Geology Reference
In-Depth Information
In the above section, we have described the response of different models on
evaporation data of the Chahnimeh reservoirs region, Iran. It was found that the
predictions are better in terms of numerical values when we used a complex
modeling structure. It is necessary to evaluate critically the utility of different
models in evaporation studies in terms of different attributes such as model error,
model sensitivity, and model complexity. The classi
cation of models in terms of
model utility is performed for this case study and is elaborated upon in the next
section.
7.7 Evaporation Data Model Selection in Terms of Overall
Model Utility
The evaporation data modeling was performed using models such as LLR, ANNs
(with different algorithms), ANFIS, SVM (
-SVR with linear kernel), and other
wavelet-based hybrid models such as NW, W-ANFIS, and W-SVM. In order to
สต
nd
the best utility model among these models, a model error
model
uncertainty procedure is adopted as explained in Chap. 2 . The best and most useful
data-based model for evaporation modeling in this study area was identi
-
model sensitivity
-
ed using a
new statistic called overall model utility (U). This statistic has been adopted to
measure the
of the model with respect to modeling uncertainty (assuming
that model uncertainty is connected to its sensitivity and error) and model com-
plexity (as a function of modeling time). As the model complexity is not directly
measureable,
'
utility
'
this quantity is connected to the training time after
fixing an
Search WWH ::




Custom Search