Webinar training session on Ecological Niche Modelling

By Jean-Olivier Irisson (irisson@normalesup.org)

On November 8th, CoCliME project participants gathered online for an internal webinar training session on Ecological Niche Modelling (a.k.a. habitat suitability modelling). A goal of CoCliME is to produce climate services related to harmful algal blooms, which means using the data and knowledge we currently have about these blooms to try to predict their occurrence and intensity in the future, in a changing climate.

Ecological Niche Models are one solution to provide such predictions. These models relate the occurrence or abundance of a species of interest to the environmental conditions it is found in, hence defining the range of environmental conditions the species can tolerate or thrive in, i.e., its ecological niche. Then, based on the output from models of future environmental conditions, it is possible to outline when and where these conditions will occur and infer if the species of interest will be present/abundant.

The CoCliME training started with an R-programming primer, for those who were not familiar with it. R is an open-source language, oriented towards data science and statistics, which makes it quite easy to perform niche modelling. Then we continued with a class on the general principles of niche modelling, outlined above, and the specificities of Gradient Boosted Trees, a statistical methodology used to numerically fit niche models. We concluded with a practical programming session, to actually code in R and model the niche of a target species (in this training case, an Antarctic fish, to showcase the generality of the approach).

Teaching over 15 participants remotely was challenging but the videoconferencing worked well for the theoretical parts and using RStudio Cloud, which is free to use, greatly facilitated the practical parts; participants just needed a web browser and then cloned the data, wrote and ran their code in the cloud.

Screenshot (2).png

Picture: Result of the niche model computed by the attendees, in R through RStudio Cloud

JPI Climate Central Secretariat, Wednesday 4 December 2019