Automatic morphing using image registration: Application to continuous tracking of radar reflectivity and rain fields

Rainfall is one of the most important natural phenomenon that influences human life. Accurate rainfall estimation and prediction are crucial for flood forecasting, flood control, climate diagnostics, and water resource management.; Rain data may be collected from numerous sources. Conventional rain gauge networks or meteorological radars provide continuous coverage in time. Satellite observations provide snap-shots of precipitation fields at poor temporal resolution. While a number of spaceborne platforms have been deployed for rain observation, the development of continuous space/time rainfall remains a major challenge.; This dissertation seeks alternative techniques to automatically generate continuous data streams of rainfall data from sparse or intermittent observations. In order to avoid human intervention in the process, an automatic procedure is needed for real-time operations. For this purpose, Automatic Morphing Using Image Registration (AMIR) model is developed by integrating automatic image registration and image morphing algorithm. The new AMIR technique uses automatic image registration as the basis for finding control points for the morphing process. In the study of data assimilation for weather forecasting, there is a need to generate continuous streams of rainfall data to alleviate the so-called “spin up” problem, or the inability to provide short-term forecasts [Road90]. The proposed algorithm has been tested using remote sensing images from Next Generation Weather Radars (NEXRAD) and Tropical Rainfall Measuring Mission (TRMM). Three cases of rainfall data have been used. These include the passage of a storm in Florida, hurricane Floyd, and scattered rain in the southwestern of the United States for the same period using NEXRAD radar data as surrogate for spaceborne observations.; These cases have drastically different spatial and temporal characteristics and hence provide tests on the applicability of the AMIR method. Comparative experimental results have shown that AMIR advance the current state of the art as it is comparable to manual morphing and outperforms automatic morphing without control points proposed in literature.