Test cases for binary image metrics with weather forecast verification in mind

What is it about?

Several test cases are presented to be used to compare binary image metrics/measures with an aim at representing difficult comparisons in weather forecast verification. Several such measures are also applied to the cases, including the Hausdorff distance, mean-error distance, Baddeley's delta metric, dFSS, among others.

Featured Image

Photo by Kai Dahms on Unsplash

Photo by Kai Dahms on Unsplash

Why is it important?

In the last couple of decades, many new weather forecast verification methods have been developed that better address forecast performance spatially. These methods were developed in response to a need for better summaries of forecast performance as the resolution of weather models increased to a point where the high-resolution models had worse verification summaries than the coarser ones, even when a visual inspection of the two revealed that the high-resolution model was much better. New spatial verification methods are still being developed, and there is still a lack of information about which ones are most appropriate. These test cases go a long way towards identifying properties of the so called distance-based methods.