Abstract
Organised tropical convection is strongly influenced by intraseasonal convectively coupled waves (CCWs), particularly Kelvin waves and the Madden-Julian Oscillation (MJO). Kilometre-scale models simulating deep convection explicitly offer new possibilities for investigating the multiscale dynamics of these waves. Due to the high computational cost of running kilometre-scale models, the output is often limited in duration. To evaluate the uncertainties in detecting CCWs using FFT-based and wavelet-based methods from datasets shorter than one year, we employ a measurement known as the Signal Ratio (SR), defined within the zonal wavenumber-frequency spectrum. The wavelet-based detection method slightly outperforms the FFT-based method in identifying CCWs across dataset lengths. Detectability is affected by the variable used: OLR provides superior detection for Kelvin waves, while precipitation is more useful for the MJO. Applying the wavelet-based method, Kelvin waves can be robustly detected with at least 90 days. The MJO can be robustly detected with at least 30 days of precipitation data. A SR-based criterion is applied to a case study of six tropical kilometre-scale simulations by the Unified Model (UM). Across the simulations, skill in representing Kelvin waves and the MJO appears linked, suggesting that both can be affected by similar aspects of model design. Our results encourage the use of wavelet-based detection, especially with relatively short datasets, and emphasize the importance of selecting suitable variables for detection. They also provide a benchmark for CCWs in numerical model runs of limited duration.