We are trying to make distribution models of three recently discovered species (SDM) in my research area, since 2020. Very little is known about the biotic relationships (intra and interspecific) of these species, so in a BAM diagram (Peterson and Soberón 2012, Melo-Merino et al. 2020) we do not have the biotic factors, therefore we are not in a position to carry out ecological niche modeling (ENMs).
The province we work in has been sampled quite a lot since 2020 and our objective is to try to deduce the bioclimatic variables that intervene on the distribution area of each species (they are parapatric) and to see their possible potential spatial distribution in the province, in order to carry out future sampling and to know the spatial limits of each species and their intersection of areas. We think that Maxent is a good tool, together with QGIS, to establish this type of modeling. There are numerous tools that have been developed and complemented with the R software, which allow us to establish correlations between variables and eliminate them, to avoid collinearity with packages such as dismo or ENMTools (Warren et al 2024). However, due to the great development of these and many other tools, such as Wallace and others, we find a great debate about modeling (Araújo and Guisan, 2006; Domínguez, 2010), criticisms about the use of AUCs of ROC curves (Lobo et al 2013) or even whether Chelsea's bioclimatic variables are better than Worldclim (Karger et al 2017). We understand that it is currently conceivable to use several models and establish a "consensus model". We still have little experience with R, and given the great variety of methodologies used, we are somewhat lost in relation to the workflow to follow, although the new tutorial presented by Zoe et al (2024) provides us with great help. In conclusion, our question to the forum is: What would be your advice or the best workflow, depending on our objective, from the choice of variables (worldclim vs Chelsea) to what we want to achieve? Thank you for your attention.
Ensemble modeling with biomod2 package. I can help voluntarily in this matter. Worldclim and envirem have great, uptodate climatic variable sets. In some modeling, very good results can be obtained even when only bioclimatic variables are used. If I were you, I would first identify the most important climatic variables influencing their distribution, and only after making sure that these important climatic variables can be linked to their physiological, ecological or other biological traits, would I do ensemble modeling. Different models can be constructed using both chelase and worldclim variables separately and the results can be discussed. By the way, I am puzzled why researchers do not give a chance to the most recent scenarios (CMIP6; SSPs) but continue to pay attention to variables generated from older scenarios (CMIP5; RCPs).
Thank you very much for your contribution. As you have very well indicated, first of all I am working to identify the bioclimatic variables that can be linked to the species under study and physiological relationship, hence my doubts about which ones to use (worldclim vs chelsea), so it seems appropriate to make a comparison with both. I didn't know CMIP6; SSP, although for the moment we only intend to identify the climatic variables that affect our species to predict potential presence maps for future sampling. If you could tell me where I can download the CMIP6 climate variables, I would appreciate it.
CHEALSE data was created from old Worldclim data: here it is new Worldclim data : https://www.worldclim.org/data/worldclim21.html
Bioclimatic variables (30 seconds) for more preciese determining the important variables for the species..
Correct, I have this data. Sorry, I thought you were referring to another system. The overview paper on the CMIP6 experimental design and organisation is published in GMD (Eyring et al., 2016).
Hi, professor, I work with maxent, environmental suitability of plants. From geographic coordinates of species location (presence). Using wordclim data, it is possible to predict suitability for the future and simulate the present as well. If this is the objective of your work, it is available (2021-2040, 241-2060, 2061-2080, 2081-2100). for scenarios ssp126- ssp245 -ssp370- ssp585 from wordclin (some of these scenarios increase greenhouse gases, more moderate to more critical scenarios). If you have any questions, contact me in the chat so I can try to help you or, if you prefer, by email.
The data available here are future climate projections from CMIP6.
From modeling in Maxent it is possible to know which main climatic variables influence the distribution of species.
Worldclim has 19 climate variables, I apply it in R studio, where it is possible to remove multicollinearity (variables that correlate and select those that best explain their distribution based on geographic coordinates together with climate coordinates)
Today, professor, 90% of the works I read in literature. During my dissertation, several works that I read, MAXENT (WORDCLIN data, AUC and TSS test) are mostly used. They are effective for analysis, with other methods available that are the researcher's preference.
Thank you very much Marcelo for your response and disposition. Indeed, I also work with Maxent using geographical distributions of arachnids and with RStudio the statistical analysis is carried out to eliminate multicollinearity and select the bioclimatic variables that influence the study species. My debate is whether it was better to use worldclim vs CHELSA for European data. Both climate models could be used and comparisons made, but it is cumbersome, and CHELSA gives the temperature in tenths of ºC, so a temperature conversion must be made. Little by little I am solving some problems. Thank you very much for your response
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology