TY - GEN
T1 - Clustering on SNMR and TEM parameters over a saline-freshwater interface to resolve hydrogeological layers
AU - Vang, Mathias Østbjerg
AU - Grombacher, Denys
AU - Larsen, Jakob Juul
AU - Christiansen, Anders Vest
PY - 2024/9/8
Y1 - 2024/9/8
N2 - Coastal aquifer vulnerability to saline intrusion is a major risk for drinking water supply. Transient electromagnetics (TEM) can identify vulnerable aquifers yet areas where the geology consists of saline filled sands and clays, it is not possible to distinguish only based on salinity. Instead, we propose a combined interpretation of surface nuclear magnetic resonance (SNMR) and TEM inversion to resolve the boundary between clays and sands with saline water. A field case from an island, Endelave, in Denmark is shown with possible saltwater intrusion and a geology consisting of sandy tills to thick clays. A K-means-clustering approach is used to group the data into separate clusters based on three parameters, SNMR water content, relaxation time and TEM resistivities. The number of clusters is evaluated based on their silhouette index to ensure maximum uniqueness and avoid overfitting. We show that clustering can separate the multivariate data into important hydrogeological layers, including saline filled aquifers, freshwater aquifers, and clay rich materials. Future work will try to implement a translator function from SNMR data to TEM, to populate the full TEM space with SNMR parameters, for further clustering analysis.
AB - Coastal aquifer vulnerability to saline intrusion is a major risk for drinking water supply. Transient electromagnetics (TEM) can identify vulnerable aquifers yet areas where the geology consists of saline filled sands and clays, it is not possible to distinguish only based on salinity. Instead, we propose a combined interpretation of surface nuclear magnetic resonance (SNMR) and TEM inversion to resolve the boundary between clays and sands with saline water. A field case from an island, Endelave, in Denmark is shown with possible saltwater intrusion and a geology consisting of sandy tills to thick clays. A K-means-clustering approach is used to group the data into separate clusters based on three parameters, SNMR water content, relaxation time and TEM resistivities. The number of clusters is evaluated based on their silhouette index to ensure maximum uniqueness and avoid overfitting. We show that clustering can separate the multivariate data into important hydrogeological layers, including saline filled aquifers, freshwater aquifers, and clay rich materials. Future work will try to implement a translator function from SNMR data to TEM, to populate the full TEM space with SNMR parameters, for further clustering analysis.
UR - http://www.scopus.com/inward/record.url?scp=85214788186&partnerID=8YFLogxK
U2 - 10.3997/2214-4609.202420102
DO - 10.3997/2214-4609.202420102
M3 - Article in proceedings
T3 - EAGE Conference Proceedings
SP - 1
EP - 5
BT - NSG2024, 30th European Meeting of Environmental and Engineering Geophysics
PB - European Association of Geoscientists and Engineers
ER -