Abstract
The PhD thesis presents a new method for analyzing the relationship between resistivity and lithology, as well as a method for quantifying the hydrostratigraphic modeling uncertainty related to Multiple-Point Statistical (MPS) methods. Three-dimensional (3D) geological models are im-portant to securing, managing and protecting clean drinking water. The distribution of hydraulic conductivity, or model structure, of groundwater models is defined based on such 3D models, and it has been proven that the hydrological predictions are sensitive towards the model structure. The focus of this thesis is to improve analysis and research of the resistivity-lithology relationship and ensemble geological/hydrostratigraphic modeling.
The groundwater mapping campaign in Denmark, beginning in the 1990’s, has resulted in the collection of large amounts of borehole and geophysical data. The data has been compiled in two publicly available databases, the JUPITER and GERDA databases, which contain borehole and geophysical data, respectively. The large amounts of available data provided a unique opportunity for studying the resistivity-lithology relationship. The method for analyzing the resistivity-lithology relationship was presented in Paper I, along with an elaborate Resistivity Atlas of Denmark compiled from ground-based and airborne electromagnetic data. The presented method can easily be integrated in semi-automatic MPS approaches, for the purpose of stochastic 3D hydrostratigraphic modeling (e.g. Paper II and Paper III).
A common approach to geological modeling is deterministic modeling. A ‘best’ model is gen-erated manually by a geoscientist using prior knowledge on regional geology, borehole data, and geophysical data. Such models are uncertain, commonly without quantified uncertainty. Alterna-tively stochastic MPS methods can be utilized for creating ensembles of equiprobable models. The models are generated from soft geophysical data, an AEM dataset, and hard borehole data from a study area in eastern Jutland, Denmark. The geological conceptualization is implemented via a Training Image (TI) generated from a deterministic 3D geological model of the study area. The stochastic ensemble modeling approach is used to compare three different MPS methods (Paper II). However, visually comparing a large set of 3D hydrostratigraphic models is no trivial task. Therefore, a quantitative comparison technique is used to exhaustively compare the 3D models. The presented comparison techniques use so-called distance measures for comparing the similarity between two models. If two models are similar the distance is small, if they are dis-similar the distance is large. The study revealed the advantages/disadvantages related to using three different MPS methods for modeling airborne electromagnetic data.
The usage of stochastic MPS methods unlocks the ability to quantify ensemble hydrostrati-graphic modeling uncertainty. Each model in the ensemble is different, and together the ensem-ble quantifies the uncertainty. In Paper III the ensemble modeling uncertainty was studied by varying the MPS setup. The study was divided into 8 sub-cases, of which each case was con-cerned with a specific type of uncertainty related to a couple of overall topics: reconstructing incomplete soft data, conceptual geology via the TI, and borehole data. The results revealed the importance of the geophysical data for conditioning the MPS simulations. It was also illustrated in practice how a 3D geological model from another area with a similar geological setting could be used as a TI for MPS simulation of the actual survey.
The research of this thesis focuses on 3D hydrostratigraphic modeling, and future research would entail testing MPS in relation to groundwater modeling and flow predictions.
The groundwater mapping campaign in Denmark, beginning in the 1990’s, has resulted in the collection of large amounts of borehole and geophysical data. The data has been compiled in two publicly available databases, the JUPITER and GERDA databases, which contain borehole and geophysical data, respectively. The large amounts of available data provided a unique opportunity for studying the resistivity-lithology relationship. The method for analyzing the resistivity-lithology relationship was presented in Paper I, along with an elaborate Resistivity Atlas of Denmark compiled from ground-based and airborne electromagnetic data. The presented method can easily be integrated in semi-automatic MPS approaches, for the purpose of stochastic 3D hydrostratigraphic modeling (e.g. Paper II and Paper III).
A common approach to geological modeling is deterministic modeling. A ‘best’ model is gen-erated manually by a geoscientist using prior knowledge on regional geology, borehole data, and geophysical data. Such models are uncertain, commonly without quantified uncertainty. Alterna-tively stochastic MPS methods can be utilized for creating ensembles of equiprobable models. The models are generated from soft geophysical data, an AEM dataset, and hard borehole data from a study area in eastern Jutland, Denmark. The geological conceptualization is implemented via a Training Image (TI) generated from a deterministic 3D geological model of the study area. The stochastic ensemble modeling approach is used to compare three different MPS methods (Paper II). However, visually comparing a large set of 3D hydrostratigraphic models is no trivial task. Therefore, a quantitative comparison technique is used to exhaustively compare the 3D models. The presented comparison techniques use so-called distance measures for comparing the similarity between two models. If two models are similar the distance is small, if they are dis-similar the distance is large. The study revealed the advantages/disadvantages related to using three different MPS methods for modeling airborne electromagnetic data.
The usage of stochastic MPS methods unlocks the ability to quantify ensemble hydrostrati-graphic modeling uncertainty. Each model in the ensemble is different, and together the ensem-ble quantifies the uncertainty. In Paper III the ensemble modeling uncertainty was studied by varying the MPS setup. The study was divided into 8 sub-cases, of which each case was con-cerned with a specific type of uncertainty related to a couple of overall topics: reconstructing incomplete soft data, conceptual geology via the TI, and borehole data. The results revealed the importance of the geophysical data for conditioning the MPS simulations. It was also illustrated in practice how a 3D geological model from another area with a similar geological setting could be used as a TI for MPS simulation of the actual survey.
The research of this thesis focuses on 3D hydrostratigraphic modeling, and future research would entail testing MPS in relation to groundwater modeling and flow predictions.
| Bidragets oversatte titel | Undersøgelse af lithologiske og geofysiske relationer med anvendelser til analyse af geologisk usikkerhed ved brug af Multiple-Point Statics metoder |
|---|---|
| Originalsprog | Engelsk |
| Udgiver | |
| Status | Udgivet - 2 nov. 2017 |