Το έργο με τίτλο Ενισχυμένη ανάλυση Kriging για την κατάστση των υπογείων υδάτων στο Τυμπάκι, Κρήτη με αυτό οργανομένους χάρτες από τον/τους δημιουργό/ούς Spyropoulos Fotios διατίθεται με την άδεια Creative Commons Αναφορά Δημιουργού 4.0 Διεθνές
Βιβλιογραφική Αναφορά
Φώτιος Σπυρόπουλος, "Ενισχυμένη ανάλυση Kriging για την κατάστση των υπογείων υδάτων στο Τυμπάκι, Κρήτη με αυτό οργανομένους χάρτες", Μεταπτυχιακή Διατριβή, Σχολή Χημικών Μηχανικών και Μηχανικών Περιβάλλοντος, Πολυτεχνείο Κρήτης, Χανιά, Ελλάς, 2023
https://doi.org/10.26233/heallink.tuc.99319
The main objective of this work is to investigate the pairing of Self-Organizing Maps with Ordinary Kriging techniques as applied to hydrogeology. Ordinary Kriging predicts the values of hydraulic head in a case study. The use of Self-Organizing Maps aims to create groups of observations in the case study, to which Ordinary Kriging is applied. The implementation of the proposed methodology was carried out on the case study of the aquifer of Tympaki, Crete. The Tympaki aquifer is a generally homogenous porous aquifer with local differences in hydraulic properties. The pairing is evaluated using the following validation criteria: mean absolute error, maximum absolute error, root mean square error and correlation coefficient. Each group of observations is called a cluster and a whole set of clusters is called a topology. Different configurations of clusters were investigated to select the best topology. In order to assess the improvement of prediction using the Self-Organizing Map algorithm the Ordinary Kriging prediction was performed with the following results: Mean Absolute Error 6.9 m, Maximum Absolute Error 56.5 m, Root Mean Square Error 11.7 m and Correlation Coefficient 92%. The best performing topology consisted of 6 observation groupsdivided by location and hydraulic properties by the Self-Organizing Map algorithm. The best topology resulted in the following ranges of validation criteria: Mean absolute error 0.39-2 m, maximum absolute error 1.7-33 m, root mean square error 0.7-8.7 m and correlation coefficient 81-93% with an outlier of -14% due to linear and non-stochasticprediction. In addition, the grouping provided insight on the properties of the heterogeneities of the case study. The proposed methodology yielded improved results, even in the initial configurations and it applicable to other case studies with very few modifications due to its generic structure.