An error analysis for the hybrid gridding of Texas daily precipitation data

✍ Scribed by Andreas J. Rupp; Barbara A. Bailey; Samuel S.P. Shen; Christine K. Lee; B. Scott Strachan

• Publisher: John Wiley and Sons
• Year: 2009
• Tongue: English
• Weight: 838 KB
• Volume: 30
• Category: Article
• ISSN: 0899-8418
• DOI: 10.1002/joc.1917

✦ Synopsis

Abstract

This paper reports the error analysis results for the gridded daily precipitation data over the state of Texas of the United States from January 1, 1901 to December 31, 2000. The Global Daily Climatology Network dataset is used for both the data gridding and error analysis. The station data have been interpolated onto a 0.2° × 0.2° grid which starts at the base point (25°50′N, 106°38′W). The data gridding approach is a hybrid method, which is a blend of two simple methods: inverse distance weighting and nearest station assignment. Our gridding results are compared with those obtained by other gridding methods. The cross‐validation method is used for the error analysis. Our error analysis of the interpolated products includes not only the conventional errors, such as the mean bias error, but also the probabilistic distribution of the relative errors of precipitation frequency and the spatial distribution of a major Texas historical storm. The following results have been found: (1) a simple arithmetic average of station data usually overestimates Texas' average precipitation by 2.4 mm per day, (2) the relative error of the precipitation frequency follows a lognormal distribution, and (3) the hybrid gridding data do not have obvious bias and can reasonably display storm‐covered areas in Texas. The gridded data and error results are useful for the validation of climate models, calibration of satellite borne remote sensing devices, and numerous agricultural and hydrological applications. The statistical methods of our analysis and some of our results are applicable to other regions of the world. Copyright © 2009 Royal Meteorological Society