A critical comparison of using a probabilistic weather generator versus a change factor approach: irrigation reservoir planning under climate change

Abstract

In the UK, there is a growing interest in constructing on-farm irrigation reservoirs, however deciding the optimum reservoir capacity is not simple. There are two distinct approaches to generating the future daily weather datasets needed to calculate future irrigation need. The change factor approach perturbs the observed record using monthly change factors derived from downscaled climate models. This assumes that whilst the climate changes, the day-to-day climate variability itself is stationary. Problems may arise where the instrumental record is insufficient or particularly suspect. Alternatively, probabilistic weather generators can be used to identify options which are considered more robust to climate change uncertainty because they consider non-stationary climate variability. This paper explores the difference between using the change factor approach and a probabilistic weather generator for informing farm reservoir design at three sites in the UK. Decision outcomes obtained using the current normal practice of 80% probability of non-exceedance rule and simple economic optimisations are also compared. Decision outcomes obtained using the change factor approach and probabilistic weather generators are significantly different; whether these differences translate to real-world differences is discussed. This study also found that using the 80% probability of non-exceedance rule could potentially result in maladaptation.