Investigating the similarity of satellite rainfall error metrics as a function of Kِppen climate classification

This study addressed the question: How much similarity exists in uncertainty of space-borne precipitation products for similar Köppen climate zones in different and distant landmasses? Various metrics of satellite rainfall uncertainty were identified using a six year (2002–2007) archive of NASAʹs TRMM Multi-satellite Precipitation Analysis (TMPA) data product called 3B42V6 for two large distant landmasses that share many similar Köppen climate zones: 1) United States and 2) Australia. The level of quantitative similarity in error metrics for the same Köppen climate zones was then investigated. It was found that the bias and root mean squared error exhibited very close levels of similarity for similar Köppen climate zones in the US and Australia. However, similar inferences could not be drawn for other (higher-ordered) error metrics such as Probability of Detection (POD). The contrasting nature of the ground validation (GV) data (i.e., NEXRAD-radar in US and point gauge in Australia) for characterizing uncertainty may be one of the reasons for this observed lack of similarity. Using a dense gauge network of 42 gauges over a standard 3B42V6 grid box (~ 0.25°) as a ground validation benchmark, the dependence of uncertainty as a function of gauge density was quantified. These relationships were then cast in the context of our Köppen climate similarity experiment to identify the minimum level of gauge density that would be needed to resolve more accurately the actual level of similarity of error metrics for distant landmasses.