Gyuwon Lee - Toward quantitative estimate and forecast of precipitation
The observational capability has been significantly improved for the last two decades. High resolution surface observations and remote sensing measurements, in particular Doppler radar networks, provide detail information on the quantitative precipitation amount, storm evolution, and environment. Although active research is underway to integrate this abundant information into the weather forecast, the forecast skill in short-term (0-12 hours) is still low particularly for precipitation forecast. The improvement in short-term precipitation forecasts has significant societal and economic benefits. There are increasing demands on improving the short-term precipitation forecast from various user groups such as aviation, construction, transportation, public service, etc. Scientifically, the accurate short-term forecasts require the optimal use of the abundant observational data in fine scales and the understanding of the physical processes that governs the storm evolution. In addition, it is very important to understand and overcome the shortcomings of the used models (numerical models or radar extrapolation) and their optimal use is essential to improve the overall predictability.
Recent advance in data assimilation techniques facilitates the use of observations in numerical models and leads to the improved precipitation forecast. However, they still suffer from phase (spatial and temporal) and intensity errors, possibly due to imperfect parameterization of various physics. Small phase errors can easily undermine the use of the traditional point-by-point skill. Thus, the separate evaluation of phase and intensity errors is extremely important to better represent the model performance. In addition, their consistency should be evaluated to better understand dynamics and parameterization in numerical models. I have been developing a new way of evaluating the consistency of these model errors and furthermore a scheme to correct model errors. This correction algorithm assumes the consistency of model errors in time and utilizes radar extrapolation techniques to advect model errors into the future. This correction significantly improves precipitation forecast up to 6-8 hours at the scale of North America. I have been implementing this correction for the Consolidated Storm Prediction for Aviation (CoSPA) initiated by the Federal Aviation Administration and testing with a meso-scale model (Fig 1). Eventually, the corrected model forecasts will be blended into the radar extrapolation with three (or four)-dimensional data assimilation methods. Currently, I am investigating several ways of integrating radar extrapolation with numerical models as a pre- or post-processing.
Fig.1: Precipitation rate from radar and MM5 forecast at t = 14z (t0 + 3 h). MM5 is initialized at t0 = 11z by assimilating radar data from 08z. (a) Radar rain rate, (b) original MM5 precipitation forecast with Lagrangian phase errors (vectors) predicted from the tendency between 11z and 12z and then advected to 14z, (c) MM5 forecast after correcting Lagrangian phase errors, and (d) MM5 forecast with the correction of Lagrangian phase errors and intensity errors.
The precipitation forecast highly relies on the accuracy of the precipitation estimate. The quantitative precipitation estimate is in particular crucial for the cold-season precipitation and its application into radar extrapolation and radar data assimilation. This requires the proper understanding and description of the microphysical processes, such as deposition, riming, and aggregation. I have been involved in the Canadian Cloud-Sat Validation project in which I try to develop a solid method to estimate water equivalent snow rate. The optical disdrometers provides snow shape, size, and fall velocity of individual snowflakes from which the snow mass (or density) – size relationships are derived from the aero-dynamics. Using these relationships, the water equivalent snow rate (S) and equivalent radar reflectivity (Ze) are derived from measured snow size distributions and are verified against snow gages and radars (Fig. 2). Currently, I am trying to parameterize these relationships and to connect them with underlying microphysical processes.
Fig. 2: Comparison of (a) equivalent radar reflectivity and (b) water equivalent snow rate from the optical disdrometer (HVSD) and other measurements (C- and X-band radars and snow gages). The total water accumulation is 12.1 and 11.2 mm from the optical disdrometer and double-fence international reference, respectively.
ASP Spotlight June 2007
For more ASP spotlights click here http://www.asp.ucar.edu/spotlight/archive.jsp