U.S. Department of Commerce

 

Date of this Version

2012

Citation

Journal of Hydrology 428–429 (2012) 129–141; doi:10.1016/j.jhydrol.2012.01.030

Abstract

It is common that bias trends of long term precipitation data change over time due to various factors such as gauge relocation and changes in data processing methods. Temporal consistency of this error characteristic of precipitation data is as important as accuracy itself for reliable streamflow prediction with a hydrologic model. The main goal of this paper is to illustrate the detection and adverse effect of inconsistent precipitation data on distributed hydrologic model simulations over a mountainous basin in the Sierra Nevada Mountains of California. We used 1-h 4 km gridded precipitation time series intended for the second phase of the Distributed Model Intercomparison Project (DMIP-2), sponsored by the US National Weather Service (NWS). We present various analyses to investigate the consistency of an hourly gridded precipitation time series from October 1988 through September 2006. First, hourly gridded precipitation data were aggregated into monthly mean areal precipitation totals over the basin and compared with basin average totals derived from the Parameter-Elevation Regressions on Independent Slopes Model (PRISM) monthly values. Second, double mass analysis was preformed on several discrete locations within the basin using both the DMIP-2 gridded precipitation and PRISM data. In addition, we performed a statistical consistency test on the DMIP-2 time series. Both analyses lead to the conclusion that over the entire analysis time period a clear change in bias in the DMIP-2 data occurred in the beginning of 2003. Moreover, it was found that the PRISM data also exhibit some inconsistency. The inconsistency of two elevation zone mean area precipitation (MAP) time series computed from the DMIP-2 gridded precipitation fields was corrected by adjusting hourly values based on the result from double mass analysis. Model simulations using the adjusted MAP data are improved compared to simulations with the inconsistent MAP input data.

Share

COinS