ABSTRACT
In this study, we used fire ignitions detected from the Visible Infrared Imaging Radiometer Suite (VIIRS) 375 m I-Band satellite data with the event location information to assess spatial point patterns and scale in Laos and Cambodia. The simplest theoretical model in deciphering spatial point patterns is an assumption of complete spatial randomness (CSR). Thus, rejection of CSR is a minimal prerequisite to model the spatial patterns through various statistical tests. The CSR acts as a dividing hypothesis to distinguish between spatial patterns. In this study, we used a variety of dispersion indices to quantify the spatial pattern which included the Index of Dispersion (ID), Index of Clumping (IC), Green’s index (GI), Index of Cluster Frequency (ICF), Index of Mean Crowding (IMC), Index of Patchiness (IP), and Morisita’s index (MI). We also used the geostatistical tool of semivariance to infer spatial dependence in fire datasets. In addition, correlogram analysis using Moran’s I was used to infer autocorrelation and scale. Results suggested the clustering pattern of fires in both Laos and Cambodia. All indices, ID, IC, GI, IMC, IP, and MI, suggested relatively higher clumping pattern for Cambodia than for Laos, except ICF, which was higher for Laos than for Cambodia. Semivariogram analysis revealed distinct spherical and exponential models for Laos and Cambodia, respectively. Also, a relative comparison of semivariance of fire events suggested higher values for Cambodia than for Laos. Further, fire datasets showed a positive Moran’s I at 50,000 m in Laos in contrast to less than 120 m in Cambodia. We discussed the implications of these results specific to fire management and mitigation efforts.
