In recent years, several sets of analyzed fields of global land precipitation have been constructed by interpolating historicalgauge observations (e.g. Dai et al. 1997, New et al. 2000, and Chen et al. 2002). Covering extended time periods of multiple decades, these data sets have been utilized to detect long-term trends in precipitation over various portions of the global land, in addition to their wide applications in analysis of climate variations of seasonal to inter-annual time scales. Uncertainties, however, exist, in these gauge-based analyses due to changes in the density and configuration of gauge networks. In particular, over regions where natural long-term variability of precipitation is relatively small compared to the spatial gradients of precipitation fields, shifts of gauge locations over the recording periods will yield temporally changing bias in the gauge-based analyses, producing an artificial trend of long term precipitation.

In this study, we examine and quantify the uncertainties of the published data sets of gauge precipitation in detecting long-term trend of global land precipitation. Quantitative comparisons are performed between gauge-based analyses derived from fixed and changing gauge networks to examine the magnitude of the uncertainties on each grid box of 0.5 deg lat/lon over the global land areas. The results are then compared against the long-term trend derived from the gauge-based analyses on various spatial scales to quantify their relatively importance. Detailed results of this study will be reported at the workshop.