Evaluating geostatistical methods of blending satellite and gauge data to estimate near real-time daily rainfall for Australia
Rain gauges provide valuable information about the amount and frequency of rainfall. In Australia, the majority of rain gauges are located in populated, wet coastal regions. Approximately 2000 gauges reporting within 24 h of a target day were used to make near real-time (NRT) estimates of daily precipitation. The remaining approximate to 4000 gauges for the same target day were used to evaluate bias and estimation performance using several traditional statistics. There is considerable potential to improve the estimation of rainfall in Australia using related ancillary data, particularly in sparsely gauged areas. The Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TMPA-RT) near real-time product (3B42RT) provided images (0.25 degrees resolution) of precipitation across Australia. Daily precipitation was estimated in 2009110 approximately every 5 km across Australia. This study evaluated selected geostatistical methods for estimating daily rainfall maps across Australia. It tackled the change of support problem and spatial intermittency of daily rainfall data in blending satellite and gauge data. Dissension occurred amongst traditional global statistical measures of performance which were compounded by extremes of gauge density. Overall, our assessment is that blending the 3B42RT satellite and rain gauge data was not worthwhile. However, the blending considerably reduced the estimation variance indicating that uncertainty of the map estimates was a neglected property necessary to detect change and difference in patterns. (C) 2013 Elsevier B.V. All rights reserved.