Expanding the Utility of GOES-R with Improved Assimilation of Lightning and Infrared Satellite Observations

This study develops nudging methods of data assimilation that will expand the utility of the forthcoming GOES-R satellite series by including data from the Geostationary Lightning Mapper (GLM) and Advanced Baseline Imager (ABI). We develop a method to help trigger storms where lightning is observed, but no storm in simulated. Another method suppresses spurious simulated storms based on estimates of satellite-derived cloud top height (CTH) that will be improved due to the increased number of infrared channels on the ABI. In lieu of GOES-R data, Earth Networks Total Lightning Network (ENTLN) observations and GOES-13 10.7-micron observations are used in the current research. The assimilation methods are developed and evaluated using the Weather Research and Forecasting (WRF) model at convection-permitting scales (3-km horizontal grid spacing). We verify precipitation simulations against NCEP Stage IV hourly precipitation observations by computing fraction skill score (FSS; a neighborhood approach) and frequency bias for three case days. Simulated temperature, winds, humidity, and surface pressure are also verified against METAR surface observations. Our new method of assimilating lightning observations nudges low-level vertical velocity to trigger storms. This method is compared to a previous method that nudges low-level temperature (MU). Both lightning assimilation methods then are combined with the assimilation of CTH. CTH assimilation removes hydrometeors above the CTH estimate and applies an amount of cooling that is proportional to the latent heat of the removed hydrometeors to suppress spurious convection. Without applying CTH assimilation, the MU method produces better precipitation forecasts in terms of FSS than our optimal configuration of the vertical velocity nudging (WNO) method. However, if WNO is applied together with CTH assimilation, WNO produces FSS during the forecast period similar to MU applied with CTH assimilation. MU generally produces stronger storms than WNO that cause more mesoscale subsidence and indirect suppression of spurious storms. Direct suppression during CTH assimilation diminishes the impact of the indirect suppression. CTH assimilation also provides greater convective available potential energy that supports the weaker WNO storms. Regardless of whether CTH assimilation is applied, WNO generally produces superior forecasts of surface fields relative to MU and a control that employs no assimilation. This improvement is 1-6% of root mean square error during a 12-h forecast period subsequent to assimilation. Lightning assimilation (WNO or MU) combined with CTH assimilation typically provides the best precipitation forecasts. These are better than the control during the first 6-12 h of the forecast period for 1-mm and 10-mm precipitation thresholds. This combined assimilation method with the operation of GOES-R enables assimilation of clouds and storms over areas devoid of quality radar, including mountainous terrain, ocean basins, and Central and South America.