The field of species distribution modelling has experienced fast growth in the last decade. With so many R packages available (sdm, dismo, ENMeval, BIOMOD, SDMtune, ssdm, esdm, ENMTML, ...) it is difficult to find the "best" approach. Which is the most comprehensive platform to fit, evaluate and project/predict species distributions across space and time, in addition to assessing variable's importance and response curves?
Dear Lilian,
You can find within the following tutorial of Babak Naimi (see the link), using SDM package, a very nice explanation of best approaches to species modelling.
https://www.youtube.com/watch?v=83dMS3bcjJM&list=PL6fy2AEPDkHRkIHPJ0BR3V1cEZ5UuTkUY&index=3.
Kind regards.
In my opinion SDMtune , is the best, User-friendly framework that enables the training and the evaluation of species distribution models
https://CRAN.R-project.org/package=SDMtune
I am in agreement with Nasir Hameed .
Jalal Kassout
Dr Sileesh Mullasseri
Ahnaf Ilman
Hello Lilian, I've been using BIOMOD2 for quite a while and I find it very useful, since you can perform all modelling procedures in a single package. Also, you have several algorithm options, as well as different ways to create pseudo-absences.Here you can have a video tutorial about it: https://www.youtube.com/watch?v=QrwqhJgRbnY.
Hi Lilian,
I totally agree with Nasir Hameed, the answer to your question depends on what type of SDM you are building. Each SDM has its strength and drawbacks based on the purpose and the goal of the model itself. For example, if you aim to build a rare insect species distribution model, you should consider using a presence-absence model rather than a presence-only model. However, I would suggest using the abundance model and include other abundant insect species that may affect the distribution of the rare species you are studying.
I present you a new R package for sdm that we just put online a few days ago. It is simple and efficient with a R shiny interface
https://cran.r-project.org/web/packages/sdmApp/index.html
SSDM and mopa packages are very good options.
https://cran.r-project.org/web/packages/SSDM/vignettes/SSDM.html
https://github.com/SantanderMetGroup/mopa/wiki
You can also check kuenm package, which helps to calibrate your model, to find the best candidate model, and to create and then evaluate final models into different areas or climate scenarious.
Once you get the folder structure, I think it is very easy to use.
https://github.com/marlonecobos/kuenm
Lilian Sales , I have two recommendations based on the ones i've personally used
1. SSDM package if you want to model multiple species together, the package is quite detailed and easy to implement. You will need to read their paper though (ssdm: An r package to predict distribution of species richness and composition based on stacked species distribution models). Here is the tutorial https://cran.r-project.org/web/packages/SSDM/vignettes/SSDM.html
2. If you're interested in single species modelling and will like to make future projections, then look at ‘MIAmaxent’. Also, it very important to read the paper to understand some of the underlaying assumptions. The paper is titled. "The MIAmaxent R package: Variable transformation and model selection for species distribution models".
Let me know if you need help getting the papers.
When properly tuned, the Random Forest algorithm 'library(randomForest)' seems to outperform even the biomod full ensembles. And it's fast (hence, cost-effective). See the super nice article from Valavi with Elith. Biomod is still very good (but need to select carefully the algorithms because not all are recommended).
Article Predictive performance of presence‐only species distribution...
I've just stumbled into this (very interesting) question, so probably my answer won't be relevant for Lilian's original aims... nevertheless, this is a never-ending topic and both the original question and all the replies provide useful thoughts and reading. I agree with answers saying that the right choice is somehow question-dependent; the best approches to be selected depend on the nature of the data, as well as on the purposes of the modelling. Different packages have different advantages for the different algorithms and purposes... in any case, as people working with distribution models, we should all be very grateful to all those who put their efforts in making available so many nice tools!
Actually, one could also build her/his own approach by taking hints, functions and commands from different packages and sources, maybe adding something more, to get to the approach that best mirrors the specific needs. We did something similar in this recent work:
Article Identifying climate refugia for high‐elevation Alpine birds ...
All the best,
Mattia
I suggest having a look at this just-published paper: Article Assessing the Effectiveness of Correlative Ecological Niche ...
This really new suite seems promising: Article flexsdm: An R package for supporting a comprehensive and fle...
SDMtune, a new R package that aims to facilitate training, tuning, and evaluation of species distribution models in a unified framework.
(13) (PDF) SDMtune: An R package to tune and evaluate species distribution models (researchgate.net)
I also found quite confusing to select the right packages for my needs. Sillero et al. (2023) provides some useful insights and a curated list of packages. I found it useful!
Sillero et al. (2023). A curated list of R packages for ecological niche modelling. Ecological Modelling. Article A curated list of R packages for ecological niche modelling
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling