Reproducibility and Trends of Extreme Climate Indices in Japan: Insights From Dynamical JRA‐55 Downscaling

ABSTRACT Dynamical downscaling (DDS) datasets play a crucial role in understanding regional climate patterns and extreme weather events. This study evaluates the reproducibility of extreme climate indices in Japan using two DDS datasets based on the JRA‐55 reanalysis for the period 1979–2012. A total of 48 extreme climate indices were analysed to assess biases, interannual variability, and trends in precipitation and temperature by comparing the DDS datasets with AMeDAS observations, a high‐resolution automated meteorological observation network in Japan. Both DDS datasets reasonably captured interannual variability, with correlation coefficients exceeding 0.6 for many indices. However, systematic biases and underestimations of trend magnitudes were observed. For precipitation indices, DS‐run (DDS using the Non‐Hydrostatic Model, NHM) generally exhibited a consistent tendency toward negative biases across most areas, while RC‐run (DDS using the Non‐Hydrostatic Regional Climate Model, NHRCM) showed relatively smaller biases in some regions but larger negative biases in the Southwest Islands (area 7). For temperature indices, both runs successfully reproduced interannual variability. However, the RC‐run showed pronounced negative biases in TX‐related indices, particularly TXm and TXn, while the DS‐run exhibited slightly larger biases for TXx. Positive biases were more common in TN‐related indices, especially in area 1. Trend analyses revealed regionally varying patterns. Both DDS runs captured the direction of observed trends for most indices across all regions, with high agreement in trend sign. However, agreement in trend magnitude and statistical significance varied depending on index type and region. Although each DDS run exhibited distinct characteristics, both shared common biases, highlighting the need for further improvements in model performance. These findings suggest the importance of careful model evaluation when using DDS outputs for climate impact assessments and offer useful insights for future model improvement and the development of downscaling strategies in Japan.