![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Geostatistical analysis of groundwater level using Euclidean and non-Euclidean distance metrics and variable variogram fitting criteria |
VerfasserIn |
Panagiota G. Theodoridou, George P. Karatzas, Emmanouil A. Varouchakis, Gerald A. Corzo Perez |
Konferenz |
EGU General Assembly 2015
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 17 (2015) |
Datensatznummer |
250114552
|
Publikation (Nr.) |
EGU/EGU2015-15340.pdf |
|
|
|
Zusammenfassung |
Groundwater level is an important information in hydrological modelling. Geostatistical
methods are often employed to map the free surface of an aquifer. In geostatistical analysis
using Kriging techniques the selection of the optimal variogram model is very important for
the optimal method performance. This work compares three different criteria, the least
squares sum method, the Akaike Information Criterion and the Cressie’s Indicator, to assess
the theoretical variogram that fits to the experimental one and investigates the impact on the
prediction results. Moreover, five different distance functions (Euclidean, Minkowski,
Manhattan, Canberra, and Bray-Curtis) are applied to calculate the distance between
observations that affects both the variogram calculation and the Kriging estimator. Cross
validation analysis in terms of Ordinary Kriging is applied by using sequentially a different
distance metric and the above three variogram fitting criteria. The spatial dependence of the
observations in the tested dataset is studied by fitting classical variogram models
and the Matérn model. The proposed comparison analysis performed for a data
set of two hundred fifty hydraulic head measurements distributed over an alluvial
aquifer that covers an area of 210 km2. The study area is located in the Prefecture
of Drama, which belongs to the Water District of East Macedonia (Greece). This
area was selected in terms of hydro-geological data availability and geological
homogeneity. The analysis showed that a combination of the Akaike information
Criterion for the variogram fitting assessment and the Brays-Curtis distance metric
provided the most accurate cross-validation results. The Power-law variogram model
provided the best fit to the experimental data. The aforementioned approach for the
specific dataset in terms of the Ordinary Kriging method improves the prediction
efficiency in comparison to the classical Euclidean distance metric. Therefore, maps
of the spatial variability of the hydraulic head and of prediction uncertainty were
constructed with the different approaches to indicate the prediction differences. |
|
|
|
|
|