Evaluation of a Mesoscale Model with Different Surface Parameterizations and Vertical Resolutions for the Bay of Valencia

Abstract

Two different setups of the fifth-generation Pennsylvania State University?NCAR Mesoscale Model (MM5) are used and the results of 71 forecasts are evaluated with a focus on the accuracy of meteorological surface data including wind, temperature, dewpoint temperature, and pressure. The forecasts, which cover the Valencia coastal region with a 3-km grid, are nested into two coarser grids of 9- and 27-km mesh size. The integration time is 18 h. The ability to correctly simulate the sea breeze plays a major role for this area. Different model setups are used: in the M1 simulations 34 atmospheric vertical sigma levels combined with a five-layer land surface model (5L LSM) are applied for the first half of the year 2006. In total, 52 atmospheric vertical sigma levels and the Noah LSM are used for the second half of 2006 (setup M2). The accuracy improves for all meteorological parameters except pressure with the M2 setup. However, negative biases of temperature, dewpoint temperature, and wind speed worsen. Detailed analysis of results for 4 days simulated with all combinations of vertical resolutions and LSMs shows that the increase of vertical resolution is responsible for the improved forecast performance. The Noah LSM generally improves temperature performance, but reduces the performance of all other parameters, especially for days with sea-breeze conditions. The Noah LSM performs better than the 5L LSM scheme only for a day driven by synoptic changes.