Evaluation of Heat Wave Predictability Skills of Numerical Weather Models

Ademola Akinbobola, Ibraheem Ayomide Raji

Research Square · 10 May 2023 · ECMWF, NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)

In one sentence. S2S models overestimate heatwave frequency over northern Nigeria at 30-day lead; deep-learning bias correction improves forecast reliability.

Abstract

Abstract Significant changes is being experienced in the climate system due to the unprecedented rate of global warming. This has resulted in the increased frequency of weather extreme events which such as heatwave occurrence in the northern Nigeria. In order to mitigate the effects of heatwaves, early warning systems are needed to be implemented. Insufficient knowledge about the performance of the models is partly a factor that hinders the development of such systems. This study thus, addresses the gap by assessing the predictability skills of sub-seasonal to seasonal numerical weather model over different time lead and as well improves the predictability skills through the incorporation of deep learning to bias correct the model output at a 30-day lead period. The Excess heat index (EHI) was used to detect heatwave occurrence over the study area, using both observational and forecast data from selected S2S models at 5 -, 7 -, 15 -, and 30 – days lead time. Metrics employed to evaluate the skills of the models are; the Anomaly corelation coefficient (ACC), Symmetric External Dependency Index (SEDI) with each evaluating different strength of the models. The result of the analysis shows that the three models considered in this research overestimates the heat wave frequency in the region. This results in reduced reliability of the models in the region. Further analyses shows that the use of deep learning to bias the model output increases the forecast reliability in the region significantly.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.