Adaptive irrigation method for closed cultivation in greenhouse

Abstract: At present, the research of precision irrigation are mostly concentrated on combining the Penman-Monteith(P-M) formulate and sensor data information to calculate the crop evapotranspiration, and then the irrigation operations are performed in accordance with the calculation results. But the calculation of crop evapotranspiration based on the P-M formula not only requires a lot of environmental parameters, but also is an approximated calculation for the greenhouse environment, and the results of calculation cannot accurately reflect the crop evapotranspiration. So a more practical and simple method is required for the closed cultivation in greenhouses. Based on the characteristics of closed cultivation in greenhouse, the excess irrigation water can be collected and recycled easily by drainage system, we established a relatively accurate crop evapotranspiration computational model using artificial neural network according to the difference between the amount of irrigation and drainage based on the microclimate data in greenhouse, and tested for its feasibility in the adaptive control for irrigation in greenhouse.The difference between irrigation and drainage was calculated, and then an artificial neural network model was trained based on that difference and environmental information (temperature, relative humidity and solar radiation) in greenhouse so as to establish the crop transpiration model. The crop transpiration was calculated using this model according to the current environmental information until the set of value was reached, and then, the irrigation program was executed. Then, the difference between the amount of irrigation and drainage was calculated again, and the crop transpiration model was modified according to the value. After that, the amount of crop transpiration was calculated again by the modified model and the irrigation operation was executed when the set of value was reached again, so the crop transpiration calculation model was adjusted again and again, finally the crop transpiration model would be infinitely close to the actual evapotranspiration process, and the crop irrigation amount and frequency of days were adaptively adjusted in accordance with the method described above. The research data used in this work had been obtained from greenhouses located in the Agricultural Unisversity of Athens, Greece, at November 3-13, 2014. In the greenhouses, tomato was planted in the bags of substrate, and the bags were placed on the brackets that height of 0.5 m, and the drainage system were designed in the brackets. The drainage was collected into the recycle tank through the pipe, and a water level sensor was placed in the recycle tank. The drainage was measured after each irrigation and then the actual crop water requirements could be obtained by the combination of irrigation and drainage, which was the actual crop transpiration between the two irrigation processes. In addition, greenhouses temperature, relative humidity and solar radiation were also collected. The results showed that the calculated value of the model was more accurate under the condition that the information on greenhouse environment parameter was sufficient, such as the use of environment data within 216 hours. In the experiment, the 10 days irrigation control was implemented using the calculated the value of crop transpiration model, the total amount of irrigation water was 76517.59 L, and the actual amount of transpiration was 78238.85 L, which was obtained using the difference between irrigation and drainage. The error between the theoretical value and the actual value was 2.2%, and the daily average difference between the irrigation and actual transpiration was 172.126 L, or 0.043 L/(m2·d), suggesting that the irrigation amount was consistent with actual crop evapotranspiration. So the adaptive irrigation method is reliable to meet actual crop water demand for irrigation control purpose.