Species occurrence data are always incomplete snapshots of actual species distributions, which are often elusive, dynamic and constantly changing. Predicting species distributions and how they may evolve is crucial for ecosystem management and conservation planning, but the intrinsic unpredictability that is inherent to living organisms prevents accurate categorical forecasts of species’ occurrence sites, or accurate delimitations of their distribution ranges.
Species distribution models are often applied, but their predictions are then typically converted to binary predictions of presence or absence, using artificial thresholds whose choice can have substantial impacts on the results, conclusions and recommendations arising from such studies. The problem is compounded when these models need to be combined to forecast species diversity patterns, range overlaps, and distributional changes such as range shifts, expansions or contractions.
In this course, we will:
1) Recognize the intrinsic uncertainty associated with all species occurrence data
2) Employ easy-to-use tools for formally dealing with this uncertainty using species distribution models, favourability, fuzzy logic, and an analogy with quantum physics
3) Obtain more realistic estimates of species distribution, diversity (including vulnerability and endemicity) and change, without using thresholds.
The course includes theoretical and practical sessions, which are carried out entirely in R. We will use R packages such as fuzzySim and modEvA to 1) model species distributions based on available occurrence records, and obtain favourability from species presence probability; 2) evaluate the predictive capacity of these models with a varied set of metrics that address different facets of model performance, including discrimination and calibration; and 3) incorporate natural uncertainty in the analysis of species distributions, avoiding thresholds and instead using fuzzy logic to assess potential diversity, vulnerability, endemicity, overlap and change in biogeographical communities.
We will provide plenty of example data for the practicals, but participants are also encouraged to bring their own species distribution data to analyse under our guidance.
Participants will need a computer with R installed, a good internet connection, and preferably a webcam to keep the live sessions as dynamic and interactive as possible. Basic knowledge of R is required.
Click here to see the full Program
- Introduction to species distributions.
- Types of distribution data: Presence-only, presence-absence, range maps.
- Facing uncertainty in species distributions.
- Introduction to modelling species distributions.
- Presence-only, presence-background and presence-absence modelling methods.
- Overview of R packages for modelling species distributions.
- fuzzySim introductory practice: Obtaining distance-based fuzzy distribution data.
- Species distribution modelling with the favourability function.
- fuzzySim modelling.
- Calculating “quantum” measures of diversity and distributional change.
- Evaluating the fit and predictive capacity of distribution models.
- modEvA practice: Analysing and evaluating fuzzy distribution data.
- Course Fee
- Early bird (until November 30th, 2021):
- 475 €
(380 € for Ambassador Institutions)
- Regular (after November 30th. 2021):
- 620 €
(494 € for Ambassador Institutions)
- This includes course material (VAT included).
After registration you will receive confirmation of your acceptance in the course. Payment is not required during registration.