Improvement of the Optimal Blended NLDAS Drought Index (OBNDI) by Incorporating USGS Runoff Percentile and NCEI Operational Drought Indices

This paper presents the improvement of the Optimal Blended NLDAS Drought Index (OBNDI, a combined drought index) by incorporating USGS operational runoff percentile and NCEI's drought indices, such as the Standardized Precipitation Index (SPI) and the Standardized Precipitation-Evapotranspiration Index (SPEI). The North American Land Data Assimilation System (NLDAS) is a multi-institutional collaboration project. One of its major purposes is to provide long-term (from 1979-present) hourly and fine-scale (i.e., 0.125 degree) land surface hydrometerological products to support U.S. operational drought monitoring and prediction endeavors, such as the U.S. Drought Monitor (USDM) and the Climate Prediction Center (CPC) monthly and seasonal Drought Outlook. Eventually, NLDAS will support the National Integrated Drought Information System (NIDIS) and its Drought Early Warning System (DEWS). The OBNDI has been developed based on NLDAS phase 2 (NLDAS-2) multi-model ensemble mean soil moisture percentile, total runoff percentile, evapotranspiration percentile, and USDM monthly spatial extent of drought for six USDM regions (Northwest, West, High Plains, South, Southeast, and Northeast). The results show that the OBNDI can capture drought signals well for the USDM South region. For most of the other regions, the OBNDI can generally capture the broad drought features for the less severe USDM drought categories, known as “abnormally dry” and “moderate”. However, for the more pronounced USDM drought categories of “severe” and above, the OBNDI has difficulty in capturing the spatial extent and pattern of these more severe categories for most of the six regions.

Recent studies cited from the NLDAS-2 user community have shown that NLDAS-2 monthly total runoff percentile has weaker drought signals than the USGS monthly total runoff percentile. In addition, the NCEI SPI and SPEI products have not heretofore been used to derive OBNDI. It remains unclear if these products can improve the performance of the OBNDI to enhance drought monitoring capacity. In this study, we execute the following three experiments to explore if the USGS and NCEI products would improve OBNDI performance: (1) the current OBNDI algorithm as the default benchmark, (2) incorporating both the USGS total runoff percentile and the NCEI SPI percentiles into the OBNDI (Test1), and (3) incorporating both the USGS total runoff percentile and the NCEI SPEI percentile into the OBNDI (Test 2). The results show the Test1 and Test 2 largely improve the performance of the OBNDI for the West, High Plains and Southeast regions based on analysis of anomaly correlation, bias, Root-Mean-Square Error (RMSE), and Nash-Sutcliffe efficiency, while near neutral improvement is evident for the other three regions (albeit a slight deterioration in the Northeast). For the drought category of “exceptional” (most severe category), the OBNDI still has difficulty in capturing the spatial extent, pattern, and extremity of USDM drought features, partially due to few cases (small sample size). Therefore, further investigation is needed, for example, adding NCEI's Palmer drought index (PDSI) or satellite-based products such as the USGS Vegetation Drought Response Index (VegDRI), USDA Evaporative Stress Index (ESI), and NASA GRACE-based ground water product into the OBNDI. In addition, ongoing NLDAS-2 upgrade efforts such as (a) more accurate surface meteorological forcing data, (b) upgraded land surface models, and (c) applicable land data assimilation techniques, is expected also to improve OBNDI performance. These efforts are underway.