Data Requirement for Evapotranspiration Estimation

Abstract

Various approaches based on climatological data (CD) have been used to estimate evapotranspiration (ET). A multiple regression with ordinary least square analysis (OLS) has been commonly used to assess the relatively important CD used as predictors. An optimum ridge regression analysis (OPT) is introduced to overcome the problem caused by the OLS analysis when multicollinearity exists among the CD. An ideal method used for ET estimation should be chosen based as minimally as possible on the input of the CD variables without affecting the accuracy of estimation, so that not only can the multicollinearity problem be eliminated but also the CD availability can be improved considerably. Particularly, a model based on two variables of air temperature and solar radiation can provide a quite satisfactory estimation of the ET in southern Florida for irrigation requirement prediction. The technique developed can be used to improve the assessment of the relative importance of the CD used not only for developing a better ET equation but also for improving future programs of the weather data collection.