The Research Of Estimation Model With Fewer Climate Factors On Reference Crop Evapotranspiration

Irrigation areas have a vital position in the national economy and social development of our country. High-levels of irrigation water management plays an important part to improve the yield of irrigated agriculture and fully use of irrigation projects benefit. The accurate estimation and universal analysis of reference crop evapotranspiration(ET0) via a few numbers of core meteorological factors, could formulate the crop irrigation system reasonable in order to achieve the goal of high-yield crops and water conservation. It has a very potential practical significance for Chinese agricultural water rational planning and effective management of irrigation water.Based on the National Science and Technology Support Program(2012BAD08B01), longtime meteorological data from some representative stations in Shaanxi province would be taken as an example. Then using the P-M formula to calculated ET0 values, choosing the appropriate meteorological combination with correlation analysis, and establishing ET0 estimation model via artificial neural network technology. Also this model should be certificated and made universal analysis in other stations. Finally this paper proposed a double-factor(the average temperature Tmean and the average relative humidity RH) estimation model in a seasonal scale, which could be promoted into large-scale irrigation areas based on the simple practicality and robust reliability in real-time accurate ET0 estimation. The main results and conclusions during the stages of model establishing, analyzing and popularizing are as follows:(1) Obvious seasonal variations were found via a simple descriptive statistics from each meteorological data in stations and ET0 values calculated by P-M method, and the correlation coefficients of temperature and humidity between ET0 were more significant than other climate factors. In addition, three temperature items existed a strong coupling. Then the multiple linear regression model was introduced to establish ET0 estimation model from time scales and spatial scales separately, in order to provide a sufficient theoretical basis of determining the seasonal scale and selecting the suitable meteorological factors.(2) Artificial neural network technology was introduced to establish seasonal ET0 estimation model, using temperature and humidity as input variables. The statistical error analysis results showed that whether the estimation models based on four factors or two factors, the estimation accuracy had significant differences in different seasons, instead of how much the hidden layer nodes selected. It provides a sufficient scientific basis to establish the estimation model with few amount of hidden nodes in seasonal scale.(3) The above concise estimation models were certificated in Xi’an, Hanzhong, Yan’an stations respectively, in order to strengthen the applicability of these models. Comparing the estimation results between four-factor model and the two-factor model, finding that the estimation performance got slightly reduced but the model structure parameters halved, with the number of input factors halved, which was benefit for the model to promote. Anyway, the double-factor model shows more potential value than the four-factor model in actual application.(4) The adjacent universal analysis of the four-factor estimation model was started by the data in other three stations(Baoji, Ankang and Yulin). Results showed that the average absolute error is less than 0.20 mm in autumn and winter; the average relative error is within 10% during spring and summer; with high uniform accuracy and stability for each corresponding reference stations and adjacent stations. These conclusions indicated that the above four-factor model has a strong practical value in actual irrigated water demand estimation.(5) Then the more general universal analysis of the double-factor estimation model, which established in Xi’an station, was developed by the data in all five stations, and the estimation results maintained a very strong consistency with the results by the four-factor model. Furthermore, the double-factor estimation performances during the universal analysis showed more stable and reliable than the four-factor’s, indicating that the above double-factor estimation model is more suitable for ET0 real-time estimation into a large scale of irrigation areas.(6) Finally, the secondary correction process was carried out against the original universal estimation results, which came from short-time meteorological data in Yulin station. After the correction processing by linear fitting, some estimation results, which were not satisfactory during some seasons, had been improved significantly. This process could promote the estimation accuracy and universal applicability in some special districts effectively, and strengthen the popularizing of the model ability significantly.