Short Range Forecasting Of Cloudiness And Precipitation Through Extrapolation Of Goes Imagery PDF Download

(Very) short range quantitative precipitation forecasting (QPF) plays an important role in both meteorological and hydrological risk management. Since New Zealand is an island country, which is surrounded by the Tasman Sea and South Pacific Ocean, most high impact weather systems, especially heavy rainfall systems, usually initiate and develop in the regions where there are no direct high resolution observations. Using satellite rainfall and cloudiness estimates to couple with the observations from the National Radar Network becomes crucial. This thesis makes use of satellite data coupled with observations from the National Radar Network for the initialization of a mesoscale forecast model for the region. To achieve this we employed a technique called “RainSat” to delineate precipitation maps in the regions beyond radar range. The errors associated with the “RainSat” technique include the accuracy of the statistical technique itself, sampling errors, height assignment, and the estimates of rain rates. These errors and the impacts on the forecast model have been investigated in Chapter 2 and 3 of the thesis. It has been found that, in spite of these significant errors, the “RainSat” technique is able to provide relatively useful estimates of precipitation out to a range of 200 km beyond radar maximum range. Besides the capability of extending the availability of the precipitation observations to the Tasman Sea, the “RainSat” technique has been used as additional data with the observed radar reflectivity for improving nowcasting in New Zealand (Chapter 4). The results showed that the combination of radar reflectivity and satellite retrieved rain rates can significantly reduce the uncertainties in the extrapolation based techniques that are caused by the incomplete echoes observed by radar alone in areas near the edge of the radar coverage area. According to our experiments, the improvements led by using the additional “RainSat” analysis became more obvious as the lead time increased. However, the skill was still very limited after 2-3 hours. Data assimilation experiments with radar and satellite data in New Zealand are introduced in Chapters 5-8. In order to incorporate radar (satellite) observed rainfall information with modest computing facilities, a new nudging based scheme has been introduced in Chapter 5. The new scheme uses the reverse Kessler warm rain processes and the associated saturation adjustment. The statistical scores showed that, by assimilating radar reflectivity data in the model using the new scheme, precipitation forecasts could be improved up to 7-9 hours ahead on average compared to the dynamic downscaling experiments. Since the assimilation operator developed in this thesis only uses a simplistic liquid phase microphysics scheme, the skill of the operator with more complicated model microphysics in the model were presented (Chapter 6). The results showed that different cloud physics schemes adopted within the time window have significant effects on the precipitation forecasting whilst showing minimal effects on wind corrections. According to our experiments, the use of the WRF Lin et al. scheme coupled with the RK-nudging approach might give the highest skill score on average during the nudging time window. . For New Zealand, high impact weather systems usually initiate and develop in regions that are beyond radar range, which means that some sort of satellite technique is particularly important for these events. In addition, the model background usually presents inaccurate estimates over the oceanic areas. Therefore, the incorporation of satellite retrieved moisture fields over the Tasman Sea is expected to be beneficial to the (very) short range precipitation forecasting in New Zealand. The assimilation experiments of the “RainSat” analysis are presented in Chapter 7. The newly developed scheme and the Water Vapour Correction (WVC) scheme have been employed and the verifications were carried out against to both radar and TRMM Multi-Satellite Precipitation Analysis using different objective scoring schemes. The results indicated that by using the satellite rainfall and cloudiness estimates to adjust the moisture fields out of the radar range, the precipitation forecasts could be further improved. In Chapter 8, the extrapolated rain rates generated from both radar and satellite data were used to adjust the corresponding model background. The results showed that the assimilation of radar and satellite based nowcasting data could effectively prolong the effects of the initial conditions in the NWP model and thus improve the precipitation forecasts even further. A brief conclusion is given in Chapter 9.