Geology Reference
In-Depth Information
Fig. 7.35 NW modeled and observed evaporation data at the Chahnimeh reservoirs region:
(a) training phase as scatter plot; (b) validation data as line diagram
representation of the general NW model is shown in Chap.
4
. The daily data of air
temperature (T), wind speed (W), saturation vapor pressure de
cit (Ed), and relative
humidity (RH) were decomposed into three series of detailed coef
cients (D) and
three series of approximation (A) sub-time series. These three resolution levels of
three GT identi
ed inputs (i.e., air temperature details (DT1,
T
1, D
T
2, and D
T
3), wind
speed details (D
W
1, D
W
2, and D
W
3), saturation vapor pressure de
cit details (D
Ed
1,
D
Ed
2, and D
Ed
3), and approximate modes (A
T
3, A
W
3, and A
Ed
3). DWT is used to
decompose the input data into three wavelet decomposition levels (2
4
8). The
-
-
three detailed coef
cient series and the third approximate series of the original data
in this study are presented in Fig.
7.34
. These sub-series of the original data from
the Chahnimeh reservoirs region were used as the inputs of W-ANFIS and
W-SVMs. The technical details of SVM and ANFIS are identical to those discussed
earlier. The performance analysis of NNARX, NW, W-ANFIS, and W-SVM
models is summarized in Table
7.9
.
It can be observed from Table
7.9
that the NW model in predicting the evap-
oration data is better than all other models with superior values of statistical
parameters. The observed and predicted evaporation values of the NW model for
the training data are given in Fig.
7.35
a as a scatter plot and the corresponding
values for the validation data in Fig.
7.35
b as a line diagram. For all the models, the
evaporation is underestimated for the validation phase as indicated by the MBE
values in Table
7.9
. A sight overestimation is observed with the NW model during
Fig. 7.36 W-SVM modeled and observed evaporation data at the Chahnimeh reservoirs region:
(a) training phase as scatter plot; (b) validation data as line diagram
