Simplified mechanistic model for estimating leaf wetness

Abstract

This study aimed to develop a simplified mechanistic leaf wetness estimation model (SMLW) to support risk assessment of insect pests and diseases in irrigated specialty crops grown in the state of Washington, and around globe. Employing the energy balance and water budget approach, two variants of SMLW models (M1: SMLW and M2: SMLW_DPD) were developed using historical data from ten randomly selected automated weather stations in the Washington State University — AgWeatherNet ecosystem. Both variants simulate dewfall, rainfall interception, and evaporation processes; however, SMLW_DPD incorporates dew point depression (DPD) as an additional constraint. Both models enabled estimation of leaf wetness (LW, mm) and leaf wetness duration (LWD, h, defined as the duration when LW > 0). The model input parameters were optimized through the Bayesian method and Morris's sensitivity index. For comparison, LWD was also estimated following existing empirical approaches based on constant DPD (M3) and relative humidity (M4). The LWD estimates from all four models were finally evaluated against actual historical leaf wetness sensor data. Results indicated that SMLW and SMLW_DPD outperform M3 and M4 in estimating daily LWD. With relatively higher precision and recall, SMLW_DPD exhibited higher coefficient of determination (0.84) with root mean squared (RMSE) and absolute error (MAE) of 1.13 h and 0.34 h, respectively. Whereas, both M3 and M4 had MAE of 10.16 h and 7.03 h, respectively. Overall, SMLW_DPD model could be a viable option to reliably estimate leaf wetness using typical weather variables and reducing reliance on intricate inputs such as net radiation and leaf area index. Tied with other weather variables like degree days, LWD estimated using SMLW_DPD can be an effective decision support for growers in determining optimal timing and frequency of sprays to manage insect pests and disease pressure.