Spatial forecast verification

A recent wave of new methods for accounting for spatial information when verifying forecasts have been published in the past several years, and a new special collection of papers in the Weather and Forecasting journal is underway. This talk will give an overview of these methods, which can generally be classified into four categories: (i) neighborhood, (ii) scale-separation, (iii) features-based and (iv) field deformation. The first two categories are filtering approaches and the latter two are spatial displacement approaches. Some methods provide distributional information about forecast performance within the general bounds of these categories. The methods will be described and some discussion about where they belong in terms of these categories will be given.