Search

Identification of spatial and spatiotemporal precipitation variations plays an important role in different hydrological applications such as missing data estimation. In this paper, the results of Bayesian maximum entropy (BME) and ordinary kriging (OK) are compared for modeling spatial and spatiotemporal variations of annual precipitation with and without incorporating elevation variations. The study area of this research is Namak Lake watershed located in the central part of Iran with an area of approximately 90,000 km. The BME and OK methods have been used to model the spatial and spatiotemporal variations of precipitation in this watershed, and their performances have been evaluated using cross-validation statistics. The results of the case study have shown the superiority of BME over OK in both spatial and spatiotemporal modes. The results have shown that BME estimates are less biased and more accurate than OK. The improvements in the BME estimates are mostly related to incorporating hard and soft data in the estimation process, which resulted in more detailed and reliable results. Estimation error variance for BME results is less than OK estimations in the study area in both spatial and spatiotemporal modes. [ABSTRACT FROM AUTHOR]

As the largest component of the global water cycle, precipitation plays a critical role in determining available water in the terrestrial ecosystems. Numerous precipitation products have been generated during the last decades, but they differ substantially in terms of their spatial and temporal variations. In this study, we compared seven precipitation products covering ten major river basins of China from 1980 to 2007 and found large differences among the products. Overall, the Global Precipitation Climatology Center (GPCC) product shows the highest correlation with rain gauge measurements and lowest estimation errors over the ten basins. The performances of other products also clearly differ over the ten basins, and the maximum estimates were 1.27 ± 0.21 times the minimum estimates in terms of interannual variability in precipitation. Based on the seven products, we generated a new precipitation dataset using the Bayesian model averaging method. Compared with all the individual products, the new dataset indicates the decreased root mean square error and increased coefficient of correlation. In the statistical analysis of the weight for each of the products, the GPCC product contributed the maximum weight among almost all river basins. Both the ERA-Interim and MERRA products contributed the minimum weights, being less than 10 %. Our results highlight the need to investigate and improve the performance of precipitation products when evaluating the regional water balance. [ABSTRACT FROM AUTHOR]