Sensitivity of rainfed rice yields to historical and projected monsoon precipitation and temperature over South Asia

Abstract:

South Asian countries occupy less than 15% of the globe’s arable land but supply food to 25% of the world’s population. Improved crop varieties and inputs have dramatically increased yield. However, uncertainties still exist since yields vary strongly due to fluctuations in climate. Understanding such fluctuations is crucial for developing resilient food systems. This study focuses on how interannual variability of monsoon climate impacts rice yields in South Asia, recognizing monsoon kharif rice as the region’s main crop. We use gridded observed rainfed rice yields (1965–2005) to generate empirical statistical crop models able to capture the effects of climate. We characterize the monsoon climate with agriculturally relevant climate metrics covering water availability (precipitation, dry spells), timing of monsoon onset and withdrawal, planting date, seasonal temperature (growing degree days) and extreme events (warm days and nights). Normalized anomalies of climate and rice yields are used to generate empirical regressions that additively incorporate the effect of individual climate metrics and interactions. We use daily precipitation and temperature data from an ensemble of 21 General Circulation Models for historical climate (1950–2005) and future projections (2006–2099; RCP 4.5 and RCP 8.5). The Princeton Global Forcing product, which provides data at the same time and spatial resolution, was used as an observational reference. Once we have run the models and shown that it is able to satisfactorily simulate yield anomalies, we apply dominance analysis to quantify the relative importance of climate predictors. Statistical models help us investigate the historical association between climate variables and their importance as determinants of rice yield variability, spatial and interannual/inter-decadal variability, and uncertainties associated with climate models and observations. We also use future climate scenarios to examine the sensitivity of rice yields in response to projected climate and in relation to model uncertainty.