Research And Application Of Chaotic Time Serie Prediction Method On Reference Crop Evapotranspiration

In recent years, with the impact of the climate, environmental damage and the process ofindustrialization increasing, the characteristics of the evaporation trend in each region ischanging. According to the climate warming, the changes of reference crop evapotranspirationtrend under human activities, the analysis of factors and the simulation and prediction ofclimate change, a scientific basis is provided for exploring the regularity of the globalagricultural water using, studying the influence of the agricultural water using under the effectof climate changes, the integrated management and planning of agricultural irrigation anddraining technology, and setting down the mode of the sustainable regional agriculturaleconomic development.This paper selected meteorological data of six sites in Shaanxi Province, andsystematically analyzed the trend of reference crop evapotranspiration and its influentialfactors. Then, in case of shorting for information, comparisons of different calculationformulas’ practicality were done. Also, it tried to combine chaotic nonlinear theory, which hada wider research field in recent years, and analyzed the phase space reconstruction ofreference crop evapotranspiration. At last, the prediction of reference crop evapotranspirationbased on neural network system was done. The results and conclusions as the following:（1）Data of six sites was selected for modeling and ET₀calculations. Then, this paperanalyzed the trends of factors’ changing since the1950s, including temperature, humidity,solar radiation, wind speed and average annual ET₀. After comparing the results of a varietyof trends, different degrees of changes in each station were found, where the temperature wassubstantially increased.（2）Many meteorological factors needed to be considered in the ET₀calculation process,so another three formulas were chosen, which had less variables, to make a comparison withP-M formula about the calculating results. The Hargreaves formula, Markkink formula andPriestley-Taylor formula were used to calculate the yearly and the monthly ET₀value thatcame from the selected six sites, and compared with the P-M formula calculate results.Through the comparative analysis and error analysis, that the results of Priestley-Taylor andHargreaves methods were found closer to the results of P-M formula, and the relative errorwas smaller than the P-M method. The error of Markkink method was a little larger than theother methods, and the consistency of it was poor. The three methods had stronger applicability in arid regions.（3） As many variables needed to be considered during the calculation process, themeteorological elements of the selected six sites including temperature, relative humidity,sunshine hours, wind speed and atmospheric pressure, were made principal componentanalysis. The results of analysis showed that the main influential factors of the reference cropevapotranspiration in most regions are temperature, relative humidity and sunshine hours. Inorder to estimate the reference crop evapotranspiration accurately, reference cropevapotranspiration time series prediction model based on fuzzy neural network was made.According to the monthly meteorological data of Baoji1954-2004, the main impact factors ofreference crop evapotranspiration were found under the use of principal component analysis,and4variables were found as input vector. Then the reference crop evaporation transpirationcalculated with the Penman-Monteith formula was collected as the target vector. Fuzzyneural network model was used to predict reference crop evapotranspiration through matlabprogramming. The results showed that: the absolute value of12groups of test samples’ meanrelative error was5%, at the same time, the maximum relative error was11.4%while thesmallest relative error was0.4%; the fuzzy neural network model and Penman-Monteithformula calculated results had a very high consistency.（4） According to the existing literature that the long-term evolution of the reference cropevapotranspiration had chaotic characteristic. This study attempted to use chaotic phase spacereconstruction method to reconstruct it, and established a new forecasting model. In order togain a more suitable phase space reconstruction parameters, this paper mainly applied timedelay and C-C method that related to embedding dimensionality to select two parameters, andtried to improve the C-C method for getting more accurate parameters. Then Yan’an monthlydata was used as case study, and the method of correlation dimension was applied to test it.The test results showed that the parameters selected through the improved C-C method werethe more appropriate.（5） For example of Baoji site in Guanzhong area in Shaanxi Province, the data （1955～2004）of the reference crop evapotranspiration was used, and phase space reconstructionparameters were selected through the improved C-C method. Then, phase spacereconstruction was made for the reference crop evapotranspiration, and radial Basis Function（RBF） artificial neural network model was used for predicting simulation of it. the method hascertain research value.