The fluctuations in water-table depths in agricultural fields depend on various parameters such as rainfall, evapotranspiration etc. Efforts to simulate the water-table depths in subsurface drained farmlands using artificial neural networks (ANNs) have been made earlier without much success. In this paper, these efforts are reviewed and the sensitivity of the performance of the ANNs on the type of input variables used is demonstrated. Measured data on rainfall, evapotranspiration, and midspan water-table depth were used. Input and output variables were variants of these measured data. Two sets of input/output variables were tried, one a minor variation from an earlier attempt, and the other a set of variables based on physical reasoning. The results showed that the performance of the ANNs depends on the type of input variables used, and if the input variables are chosen so as to mimic the physical dependence of the output, the ANNs perform very well.
[1]
I. Seginer,et al.
Mimicking an Expert Greenhouse Grower with a Neural-net Policy
,
1996
.
[2]
Jehng-Jung Kao,et al.
Neural net for determining DEM-based model drainage pattern
,
1996
.
[3]
R. W. McClendon,et al.
Optimal Control and Neural Networks Applied to Peanut Irrigation Management
,
1996
.
[4]
S. O. Prasher,et al.
Use of Artificial Neural Networks in Transient Drainage Design
,
1996
.
[5]
Simon Haykin,et al.
Neural Networks: A Comprehensive Foundation
,
1998
.
[6]
Chun-Chieh Yang,et al.
Applications of Artificial Neural Networks to Land Drainage Engineering
,
1996
.
[7]
Shiv O. Prasher,et al.
A NEURAL NETWORK APPROACH FOR THERMAL PROCESSING APPLICATIONS
,
1995
.