A warming climate may influence forecast performance : analysing the skill of maximum temperature seasonal climate forecasts over southern Africa

Abstract

Understanding seasonal variability and development of skilful seasonal climate forecasts (SCFs) is key in mitigating climate-related risks, including helping to support adaptation to climate change and variability. The purpose of this study is to consider possible factors influencing the predictability of maximum temperature SCFs in southern Africa. To address this question, two hypotheses are tested: namely (1) There is skill in making maximum temperature forecasts in the Southern African Development Community (SADC); and (2) The skill is contributed by two main attributes—ENSO-related climate variability and anthropogenic climate change—as a result, temperature forecasts are worth taking into account in pre-season decision-making. A state-of-the-art global climate model's atmospheric thickness fields are statistically downscaled to maximum temperatures for the austral spring to autumn period. Forecast performance over a 24-year period is evaluated for both original and for linearly detrended temperature data. The verification results indicate that predictive skill for maximum temperatures reflects the combined influence of ENSO-related variability and long-term anthropogenic warming trends. The majority of the skill is not, however, a consequence of warming trends, since the climate model is able to predict the seasonal-to-interannual maximum temperatures variation skilfully, without assistance from temperature trends. Detrending data improves probabilistic skill, suggesting that removing trends helps isolate the seasonal signal, enhancing the models' reliability and discrimination of probabilistic maximum temperature SCFs. However, deterministic skill declines, revealing long-term climate trends' influence on the apparent accuracy of deterministic forecasts. The trend thus influences understanding of forecast performance and needs to be considered when conveying how good a forecasting system is.