Statistical Method For Reducing The Number Of Climatic Predictors In Species Distribution Modeling

Nineteen bioclimatic parameters from BIOCLIM are widely used in Species Distribution Modeling (SDM). To improve modeling quality, it is essential to reduce the number of parameters. Several approaches have been proposed to solve this challenge, but each has its own limitations. In this study, we aimed to develop an effective statistical method based on identifying correlation groups of parameters and selecting the least correlated ones. Several statistical techniques were used to ensure a reliable parameter selection: simple correlation matrix analysis, cluster analysis (HDBSCAN), and factor analysis (varimax and quartimax). As an example, bioclimatic parameter values for the period 1991–2020 were analyzed for the whole globe. The results obtained using different methods show good consistency. Several correlation groups were identified, ranging from four to five, depending on the interpretation of the negative correlations. One group of two parameters, BIO14 and BIO17, can also be identified based on the results of the varimax factor analysis, although this correlation group was not identified by other methods. Finally, six bioclimatic parameters were selected (BIO2, BIO5, BIO7, BIO14, BIO15, and BIO18), one from each group that demonstrated the minimum average value of the correlation coefficient with parameters from other groups. The average correlation between the selected parameters was significantly lower than in the case of using previously applied methods with the same number of selected parameters.