Skip to contents

Videos

[author name]

2026-07-24

Source: vignettes/articles/Videos.Rmd
Videos.Rmd

Introduction to caretSDM package

This video introduces caretSDM, an R package designed to unify the Species Distribution Modeling (SDM) workflow into an integrated process, leveraging machine learning and GIS (Geographic Information System) tools.

Key features shown in the video:

  • Unified Approach: caretSDM addresses the fragmentation common in biodiversity research by connecting GBIF data, environmental variables, model training (based on the caret package), evaluation, and projections into a single, coherent pipeline (1:00 - 1:25).

  • Workflow: I demonstrate a practical example where, using just a few functions, one can download data, perform cleaning, define a study area, train multiple algorithms, and generate projections for future scenarios (2:35 - 3:35).

  • Accessibility and Open Science: The project is open-source, focused on reproducibility, and features extensive documentation, including 10 vignettes available on the official website for users of all levels (4:28 - 4:45).

The goal of caretSDM is to allow researchers to spend more time solving biogeographical hypotheses rather than fixing technical GIS issues or implementing standard methods, making the process more efficient and accessible.

Beginner’s guide

This video is a beginner’s guide on how to perform Species Distribution Modeling (SDM) using the caretSDM package. I demonstrate a unified workflow that allows users to complete the entire process, from data import to final visualization, within a single package.

Key Steps in the caretSDM Workflow, shown in the video:

  • Data Preparation: I explain how to obtain species occurrence data (using local files or GBIF) (0:05:34) and how to define the study area (sdm_area) using buffers, vectors, or rasters (0:15:44 - 0:29:29).

  • Structuring (input_sdm): Creation of the primary object that integrates occurrences and environmental/predictor data (0:51:15).

  • Cleaning and Processing: Automated data cleaning routines (0:55:35) and assessment of multicollinearity between predictor variables (1:00:40).

  • Pseudo-absences and Background: Selecting background or pseudo-absence data for model training (1:03:28).

  • Training and Prediction: Using machine learning algorithms to train models (1:11:15) and make predictions across different climate scenarios (1:23:28).

  • Results and Ensemble: I show how to create an ensemble of predictions (1:25:09) and generate outputs such as occurrence probability maps and partial dependence plots (1:27:57 - 1:29:06).

The video emphasizes that caretSDM is ideal for those seeking a closed, reproducible framework, while still offering the flexibility for advanced users to implement custom functions.