Estimation of Evapotranspiration Rate Using Neural Network with Plant Motion
Access this Article
Search this Article
Two neural network (NN) models were developed to estimate evapotranspiration (ET) rate of New Guinea Impatiens (<I>Impatiens New Guinea Hibrid</I>). Inputs of one NN model were canopy temperature, environmental factors (air temperature, relative humidity, radiation), and the plant motion (optional). The plant motion was calculated using the top projected canopy area. The mechanistic model was used in order to provide a baseline with which to compare performances of the NN models. In non-drought stress condition, root mean square error (RMSE) between estimated and measured ET rate of the NN model with the plant motion (NNP), the NN model without plant motion (NN), and the mechanistic model were 21.80%, 22.04%, and 29.94%, respectively. In drought stress condition, RMSE of the NNP, the NN, and the mechanistic model were 39.02%, 49.81%, and 72.09%, respectively. The plant motion could contribute the better performance when the plants were in drought stress condition. The NN model could estimate the ET rate without parameters used in the mechanistic model.
- Environment Control in Biology
Environment Control in Biology 46(1), 13-19, 2008-03-30
Japanese Society of Agricultural, Biological and Environmental Engineers and Scientists