Wavelet-based spatial comparison technique for analysing and evaluating two-dimensional geophysical model fields

Abstract

Complex numerical models of the Earth's environment, based around 3-D or 4-Dtime and space domains are routinely used for applications including climatepredictions, weather forecasts, fishery management and environmental impactassessments. Quantitatively assessing the ability of these models toaccurately reproduce geographical patterns at a range of spatial and temporalscales has always been a difficult problem to address. However, this iscrucial if we are to rely on these models for decision making. Satellite dataare potentially the only observational dataset able to cover the largespatial domains analysed by many types of geophysical models. Consequentlyoptical wavelength satellite data is beginning to be used to evaluate modelhindcast fields of terrestrial and marine environments. However, thesesatellite data invariably contain regions of occluded or missing data due toclouds, further complicating or impacting on any comparisons with the model.This work builds on a published methodology, that evaluates precipitationforecast using radar observations based on predefined absolute thresholds. Itallows model skill to be evaluated at a range of spatial scales and rainintensities. Here we extend the original method to allow its genericapplication to a range of continuous and discontinuous geophysical datafields, and therefore allowing its use with optical satellite data. This isachieved through two major improvements to the original method: (i) allthresholds are determined based on the statistical distribution of the inputdata, so no a priori knowledge about the model fields being analysed isrequired and (ii) occluded data can be analysed without impacting on themetric results. The method can be used to assess a model's ability tosimulate geographical patterns over a range of spatial scales. We illustratehow the method provides a compact and concise way of visualising the degreeof agreement between spatial features in two datasets. The application of thenew method, its handling of bias and occlusion and the advantages of thenovel method are demonstrated through the analysis of model fields from amarine ecosystem model. © Author(s) 2012.