Promises and challenges of trend detection and prediction for adaptation planning in the water sector – the Bayesian perspective

Recent development in water resources analysis and decision making under deep climate uncertainty brought various methods to help decision makers choose and sequence adaptation measures when key indicators of risk reach specific threshold values. For instance, a measure for enhancing storm surge protection could be triggered when sea level rises by one foot above the historical record. While the threshold values can be identified through stress testing procedures, there is no consensus for assess or confirm the risk has reached the risk thresholds, especially for variables where interannual variability is large, like precipitation. Extrapolating from the historical record is possible, although the often limited length reduces confidence in the trend estimate. Even if a significant signal were detected, one cannot be certain the trend will continue without complete knowledge of the physical mechanism underlying the trend. This study aims to combine insights from the observed record with climate projections within a Bayesian inference framework that relies on the assumption that information about future trends from climate projections (prior) can inform the estimate of the trend inferred from the historical period (posterior). The approach is illustrated for the case of the Santa Cruz water system (California) where precipitation is the used as the main predictor of risk. Model performance is discussed while using a single climate projection as a proxy for observed future climate.