University of Wollongong
Exceedance set prediction for spatial environmental processes
For many environmental processes, thresholds are used to define anomalous or noteworthy conditions, and there is widespread interest in predictions of when, where, and by how much the threshold might be exceeded. The problem is challenging: Spatial predictive inference is based on sparse, noisy observations or proxy measures; it should account for spatial dependence in the process; and it should quantify prediction uncertainty. Hence, spatial statistical methods are increasingly being used to make inference on exceedances.
In this talk, a spatial hierarchical model is used to obtain exceedance-set predictions with known statistical properties. The geometric and probabilistic properties of the exceedance sets are used for their prediction and for quantification of their uncertainty. The methodology is illustrated for spatial fields of monthly Australian maximum-temperature data.