Abstract
The statistical evaluation of the spatial similarity of human caused fire patterns is an important issue for wildland fire analysis. This paper proposes a method based on observed data and on a statistical tool (homogeneity test) that is based on non-explicit spatial distribution hypothesis for the human caused fire events. If a tessellation coming from a space filling curve is superimposed on the spatial point patterns, and a linearization mechanism applied, the statistical problem of testing the similarity between the spatial point patterns is equivalent to the one of testing the homogeneity between the two multinomial distributions obtained by modeling the proportions of cases on each cell of the tessellation. This way of comparing spatial point patterns is free of any hypothesis on any spatial point process. Because data are spatially over-dispersed, the existence of many cells of the grid without any count is a problem for classical statistical homogeneity tests. Our work overcomes this problem by applying specific test statistics based on the square Hellinger distance. Simulations and actual data are used in order to tune the process and to demonstrate the capabilities of the proposal. Results indicate that a new and robust method for comparing spatial point patterns of human caused fires is available.
Highlights
The understanding of spatial point patterns is important in many scientific scenarios and for wildland fires understanding and management
For a given period of time, an human caused fire (HCF) occurrence can be considered as a single point in a specific location on the Earth, and a spatial pattern of HCFs, the set of such points during the same period
This paper focuses on the comparison of two observed HCFs in order to determine if there is spatial patterns across census units, neighborhoods, counties, provinces/states, etc. [11]
Summary
The understanding of spatial point patterns is important in many scientific scenarios (e.g., biogeography, ecology, geology, pedology, epidemiology, etc.) and for wildland fires understanding and management. The improvement introduced in this paper is a motivated by the presence of empty cells in similarity the point patterns test statistic belonging to the family This fact may occur when the order of φ-divergence the curve increases or whenTwo we have a small or moderate number of events.have. HCFwe under theused null ahypothesis (H0) and Second, in relation to its to real of data, have dataset from [16]several that comprises parameters defining the base case are changed in order to carry out a study of the power of the the distribution of fires by different causes and years in the region of “Castilla La Mancha” (Spain) According to this dataset, 3095 to HCFs happened of them in 2004, 917 in 2005, Second, in relation its application to during real data,2004–2007, we have used a dataset from [16] that 493 in 2006comprises and 608the in 2007, respectively.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
