Application of Topographic Impact Horizontal Correlation Model to CMA-MESO System

Abstract

The impact of near-surface observations on analysis and forecasting in complex terrain is studied by introducing the role of terrain in the background error horizontal correlation model. Observation information propagates isotropically on the model level in the height-based terrain-following coordinates since the background error horizontal correlation in CMA-MESO 3DVar system is characterized by an isotropic Gaussian correlation model. However, in the near-surface layer with complex topography, the propagation of observation information is blocked by mountain ranges, and thus its background error covariance is inhomogeneous and anisotropic, and furthermore, the propagation of observation information should vary with topography. The background error horizontal correlation coefficients in complex terrain are computed using NMC method by National Meteorological Center of the USA. Results show that the blocking of large terrain causes the background error horizontal correlation coefficients to decrease more rapidly across mountain ranges, where the near-surface wind field is more localized than the temperature and humidity fields, with smaller horizontal correlation characteristic length scales, and the wind field information propagates over a closer distance. Based on the actual statistical structure, a Gaussian correlation model that includes effects of terrain height and terrain gradient is constructed, and the newly constructed horizontal correlation model accurately characterizes the decrease after mountain ranges are blocked. In CMA-MESO 3DVar analysis, the impact of terrain on the propagation of observational information is effectively incorporated by including a terrain height error term in the background error level correlation model. Idealized experiments show that the horizontal correlation modeling scheme considering the terrain height error term allows the observation information to propagate in an anisotropic manner that varies with terrain height，and significantly reduces the influence of observation information across large terrain features, thereby achieving more reasonable analysis increments. Results of a forecast experiment for a heavy precipitation process in northern China indicate that the correlation modeling scheme varying with the terrain height propagates the anisotropy of the ground observation information and weakens the analytical increment near the ground with large terrain, and thus makes a slightly biased and positive contribution to the precipitation forecast neutrality. Results of a 5-day hourly cycle rapid updating analysis and forecast for precipitation processes in East China show that the horizontal correlation modeling scheme with terrain elevation makes a positive contribution to 10-m wind field at the ground level and the precipitation forecast within 24 hours.

Keywords

• background error
• horizontal correlation model
• terrain