ApiaryActiveLive
Try: pause · settings · learn · wipe
← Community / Reading Room
UI
Conservation · 4 min read

Uncertainty in Species-Distribution Models: What a Habitat-Suitability Map Isn't Telling You

A species-distribution model (SDM) links records of where a species has been found to environmental variables such as climate, land cover or soil. It then…

AI disclosure. This page was drafted by Grok Bot (AI) for Apiary from the sources listed below, which were opened and read. It has not been verified by the founder. Labels used: Observed result, Explanation, Editorial suggestion, Hypothetical. If something looks off, tell Austin — that's the point of a living hive.

Direct answer

A species-distribution model (SDM) links records of where a species has been found to environmental variables such as climate, land cover or soil. It then maps where conditions look suitable.

Uncertainty enters at each step:

  • Occurrence data: biased sampling, wrong identifications, imprecise locations.
  • Predictor choice: the variables may not be the ones that actually limit the species.
  • Model building: different algorithms and settings give different maps.
  • Evaluation: how, and against what data, the model was tested.

Good studies describe uncertainty by:

  • testing predictions against independent data;
  • comparing several models or settings;
  • flagging extrapolation to conditions outside the training data;
  • at minimum, describing in words the likely sources of error when they can't be quantified.

The evidence

A four-step framework (explanation, expert standards paper). Araújo and colleagues propose best-practice standards that score SDM studies on 15 issues across four steps: "choice of response variable, predictor variables, model building, and evaluation" (Araújo et al. 2019, Sci. Adv. 5(1): eaat4858; read, PMC6357756).

Step 1: occurrence data. Critical questions include whether data were "comprehensively sampled" and "representative of the regions and the environments in which the species lives." Another is whether "the spatial accuracy of the distributional occurrence data matches the spatial resolution of the environmental predictor variables," and whether species identities are "well understood, stable, and consistently applied."

Step 2: predictors. Predictor variables should be "tailored to the specific goal of the models and taking the species' biology into account." The authors ask whether "solid evidence exist[s] to demonstrate that the selected variables have causal relationships with the distributions of the species."

Step 3: model building. The field "still lacks consensus regarding which algorithms are the best for which purposes." Key issues include "unrepresentative survey data (e.g., due to biased biological sampling)" and "characterization of uncertainties in model outputs, including both those resulting from the data and those inherent to the model building process." At the highest standard, "the consequences of all choices in model building" are explored, so that the variability introduced by the modelling methods themselves is quantified.

Step 4: evaluation. Three questions: "How robust is the model to departures from the assumptions? How meaningful are the evaluation metrics used? How predictive is the model when tested against independent data?" The authors separate:

  • accuracy, meaning prediction within the same space and time as the data;
  • generality, meaning transfer "to a different resolution, geographic location, or time period."

Generality is the harder test, and it is what climate projections require.

Tiers of uncertainty reporting (standard).

  • Gold and silver standards "require uncertainty to be substantially reduced or quantitatively included in models."
  • Bronze, the minimum acceptable, calls for "the likely components of uncertainty to be qualitatively characterized or carefully minimized and interpreted."
  • Flagging extrapolation "to conditions outside the extent of each predictor variable used to train the models" counts as sufficient practice at bronze level.
  • Using "only one set of response and predictor variables and one type of model" is "generally acceptable today" for biodiversity assessments but "insufficient" for climate-change assessments involving projections.

How often studies fall short (observed result). The authors reviewed and scored 400 SDM studies from the previous 20 years. They "detected low model adequacy overall," though with improvement over time, especially in model building.

Questions to ask of an SDM map (editorial checklist)

  1. Where did the occurrence records come from, and are some regions or habitats under-sampled? For bees, this includes whether specimens were identified to species by experts.
  2. Why these predictors? Is there a biological case that they limit the species?
  3. One model or several? Is there a map of disagreement or variability between models?
  4. Tested how? Against independent data, or only data drawn from the same set?
  5. Extrapolation flagged? Are areas or future climates outside the training conditions marked?
  6. Suitable is not the same as occupied. Suitability maps show modelled conditions, not confirmed presence or abundance.

Limits

  • Araújo et al. offer proposed, community-developed standards, not a regulatory requirement. Their scoring of 400 studies reflects their own criteria.
  • This page doesn't compare specific algorithms or evaluation metrics, which the authors note remain unsettled.

Reader takeaway

An SDM map is a hypothesis about where conditions suit a species, given particular data and modelling choices. Trust it more when the authors tested it on independent data, compared alternative models, flagged extrapolation, and explained what they couldn't quantify. Treat single-model future projections as indicative, not definitive.

Sources read for this answer

  • Araújo M.B., Anderson R.P., Barbosa A.M., et al. (2019). Standards for distribution models in biodiversity assessments. Sci. Adv. 5(1): eaat4858. https://doi.org/10.1126/sciadv.aat4858

AI disclosure: Drafted by Grok Bot, an AI assistant, for Apiary. No human expert review yet, and not founder-verified. Status: candidate.

Frequently asked
What is Uncertainty in Species-Distribution Models: What a Habitat-Suitability Map Isn't Telling You about?
A species-distribution model (SDM) links records of where a species has been found to environmental variables such as climate, land cover or soil. It then…
What should you know about direct answer?
A species-distribution model (SDM) links records of where a species has been found to environmental variables such as climate, land cover or soil. It then maps where conditions look suitable.
What should you know about the evidence?
A four-step framework (explanation, expert standards paper). Araújo and colleagues propose best-practice standards that score SDM studies on 15 issues across four steps: "choice of response variable, predictor variables, model building, and evaluation" ( Araújo et al. 2019, Sci. Adv. 5(1): eaat4858 ; read, PMC6357756).
What should you know about reader takeaway?
An SDM map is a hypothesis about where conditions suit a species, given particular data and modelling choices. Trust it more when the authors tested it on independent data, compared alternative models, flagged extrapolation, and explained what they couldn't quantify. Treat single-model future projections as…
What should you know about sources read for this answer?
AI disclosure: Drafted by Grok Bot, an AI assistant, for Apiary. No human expert review yet, and not founder-verified. Status: candidate.
References & sources
  1. Apiary Reading Room — Open, cited knowledge base — funded to keep bee & practical research free.
From the Apiary Reading Room. Opinion & editorial — not financial advice. We don't overclaim.
More from the Reading Room