Leaf area index estimation for a greenhouse transpiration model using external climate conditions based on genetics algorithms, back-propagation neural networks and nonlinear autoregressive exogenous models

A precise transpiration prediction model thus becomes an important tool for greenhouse automatic irrigation management. Moreover, leaf is an organ of transpiration, and leaf area index is a basic variable to estimate this water lost, but it is still a weak spot in the crop growth estimation. In this paper, two different leaf area index models are established and compared with the evolution of the real crop determined with an electronic planimeter: (1) Considering the temperature and photosynthetically active radiation (PAR) as the main impact factors over crop growth, a TEP-LAI model based on product of thermal effectiveness and PAR is built to estimate the leaf area index dynamics; and (2) TOM-LAI model based on a tomato growth model is also used to estimate the leaf area index as an explicit function of the number of leaves and vines. Finally, the results of both simulation models (TEP-LAI and TOM-LAI) are compared with the measured values. Moreover, a crop transpiration model is established using the empirical data sampled in a multi-span greenhouse in Almeria (Spain). In this greenhouse, a microlysimeter (two different weight scales) was used to obtain the transpiration and the drainage values. Thus, the data collected is used to obtain a model of the estimated water lost by transpiration, that it is based on Back Propagation-Neural Network was optimized using genetic algorithm and Nonlinear Auto-regressive model with Exogenous Inputs model. Once described the different models, the estimated values of leaf area index are compared satisfactorily with the measured ones. TEP-LAI is the model chosen to be introduced as input of the final transpiration model. As expected, the transpiration estimation with inside conditions generates better results, but the outside climate based model shows that it could be used as an irrigation predictor with data from cheaper outside meteorological stations.