The Vegetation Canopy Interception Model Of Multiple Regression Analysis

In the21st century, with the high-speed development of information technology,data piled up, and it’s more difficult to find the knowledge in the huge data. So, peoplepaid more and more attention to Data Mining, which has been applied extensively inmany areas, including hydrology. Study of canopy interception model is always a hotspot of hydrology and many researches has been made on this problem. Many canopyinterception models have been built, but all of them have defects. Some of them lack ofportability and hydrologic meaning, some are hard to work out and some areover-reliance on assumptions. A model, which is portable, with hydrologic meaning,easy to work out and has little reliance on assumptions, has never been built.This thesis makes some study on canopy interception model with the method ofmultiple regression analysis. At first, the influence factors are summed up to make aalternate independent variable set X. Then,3canopy interception models have beenbuilt with the method of multiple regression analysis.1) After the analysis of physical meaning, redundant independent variables havebeen select out from set X. Thus we get a new X with only8members. Then, the firstinterception model M1has been built on set X with the method multiple linearregression analysis. However, in the significance test, M1’s significance factor is0.32,much bigger than the significance level0.05.2) From the view that the reason why M1failed is because linear relationship exists among the independent variables in set X,stepwise regression and stepwise elimination have been used to improve M1to a newmodel M2. M2has a significance factor of0.001, so it’s linear. M2plays good inapplication test. However it is lack of physical meaning.3) After a reanalysis, the17influence factors have been included in set X again. Redundant independent variableswere selected out according to the relationship among independent variables. After theselecting, there were only6members remained in X. Then, the relationship betweencanopy interception and single member of X is analysed, and a multiple nonlinearregression model M3were built. Experimental results show that M3is more accuratethan Zhang’s model on short-time forecasting, and it has an acceptable accuracy on long-time forecasting, which is close to Gash model.