Evaluating the role of observational uncertainty in climate impact assessments: Temperature-driven yellow fever risk in South America

Sally Jahn , Keith Fraser, Katy AM Gaythorpe, Caroline M Wainwright, Neil M Ferguson

Global gridded temperature data sets (GGTDs) vary in their information sources, quality control procedures, generation techniques, and spatial-temporal resolutions, introducing observational uncertainty. This uncertainty is critical not only for studies on current climate conditions but also for future climate change projections, where observational data sets are used for bias correction and downscaling of global climate model (GCM) outputs. To minimize the impact of biases on current assessments and future projections, it is essential to ensure that the reference data set accurately represents the true climate state and spans a sufficiently long period to filter out internal variability. The selection of appropriate GGTDs is hence a crucial yet often overlooked factor in research that examines the impact of climate variability and change on vector-borne diseases such as yellow fever (YF). YF, an arboviral disease endemic to tropical regions of Africa and South America, has transmission dynamics that may be significantly influenced by climate change. In this study, we evaluated four GGTDs, namely the Berkeley Earth Surface Temperatures (BEST), the Climatic Research Unit Time-Series (CRUTS) and the ERA5 and ERA5Land reanalysis data sets, for health-related impact research, specifically examining YF transmission in South America. Each dataset was evaluated via grid-based analysis and validated against national weather station data, focusing on Brazil and Colombia, where YF outbreak risk remains. Our findings show that substantial differences among GGTDs affect the spatial representation of climate change indices, bioclimatic variables, and spatially aggregated temperature estimates at the administrative unit level, which generally serve as inputs for transmission models. In particular, while the reanalysis data sets generally outperformed the lower-resolution products, ERA5 demonstrated a slight advantage over ERA5Land despite the latter’s higher spatial resolution. Most importantly, our results highlight that variations among GGTDs can lead to markedly different estimates of key disease transmission parameters.