Ensemble‐based singular value decomposition analysis to clarify the causes of heavy rainfall

Abstract

The complex relationships between rainfall amounts and their causes require further clarification through analytical research. This study utilizes ensemble-based singular value decomposition (ESVD) analysis that decomposes the ensemble-based cross-covariance matrix between datasets related to atmospheric states and hydrometeors. ESVD analysis is applied to the "Heavy Rain Event of July 2018 in Japan." The initial states of 301-member ensemble forecasts are created using a local ensemble transform Kalman filter, and the ensemble forecasts are obtained by the regional nonhydrostatic model (2 km horizontal grid interval). The ESVD analysis results indicate that the heterogeneous correlation maps of the first mode (maximum squared singular value) and second mode exhibit high correlations between the synoptic-scale atmospheric states (the location of the stationary front and the baroclinically enhanced updraft) and rainfall characteristics. The results obtained using the sixth mode show that local water vapor fluxes in the lower layer are correlated with mesoscale rainfall characteristics. Therefore, ESVD analysis can be used to clarify multiple, independent relationships between multiscale atmospheric states and heavy rainfall.