Improving the Drought Monitoring Capabilities of Land Surface Models by Integrating Bias-Corrected, Gridded Precipitation Estimates

Abstract

The “SPoRT-LIS” product is used by the NWS forecast offices in situational awareness (e.g., drought monitoring; White and Case 2015) as well as in local numerical modeling applications for enhanced initialization of land surface fields (Medlin et al. 2012). One of the limitations of the SPoRT-LIS accuracy is the input atmospheric forcing that drives the offline Noah land surface model simulation. The quality of the land surface model output is dependent on the input forcing fields, particularly the quantitative precipitation estimate (QPE). The current configuration of the SPoRT-LIS relies on long-term input forcing from the North American Land Data Assimilation System – Phase II product (NLDAS-2; Xia et al. 2012a; Xia et al. 2012b). The NLDAS-2 is the top choice for forcing due to its long duration (30+ years) and hourly output frequency. Due to real-time latency of 4 days in NLDAS-2, SPoRT-LIS bridges the gap for real-time output with supplemental hourly QPE from the Multi-Radar Multi-Sensor high-resolution radar/gauge blended product (Zhang et al. 2016) and global analyses from the National Centers for Environmental Prediction (NCEP)/Environmental Modeling Center (EMC). The limitations of NLDAS-2 forcing include real-time latency, coarse resolution grid (~15 km grid spacing), periodic quality issues arising from gauge quality control, and improper blending with the North American Regional Reanalysis (NARR). The primary limiting factors of implementing higher-resolution Multi-Radar/Multi-sensor (MRMS) QPE in a longer-term SPoRT-LIS are radar gaps and blending challenges on the peripheries of data coverage. Meanwhile, satellite QPE often introduces large biases due to inherent uncertainties in retrievals (e.g., Prat and Nelson 2015). SPoRT documented each of these QPE limitations and impacts on the soil moisture solution in a series of offline LIS simulations (Case et al. 2013). Problems with grid coverage and radar gaps are evident in the Stage IV and NMQ (currently MRMS) products, and the large over-estimation of satellite QPE was evident with the GOES-based algorithm. In order to provide an adequate spin-up of land surface model features, improved and bias-corrected QPE inputs are needed to preserve high spatial resolution features while offering seamless continuity in both space and time over a long duration simulation.