Impact of Total Precipitable Water Assimilation into North American Moonsoon Precipitation Forecasts

In our previous work, we concluded that, in days where strong synoptic forcing is lacking (weak days) in the North American Monsoon (NAM) region, the sensitivity of the initial specification of Total Precipitable Water (TPW) is crucial in precipitation forecasts in a weather model using a convective-resolving scale. Here, we assess retrospectively the impact of assimilating TPW observations from stations in the GPS Transect Experiment 2013 across northwest Mexico into the Weather Research and Forecasting Model (WRF) on precipitation hindcasts of the aforementioned weak days. We hypothesize that the correction of the initial conditions of TPW in WRF will improve convective forecasts. To do this, we carried out a suite of assimilation experiments using an ensemble-based data assimilation community software package (Data Assimilation Research Testbed, DART), in conjunction with a high-resolution WRF-ARW (WRF Advanced Research) 3-nested grid configuration with an innermost convective-resolving grid (2.5 km). The ensemble consisted of 20 members that have consistent physics and dynamics, but have different initial and boundary conditions. The initial ensemble was generated from perturbations in initial and boundary conditions of the outer domain from NCEP's 0.5-degree Global Forecast System (GFS), using the background error covariance (random cv=3) in the WRF data assimilation system (WRFDA). A spin-up period of 1 day was used to propagate the perturbations to the convective-resolving innermost domain prior to assimilating the 5-minute TPW data (with an observation error of 1 mm) at hourly assimilation cycles. We also use covariance localization and adaptive inflation in DART to account for sampling errors and insufficient variance due to deficiencies in the forecast model. We then conducted 24-hour forecast and analysis experiments for each weak day of the NAM 2013 period and assess the improvements in TPW and precipitation forecasts. Our initial results show that there are improvements of TPW and precipitation forecasts, which is consistent with our hypothesis. We also find that TPW data from the 9 stations in the GPS Transect Experiment 2013 may need to be augmented by other measurements, such as GPS stations from Suominet, to fully constrain TPW across the region. We present our initial results on our OSSE/OSE studies and discuss future directions.​