Quantisation Analysis of Synthetic Rainfall

The advent of weather radar and satellite has been able to provide useful precipitation information for various water resources projects. However, in comparison to the traditional raingauge networks, the remote sensed rainfall suffers from a variety of unique error sources and it is very important for the data users to understand the uncertainties in the data and quantify them. This paper focuses on one of the error sources in the remote sensed data: quantisation. Basically, quantisation is a process by which continuous signals are transformed into discrete values. It is an important part of the signal processing involved in using remote sensed data since they don’t measure rainfall directly. Currently technological advances have made it easier to increase the number of quantisation levels, as witnessed by the replacement of a three bit system with an eight bit system by the UK Meteorological Office. However, there are still long valuable records of rainfall in the UK which are in three bits and they would be useful in statistical rainfall analysis such as frequency calculations. Therefore, it is important to explore the quantisation effects on different rainfall characteristics under different quantisation lengths. In this paper, a novel approach of using synthetic rainfall is adopted (based on a Poisson Cluster model). The benefit of this is the ability to vary the rainfall parameters, and thus the characteristics of the generated rainfall. Applying systematic errors and random noise to the synthetic rainfall, the effect of different quantisation schemes under various rainfall characteristics has been tested, which would provide valuable error information for the users of the quantised data.