I would like to express my thanks for the interest shown by the discussers and for their comments on the paper (Kişi, 2009). I have tried to clarify all the points raised by them in this reply.
The discussers claim that '. . . the accuracy of estimated pan evaporation as presented by Kişi ( 2009) was unrealistically high' and compare with the studies of Keskin and Terzi (2006), Bruton et al. (2000) and Sudheer et al. (2002). Keskin and Terzi (2006) used meteorological parameters including air and water temperature, relative humidity, solar radiation, wind speed, air pressure and sunshine hours, which were logged every 2 h. They integrated two hourly data subsequently to obtain daily data because the pan evaporation values used as output in the ANN models are daily measurements obtained from the Directorate of State Hydraulic Works, Turkey (Keskin and Terzi, 2006). Their data include 490 daily observations from 1 March to 31 October in the years 2001 and 2002. They used the standard back propagation algorithm, gradient descent for the training of ANN models and 10 000 iterations. They used 1 year data (2001, only 4 months were included) in the training period and 1 year (2002) in the validation period. Sudheer et al. (2002) used four daily climatic parameters, that is, the air temperature, humidity, sunshine hours and wind speed to estimate daily pan evaporation. They used the gradient descent algorithm for the training of ANN models and 50 000 iterations. They used 1 year data (1991) in the training period and 3 years data (1990, 1992 and 1993) in the validation period. In Kişi's study (2009), the data sample consisted of 4 years (1987)(1988)(1989)(1990) of daily records. Air temperature, solar radiation, wind speed, pressure and humidity (H) were used to estimate daily pan evaporation. He used the