Extended Abstract (164K)P1.4
The impact of assimilating AVHRR–derived humidity on high latitude MM5 forecasts
Xingang Fan, University of Alaska, Fairbanks, AK; and J. S. Tilley
High resolution AVHRR (Advanced Very High Resolution Radiometer) infrared imaging data is applied to MM5 Forecasts in an attempt to improve mesoscale cloud and precipitation representations. A technique is developed to adjust moisture fields from an initial first-guess. The adjusted moisture fields are then assimilated via two assimilation approaches previously used for the study of an Alaska heavy rain event in August 2000.
The previous studies showed the benefits of both a Newtonian nudging scheme (NN) and an intermittent data assimilation scheme using a Bratseth analysis (IDA). However, some problems were revealed by the aforementioned studies, including terrain-associated wet biases and a lack of systematic domain-wide improvements in the forecasts. A primary contributing factor to the above problems may well be the spatially sparse nature of available observational data. Available surface observations and upper air soundings are too sparse to give a good representation for meso- or micro-scale atmospheric structures, especially over areas of complex terrain.
AVHRR infrared data has 4km resolution, which is very good for application to mesoscale simulations and forecasts. In this study, the AVHRR channel 4 (10.3-11.3 m) brightness temperature is used to derive cloud information by a simple method. Under a one-layer cloud assumption, the original 3-dimensional MM5 analysis fields are used to determine cloud top and cloud base. Then the 3-dimensional humidity is adjusted according to empirically-derived relative humidity thresholds for cloudy or clear grid cells. The adjusted humidity fields are used in both NN and IDA assimilation experiments.
For comparison with our previous results, we assimilate the adjusted humidity fields at 6-hour intervals. The results of the NN method show improvements in the position and magnitude of the heavy rain centers. Additional experiments using 3-hourly AVHRR data show additional improvement.
Time series of 6-hourly and 3-hourly accumulated precipitation from both observations and model forecasts are presented at several stations as part of our validation of the simulations. The time series show that the NN approach simulates a better average trend of rainfall while the IDA approach produced maximum precipitation rates in better agreement with the observations. A final experiment combining the two methods is also conducted; the results show further improvement in the precipitation forecast.
Poster Session 1, Weather Analysis, Forecasting and Numerical Prediction
Monday, 12 August 2002, 3:00 PM-4:30 PM
Previous paper Next paper