The Recent Californian Drought, Weather Variability, and Climate Variability

Abstract:

To determine the explanatory necessity of climate variability (including secular components) in manifesting the recent California drought, we create empirical stochastic climatological null models at multiple spatial scales against which to compare the observed precipitation record. These models are chosen as the optimal means to generate the observed precipitation data when marginalizing across certain types of variability, and thus allow us to place minimum bounds on the effect of climate variability in creating observed conditions. Using approximately 100 years of data for 13 California weather stations, seasonal Markovian models of daily occurrence and depth are constructed using a generalized linear modeling framework. These models are flexible enough to capture the variability of daily precipitation, and their parameters are readily estimated using standard methods. Simulated realizations of centuries of daily data, however, do not capture the measured variance of precipitation at seasonal to annual time scales, indicating that low frequency variations in the underlying processes are present, which by construct are not accounted for. To account for this, a likelihood-based clustering algorithm is used to distinguish differing data generation regimes; the final estimated models are again used to simulate ensembles of the last century of rainfall, and to assess how unusual, or not, the 2013 drought was. At 12 of 13 stations, the 2013 rainfall was not statistically unusual, in that a year as dry would be expected to occur at least once per century in 10% of the ensembles. At 1 station, however, the simulations reproduced 2013 levels of rainfall (in each simulated century) less than 10% of the time, from which we conclude that the hypothesis of 2013 precipitation being produced by the same stochastic process all other years is rejected at the p=0.1 level at 1 station. Similarly, for gridded precipitation data, models are constructed which optimally represent the daily precipitation statistics, but do not represent interannual climate variability, and thus provide a means to establish locations where climate variability is necessary to explain the 2013 drought. Further, the differences between "weather" and "climate" variability provide upper limits for the predictability of future drought events.