Modelling rice yield with temperature optima of rice productivity derived from satellite NIRv in tropical monsoon area

Abstract

Tropical rice production is at risk from rising temperature. Understanding regional and seasonal heterogeneity of optimum temperatures for rice production is important for model simulation to predict rice yield change under climate change. However, studies or tools for widely observation of crop responses to temperature over broad spatial scales with long time spans are limited. In this study, we detected optimum temperature range for rice gross primary production (ToptGPP) in the lower Gangetic plains and delta region using the near-infrared reflectance of vegetation (NIRV), which is a new photosynthetic proxy, to improve ORYZA model performance in high-temperature season and assessed how tropical rice would respond to temperature increase in the study area. According to satellite observations of NIRV from 2001 to 2015, current ambient air temperature has exceeded the mean ToptGPP of Boro rice (24.8 ± 1.8 °C) and Aman rice (26.7 ± 1.2 °C) in the lower Gangetic plains and delta region, suggesting a downtrend of rice production under future warming. The detection results show that rice has lower ToptGPP in the regions with more drought stress and lower background temperature under water-limited conditions. Furthermore, the model modified by NIRv-ToptGPP shows better performance in potential yields, especially in high-temperature seasons on the region scale. Without CO2 fertilization effect, each degree-Celsius increase is expected to reduce rice potential yields by 4.9 ± 1.6% based on the default ToptGPP range in ORYZA model and by 7.0 ± 1.2% based on the detected NIRv-ToptGPP range in the study area. This study implies that global grid-based model simulation may underestimate sensitivity of tropical rice yield to temperature rise due to the neglect of regional and seasonal heterogeneity of ToptGPP. NIRV makes it possible to determine local optimal temperatures for crop production, and to improve grid-based modelling across various agricultural systems in different growing seasons at the regional scale.