Estimating the transpiration of kiwifruit using an optimized canopy resistance model based on the synthesis of sunlit and shaded leaves

Accurate estimation of transpiration (T) in kiwifruit trees is essential for effective irrigation and water management. Canopy resistance (r c ) is crucial for estimating T, but existing models do not fully consider the unique canopy structure and microclimate variations in kiwifruit trees. This study established a r c estimation model based on a synthesis of sunlit and shaded leaves (SSL) and optimized it using Ant Colony Optimization (ACO), Grey Wolf Optimizer (GWO), and Whale Optimization Algorithm (WOA). Using the r c value inverted by the Penman-Monteith model as a standard, we compared the simulation accuracy of the SSL and Jarvis models to identify the optimal model for accurate T estimation under various data availability conditions. The results indicated significant physiological differences between sunlit and shaded leaves, with shaded leaves showing lower net photosynthetic rates and higher stomatal resistance. The optimization SSL model demonstrated improved accuracy over the Jarvis model. The simulation accuracy of the SSL model optimized by the WOA algorithm was the highest, yielding R 2 , RRMSE, and MAE of r c and T are 0.83, 0.12, 82.55 s m −1 , and 0.81, 0.09, 0.23 mm d −1 , respectively. In the Jarvis model with different restriction functions the highest accuracy for r c and T, achieved after optimizing by ACO algorithm, yielded R 2 , RRMSE, and MAE of 0.71, 0.33, 305.94 s m −1 , and 0.72, 0.23, 0.65 mm d −1 , respectively. Therefore, the SSL model can more accurately estimate the r c and T, and it provides a valuable way for scientific water use and precise irrigation in kiwifruit orchards. • Establish a sunlit and shaded leaf canopy resistance (SSL) model of kiwifruit. • The accuracy of the SSL model was significantly higher than that of the Jarvis model. • The intelligent optimization algorithms could improve the accuracy more than the least square method. • The SSL model optimized by the Whale Optimization Algorithm has the highest simulation accuracy.