The link between North Atlantic tropical cyclones and ENSO in seasonal forecasts
Abstract:
This study assesses the ability of six European seasonal forecast models to simulate the observed teleconnection between ENSO and tropical cyclones (TCs) over the North Atlantic. While the models generally capture the basin-wide observed link, its magnitude is overestimated in all forecast models compared to reanalysis. Furthermore, the ENSO-TC relationship in the Caribbean is poorly simulated. It is shown that incorrect forecasting of wind shear appears to affect the representation of the teleconnection in some models, however it is not a completely sufficient explanation for the overestimation of the link.Observed tropical cyclone-driven cold wakes in the context of rapid warming of the Arabian Sea
Validation of boreal summer tropical–extratropical causal links in seasonal forecasts
Abstract:
Much of the forecast skill in the mid-latitudes on seasonal timescales originates from deep convection in the tropical belt. For boreal summer, such tropical–extratropical teleconnections are less well understood compared to winter. Here we validate the representation of boreal summer tropical–extratropical teleconnections in a general circulation model in comparison with observational data. To characterise variability between tropical convective activity and mid-latitude circulation, we identify the South Asian monsoon (SAM)–circumglobal teleconnection (CGT) pattern and the western North Pacific summer monsoon (WNPSM)–North Pacific high (NPH) pairs as the leading modes of tropical–extratropical coupled variability in both reanalysis (ERA5) and seasonal forecast (SEAS5) data. We calculate causal maps based on the Peter and Clark momentary conditional independence (PCMCI) causal discovery algorithm, which identifies causal links in a 2D field, to show the causal effect of each of these patterns on circulation and convection in the Northern Hemisphere. The spatial patterns and signs of the causal links in SEAS5 closely resemble those seen in ERA5, independent of the initialisation date of SEAS5. By performing a subsampling experiment (over time), we analyse the strengths of causal links in SEAS5 and show that they are qualitatively weaker than those in ERA5. We identify those regions for which SEAS5 data well reproduce ERA5 values, e.g. the southeastern USA, and highlight those where the bias is more prominent, e.g. North Africa and in general tropical regions. We demonstrate that different El Niño–Southern Oscillation phases have only a marginal effect on the strength of these links. Finally, we discuss the potential role of model mean-state biases in explaining differences between SEAS5 and ERA5 causal links.Deep learning for downscaling tropical cyclone rainfall to hazard-relevant spatial scales
Abstract:
Flooding, driven in part by intense rainfall, is the leading cause of mortality and damages from the most intense tropical cyclones (TCs). With rainfall from TCs set to increase under anthropogenic climate change, it is critical to accurately estimate extreme rainfall to better support short-term and long-term resilience efforts. While high-resolution climate models capture TC statistics better than low-resolution models, they are computationally expensive. This leads to a trade-off between capturing TC features accurately, and generating large enough simulation data sets to sufficiently sample high-impact, low-probability events. Downscaling can assist by predicting high-resolution features from relatively cheap, low-resolution models. Here, we develop and evaluate a set of three deep learning models for downscaling TC rainfall to hazard-relevant spatial scales. We use rainfall from the Multi-Source Weighted-Ensemble Precipitation observational product at a coarsened resolution of ∼100 km, and apply our downscaling model to reproduce the original resolution of ∼10 km. We find that the Wasserstein Generative Adversarial Network is able to capture realistic spatial structures and power spectra and performs the best overall, with mean biases within 5% of observations. We also show that the model can perform well at extrapolating to the most extreme storms, which were not used in training.