Conditioning stochastic properties of daily precipitation on indices of atmospheric circulation

Abstract

Time series of daily precipitation, at Valentia, Ireland, are examined using conditioned chain-dependent processes. The conditioning is carried out using two different indices of large-scale atmospheric circulation. The first conditioning index relates to mean monthly sea level pressure (SLP), conditioned as either above or below the historic monthly mean. The second conditioning index is based on the geostrophic wind direction (GWD), conditioned as being inside or outside the historic storm quadrant for the month. For each conditioning index, there are two conditioned daily models and they differ in the parameters that relate to both occurrence and intensity and therefore to changes in the distribution of monthly total precipitation. The results of conditioning the daily precipitation on SLP are compared with conditioning on GWD. The GWD conditioned model gave similar results to the SLP conditioned model for the wetter months, but performed poorly for the summer season. Combining the two SLP conditioned daily models into an induced model, which is a mixture of the two chain dependent processes, produces a variance for monthly total precipitation that is closer to the observed value than the unconditioned chain-dependent model.