Comparison of the performance of leaf wetness duration models for rainfed jujube (Ziziphus jujuba Mill.) plantations in the loess hilly region of China using machine learning

Abstract

AbstractLeaf wetness duration (LWD) affects the ability of tree canopy to regulate eco‐hydrological processes. Jujube (Ziziphus jujuba Mill.) is one of the main plants with significant impact on water cycle process and flux characteristics in the Loess Plateau region of China. However, little is known about LWD in rainfed jujube plantations in the loess hilly region of China. For the causes of leaf wetness in rainfed jujube plantations, dew‐only, rain and the combined scenario were investigated in terms of LWD and meteorological variables for the 2012, 2013, 2017 and 2018 jujube growing seasons. The results were as follows: The leaf wetness events occurred frequently, and the days of leaf wetness driven by dew‐only and rain accounted for 88.3% of the total number of days in a regular jujube growing season. LWD driven by dew‐only was significantly lower than that driven by rain (p < 0.05). The correlation between LWD and meteorological factors varied with different scenarios. In dew‐only scenario, the four calibrated models (RH, DPD, CART and NN) accurately predicted LWD. In rain scenario, only the calibrated NN model performed well. In the combined scenario, the calibrated RH, CART and NN empirical models performed nearly equally; all well estimated LWD. On the basis of accuracy, practicability and variable count, the calibrated RH, NN and RH models for dew‐only, rain and the combined scenario can reliably estimate LWD in rainfed jujube plantations in the loess hilly region of China.