An investigation of satellite-radar correlation functions.

Abstract

Several Montana convective storm complexes have been studied to determine the correlation fields (auto and cross) which are required for the application of optimum interpolation objective analysis and for the classification of the storms. It is shown that suitable correlation fields can be obtained with respect to the auto-correlation fields; however, it is also shown that a more general correlation model may be required for the cross-correlation fields because these tend to be multimodal. The correlations between the satellite data and radar derived rain rates at the modes were found to range from an order of 0.2 to 0.53 at the space-time lags appropriate to maximizing the correspondence between patterns; consequently, the potential value of using the multivariate objective analysis methodology is established.

These correlation fields represent the storm structure as implied by the observed data. The correlation fields are modeled using a four-dimensional Gaussian damped cosine function whose parameters reflect the storm characteristics. The function fitting also gives an estimate of the noise present in the data. The parameters of the model correlation function can also be used to classify storms with respect to their structural characteristics.

The estimation of surface rainfall using satellite imagery has been the topic of much discussion in recent years. Several approaches to this problem have been developed by other researchers, and several of these are briefly reviewed. A multivariate optimal interpolation objective analysis which combines satellite, radar, and gage data will give improved results. This report addresses the problems of combining the radar and satellite data sets through the use of space time correlation fields.