Geology Reference
In-Depth Information
Fig. 7.10 Daily reference evapotranspiration modeled by ANN model during 1995
-
1999
(training and validation) in Brue catchment (the gaps are missing/erroneous input data)
relative humidity (RH), and wind speed. In this attempt, we did not apply any input
selection procedures. Instead, data from the
1997) is used to
train the ANN model and the remaining data is used for testing. The ANN model
used is multi-layer perceptron (MLP) with the presence of one hidden layer whose
computation nodes are correspondingly called hidden neurons or hidden units. In
this hourly modeling for the Beas basin, we have used nine neurons, which were
selected systematically after several trial and error modelings. The Levenberg
rst 3 years (1995
-
-
Marquardt technique was used to model the ANN, as it is considered to be a more
powerful and faster training algorithm in comparison to conventional gradient
descent technique. The FAO56-PM was used as the output data. The Slope, R2,
RMSD, and SEE statistics of the ANN model are shown in Table
7.3
during the
training and validation phase. The hidden layer neurons from nine started showing
relatively low mean squared errors (MSEs) during the modeling process. Therefore,
we chose nine as the optimum number of hidden neuron with least complexity. The
comparison has shown that the ANN model produced a better performance than the
CIMIS-PM and COPAIS approach in the study area. The ET
0
estimates produced
by the MLP type ANN are shown in the Fig.
7.10
.
-
7.4.2.2 Daily ET
0
Comparison
In daily comparison studies we focused on the 1995 data at the Brue catchment as
there is data discontinuity in other years. Figure
7.11
shows daily reference evapo-
transpiration comparison of the standardized ASCE Penman
Monteith equation
(ASCE-PMET
0
), CIMIS-PM equation, and daily Copais Approach equation with the
FAO56 Penman equation (FAO56-PM ET
0
) for the Brue catchment.
This comparison of the estimated daily ET
0
demonstrates the reliability of the
proposed Copais model in daily evapotranspiration estimation. The plot of the daily
-
