Enhancing Extreme Weather Forecasts: Diagnostic Strategies in ECMWF's Destination Earth Initiative

This presentation will showcase diagnostic activities conducted at the European Centre for Medium-Range Weather Forecasts (ECMWF) as part of the Destination Earth initiative of the European Commission. This initiative aims to develop a continuous global component of Earth's Digital Twin on Weather-induced and Geophysical Extremes (Extremes DT) to forecast and monitor extreme weather events worldwide within a five-day range with unprecedented precision. Currently, the Continuous (Global) Extremes DT uses ECMWF's Integrated Forecasting System cycle 48r1 with approximately 4.4 km grid-spacing (TCo2559). 

Analysing continuous forecast runs and emerging extreme weather events (e.g., mid-latitude storms, tropical cyclones, medicanes, heatwaves, strong precipitation/flooding events) in the daily Extremes DT runs allows us to evaluate the performance of the km-scale Extremes DT model, particularly for extreme events. We compare the performance of DestinE with ECMWF's operational deterministic 9km forecasts and. All forecast resolutions are validated with high-density observations and model analyses to assess quality and physical realism. This analysis provides insight into which types of extreme events benefit from higher model resolutions for more accurate prediction. But it also highlights aspects in the model where improvements have not yet been achieved, or even led to some degradation in model skill compared to our current deterministic model with 9km resolution. Those issues underscore the need for improvements in other components of global models alongside increased resolution. 

Diagnostic work on these model shortcomings is critical for identifying causes and testing possible solutions. Variables such as 2m temperature, 10m wind speed, or precipitation have been diagnosed using various tools to understand general underestimation of extreme values when compare with observations (i.e. extreme cold temperature or large precipitation accumulations). As part of the diagnosis and proposal of improvements, different experiments have been conducted to analyse small changes in the model (i.e. small modifications in the parametrizations or the initial analysis) that could lead to significant performance improvements. 

This presentation will showcase some shortcomings of the Extremes DT forecasts and the diagnostic techniques applied to identify the underlying causes, especially related to extreme event prediction. This evaluation and diagnostic work aim to provide a better understanding of necessary model developments required to reduce identified model errors, which are crucial at km-scale for representing extreme events and thereby improving the quality of the Continuous Extremes DT simulations.