Abstract:Crop transpiration was the main driving force of substrate water transfer. Aiming to establish a greenhouse tomato crop transpiration estimation model and prediction model based on the change of substrate water content, and make a comparative analysis.The calibrated EC5 matrix moisture content sensor was used to record the realtime change of matrix moisture content after the first irrigation and before the second irrigation. Realtime crop transpiration was measured by weighing method. The estimation model of daily transpiration per plant of tomato was established by multiple linear regression calculation of variation of substrate moisture content and volume of substrate cultivation tank. Taking the variation of substrate moisture content, air temperature, air humidity and illuminate intensity as input, the prediction model of daily transpiration per plant of tomato was established by GABP neural network algorithm. The greenhouse crop transpiration estimation model and predictive model were tested respectively with the greenhouse crop’s daily transpiration by linear regression analysis, the results showed that the prediction accuracy of the estimation model based on the variation of water content in the matrix was 0.9729 and 0.9796, respectively, in the seedling stage and florescence, and the prediction accuracy of the prediction model was 0.9915 and 0.9890, respectively. The differences between the two was not big, but the estimate model operation speed was much higher than predictionmodel of operation speed. In practical application, the estimation model had good robustness to environmental changes, and the relative error was less than 5% at seedling stage and flowering stage. The estimation model had the value of popularization and application for greenhouse irrigation management.