Time-Series Modeling of Rainfall Density Information.
Abstract
In the literature there are basically two approaches to the prediction of climatological occurrence (rainfall) statistics: (a) One can use large numbers of observations of climatological phenomena to compile statistics on the probability of occurrence. (b) One can use computational models and available climatological data (past history) to calculate occurrence statistics. Recent modeling attempts start with very small point rainfall distribution functions which are transformed into specific attenuation data. The point rainfall rates are based upon a large history of data which does not indicate the spatial variation of rain. There are, however, indications that outage time on line-of-sight communication links can be estimated from distributions of point rainfall rates. Due to the random nature of rainfall, and due to the time dependence of such information, a logical approach to forecasting this phenomenon and interpreting the results with respect to systems performance seems to lie within the realm of non-stationary time-series modeling. This report presents an attempt to develop statistical models which can be used to forecast in near-real-time and to characterize the underlying stochastic processes of short-term rainfall density information. It should be mentioned that the analysis and modeling are directed towards the accurate characterization of precipitation density and not the probability of precipitation occurrence. Due to the wavelengths involved in Line-of-Sight Communication links, i.e., 5-30 GHz, the size of raindrops have a definite dispersive and absorptive effect on propagated electromagnetic energy. The importance, therefore, of this work is self-evident.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1978
- Accession Number
- ADA058699
Entities
People
- Richard J. D'accardi