Abstract:
In Iran, applying geostatistics to regional analysis is said to be in its early stages. The
fundamental principle of this technique emphasizes the interpolation of hydrological
variables in physiographical, instead of geographical, spaces. This paper deals with the
adaptation, application, and comparison of two regional analysis methods based on
geostatistics. In this study, data from 38 gauging stations located in the north of Iran were
used to investigate the performance of geostatistical methods in two physiographical
spaces. Two multivariate analysis methods, namely, Canonical Correlation Ana
lysis
(CCA) and Principal Components Analysis (PCA), were used to identify physiographical
spaces. Gaussian and exponential models were selected as the best theoretical variogram
models in CCA and PCA spaces, respectively. Ordinary and simple kriging geost
atistical
estimators were also used for regional estimations in both physiographical spaces. Using
the interpolation methods in CCA and PCA spaces, regional flood estimations were made
for different return periods (10, 20, 50, and 100 years). Finally, perf
ormance of both
models was studied using five statistical indices. The results showed that both methods
had similar and satisfactory performance; however, regional estimations in CCA had
higher accuracy and less uncertainty than those in PCA
-
space. Further
more, the results
indicated that the ordinary kriging method had better performance than the simple
kriging method in both spaces and the best interpolation efficiency was observed in the
CCA space.
Keywords:
Interpolation, Kriging, Physiographical space,
Principle Component Analysis
(PCA).
