Targeted model evaluations for climate services: A case study on heat waves in Bangladesh

Abstract

Though not a sufficient condition, the ability to reproduce key elements of climate variability over the historical record should be a minimum requirement for placing any confidence in a model’s climate forecasts or projections of climate change. When projections are used to guide practical adaptation, model evaluations should focus on the weather and climate events of interest to decision-makers, their physical drivers in the climate system and their variability on decision-relevant timescales. This paper argues for a greater emphasis on such targeted model evaluations to enable useful climate services. We illustrate this approach through a case study on heat waves in Bangladesh, but draw wider conclusions that are applicable to climate services development more broadly.The simulation of heat waves in Bangladesh is evaluated in several climate models, focusing on timescales relevant to the long-term viability of a heat action plan: the average, interannual variability and seasonality of temperature and heat-wave frequency. Where the physical drivers of variability are broadly captured, a considered interpretation of the models could provide insights into future heat-wave behaviour. However, substantial biases are found in the statistics and in some physical drivers of heat, raising questions about the suitability of some of the models for determining certain aspects of future risk. Specifically, simple bias corrections cannot be used to make inferences about possible future changes in various weather statistics such as timing of heat waves during the year. Results emphasize the potential pitfalls of performing only perfunctory climatological evaluations and highlight areas for model improvement in the simulation of South Asian climate variability.