Geology Reference
In-Depth Information
original signals. The comparative case study of this kind is very
first in the
hydrological modelling literature and the results show that the wavelet hybrid
models are a useful tool in solving a speci
c problem in solar radiation modelling.
The study has used the GT approach to evaluate the best and effective wavelet
subseries for modelling through extensive modelling. The evaluation has show that
the WinGamma package failed in some cases of combination of unscaled inputs.
The GT evaluation results were compared with cross-correlation, static values in
“
and extensive modelling with LLR and ANN model. The
positive results have con
no input scenario
”
rmed the effectiveness of GT in identifying the best
subseries. However, the comprehensive controlled experiments helped to reveal
some pitfalls of Gamma Test. In some cases the modelled overall MSE values were
better (smaller) than that of the Gamma static value. This
finding was contradicting
with the claimed feature of GT to give the best possible MSE for an unseen data
with any smooth model. The analysis has shown that the model performance could
be better than that of GT and which purely depends on how the modellers tuning
the model for a particular case study and data set. Another negative point associated
with GT revealed from the comprehensive modelling is associated with suggestion
of best input combination in terms of least value of gamma static. It has observed
that the best input combination is not giving least modelled error in comparison to
other modelling scenarios, when performing controlled experiments [
2
]. But the
error values of GT suggested model inputs are comparable to that of best modelled
input structure. Another interesting point observed is that comprehensive modelling
with LLR and ANN models identi
es different input structures as the best com-
bination for same training data length. However, the overall performance shows
that GT approach has a scienti
c rigor (because of good comparison results with
entropy, BIC, AIC and traditional approach) even in above pitfalls identi
ed
through controlled experiments.
References
1. AbdulAzeez MA (2011) Arti
cial neural network estimation of global solar radiation using
meteorological parameters in Gusau, Nigeria. Arch Appl Sci Res 3(2):586
-
595
2. Abrahart RJ, Heppenstall AJ, See LM (2007) Timing error correction procedure applied to
neural network rainfall-runoff modelling. Hydrol Sci J 52(3):414
-
431
3. Abrahart RJ, See LM (2007) Neural network modelling of non-linear hydrological
4. Abrahart RJ, See L, Kneale PE (2001) Investigating the role of saliency analysis with a neural
network rainfall-runoff model. Comput Geosci 27:921
928
5. Abrahart RJ, White S (2000) Modelling sediment transfer in malawi: comparing
backpropagation neural network solutions against a multiple linear regression benchmark
using small data sets. Phys Chem Earth B 26(1):19
-
24
6. Ahmadi A, Han D, Karamouz M, Remesan R et al (2009) Input data selection for solar
radiation estimation. Hydrol Process 23(19):2754
-
2764
7. Ahmed EA, Adam ME (2013) Estimate of global solar radiation by using arti
cial neural
network in Qena, Upper Egypt. J Clean Energy Technol 1(2):148
-
150
-
