Change of the wintertime multidecadal land precipitation variability in the mid‐1970s in the observation and CMIP6 simulations

The wintertime multidecadal land precipitation (WMLP) variability is obtained using the ensemble empirical mode decomposition, and then the leading spatial pattern and temporal evolution of the WMLP in the northern hemisphere during the period of 1891–2014 are investigated. The time series shows a pronounced enhancement of the variability since the mid‐1970s, coincident with the 1976/1977 climate transition, with the variance increasing from 0.33 in the former period (P1) to 2.57 in the latter period (P2). The spatial structure in P2 resembles that in the entire period. The precipitation anomaly is positive in the North America between 30° and 50°N, South America and southern Europe; while it is negative in the southern North America, majority of Russia, Arabian Peninsula, Iran and South China. However, the spatial pattern in P1 differs from the entire period. An Interdecadal Pacific Oscillation (IPO)‐like sea surface temperature (SST) variability in cold phase and an Atlantic Multidecadal Oscillation (AMO)‐like SST variability in warm phase are associated with the leading mode of WMLP in P2 and the entire period. Yet, the relationship of IPO and AMO on the WMLP is barely found in P1. The in‐phase change of IPO and AMO in P1 may contribute to the smaller amplitude of WMLP variability in P1. In the end, the coupled models from CMIP6 historical runs are evaluated. The good‐models average is superior to any single good model in capturing the observed spatial features of WMLP in both P1 and P2. The poor‐models average performs better than any single poor model. Both good‐ and poor‐models averages are able to reproduce the IPO‐ and AMO‐like SST variabilities in P2. The variability of the wintertime multidecadal land precipitation (WMLP) in northern hemisphere is remarkably enhanced since the mid‐1970s. Interdecadal Pacific Oscillation (IPO) and Atlantic Multidecadal Oscillation (AMO) are largely in (out of) phase before (after) mid‐1970 s, inducing an opposite effect (a joint effect) on the dominant mode of WMLP. Both good‐ and poor‐models averages derived from CMIP6 simulations are able to better reproduce the IPO‐ and AMO‐like SST patterns and WMLP variability during P2 than P1.