Assimilation Of Hyperspectral Satellite Radiances And Its Application In Regional Numerical Weather Prediction

Under the background of global climate change, the increase of extreme severe weather, such as typhoon and rainstorm, has intensified the damage on natural and human environment. Assimilation of hyperspectral infrared radiance data, which provides high resolution of temperature and humidity profiles, has significantly improved the forecast accuracy in global model. With the growing attention to meso-small scale severe weather simulation and forecast, more studies on application of hyperspectral data in regional model can be seen. However, many difficulties prevented the application of hyperspectral infrared radiance data on regional model assimilation. These difficulties come from the fact that regional model used forecasting/analysis field from global model as initial and boundary conditions. Therefore, spatial coverage of satellite observations varies in different time span. What’s more, available assimilation data is limited in regional model compared to global model. In this study, mesoscale model WRF-ARW (Advanced Research Weather Research and Forecasting) was chosen as forecast model, community model GSI (Gridpoint Statistical Interpolation) as analysis system, and infrared radiance from AIRS (Atmospheric Infrared Sounder) was chosen as input. It was an application of assimilation basing on AIRS infrared radiances that took terrain and climate of China and surrounding regions into account, and it also included the tuning of background error covariance matrix (B matrix) and bias correction. This study helped to deepen the understanding of the regional assimilation procedure and mechanic on hyperspectral infrared radiance data. Meanwhile, it explored the influence of tuning of B matrix and bias correction on typhoon track and intensity forecast.Main topics and contributions of this study are as follows:(1) Based on AIRS channels used in the assimilation of NCEP (National Centers for Environmental Prediction) GDAS (Global Data Assimilation System), a temperature sensitivity analysis was used to exclude channels that contributed most to weight function at atmosphere above lOhPa in regional model. Also, quality control on cloud detection in GSI, thinning, and threshold control in Asian regional data was diagnosed and analyzed. These steps enable the qualified AIRS radiance measurement data to be assimilated in the regional model.(2) Based on NMC (National Meteorological Center) method with two-month forecast, and combined the regional weather conditions, the regional B matrix was estimated. The B matrix was then compared with the global B matrix predefined in GSI so as to know more about the structure of regional B matrix. On overcoming the drawbacks of current B matrix estimation using NMC method, this study creatively used sensitivity method to optimize horizontal length-scale and standard deviations, which are important parameters in B matrix. The optimized B matrix was then applied to WRF. The24-hour forecasted temperature field, humidity field, and wind field were validated using radiosonde observations. The result showed that length-scale of regional B matrix was underestimated; while the standard deviations were overestimated. The tuned B matrix was then applied to typhoon forecast, and it significantly improved the accuracy of typhoon72-hour track forecast.(3) Using a scheme combining scan bias correction and air-mass bias correction, characteristics of bias in a one-month time-series was summarized. It was found that AIRS channels located in15μm CO2absorption band had large scan bias, and nadir bias has strong time dependence. By contrast, other channels had small scan bias and weak time dependence. In air-mass bias correction, predictors of zenith and temperature lapse rate had huge oscillatory due to regional model various data coverage. The effect of this scheme on correction in regional model was verified with the help of the histogram analysis on innovation (O-B, observed radiance subtracted by background simulated radiance). The verification showed that correction on most of the channels got satisfied results except several land surface channels. The corrected histogram satisfied the requirement of un-bias and a normal distribution. In typhoon forecast experiment, the influence of radiance bias correction on forecast result was tested. It showed that, compared to parameters from GDAS, regional radiance correction parameters from this study improved the prediction of typhoon72-hour forecast.(4) After the tuning of B matrix and bias correction, a case of typhoon landing was studied. It exemplified the convergence of cost function during the GSI3DVAR (Three-dimensional Variation) assimilation of AIRS radiances. It also summarized the contrast of O-B/O-A (Observation-Background/Observation-Analysis) between pre-assimilation and post-assimilation in order to justify analysis increment. As a result, the assimilation of AIRS radiances improved both the typhoon track and intensity of the typhoon in a72-hour forecast.(5) In order to widen the application of this study, a combination of direct broadcast satellite data and regional assimilation model was undertaken. The feasibility of local real-time forecast based on WRF/GSI regional assimilation model was tested. The direct broadcast system was X/L Dual Band Polar Orbital Satellite Data Receiving System in East China Normal University, and preliminary simplified model was developed by Cooperative Institute for Meteorological Satellite Studies (CIMSS) at University of Wisconsin-Madison, called CIMSS Regional Assimilation System (CRAS). The realization of this local real-time forecast was the basis of WRF/GSI-based local real-time forecast application.