Climatic influences on Markovian transition matrices for Vancouver daily rainfall occurrence

Abstract

Two-state, first-order, single-site Markov models for daily precipitation occurrence were developed for each winter rainy season over the historical period of record at five long-term meteorological stations in the lower Fraser Valley of British Columbia, Canada. Monotonic temporal trends in the independent elements of the transition matrices were then assessed. Although the results remain tentative, there is some evidence for a regionally coherent long-term negative trend in the probability of wet-to-dry state transitions, P10 (or a positive trend in the probability of a wet day being followed by another wet day, P11). In contrast, there is no evidence for a regionally coordinated and consistent trend in the probability of dry-to-wet state transitions, P01 (or, therefore, in the probability of a dry day being followed by another dry day, P00). These results appear loosely consistent with previous statistical climate change impact studies in the region, and might be physically interpreted as suggesting a gradual increase in the local typical duration of a Pacific frontal storm during hydrologic winter, with no systematic trend in the average duration of a dry-day interlude. Additionally, the probability of any day-to-day precipitation state transition (from wet to dry, or from dry to wet), PST, has been tentatively interpreted to exhibit an area-wide negative long-term trend, suggesting an overall increase in precipitation memory. The findings provide some additional regional context for several issues in hydrometeorological modelling, climatology, and environmental impact assessment.