An Extreme-Preserving Long-Term Gridded Daily Precipitation Data Set for the Conterminous United States

Abstract:

Extreme daily precipitation drives flooding that can threaten lives and cause significant economic damages, and so is important to properly capture in gridded meteorological data sets. This work evaluates extreme precipitation in two widely used gridded data sets over the conterminous U.S.: Livneh et al. 2013 and Livneh et al. 2015. Compared to the underlying station data, they show a 27% reduction in annual 1-day maximum precipitation, 25% increase in wet day fraction, 1.5 to 2.5 day increase in mean wet spell length, 30-40% low bias in 20-year return values of daily precipitation, and 25% decrease in mean precipitation on wet days. These changes arise primarily from the time-adjustment applied to put the precipitation gauge observations into a uniform time frame, since most stations record either in the morning (6-8 AM) or evening (4-6 PM). The gridding process playing a lesser but still discernible role in reducing extremes. A new daily precipitation data set is developed that omits the time-adjustment (as well as extending the data by 7 years) but is otherwise similar to the earlier data sets. Omitting the time adjustment results in significantly better reproduction of observed precipitation metrics. When the new data are used to force the Variable Infiltration Capacity (VIC) land surface model, annually averaged 1-day maximum runoff increases 38%, annual mean runoff increases 17%, evapotranspiration drops 2.3%, and fewer wet days leads to a 3.3% increase in estimated solar insolation. These changes are large enough to affect portrayals of flood risk and water balance important for ecological and climate-change applications. The new data were developed as training data for statistical downscaling, where capturing extreme precipitation statistics is important for flood risk and water management. However, the new data may be applicable to other hydrological, ecological, and agricultural studies as well given the effect of the changes on annual mean runoff efficiency, evapotranspiration, and surface radiation. These results also suggest the importance of further pursuing the topic, since there is a surprisingly large impact on both daily and annual mean hydrology. Such efforts could, for example, combine hourly reanalyses and once-daily observations to better reconcile different station observation times.