This work was conducted for predicting regional evapotranspiration and biomass density by artificial neural network (ANN) method under varying climate conditions. A feed forward neural network with a set of landsat data including temperature and short wave radiation data was trained from the North part of the Jordan River Basin in Israel to predict actual evapotranspiration. The fitted results manifested that comparative analysis of the backpropagation (BP) and radial basis function (RBF) networks provided very similar data. BP would spend much more training time than RBF, which could get convergences rapidly. Nevertheless, the RBF still had some problems in the prediction, as it was not prompt in cases dealing with highdimensional input spaces, especially when the original data set contained some invalid variables. Therefore, principle component analysis was used to correct the input data. Above all, the BP and RBF can be used as perfect tools for taking the place of other mathematical models to predict the evapotranspiration (ET) and weighted difference vegetation index (WDVI). This work approved that ANN method can effectively predict regional evapotranspiration and distribution of vegetation under realistic conditions.
|Number of pages||4|
|Journal||Journal of Food, Agriculture and Environment|
|State||Published - 1 Jan 2012|
- Artificial neural network
- Radial basis function