Theoreticalview of the shannon index in the evaluation oflandscapediversity

Abstract

Shannon's diversity index is frequently used in the determination of landscape diversity. Its indisputable advantage is a possibility to obtain numeric values that can subsequently be easily compared. However, accurate evaluation of landscape diversity from obtained results is rather complicated. The aim of the article is (i) to take a closer look at the theoretical origin of the formula that stems from the principles of the calculation of information entropy and (ii) to draw attention to several issues connected to the Shannon index application in landscape diversity assessment. Numeric value of the Shannon's index depends on applied logarithm base that is not precisely specified by the formula. Presenting the resulting Shannon index value without stating the logarithm base is not very suitable. Nevertheless, a bigger problem is the dependence of the resulting Shannon's diversity index value on two parameters, namely the number of studied categories and evenness of spatial distribution of individual categories. The resulting value may be identical for different types of the division of the study area. Therefore, the number of categories and the evenness of spatial distribution need to be taken into consideration in the very assessment of the Shannon index result. The number of categories could also be presented along with the resulting Shannon's index value. A major drawback of the Shannon index is its inability to express spatial distribution of patches within the area; it only presents the total extent of each category. Out of existing modifications of the index that try to take spatial distribution into consideration, the most convenient is the coefficient of the distance between the extent of identical and different categories. Based on arguments deriving from theoretical basis of the Shannon index formula and its practical application, a new view of landscape diversity maximum is presented. The application of the Shannon index disregards the fact that the original relation required for entropy calculation presupposes independence of the existing state (e.g. land cover categories in case of landscape assessment). With regard to the fact that commonly defined categories of patches are independent; the index calculation should make use of the relation considering conditional probabilities of the occurrence of a certain category.