Nonlinear synergistic effects of three‐dimensional soil moisture errors on uncertainties in subseasonal heatwave onset predictions in the <scp>Yangtze River</scp> region

Abstract Heatwaves (HWs), which are increasingly frequent and intense, require improved subseasonal predictions for early warnings. Soil moisture (SM) has been suggested to play an important role in subseasonal HW predictions. However, few studies have focused on the response of subseasonal HW predictions to SM spatial variations, especially the vertical structure of SM. In this study, the conditional nonlinear optimal perturbation (CNOP) method is employed to identify the three‐dimensional (3D)‐structured initial SM errors that maximize the uncertainty in subseasonal HW onset predictions over the middle and lower reaches of the Yangtze River (MLYR) region. The numerical results reveal that the dominant CNOP‐type SM error pattern manifests as mainly negative anomalies and colocated with variations in depth. These 3D‐structured errors amplify HW intensity and advance HW onsets by reducing latent heat and amplifying sensible heat, primarily through vegetation‐related processes. Moreover, the previously underrecognized reduction of net long‐wave radiation owing to SM errors via the Stefan–Boltzmann law is emphasized here. Notably, 3D‐structured CNOP‐type errors have significantly greater impacts on the uncertainties in HW onset predictions than spatial random errors do, emphasizing the key role of the error structure in prediction uncertainty. Further experiments revealed that SM errors in deep soil layers are important in terms of the uncertainties in HW predictions and nonlinear synergistic effects exist among CNOP‐type errors in different soil layers. This study highlights the key role of the 3D structure of SM errors in the uncertainties of HW onset predictions and emphasizes the importance of land–atmosphere interactions in improving HW onset prediction ability.