell is the sum over the features of the loss of a benefit that saturates with the share kept (R^z). Ch. 12.3

Aggregation

The tendency of cells or patches of a class to be clumped together rather than dispersed. Measured by ai, clumpy, cohesion, contag. Ch. 5

Area-weighted mean (_am)

The mean of a patch metric in which every patch counts in proportion to its area: the value of the patch in which a random cell of the class falls. Ch. 5

Attrition

The disappearance of patches, typically small ones. Ch. 10

AUC

The area under the ROC curve: the probability that a random presence gets a higher prediction than a random background point or absence (the Mann-Whitney statistic); depends on the extent of the background. Ch. 13

Autocorrelation length (Sal)

The shortest distance at which a surface's autocorrelation falls to 0.2: the size of the grain of its texture. Ch. 11

Background points

A sample of the conditions available in the study area, contrasted with the presences by presence-only distribution models; not absences. Ch. 13

BAM diagram

The places where abiotic (A) and biotic (B) conditions suit a species and that it can reach (M); it occurs where the three overlap. Ch. 13

Bearing area curve

The heights of a surface sorted from the highest, against the share of the surface at or above each; its core, peak and valley zones give Sk, Spk and Svk. Ch. 11

Boundary penalty (BLM)

A term added to the objective of an optimized plan, proportional to the length of its outline (Marxan's boundary length modifier): it makes plans compact. Its value has units and must be calibrated. Ch. 12.5

Boyce index (continuous)

The rank correlation between predicted suitability and the ratio of presences to background predicted in moving windows of suitability: near 1 for a good presence-only model. Ch. 13

Branch and bound

How a MILP solver searches: it solves the linear relaxation, splits on a unit with a fractional value (in or out), and discards every branch whose bound is worse than the best plan found. Ch. 12.4

Bridge

In MSPA, a non-core strip of habitat connecting two different core areas; a structural corridor. Ch. 4

Budget

The most a plan may cost, in the units of the cost layer; the constraint of the budget objectives (maximum utility, minimum shortfall, ...). Ch. 12.4

Calibration (of a penalty)

Solving a problem with a range of penalty values and choosing one on the curve of cost against the penalized quantity, e.g. at its knee or by TOPSIS. Ch. 12.6

Circuit theory

The modelling of movement as electric current through a landscape of resistances; equivalent to random walks. Ch. 8

Clamping

Holding each predictor at the edge of its training range before projecting a model, so that responses stay flat beyond the data. Ch. 13

Class

A category of a categorical map (a land-cover type); class-level metrics summarise all patches of one class. Ch. 5

Complementarity

The value of an area for what it adds to the areas already chosen: a site with species found elsewhere adds little, one with species found nowhere else adds much. Ch. 12.1

Component

A group of habitat patches joined, directly or through others, by links shorter than a threshold distance. Ch. 7

Composition

What a landscape contains and in which amounts, regardless of spatial arrangement. Ch. 5

Configuration

The spatial arrangement of the classes: size, shape, isolation, aggregation of patches. Ch. 5

Connectivity

The degree to which the landscape facilitates or impedes movement among resource patches; structural, potential or actual. Ch. 6

Contagion

The tendency of classes to form large, contiguous areas; at the cell level, the probability that a neighbour of a habitat cell is habitat (P_ff). Ch. 3, ch. 5

Contiguity constraint

A rule that the selected planning units form one connected network. Ch. 12.5

Core area

The part of a patch further from its edge than a given depth, where interior conditions prevail. Ch. 4

Core-area Zonation (CAZ)

A Zonation rule: the marginal loss of a cell follows the features that would lose the largest fraction of what remains of them (CAZMAX; CAZ1 and CAZ2 combine the fractions as sums or sums of squares). Ch. 12.3

Corridor constraint

A rule that given areas (terminals) are selected and joined by selected units, through low resistance. Ch. 12.5

Cost distance

The accumulated resistance along the cheapest route between two places. Ch. 6, ch. 9

Cost (of a planning unit)

What it takes to include a planning unit in a plan: acquisition, opportunity, management cost, or area. Missing values leave the unit out of the problem. Ch. 12.2

Current density

In circuit models, the current through a cell: the net number of passages of random walkers. High values mark corridors and pinch points. Ch. 8

Decision type

What can be done with a planning unit: select it or not (binary), any share of it (proportion), or any share up to a limit (semi-continuous). Ch. 12.2

Directional dependency

A one-way link between planning units, such as a river reach and the reach upstream: managing the first without the second costs a penalty or is not allowed. Ch. 12.8

Dispersal kernel

The probability of moving a given distance, often a negative exponential. Ch. 6

Dissection

The subdivision of habitat by thin lines such as roads. Ch. 10

dPC (and its fractions)

The percentage of the probability of connectivity lost when a patch or link is removed; split into intra, flux and connector fractions. Ch. 7

Edge effect

Any change in conditions or processes near the boundary between habitat and another land cover. Ch. 3

Edge width

In MSPA, the depth of edge effects expressed in cells. Ch. 4

Effective resistance

The resistance between two points of a circuit, combining the length and the number of routes between them. Ch. 8

Efficiency

Reaching conservation goals with the least area or cost. Ch. 12.1

Ensemble (of models)

An average of several distribution models, often weighted by their cross-validated scores; the spread between them maps uncertainty. Ch. 13

Equivalent connected area (EC, ECA)

The area of a single patch that would provide the same reachable habitat as a whole habitat network. Ch. 7

Extent

The overall size of the area studied. Ch. 2

Feature (prioritization)

Anything a plan should conserve that has an amount in every planning unit: a species, a habitat, an ecosystem service. Ch. 12.2

Fragmentation per se

The breaking apart of habitat independently of habitat loss: more, smaller, more isolated patches for the same amount. Ch. 3

Gap analysis

The measure of how much of every feature existing protected areas hold, and of what they miss. Ch. 12.3

Gradient model

A view of landscape structure as continuous variation in space, of which patches are a special case; the alternative to the patch-mosaic model. Ch. 11

Grain

The finest spatial resolution of the data: the cell size of a raster. Ch. 2

Graph

A set of nodes (patches) and links (connections) with attributes and weights. Ch. 7

Habitat

The resources and conditions that allow a given species to live and reproduce; species-specific, not a land-cover class. Ch. 1

Habitat amount hypothesis

The proposal that species richness at a site depends mainly on the amount of habitat around it, not on how it is divided into patches. Ch. 3

Habitat availability (reachability)

The amount of habitat reachable from within the habitat, counting the habitat of each patch and of the patches connected to it. Ch. 7

Hierarchic mask

In Zonation, levels of cells removed in order (all cells of a level before any of the next): existing reserves ranked on top of the rest. Ch. 12.3

IIC

Integral index of connectivity: habitat availability on a binary graph, discounting pairs by the number of links between them. Ch. 7

Irreplaceability

How much the conservation goals depend on an area; an area holding the only population of a species cannot be replaced. Ch. 12.1

Islet

In MSPA, a patch too small or thin to contain any core. Ch. 4

Knee point

The point of a trade-off curve farthest from the line between its ends: where improving one criterion starts to cost much more of the other. Ch. 12.6

Landscape

An area spatially heterogeneous in at least one factor of interest, at a scale set by the organism or process studied. Ch. 1

Landscape mosaic

The classification of each neighbourhood by its shares of agriculture, natural and developed land. Ch. 3

Least-cost path

The route of minimum accumulated cost between two places; a least-cost corridor includes near-optimal routes. Ch. 9

Linear relaxation

An optimization problem with its whole-number decisions allowed to take fractional values: easy to solve, and its optimum is a bound on the original problem's. Ch. 12.4

Locked-in, locked-out unit

A planning unit that must (existing reserves) or must not (cities, mines) be selected. Ch. 12.2

Marginal loss

In Zonation, the loss to the features of removing a cell at the point where it is removed, given the cells still remaining. Ch. 12.3

Matrix

The dominant, most connected land cover in which habitat patches are embedded. Ch. 1

Maxent

A presence-background distribution model: a log-linear model of the relative occurrence rate with a lasso penalty, equivalent to a Poisson point process model and to infinitely weighted logistic regression. Ch. 13

Maximum utility

The budget objective that maximizes the total (weighted) amount of the features held, without targets. Ch. 12.4

MESS

Multivariate environmental similarity surface: for every cell, the similarity of its least similar predictor to the data a model was fitted to; negative where a projection extrapolates. Ch. 13

Metapopulation

A set of local populations in separate patches, linked by dispersal, persisting through a balance of local extinction and recolonisation. Ch. 1

Minimum mapping unit (MMU)

The smallest area delineated on a map; it can make the effective grain coarser than the cell size. Ch. 2

Minimum set problem

Meeting a target for every feature at the least cost: the classic formulation of systematic conservation planning. Ch. 12.4

Minimum shortfall

The budget objective that minimizes the sum of the targets' shortfalls, each as a share of its target. Ch. 12.4

Mixed integer linear program (MILP)

An optimization problem with a linear objective, linear constraints and some whole-number variables, such as a decision of 0 or 1 per planning unit; the model of every ls.prioritize problem. Ch. 12.4

MOP

Mobility-oriented parity: the distance from each cell to its nearest reference conditions, and whether it is outside their ranges. Ch. 13

Moving window

A neighbourhood of fixed size and shape centred on every cell in turn, in which a measure is computed to produce a map. Ch. 3

MSPA

Morphological Spatial Pattern Analysis: the classification of every habitat cell into core, edge, perforation, bridge, loop, branch or islet. Ch. 4

Multi-action planning

Choosing which management action to take where (monitoring, actions against each threat) to meet recovery and conservation targets at least cost, or with the most benefit. Ch. 12.8

Neutral landscape model

An artificial map with known properties, used as a baseline to test the effects of pattern. Ch. 3; tools: neutral landscapes

Niche (fundamental, realized)

The environmental conditions in which a species can maintain populations (fundamental), and the part of them it occupies given competitors, predators and dispersal (realized). Ch. 13

Objective (prioritization)

The quantity an optimized plan minimizes or maximizes: cost, shortfall, utility, extinction risk, ... Ch. 12.4

Omission rate

The share of test presences predicted below a threshold set on the training presences (minimum or 10th percentile); much above 0 and 10 % signals overfitting. Ch. 13

Omnidirectional connectivity

Connectivity mapped between all places within a moving window, without choosing start and end points (Omniscape). Ch. 8

Optimality gap

The relative difference between the plan a solver returns and the bound no plan can beat: a gap of 0.01 proves the plan is within 1 % of the best possible. Ch. 12.4

Pareto front

The plans for which no other plan is better in one criterion without being worse in another, such as cost and shortfall. Ch. 12.6

Patch

A group of connected cells of the same class; "connected" by the 4- or 8-neighbour rule. Ch. 5

PC

Probability of connectivity: the probability that two random points of the landscape fall in habitat areas reachable from each other. Ch. 7

Percolation threshold

The proportion of habitat at which a cluster spanning the whole map appears in random maps: about 59% with 4 neighbours, 41% with 8. Ch. 3

Perforation

Holes inside habitat; also the MSPA class of the boundaries of such holes. Ch. 4, ch. 10

Performance curve

The share of a feature remaining as the cells of a ranking are removed, from the least to the most important. Ch. 12.3

Pinch point

A narrow place where movement is funnelled, marked by high current density. Ch. 8

Planning unit

The smallest area a plan can select or not: a grid cell, a property, a watershed, a hexagon. Ch. 12.2

Portfolio (of plans)

A set of distinct plans of nearly equal quality; how often a unit is selected in it is a simple irreplaceability. Ch. 12.6

Presence-only data

Records of where a species was found, without records of where it was looked for and not found: museum specimens, herbarium sheets, most GBIF data. Ch. 13

ProtConn

The percentage of a region that is protected and connected. Ch. 7

Rarity-weighted richness

The sum over the features of a unit's amount divided by the feature's total: a quick score of how much rare biodiversity a unit holds. Ch. 12.7

Replacement cost

How much more a plan costs when a unit is left out and the rest optimized again; zero when an equally good substitute exists. Ch. 12.7

Resistance surface

A map of the cost, risk or reluctance of crossing each cell. Ch. 6

Scale of effect

The extent around a site at which landscape structure best predicts a biological response. Ch. 2

Shrinkage

The reduction of patch size, typically from the edges. Ch. 10

Simulated annealing

A heuristic that improves a plan by random swaps, sometimes accepting worse plans to escape local optima; Marxan's method. It gives no guarantee of quality. Ch. 12.4

Solver

A program that solves optimization models, such as HiGHS or Gurobi. Ch. 12.4

Spatial cross-validation

Cross-validation with folds made of spatial blocks, so that test records are not next to training ones; random folds overestimate how well a model predicts new places. Ch. 13

Spatial prioritization

The ranking or selection of areas for conservation actions, balancing many features, costs and constraints. Ch. 12

Species distribution model (SDM)

A model relating the records of a species to environmental predictors, projected onto maps of them; also ecological niche or habitat suitability model. Ch. 13

Species-area relationship

The number of species, or a species' persistence, grows with habitat area as a power, S ∝ A^z, z near 0.25; the base of the minimum extinction risk objective. Ch. 12.4

Structuring element

In mathematical morphology, the small reference shape used to probe the habitat (erosion, dilation). Ch. 4

Subregional targets

Targets that must be met within each of several regions, such as states, to share the effort among them. Ch. 12.10

Surface metrics

Measures of a continuous raster (ISO 25178, gradient surface metrics): amplitude, height distribution, slopes, features and spatial structure, computed without a threshold. Ch. 11

Systematic conservation planning

Choosing areas for conservation with explicit goals, measuring what exists, and selecting areas for what they add to the whole (Margules & Pressey 2000). Ch. 12.1

Target (prioritization)

How much of a feature a plan should hold: a share of its total (relative), an amount (absolute), or one that falls with the feature's range (log-linear). Ch. 12.2

Target-group background

Background points drawn with the same sampling bias as the presences, from the records of the whole taxonomic group, so that the bias cancels. Ch. 13

Texture image

A map of a surface metric computed in a moving window around every cell. Ch. 11

Thematic resolution

How finely the classes of a map are defined. Ch. 2

Thinning (spatial)

Removing occurrence records until none is closer than a given distance to another, to reduce sampling bias and pseudo-replication. Ch. 13

Threat

Anything that harms features in a planning unit, with an intensity, and that an action can remove: deforestation, fire, invasive species. Ch. 12.8

Topographic position index (TPI)

Elevation minus the mean elevation of a window: positive on ridges, negative in valleys. Ch. 11

TOPSIS

A multi-criteria method that chooses the alternative closest to the ideal and farthest from the anti-ideal, after scaling the criteria. Ch. 12.6

Transition matrix

A table of the cells that moved from each class at one date to each class at another. Ch. 10

TSS

True skill statistic: sensitivity + specificity − 1 at a threshold, from −1 to 1 (0 for a random model). Ch. 13

VIF

Variance inflation factor: 1 / (1 − R²) of a predictor regressed on the others; above 5–10, collinearity makes its effect unstable. Ch. 13