Use of exploratory spatial data analysis and geostatistics to verify the skill of atmosphere forecast models with improved data initialisation schemes

Abstract

The forecasting skill of a numerical atmosphere model with an improved initialisation scheme was evaluated against remote sensing observations using new spatial diagnostic verification methods. These methods use schemes to measure similarity/dissimilarity and spatial matching features between forecast and observations using image processing and geostatistical analysis. In this study spatial analysis showed an enhanced similarity between the experimental forecasts produced by a high resolution Eta atmosphere model and collocated remote sensing observations acquired by the orbiting Tropical Microwave Imager on board the Tropical Microwave Rainfall Measuring Mission (TRMM) satellite. This study approaches numerical model validation from a new angle by using the concepts of exploratory spatial data analysis and geo-statistics for numerical model validation. This approach has never been applied to verify improvements made to numerical atmosphere models. On the other hand, geostatistics is a fairly common approach to study and derive the distribution, spatial patterns and texture analysis of natural phenomena.