These models effectively capture the yield response of rainfed and irrigated maize to temperature as those response curves in Fig x while controlling the precipitation effect. The two models after training have good performance as each explains about xx% and xx% of yield variations of rainfed and irrigated maize from 200x to 201x, respectively.
Therefore, the cooling effect on yield can be quantified as Eq3, the difference in yields predicted by the rainfed maize model with the irrigated maize LST (f_rainfed(LST_irrigated)) and that with the rainfed maize LST (f_rainfed(LST_rainfed)). It reflects the yield increase in rainfed maize due to hypothetical cooling as irrigated maize. Similarly, the water effect on yield can be quantified as Eq4, the difference in yields predicted by the irrigated maize model with irrigated LST (f_irrigated(LST_irrigated, ...)) and rainfed maize model with irrigated LST. It reflects the yield increase in rainfed maize due to added water as irrigated maize.
The cooling effect: f_rainfed(LST_irrigated, ...) - f_rainfed(LST_rainfed, ...). eq.3
The water supply effect: f_irrigated(LST_irrigated, ...) - f_rainfed(LST_irrigated, ...). eq.4
Four different yields can be obtained by the cross-prediction between two temperature variables (LST_rainfed and LST_irrigated) and two models (rainfed and irrigated model). They are f_rainfed(LST_rainfed), f_irrigated(LST_irrigated),
The water supply effect
Results