Estimating actual ET for wheat from climatic and IRS images data using neural computing technique in the arid zone of Central Iran (Marvast Region)

This paper examines the potential of artificial neural networks (ANN) in estimating the actual crop evapo-transpiration (ET) from limited climatic and 4 bands of LISS3 from IRS images data. The study employed feed-forward (new ff) type ANN for computing the daily values of ET for winter wheat crop. From 2 transfer functions as, Levenbery and Marqwardt each using varied input combinations of climatic variables as temporal data and NDVI as temporal-spatial (TS) data at 6 step of crop growth for one year, have been trained and tested. From NDVI the studied area was classified in to separated crops and trees polygons beside of winter wheat. The model estimates are compared with combination of measured class A evaporation pan with plant coefficient (k) and measured water balance in the field. The results of the study clearly demonstrate the proficiency of the ANN method in estimating the ETo. The analyses suggest that the TS evapo-transpiration could be computed from air temperature and LISS3 bands of IRS using the ANN approach. However, the present study used the data for a limited period; therefore further studies using more data may be required to strengthen these conclusions.