Probabilistic evaluation of climate change impacts on paddy rice productivity in Japan

Projecting the impacts of climate change includes various uncertainties from physical, biophysical, and socioeconomic processes. Providing a more comprehensive impact projection that better represents the uncertainties is a priority research issue. We used an ensemble-based projection approach that accounts for the uncertainties in climate projections associated with general circulation models (GCMs) and biophysical and empirical parameter values in a crop model. We applied the approach to address the paddy rice yield change in Japan in the 2050s (2046–2065) and 2090s (2081–2100) relative to the 1990s (1981–2000). Seventeen climate projections, nine (eight) climate projections performed by seven (six) GCMs conditional on the Special Report on Emission Scenarios (SRES) A1B (A2), were included in this projection. In addition, 50 sets of biophysical and empirical parameter values of a large-scale process-based crop model for irrigated paddy rice were included to represent the uncertainties of crop parameter values. The planting windows, cultivation practices, and crop cultivars in the future were assumed to be the same as the level in the baseline period (1990s). The resulting probability density functions conditioned on SRES A1B and A2 indicate projected median yield changes of + 17.2% and + 26.9% in Hokkaido, the northern part of Japan, in the 2050s and 2090s with 90% probability intervals of (− 5.2%, + 40.3%) and (+ 6.3%, + 51.2%), relative to the 1990s mean yield, respectively. The corresponding values in Aichi, on the Pacific side of Western Japan, are 2.2% and − 0.8%, with 90% probability intervals of (− 15.0%, + 14.9%) and (− 33.4%, + 17.9%), respectively. We also provided geographical maps of the probability that the future 20-year mean yield will decrease and that the future standard deviation of yield for 20 years will increase. Finally, we investigated the relative contributions of the climate projection and crop parameter values to the uncertainty in projecting yield change in the 2090s. The choice of GCM yielded a relatively larger spread of projected yield changes than that of the other factors. The choice of crop parameter values could be more important than that of GCM in a specific prefecture.