[Impacts of future climate change on spring phenology stages of rubber tree in Hainan, China].

To explore the impacts of future climate change on spring phenology stages (first leaf storey expansion stage, spring flowering stage) of rubber tree in Hainan Island, we established a rubber tree spring phenology simulation model based on the crop clock model and developed a computer software RubberSP. The model simulation accuracy was examined with experimental observed phenology data. Five global climate models (GCMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5) were integrated using Bayesian Model averaging method (BMA) to predict the impacts of climate change on the spring phenology of rubber tree in 2020-2099 (relative to 1986-2017) under climate scenarios of RCP2.6, RCP4.5 and RCP8.5, respectively. The results showed that the RubberSP model had good simulation accuracy, with the determination coefficient (R2) values ranging between 0.73-0.87, the root mean square error (RMSE) ranging from 3.26 to 4.15 d, and the normalized root mean square error (NRMSE) of 3.4%-7.4% between measured and simulated phenology stages. The uncertainty of a single GCM could be avoided by BMA method, which could better reflect the change trend of temperature. Temperature of Hainan Island in the end of 21 century, under the scenarios of RCP2.6, RCP4.5 and RCP8.5, would increase by more than 0.3, 1.0 and 2.5 ℃ compared with the baseline, respectively. The spring phenology stages would appear earlier and yield would increase in the future climate scenario. The time isoline of spring phenology stages would move forward to northwest, which indicated that most suitable area for rubber tree plantation in Hainan Island would expand to the northwest. The spatial difference of the first leaf storey expansion stage would be more evident, but not for spring flowering stage. The amplitude of rubber tree spring phenology variations was closely related to the increases of temperature under different RCP scenarios, with the most apparent change under RCP8.5 scenario and most mild change under RCP2.6 scenario.