Space-time clustering analysis performance of an aggregated dataset: The case of wildfires in Portugal

Abstract

This study focuses on the use of space–time permutation scan statistics (STPSS) to assess both the existence and the statistical significance of clusters on aggregated datasets. The investigated case study is represented from the Portuguese Rural Fire Database (PRFD) where the fire occurrences are georeferenced to an administrative unit level. The main goals are: (i) assessing the robustness of the STPSS to correctly detect clusters on aggregated datasets; (ii) testing the existence of space–time clustering in the PRFD; and (iii) characterizing the detected clusters. A synthetic database was designed to assess the potential bias introduced by aggregation of the data on the performance of the STPSS method. Results confirmed the ability of the STPSS to correctly identify clusters, regarding their number, location, and spatio-temporal dimensions and provided recommendations about the parameters setting of the scanning window. Finally, a discussion of the identified clusters on the PRFD is presented.