High-resolution Simulations of a High-impact Rainfall Event for the Montpellier Region using WRF-ARW
IBM Research has an on-going project to develop capabilities for predicting the business impact of severe weather events, known as Deep Thunder. One of the goals is to enable customized, high spatial- and temporal-resolution forecasts in order to enable the optimization of resources and mitigate the impact of such events. Deep Thunder was deployed for the Montpellier region in an attempt to capture the 1 November 2011 event via “hindcast” analysis, based on the fine-tuning of a base model configuration. For this task, the Advanced Research Weather (ARW) core of version 3.3.1 of the Weather Research and Forecasting (WRF) community model was utilized to produce multiple 48-hour hindcasts, with output every ten minutes.
The domain configuration consisted of four, two-way feedback nests at 40.5-, 13.5-, 4.5-, and 1.5-km horizontal resolution, each with forty-five vertical levels, ten to fifteen of which were located in the planetary boundary layer, as defined via a hyperbolic function. The model was initialized with 0.5-degree data from NCEP's Global Forecasting System (GFS) spectral model, along with NASA 1-km sea surface temperatures in order to properly account for the oceanic influence on winds and convection along the coast of France. Because of the complex topography in the three outer domains, which include the Pyrenees and western Alps, and several peaks (> 1000 m) within the 1.5-km domain, particularly to the north and west of the city, 90-m NASA Shuttle Radar Topography Mission (SRTM) data were used as input for the model terrain. Given the influence of these mountain ranges, large grid meshes were required to capture such areas (i.e., the sizes of the aforementioned nests were 66x44, 88x82, 106x100, 124x115, respectively).
We will primarily focus on one of the hindcasts, initialized at 12 UTC on 31 October 2011. We will discuss the research objectives and challenges, model results and qualitative comparisons, and the potential for operational use of this capability at the IBM lab in Montpellier and the envisioned applications.