Impact of improved initial land surface conditions on HWRF simulations and HWRF coupled streamflow routing model

Thursday, 21 April 2016: 11:15 AM
Miramar 1 & 2 (The Condado Hilton Plaza)
Subashini Subramanian, Purdue University, West Lafayette, IN; and Y. Xia, Y. Wu, M. B. Ek, V. Tallapragada, and D. Niyogi

For the 2015 hurricane season, Hurricane Weather Research and Forecast model (HWRF) underwent significant changes aimed at improving tropical cyclone forecasts. One of such improvements was to change from GFDL slab model to Noah land surface model. In addition to predicting soil temperature, Noah land model also prognostically predicts soil moisture. It also produces other hydrological parameters that can be used to produce real time hurricane flood models. Though hurricane models continue to be a subject of targeted research and have come a long way in the past few years, they still remain very dependent on initial conditions supplied to the model. Currently, initial land state is obtained from GFS (Global Forecast System) input conditions. GFS soil moisture field is initialized through GDAS data and nudged to climatological values and thus may not represent the true state of land surface. NLDAS (North American Land Data Assimilation System), on the other hand is initialized using observations in precipitation and thus more realistic than GFS initial conditions. We hypothesize that by assimilating NLDAS land data into HWRF, the improved representation of land surface will drive improved performance in simulating surface parameters such as surface fluxes, soil moisture, soil temperature along with tropical cyclone related precipitation. Initial evaluation of TS Bill (2015) [Figure 1] suggests soil temperature and moisture patterns are improved upon using NLDAS data as initial land conditions when compared to the control run of using GFS as input data. Detailed diagnostics and analysis will be conducted using in-situ observations for soil moisture, soil temperature, TRMM precipitation and surface fluxes through Ameriflux data.


Since, NOAH land surface model also produces hydrology variables such as runoff, it can be potentially used as a flood prediction model when coupled with streamflow. Runoff at a point depends on soil moisture, its saturation and the amount of precipitation received at that point. Streamflow is a routing tool that is being used as part of NLDAS to provide streamflow anomalies. By coupling EMC's (Environmental Modeling Center) streamflow model and HWRF, real time streamflow forecast can be achieved. Techniques were developed to couple both the models to use runoff parameters from HWRF as input for streamflow model. Initial results for TS Bill (2015) show that the streamflow predictions in HWRF follow precipitation and soil moisture patterns and improves upon operationally produced NLDAS streamflow data [Figure 2]. We also hypothesize that by using runoff data from NLDAS assimilated HWRF runs into the streamflow model, greater accuracy in streamflow patterns can be achieved. Results will be compared against streamflow observations from USGS (United States Geological Survey) Water Data.


Figure 1: First level soil temperature (left) and soil moisture (right) for TS BILL (2015). True values (top, from NLDAS), GFS initialized experiment (center, FY15) and NLDAS initialized experiment (bottom).


Figure 2: Streamflow predictions when (top) coupled with runoff obtained from HWRF and (bottom) as obtained from the operationally run NLDAS model.


- Indicates paper has been withdrawn from meeting
- Submission entered in competition