Improving the Quality of Precipitation Datasets in a Semi-Arid Region Using Spatial Interpolation

In the semi-arid southwest United States, summer monsoon precipitation is characterized by intense and extremely localized convective storms. To capture the spatial variability in precipitation, the United States Department of Agriculture-Agriculture Research Service (USDA-ARS) established the 150 km² Walnut Gulch Experimental Watershed (WGEW) in southeastern Arizona. The WGEW is equipped with a dense network of 88 rain gauges (approximately one rain gauge per 0.5 km²) and has been making hydrologic measurements since 1953. Quality control procedures have been implemented through time to ensure the quality of the precipitation dataset. However, no systematic error assessment has been made to the analog (prior to the year 2000) chart-based data. In this study we use daily interpolated precipitation fields for the period 1954-2015 to identify potential timing issues and outliers in the database. Two main sources of errors were identified in the analysis: Error Type I: rain gauges off, and Error Type II: coding date error. The Error Type I occurred when rain gauges were not operational, whereas the Error Type II occurred when precipitation is erroneously assigned to the day before or after the actual precipitation day during the digitization process or when there is a 12-hour shift as 8:00 AM and 8:00 PM. Additionally, the analysis is extended to identify outliers in streamflow in each subwatershed. The curated database will improve process understanding, facilitate model simulations, and allow a more accurate quantification of precipitation in a semi-arid environment where human and ecosystems compete for scarce and already over-allocated water resources.