An Artificial Intelligence Model to Predict Crop Water Requirement Using Weather, Soil Moisture, and Plant Health Monitoring Data

Abstract

Although the use of new tools and technology for irrigation is becoming more popular, the use of methodology such as evapotranspiration to determine the irrigation amount is relatively old and may result in over or under irrigation. Since crop water requirement is function of many parameters, a thorough study is thus required to determine more important parameters that contribute to estimating crop water requirement. These important parameters then can be used to develop an artificial intelligence model, which is expected to be predicting the crop water requirement with minimum errors. The methodology to predict the crop water requirement presented in this paper consists of the following steps: (a) collecting field and weather data, (b) classifying soil moisture data to estimate crop water requirement, (c) identifying important input variables for the crop water requirement by correlation analysis, and (d) developing artificial intelligence models to predict crop water requirement using the important parameters identified in the previous step. The methodology has been applied to analyze olive field data at the University Agricultural Laboratory (UAL) at California State University Fresno. Results show that past PSM (positive change in soil moisture) data can be used to predict future crop water requirement. Besides PSM, soil temperature (ST) and cumulative evapotranspiration (CET) show higher correlation when they are used as a single input variable for the neural network model at lags 3 days and 11 days respectively. When two input variables are used to train the network, the combination of ST and PSM; and SR (solar radiation) and ST at lags 3 days and 6 days show better performance.