Effect of Reconstruction Parameters on Tomographic Imaging of Rainfall Fields from Multi-Parameter Microwave Observables

Abstract

Monitoring the temporal and spatial distribution of rainfall events is essential to the prevention and mitigation of natural hazards such as flash floods and landslides. Other applications include developing a database from which climatological behavior of storm information might be determined for improved understanding of storm dynamics of cloud physics, effects on radio Propagation links and rainfield statistics for rainfall prediction. Presently, spatial interpolation of point raingage measurements or weather radar rainfall estimates are employed for these purposes. However, these methods have limitations (such as low spatial and temporal resolution attained when using raingage networks or the loss of spatial coverage when using radar in complex terrain) which hamper their unconditional reliability. Alternatively, increasing reliability via denser raingage telemetry networks can lead to exorbitant costs associated with system complexity, maintenance and operation. Tomographic reconstruction algorithms have been prevalently used in the medical imaging field wherein a series of one-dimensional measurements or projections are transformed into a two-dimensional cross sectional image. The potential high resolution attained using these methods, coupled with their non-invasive nature, have made tomographic imaging attractive to widely diverse applications such as geophysical imaging, nondestructive testing in industrial manufacturing, and most recently, imaging ground-level rain intensities [ 11. Repeated over time, tomographic im-g of rainfall can monitor both spatial and temporal characteristics as rain events develop. For certain applications, this method possesses a significant advantage over current methods which use either raingages or weather radar, as it allows the observation of rain events in real-time (or near real-hf) with relatively high resolution over a reasonably large area (e.g. 500 km ). The original work by Giuli et al. [l] proved the feasibility of implementing tomographic imaging of rainfall fields using one-way specific attenuation measurements over a small fixed network. However, the use of specific attenuation measurements alone for tomographic reconstruction is not without limitation. Besides affirming the original techniques developed by Giuli et al. this paper explores the following issues: 1) transmitter and receiver siting as it affects accumulated and lnstantaneo us image formation; 2) inmduction of multi-parameter propagation observables (such as specific differential phase shift and specific differential attenuation) and their role in practical system implementation; 3) basis function selection in deference to physical storm Characteristics; and 4) the ability of the tomographic imaging process to adapt to different types and intensities of storms.