On the representation of convectively coupled Kelvin waves in operational forecast models: An object-tracking perspective
Lawton, Q. A., Rios-Berrios, R., Judt, F., Magnusson, L., Köhler, M.. (2026). On the representation of convectively coupled Kelvin waves in operational forecast models: An object-tracking perspective. Weather and Forecasting, doi:https://doi.org/10.1175/WAF-D-25-0182.1
| Title | On the representation of convectively coupled Kelvin waves in operational forecast models: An object-tracking perspective |
|---|---|
| Genre | Article |
| Author(s) | Quinton A. Lawton, Rosimar Rios-Berrios, Falko Judt, L. Magnusson, M. Köhler |
| Abstract | Accurately forecasting convectively coupled Kelvin waves (CCKWs) remains a major challenge, as many models struggle to realistically simulate their structure and propagation. However, previous studies have often focused on a limited set of models or relied on diagnostics that obscure individual wave characteristics. It also remains unclear how well models represent interactions between CCKWs and other tropical waves, such as easterly waves (EWs). In this study, an object-based tracking framework is used to evaluate forecasts of CCKWs and EWs across nine operational models. These include traditional physics-based models and ECMWF’s new Artificial Intelligence Forecasting System (AIFS). Forecast skill is assessed as a function of observed wave attributes, life cycle phase, and environmental context. All nine models are found to underestimate CCKW strength and misrepresent the vertical structure, with the largest errors occurring during wave growth. More skillful models exhibit better vertical coherence between Kelvin wave–filtered rainfall and divergence, suggesting that forecast errors are linked to deficiencies in representing convective coupling. Forecast models also frequently miss CCKW–EW interactions, and captured interactions are typically understrengthened. Despite these challenges, AIFS forecasts of EWs and CCKWs compare favorably to those of physics-based models. While the original analysis period overlaps the AIFS training window, we find that this skill persists for forecasts outside that period, highlighting the potential of data-driven systems for tropical wave prediction. These results underscore persistent challenges in CCKW forecasting and motivate further work to better understand the representation of tropical wave interactions in numerical models. |
| Publication Title | Weather and Forecasting |
| Publication Date | May 1, 2026 |
| Publisher's Version of Record | https://doi.org/10.1175/WAF-D-25-0182.1 |
| OpenSky Citable URL | https://n2t.net/ark:/85065/d7ff3xx2 |
| OpenSky Listing | View on OpenSky |
| MMM Affiliations | DPM, WMR |