Selection of coherent deposit-type locations and their application in data-driven mineral prospectivity mapping

doi:10.1016/j.oregeorev.2007.07.001

Ore Geology Reviews

Volume 33, Issues 3–4, June 2008, Pages 536-558

https://doi.org/10.1016/j.oregeorev.2007.07.001 Get rights and content

Abstract

Data-driven prospectivity mapping can be undermined by dissimilarity in multivariate spatial data signatures of deposit-type locations. Most cases of data-driven prospectivity mapping, however, make use of training sets of randomly selected deposit-type locations with the implicit assumption that they are coherent (i.e., with similar multivariate spatial data signatures). This study shows that the quality of data-driven prospectivity mapping can be improved by using a training set of coherent deposit-type locations. Analysis and selection of coherent deposit-type locations was performed via logistic regression, by using multiple sets of deposit occurrence favourability scores of univariate geoscience spatial data as independent variables and binary deposit occurrence scores as dependent variable. The set of coherent deposit-type locations and three sets of randomly selected deposit-type locations were each used in data-driven prospectivity mapping via application of evidential belief functions. The prospectivity map based on the training set of coherent deposit-type locations resulted in lower uncertainty, better goodness-of-fit to the training set, and better predictive capacity against a cross-validation set of economic deposits of the type sought. This study shows that explicit selection of training set of coherent deposit-type locations should be applied in data-driven prospectivity mapping.

Introduction

Known deposit-type locations are samples of a mineralized landscape. They are used in data-driven prospectivity mapping to predict sampling (exploration) targets for undiscovered deposit-type locations in the mineralized landscape. After creation of a GIS geoscience spatial database, there are three stages in data-driven prospectivity mapping: (1) selection of training deposit-type locations; (2) creation of prospectivity map; and (3) cross-validation of prospectivity map. In the first stage, the common practice is random selection of training deposit-type locations with the tacit assumption that they are coherent (i.e., having strongly similar characteristics). In the second stage, spatial associations between training deposit-type locations and individual evidential variables are quantified to create predictor maps, which are integrated to derive a prospectivity map. In the third stage, the prospectivity map is validated in terms of goodness-of-fit to training deposit-type locations and is cross-validated in terms of predictive capacity against deposit-type locations not used to create the predictor maps. The validation and cross-validation of a prospectivity map indicate its usefulness as a predictive tool to guide further exploration toward undiscovered deposit-type locations in the mineralized landscape.

Explicit selection of coherent training deposit-type locations is not practiced in most cases, even though it is crucial to prospectivity mapping. Every mineral deposit, even if classified into a deposit-type, is unique and has characteristics that are, to a certain extent, dissimilar to other mineral deposits of the same type. It follows that multivariate spatial data signatures of deposit-type locations are, to a certain extent, dissimilar or non-coherent. Random selection of deposit-type locations is therefore likely to result in a non-coherent training set. Dissimilarity in multivariate spatial data signatures of deposit-type locations in a non-coherent training set can undermine the quality of a prospectivity map. It is hypothesized here that uncertainty of a prospectivity map can be reduced and that goodness-of-fit and predictive capacity of a prospectivity map can be improved by using a training set of coherent deposit-type locations (i.e., with similar multivariate spatial data signatures).

This paper demonstrates a two-step methodology to select, from all known deposit-type locations in a study area, a set of coherent deposit-type locations. Then, to test and demonstrate the hypothesis stated above, performance of a prospectivity map based on the set of coherent deposit-type locations is compared and contrasted with performances of prospectivity maps based on sets of randomly selected deposit-type locations. The performance indices used are map uncertainty, goodness-of-fit to training deposit-type locations, and predictive capacity against cross-validation deposit-type locations. The proposed technique of selecting a training set of coherent deposit-type locations and the hypothesis of utility of such a training set are tested and demonstrated in data-driven mapping of prospectivity for alkalic porphyry Cu–Au deposits in British Columbia.

This paper is organized into seven sections. After this introduction, the second section describes the geology/mineralization of the test area and the geoscience spatial data sets used in the study. The third section discusses the first step in selecting coherent deposit-type locations (i.e., analysis of mineral occurrence favourability scores of univariate geoscience spatial data with respect to the deposit and non-deposit locations). The fourth section discusses the second step in selecting coherent deposit-type locations (i.e., analysis of deposit-type locations with similar multivariate spatial data signatures via logistic regression using binary deposit occurrence scores as dependent variable and multiple sets of deposit occurrence favourability scores of univariate geoscience spatial data as independent variables). The fifth section describes and compares results of prospectivity mapping using the training set of coherent deposit-type locations with those obtained using training sets of randomly selected deposit-type locations. The sixth and seventh sections provide, respectively, overall discussion and conclusions of the study.

Section snippets

The test area and geoscience spatial data sets

Most of the province of British Columbia (Fig. 1), in western Canada, forms part of the Canadian Cordillera, which in turn is part of the Cordilleran Orogenic belt in the North American continent. The Canadian Cordillera is an orogenic collage of five major fault-bounded morphogeological and physiographic belts (Fig. 1), which developed from Mesozoic to post-mid Tertiary events, and docked onto the North American craton (Gabrielse et al., 1991, McMillan, 1991a). Each belt has distinct

Mineral occurrence favourability scores of univariate geoscience spatial data

This section describes and discusses the first step in selecting coherent deposit locations by analysis of mineral occurrence favourability scores of univariate geoscience spatial data with respect to deposit and non-deposit locations.

Analysis and selection of coherent deposit locations

This section describes and discusses the second step in selecting coherent deposit locations by analysis of deposit-type locations with similar multivariate spatial data signatures.

Data-driven prospectivity mapping

Data-driven prospectivity maps can be created via [weighted] logistic regression (Chung and Agterberg, 1980, Reddy et al., 1991, Carranza and Hale, 2001). In fact, the results shown in Fig. 5, Fig. 6 indicate some non-deposit locations having multivariate spatial data signatures similar to several deposit locations and these are thus plausible prospective targets. Logistic regression was not applied here to create prospectivity maps, because of its disadvantage with missing data, which affects

Discussion

The work presented here for selection of useful training samples in predictive modeling was stimulated by the work of San Pedro and Burstein (2003) in the context of tropical cyclone prediction. They applied the concept of case-based reasoning as an effective means of solving a present problem by selecting and using past cases that closely match the current problem situation (Aamodt and Plaza, 1994). San Pedro and Burstein (2003) used fuzzy mathematics as an expert-driven tool for

Conclusions

•
Analysis of spatial associations between univariate geoscience spatial data and deposit/non-deposit locations using cumulative spatial frequency distributions is a satisfactory preliminary step in the selection of coherent deposit-type locations.
•
Logistic regression is an efficient quantitative model for classification and selection of deposit-type locations with coherent or similar multivariate spatial data signatures.
•
Application of coherent deposit-type locations reduces uncertainty and

References (63)

F.P. Agterberg et al.
Statistical pattern integration for mineral exploration
E.J.M. Carranza et al.
Evidential belief functions for data-driven geologically-constrained predictive mapping of gold potential, Baguio district, Philippines
Ore Geology Reviews
(2003)
Q. Cheng et al.
A spatial analysis method for geochemical mineral occurrence recognition
Journal of Geochemical Exploration
(1996)
A. Aamodt et al.
Case-based reasoning: fundamental issues, methodological variations and system approaches
AI Communications
(1994)
F.P. Agterberg et al.
Measuring performance of mineral-potential maps
Natural Resources Research
(2005)
F.P. Agterberg et al.
Weights of evidence modeling and weighted logistic regression for mineral potential mapping
P.D. Allison
Missing Data
(2002)
P. An et al.
An object-oriented knowledge representation structure for exploration data integration
Nonrenewable Resources
(1994)
P. An et al.
Uncertainty management in integration of exploration data using the belief functions
Nonrenewable Resources
(1994)
X. Bi et al.
The alkaline porphyry associated Yaoán gold deposit, Yunnan, China: rare earth element and stable isotope evidence foe magmatic hydrothermal ore formation
Mineralium Deposita
(2004)

R. Black et al.

Cratons, mobile belts, alkaline rocks and continental lithospheric mantle; the Pan-African testimony

Journal of the Geological Society

(1993)

G.F. Bonham-Carter

Geographic Information Systems for Geoscientists, Modelling with GIS

(1994)

G.F. Bonham-Carter et al.

Weights of evidence modeling: a new approach to mapping mineral potential

B.N. Boots et al.

Point Pattern Analysis

W.M. Brown et al.

Artificial neural networks: a new method for mineral prospectivity mapping

Australian Journal of Earth Sciences

(2000)

N.E. Breslow et al.

Logistic regression for two-stage case-control data

Biometrika

(1988)

E.J.M. Carranza et al.

Logistic regression for geologically constrained mapping of gold potential, Baguio district, Philippines

Exploration and Mining Geology

(2001)

E.J.M. Carranza et al.

Where are porphyry copper deposits spatially localized? A case study in Benguet province, Philippines

Natural Resources Research

(2002)

E.J.M. Carranza et al.

Application of data-driven evidential belief functions to prospectivity mapping for aquamarine-bearing pegmatites, Lundazi district, Zambia

Natural Resources Research

(2005)

C.F. Chung et al.

Regression models for estimating mineral resources from geological map data

Mathematical Geology

(1980)

D.P. Cox et al.

Distribution of Gold in Porphyry Copper Deposits

(1988)

D.R. Cox et al.

Analysis of Binary Data

(1989)

J.C. Davis

Statistics and Data Analysis in Geology

(1973)

K.M. Dawson et al.

Regional metallogeny

A.P. Dempster

Upper and lower probabilities induced by a multivalued mapping

Annals of Mathematical Statistics

(1967)

A.P. Dempster

Generalization of Bayesian inference

Journal of the Royal Statistical Society

(1968)

P.J. Diggle

Statistical Analysis of Spatial Point Patterns

(1983)

L.J. Drew et al.

Application of the porphyry copper/polymetallic vein kin deposit system to mineral-resource assessment in the Mátra Mountain, northern Hungary

H. Gabrielse et al.

Morphogeological belts, tectonic assemblages, and terranes, Part A

Geoscience Data Repository

Canadian Aeromagnetic Data Base. Geological Survey of Canada, Earth Sciences Sector, Natural Resources Canada, Government of Canada

Geoscience Data Repository

Canadian Geodetic Information System. Geological Survey of Canada, Earth Sciences Sector, Natural Resources Canada, Government of Canada

Cited by (164)

Quantifying uncertainty and improving prospectivity mapping in mineral belts using transfer learning and Random Forest: A case study of copper mineralization in the Superior Craton Province, Quebec, Canada
2024, Ore Geology Reviews
Mineral prospectivity mapping (MPM) involves identifying locations with a higher potential for mineral exploration based on a set of explanatory variables. In cases where there is a scarcity or absence of unfavorable sites that adequately represent the geological context for deposit discovery, generating synthetic negative data sets becomes necessary to employ a machine learning algorithm optimally. Moreover, when favorable sites are insufficient for deposit discovery within a geological zone, machine learning methods can potentially result in large and highly uncertain prospecting areas. This article proposed a concept based on transfer learning by applying the knowledge gained from mineral belt signatures in different geological zones to a related area. The positive training data were taken from five mineral belts distanced from each other, while the negative data were sampled using geological constraints based on the distance to occurrences and spatial associativity. The results demonstrate that transfer learning, combined with geological constraints applied to the creation of negative datasets, improves model performance and prediction of known deposits while significantly reducing uncertainties. Mineral prospectivity models for predicting potential copper formations were generated using data from the Quebec Government's spatial reference geomining information system, SIGEOM. The case study for this work focused on the geological province of the Superior Craton, which encompasses the vast majority of northeastern Quebec.
Stacking: A novel data-driven ensemble machine learning strategy for prediction and mapping of Pb-Zn prospectivity in Varcheh district, west Iran
2024, Expert Systems with Applications
Various ensemble machine learning techniques have been widely studied and implemented to construct the predictive models in different sciences, including bagging, boosting, and stacking. However, bagging and boosting concentrate on minimizing variance or bias, stacking techniques aimed at reducing both by identifying the optimal integration of base learners. Moreover, while most ensemble methods simply combine identical machine learning models, stacking utilizes a meta-machine learning model to combine different base learning models, aiming to enhance the overall accuracy of generalization. Therefore, this research showed the utilization of stacking, an ensemble approach, to develop mineral prospectivity models for Pb-Zn mineralization in the Varcheh District, west Iran. To end this, various exploration evidence layers, including geochemical data, remote sensing data, geological and tectonic controls were used to construct the stacking structure. In the following, a set of five base learners were applied, containing support vector regression (SVR) using RBF, linear and polynomial kernels, the K-nearest neighbor (KNN), and linear regression. Ridge, SVR-RBF and XGBoost were used as a meta-learner to integrate the outputs of basic learners. To measure how well each model performed, ROC, F1-score and Precision metrics was carried out. Moreover, compared to the separate algorithms, the stacking-based ensemble model showed a better prediction accuracy. The findings of this study demonstrated that the ensemble model based on stacking achieved a 95% prediction rate for Pb-Zn deposits, covering only 9% of the study area. As a result, this model holds promise as an effective tool for predicting mineral prospectivity in other study areas, regardless of whether they exhibit similar or different types of mineralization.
Leucogranite mapping via convolutional recurrent neural networks and geochemical survey data in the Himalayan orogen
2024, Geoscience Frontiers
Geochemical survey data analysis is recognized as an implemented and feasible way for lithological mapping to assist mineral exploration. With respect to available approaches, recent methodological advances have focused on deep learning algorithms which provide access to learn and extract information directly from geochemical survey data through multi-level networks and outputting end-to-end classification. Accordingly, this study developed a lithological mapping framework with the joint application of a convolutional neural network (CNN) and a long short-term memory (LSTM). The CNN-LSTM model is dominant in correlation extraction from CNN layers and coupling interaction learning from LSTM layers. This hybrid approach was demonstrated by mapping leucogranites in the Himalayan orogen based on stream sediment geochemical survey data, where the targeted leucogranite was expected to be potential resources of rare metals such as Li, Be, and W mineralization. Three comparative case studies were carried out from both visual and quantitative perspectives to illustrate the superiority of the proposed model. A guided spatial distribution map of leucogranites in the Himalayan orogen, divided into high-, moderate-, and low-potential areas, was delineated by the success rate curve, which further improves the efficiency for identifying unmapped leucogranites through geological mapping. In light of these results, this study provides an alternative solution for lithologic mapping using geochemical survey data at a regional scale and reduces the risk for decision making associated with mineral exploration.
Mapping environmental crime to characterize human impacts on islands: an applied and methodological research in Canary Islands.
2023, Journal of Environmental Management
Environmental crimes are a global issue due to the damage they cause to landscapes and ecosystems. This study focused on characterizing environmental crimes in the Canary Islands (Spain). Four categories of environmental crimes related to construction, mining and tilling, solid waste, and liquid waste) were defined and analysed. A total of 28 databases were generated, corresponding to each of the 7 major islands and each environmental crime typology. Each database was linked to information on land use and the socioeconomic and physical characteristics of the territory. For each database, firstly a descriptive statistical analysis was conducted, followed by the generation of a regularized Random Forest model with the aim of identifying characteristics that may be related to the location of environmental crimes. The results showed that, in most cases, proximity to residential accommodations, agricultural areas and industrial zones act as the main explanatory features of the distribution of environmental crimes. Furthermore, a marked pattern of concentration of environmental crimes in the coastal belt of the islands was observed, mainly associated with urban-tourist development since the 1960s and 1970s.
Recognition of mineralization-related anomaly patterns through an autoencoder neural network for mineral exploration targeting
2023, Applied Geochemistry
In mineral potential mapping, supervised machine learning algorithms have shown great promise in delineating and prioritizing potential areas. However, since mineralization being a relatively rare geological event, most supervised machine learning-based models face substantial challenges in properly identifying prospective areas. Data sets with strongly imbalanced distributions of the target variable (deposits) and insufficient training data sets impose obstacles to these kinds of models which can significantly impact adversely on the performance of the models. Moreover, in some cases, negative training data sets as the non-deposit locations aren't really true negative data, which cause higher uncertainty in a mineral potential map. In this study, for handling these challenges the deep autoencoder neural network is adopted. The autoencoder can be trained to reconstruct geospatial data set in totally unsupervised manner and identify prospective areas based on the reconstruction error, where higher error corresponds with areas of higher mineral potential. In order to confirm the efficiency of the autoencoder algorithm in mineral potential modeling, the model was compared with a popular data-driven approach that assigned a weight to the evidence layer by using a concentration-area (C-A) fractal model and a prediction-area (P-A) plot, and combined them using a multi-class index overlay method. Receiver operating characteristic (ROC) curve, success-rate curve, and P-A plot were adopted to evaluate the predictive ability of Fe prospectivity models pertaining to the Esfordi district of Iran. Also, we use an area under the ROC curve (AUC) and partial AUC (pAUC) to quantitatively evaluate the overall and sensitivity performance of models, respectively.
Random forest classification for volcanogenic massive sulfide mineralization in the Rouyn-Noranda Area, Quebec
2023, Ore Geology Reviews
Random forest (RF) classification was applied to 37 predictor maps (vectors to mineralization) producing a Mineral Prospectivity Map (MPM) for volcanogenic massive sulfide (VMS) mineralization in the Noranda District, Abitibi subprovince, which is host to ∼ 20 VMS deposits and numerous subeconomic occurrences. The predictor maps were created using geological, geochemical, and geophysical data, and the known VMS deposits were used to train the RF classifier. The RF model was applied on two regions of interest (ROI) to investigate the effect of different sized ROIs on the results. Five sets of balanced and unbalanced training data were used in the experiments to examine the sensitivity of RF to different sets of training data. The probability/prospectivity maps showed very high success rate of classification with regard to training data particularly for the training sets with higher ratio of non-deposits/deposits. Accuracy assessment of classification maps using cross-validation also showed very high accuracy when assessed by the training data in all experiments again giving a higher accuracy for the training sets with higher ratios of non-deposits/deposits, however the later showed very low accuracy when applying k-fold cross-validation or when assessed by 15 VMS showings used as the test data indicating an overfitting problem with the unbalanced data. In most of the experiments, principal component 4 (PC4) of geochemical analysis, proximity to synvolcanic tonalite-trondhjemite-granodiorite (TTG) intrusions, lithology, and sericite alteration maps showed higher predictive power. The importance of individual variables changed when the ROI was changed to a smaller area suggesting that RF is sensitive to the study area.

View all citing articles on Scopus

View full text

Selection of coherent deposit-type locations and their application in data-driven mineral prospectivity mapping

Abstract

Introduction

Section snippets

The test area and geoscience spatial data sets

Mineral occurrence favourability scores of univariate geoscience spatial data

Analysis and selection of coherent deposit locations

Data-driven prospectivity mapping

Discussion

Conclusions

Ore Geology Reviews

Journal of Geochemical Exploration

Case-based reasoning: fundamental issues, methodological variations and system approaches

AI Communications

Measuring performance of mineral-potential maps

Natural Resources Research

Weights of evidence modeling and weighted logistic regression for mineral potential mapping

Missing Data

An object-oriented knowledge representation structure for exploration data integration

Nonrenewable Resources

Uncertainty management in integration of exploration data using the belief functions

Nonrenewable Resources

The alkaline porphyry associated Yaoán gold deposit, Yunnan, China: rare earth element and stable isotope evidence foe magmatic hydrothermal ore formation

Mineralium Deposita

Cratons, mobile belts, alkaline rocks and continental lithospheric mantle; the Pan-African testimony

Journal of the Geological Society

Geographic Information Systems for Geoscientists, Modelling with GIS

Weights of evidence modeling: a new approach to mapping mineral potential

Point Pattern Analysis

Artificial neural networks: a new method for mineral prospectivity mapping

Australian Journal of Earth Sciences

Logistic regression for two-stage case-control data

Biometrika

Logistic regression for geologically constrained mapping of gold potential, Baguio district, Philippines

Exploration and Mining Geology

Where are porphyry copper deposits spatially localized? A case study in Benguet province, Philippines

Natural Resources Research

Application of data-driven evidential belief functions to prospectivity mapping for aquamarine-bearing pegmatites, Lundazi district, Zambia

Natural Resources Research

Regression models for estimating mineral resources from geological map data

Mathematical Geology

Distribution of Gold in Porphyry Copper Deposits

Analysis of Binary Data

Statistics and Data Analysis in Geology

Regional metallogeny

Upper and lower probabilities induced by a multivalued mapping

Annals of Mathematical Statistics

Generalization of Bayesian inference

Journal of the Royal Statistical Society

Statistical Analysis of Spatial Point Patterns

Application of the porphyry copper/polymetallic vein kin deposit system to mineral-resource assessment in the Mátra Mountain, northern Hungary

Morphogeological belts, tectonic assemblages, and terranes, Part A

Canadian Aeromagnetic Data Base. Geological Survey of Canada, Earth Sciences Sector, Natural Resources Canada, Government of Canada

Canadian Geodetic Information System. Geological Survey of Canada, Earth Sciences Sector, Natural Resources Canada, Government of Canada