Addressing cell size selection in geodiversity quantitative assessment procedures

Within geodiversity assessment methods, the grid system is the most used approach to calculate a geodiversity index. Chosen for its simplicity and adjustment aptitude, it implies a fundamental and determinant decision, namely the selection of the cell size. This should comply with the inherent properties of the input datasets, being as well a function of the main goals of the analysis. This problematic has been studied by several authors, either by discussing and presenting analytical procedures, or by testing the effect of distinct cell sizes on their (modelling) analysis.

Being a key issue within geodiversity assessment methodologies, in this work an empirical methodology is presented to select the most appropriate cell size to assess the geodiversity of Portugal mainland. A direct quantitative method based on geodiversity indices was applied, using richness, and Simpsons’ and Shannon’s diversity and equity indices to geology (1:1000 000) and geomorphology (1:500 000) datasets of mainland Portugal, in a hexagonal analytical grid, through eight cell dimensions (1km, 2km, 5km, 10km, 15km, 20km, 25km and 30km). Several descriptive statistical parameters were analysed along the eight cell dimensions, for each map, with particular emphasis for dispersion statistical measures, namely quartile coefficient of dispersion, coefficient of variation, and skewness coefficient, range, Min, Max and IQR. The effect of cell size on the final maps of lithological and geomorphological diversity was also analysed, using the conventional representation of five classes (very high, high, medium, low and very low) based on the Jenks classification, by evaluating the area occupied per each class along the distinct cell sizes. The results from the analysis indicate that skewness coefficient, quartile coefficient of dispersion and coefficient of variation could be used as indicators of optimal cell size. All the performed analysis indicated the 5-10km dimension as the optimal cell size for the diversity and richness analysis, also considering the differences between both datasets, namely distinct scale, distinct polygon distribution area, distinct number of total features (categories). In general, the evenness and diversity indices seem to be more appropriated to be used as cell size indicators when applying statistic parameters, while richness seems to provide more clear results while using final maps and correspondent graphics that synthetize the area distribution per class. The empirical procedure here presented seems to be applicable to all type of scales and datasets.