Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The use of geothermal energy in mining : a decision-making framework Patsa, Eleni 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2018_may_patsa_eleni.pdf [ 261.36MB ]
Metadata
JSON: 24-1.0365825.json
JSON-LD: 24-1.0365825-ld.json
RDF/XML (Pretty): 24-1.0365825-rdf.xml
RDF/JSON: 24-1.0365825-rdf.json
Turtle: 24-1.0365825-turtle.txt
N-Triples: 24-1.0365825-rdf-ntriples.txt
Original Record: 24-1.0365825-source.json
Full Text
24-1.0365825-fulltext.txt
Citation
24-1.0365825.ris

Full Text

THE USE OF GEOTHERMAL ENERGY IN MINING:  A DECISION-MAKING FRAMEWORK by  Eleni Patsa M.Sc. Information Technology, University of  Glasgow, 2003B.Eng. (Hons) Aeronautical Engineering, University of  Glasgow, 2002A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Mining Engineering) The University Of  British Columbia (Vancouver) April 2018 © Eleni Patsa, 2018
Abstract This thesis aims to support the uptake of  geothermal energy by the mining industry by developing a decision-making framework that when applied, will allow decision makers to quickly and inexpensively assess whether geothermal merits further consideration as a potential source of  energy for a given mining project. The intent was to demonstrate that such an assessment could be made by non-experts, without having to resort to more complex, specialist analysis that is typically part of  traditional geothermal exploration. To this end, a 3-step approach was adopted: a) identify, rank, and map indicators of  worldwide geothermal potential; b) identify, rank, and map indicators of  worldwide mineral potential; c) combine geothermal and mineral potential and map to identify areas of  significant overlap. The calculation of  geothermal potential necessitated the creation of  a comprehensive listing of  geothermal power plants, which were individually verified for location accuracy — associated maps of  current geothermal production were also created for select areas of  interest. Geothermal potential was represented by 5 indicators, namely volcanos, earthquakes, tectonic plate boundaries, heat flow, and thermal springs. Geo-indicator subtypes were ranked with respect to their proximity to active geothermal production, and were subsequently classified as either Primary or Secondary, and used to derive a rough estimate of  geothermal potential, even in areas with no current geothermal production. An Analytic Hierarchy Process-based model used to calculate the relative geo-indicator importance weights was presented, and it can be extended to include additional indicators of  geothermal potential to the ones used in this research with future releases of  exploration data. A comprehensive map of  mineral potential was also produced, using a combination of  publicly-available and proprietary data, for a total of  148 distinct commodities. Mineral potential indicator importance weights were calculated based on a combination of  attributes, including resource size, development stage, activity state, resource value, and commodity listing order. Those were subsequently ranked based on their proximity to geothermal potential, and used to produce worldwide maps of  geomine potential. 
 iiLay Summary Mines are heavy energy consumers and they rely on electricity that is either purchased off  the grid, or produced using a variety of  sources, including traditional fossil fuels and renewables. Even though geothermal energy is clean, renewable, base-load, with very high capacity factors (of  particular importance to mining operations), the industry has been slow to adopt it. The methodology developed as part of  this research used a variety of  techniques to derive an estimate for the existence and coexistence of  geothermal and/or mineral resources around the world, even in areas of  zero geothermal or mineral production information. This methodology can help mining operators make a reasonable assessment on the viability of  integrating geothermal in mining operations, without having to resort to the analysis of  more complex, specialist data.  iiiPreface All of  the work presented henceforth was conducted at the Mining Engineering Department at the University of  British Columbia, Point Grey campus.  Parts of  Chapter 2 (Problem Background and Literature Review), and have been published (Patsa et al., 2015a; 2015b). I was the lead investigator, responsible for all major areas of  concept formation, data collection and analysis, as well as manuscript composition. Arianpoo N. was involved in the early stages of  concept formation and contributed to manuscript edits. Van Zyl D. and Zarrouk S. were the supervisory authors on this project, and were involved throughout the project in concept formation and manuscript composition.  ivTable of  Contents Abstract	 ii ..................................................................................................................................................Lay Summary	 iii ........................................................................................................................................Preface	 iv ...................................................................................................................................................Table of  Contents	 v ..................................................................................................................................List of Tables  .....................................................................................................................................viii List of Figures........................................................................................................................................ x Acknowledgements	 xv ...............................................................................................................................Dedication	 xvi ...........................................................................................................................................Chapter 1: Introduction	 1 .......................................................................................................Chapter 2: Problem Background & Literature Review	 3 .......................................................2.1 Energy use in mining	 5 ...................................................................................................................2.2 Mines, energy and decision-making	 7 ............................................................................................2.2.1 Operating costs	 7 ..............................................................................................................................2.2.2 Access, safety, and power supply security	 7 ...........................................................................................2.2.3 Environmental requirements	 8 ............................................................................................................2.3 Geothermal resources	 8 ..................................................................................................................2.4 Mineral resources	 10 ......................................................................................................................2.4.1 Metals and minerals related to volcanism	 11 .......................................................................................2.5 Geothermal production	 11 .............................................................................................................2.5.1 Power production basics	 11 ...............................................................................................................2.5.2 Geothermal power stations	 12 ............................................................................................................2.5.3 The geothermal development cycle	 14 ..................................................................................................2.5.4 The renewability of  geothermal resources	 15 .......................................................................................2.5.5 Geothermal electricity costs	 16 ...........................................................................................................2.5.6 Geothermal production around the world	 19 ........................................................................................2.6 Mineral production	 25 ...................................................................................................................2.7 Geothermal energy use in mineral production	 27 .........................................................................2.7.1 Mining, heat, and water use 	 27 ........................................................................................................2.7.2 Geothermal applications akin to mining	 28 .........................................................................................2.8 The Analytic Hierarchy Process (AHP)	 39 .....................................................................................Chapter 3: Methods and Research Design	 44 ........................................................................3.1 Assumptions and research questions	 44 .........................................................................................3.1.1 Assumptions	 44 ...............................................................................................................................3.1.2 Research questions	 46 .......................................................................................................................3.1.3 Limitations on scope and data complexity 	 46 ......................................................................................3.2 Research design	 46 .........................................................................................................................3.2.1 Process overview	 47 .......................................................................................................................... v3.2.2 Identification of  data requirements	 52 .................................................................................................3.2.3 The definition of  indicator features	 55 ................................................................................................3.3 Methods used to answer the research questions	 56 ........................................................................3.3.1 Data harvesting methods	 56 ..............................................................................................................3.3.2 Data transformation methods	 62 ........................................................................................................3.3.3 Data quality assessment methods 	 72 ..................................................................................................3.3.4 Geographical analysis methods	 74 ......................................................................................................3.3.5 Visualization methods	 75 ..................................................................................................................Chapter 4: Geothermal Production	 82 ...................................................................................4.1 Indicators of  geothermal production	 82 ........................................................................................4.2 Compiling the GPP data layer	 82 ..................................................................................................4.2.1 Data sources for the GPP data layer 	 82 .............................................................................................4.3 Assessment of  data quality for the GPP dataset 	 92 .......................................................................4.3.1 Accuracy	 92 .....................................................................................................................................4.3.2 Completion	 95 .................................................................................................................................4.3.3 Coverage	 97 .....................................................................................................................................4.3.4 Suitability	 98 ..................................................................................................................................4.3.5 Confidence	 98 ..................................................................................................................................4.4 Analysis of  the resulting GPP Dataset	 99 ......................................................................................4.4.1 Scaling the GPP data for GPP type importance and mapping	 100 .........................................................Chapter 5: Indicators of  Geothermal Potential	 113 ..............................................................5.1 Using geothermal production to indicate geothermal potential	 113 .............................................5.2 Identifying and classifying additional geo-indicators	 118 ..............................................................5.2.1 Degrees of  separation	 119 .................................................................................................................5.3 Using volcanoes to indicate geothermal potential	 121 ...................................................................5.3.1 Comparison between GPPs and volcanoes	 121 ....................................................................................5.3.2 Ranking volcano geo-indicator strength	 137 .........................................................................................5.4 Using thermal springs to indicate geo-potential	 148 ......................................................................5.4.1 Ranking the surface features layer	 162 ................................................................................................5.5 Quakes and tectonic boundaries as geo-indicators	 163 .................................................................5.5.1 Tectonic plate boundaries, earthquakes and geothermal production	 163 ...................................................5.5.2 Ranking tectonic boundaries and quakes using proximity analysis	 176 ....................................................5.5.3 Ranking earthquake data using proximity analysis	 177 ........................................................................5.6 Using heat flow to indicate geo-potential	 179 ................................................................................5.6.1 Ranking the heat flow layer	 184 ........................................................................................................5.7 Limitations in this chapter	 184 .......................................................................................................Chapter 6: Indicators of  Mineral Potential	 186 .....................................................................6.1 Mineral development data	 186 ......................................................................................................6.1.1 The USGS MRDS dataset	 186 ........................................................................................................ vi6.1.2 The InfoMine dataset	 196 ................................................................................................................6.2 Mineral potential	 217 .....................................................................................................................6.2.1 Mineral development indicator weights	 217 .........................................................................................6.2.2 Relative importance of  component layers	 221 .......................................................................................Chapter 7: Suitable Application Areas	 224 .............................................................................7.1 Finalizing the mineral and geothermal potential datasets	 224 ......................................................7.2 Selecting suitable application areas for geothermal indicators	 224 ...............................................7.2.1 Suitable Application Areas: Geothermal power plants	 225 ....................................................................7.2.2 Suitable Application Areas: Volcanoes 	 228 .........................................................................................7.2.3 Suitable Application Areas: Earthquakes	 231 ......................................................................................7.2.4 Suitable Application Areas: Tectonic plate boundaries	 232 ....................................................................7.2.5 Suitable Application Areas: Heat flow and thermal springs	 237 ............................................................7.3 Selecting suitable application areas for mineral indicators	 237 .....................................................Chapter 8: GeoMine Potential	 247 ..........................................................................................8.1 Calculating the geomine potential indicator weights	 247 ..............................................................8.2 Mapping geomine potential	 251 ....................................................................................................8.2.1. Potential layers, buffers, and fishnet grids	 251 .....................................................................................8.2.2 Overview geomine potential maps	 255 ................................................................................................8.3 Regions of  high geomine potential incidence	 261 .........................................................................8.3.1 Nevada, California, and Arizona	 266 ................................................................................................8.3.2 Chile	 271 ........................................................................................................................................8.4 Recommendation scale	 273 ............................................................................................................Chapter 9: Conclusion	 276 ......................................................................................................9.1 Overview of  research objectives, approach & outcomes	 276 .........................................................9.2 Contributions to knowledge	 281 ....................................................................................................9.3 Limitations	 282 ...............................................................................................................................9.4 Recommendations for future work	 284 ..........................................................................................Bibliography	 285 .......................................................................................................................................Appendix A: MDRS Preliminary Metadata Analysis	 295........................................................................ viiList of  Tables Table 1: Sample project cost estimate for a 70 MWe geothermal power plant in Kenya. 	 17 .........................Table 2: Summation of  original and normalized values of  GPP dataset (capacity field). 	 66 ..........................Table 3: Inverse distance to Philadelphia data.	 68 ...............................................................................................Table 4: Comparison of  cost of  living between alternatives	 68 ..........................................................................Table 5: The AHP preference scale.	 68 ................................................................................................................Table 6: City comparison w.r.t. climate.	 69 ...........................................................................................................Table 7: City comparison w.r.t. commuting and arts & recreation.	 69 ...............................................................Table 8: Comparison of  criteria w.r.t. goal.	 69 .....................................................................................................Table 9: Composite priorities for the cities.	 69 .....................................................................................................Table 10: Fuzzy classification for completion. 	 71 ................................................................................................Table 11: Common index metric scale for accuracy, completion and suitability. 	 73 .......................................Table 12: Pairwise comparison matrix & results for completion, accuracy and suitability.	 74 .........................Table 13: Fields, sample row and descriptive statistics for the thinkgeoenergy.com data.	 86 ...........................Table 14: Fields and sample row of  the scraped Global Energy Observatory data.	 87 ...................................Table 15: Fields, sample row, and descriptive statistics for the supplemented OpenEI data.	 90 ......................Table 16: Fields, sample row, and descriptive statistics for the Wikipedia data.	 91 ...........................................Table 17: Assessment of  completion for the four sources of  GPP data, compared to custom dataset.	 95 .....Table 18: Fuzzy classification for completion. 	 97 ................................................................................................Table 19: Assessment of  completion, sorted in order of  decreasing completion. 	 97 .......................................Table 20: Coverage calculations for the GPP source and resultant datasets.	 98 ...............................................Table 21: Assessment of  suitability for the source and resultant GPP datasets.	 98 ...........................................Table 22: Assessment of  confidence for the four sources of  GPP data, compared to custom dataset.	 98 ......Table 23: Fields, sample row, and descriptive statistics for the resultant GPP data layer.	 99 ............................Table 24: Summary statistics for resulting GPP dataset.	 99 ................................................................................Table 25: Average count and average capacity of  geothermal power stations per installation type. 	 101 ......Table 26: Average count and average capacity of  geothermal power stations per country.	 103 .....................Table 27: Constitution of  GPP complexes, by type. 	 115 ...................................................................................Table 28: Ranking of  GPP types based on each type’s calculated capacity density value. 	 117 ......................Table 29: Sample of  rows showing calculation of  geo-potential for the GPP layer.	 118 ..................................Table 30: Relative importance scales for all geo-indicator layers.	 119 ................................................................Table 31: GPP types per MWe-production. 	 129 ................................................................................................Table 32: Updated list of  relative importance weights. 	 133 ...............................................................................Table 33: Summary proximity matrix for volcano-GPP comparisons.	 144 ......................................................Table 34: Summary proximity matrix for volcano-GPP comparisons — attribute columns.	 144 ..................Table 35: Final importance indicator weights for volcano types.	 148 ................................................................vQQQTable 36: Proposed classification for geothermal surface features. 	 149 .............................................................Table 37: Completion of  the non-numerical values of  the original thermal springs dataset.	 153 ..................Table 38: Assessment of  completion for the numerical values of  the original thermal springs dataset.	 155 ..Table 39: NULL stats for the non-numerical values of  T in the thermal springs dataset. 	 156 .......................Table 40: Temperature range for the classified values of  the thermal springs dataset. 	 156 ............................Table 41: Assessment of  completion for the transformed thermal springs dataset. 	 159 .................................Table 42: Assessment of  confidence for the thermal springs dataset.	 159 .........................................................Table 43: Proximity analysis results and final weights for the thermal springs layer.	 162 .................................Table 44: Proximity values and final importance weights for tectonic boundary type.	 177 .............................Table 45: Row count breakdown per development stage for the MRDS dataset. 	 187 ...................................Table 46: Re-classification of  the MRDS commodities using a simplified naming scheme.	 190 ....................Table 47: Commodity groupings (including economic classes) for the MRDS commodity field.	 191 ...........Table 48: Row counts per stage for the InfoMine data (NULL & non-NULL lat/long values).	 197 .............Table 49: Row counts per activity state for the InfoMine data.	 204 ...................................................................Table 50: Coal classifications in the reserves table. 	 205 ......................................................................................Table 51: Commodities, commodity groups and economic classes defined for the InfoMine dataset. 	 206 ..Table 52: Development stage weight comparisons for the MRDS & InfoMine datasets.	 218 ........................Table 53: Activity state assignments for the MRDS data based on development stage. 	 219 ..........................Table 54: Traced and measured surface areas for selected geothermal fields. 	 227 ..........................................Table 55: Suitable application area radii corresponding to volcano types. 	 231 ................................................Table 56: Select sizes (in km) for tectonic plate boundaries per boundary type. 	 233 .......................................Table 57: Breakdown of  row counts for the geo-indicator datasets.	 237 ...........................................................Table 58: Total production and measured surface footprint of  sample gold mines in NV, USA.	 240 ............Table 59: Descriptive statistics for a) total production; and b) for traced areas.	 243 ..........................................Table 60: Measured surface areas (in km2) for sample mines. 	 245 ....................................................................Table 61: Assigned limits for search radii for the proximity analysis per geo-indicator subtype. 	 245 .............Table 62: Gold and copper mines by stage and state in NV, CA, and AZ.	 266 ................................................Table 63: Geomine potential wgeomine based recommendation level scale. 	 273 ................................................Table 64: Summarized research outcomes.	 280 ..................................................................................................ixList of  Figures Figure 1: Example schematic of  a geothermal (flash-steam) power station.	 13 ................................................Figure 2: Project development timeframe for a geothermal power plant. 	 15 ...................................................Figure 3: Total cost per energy unit, Iceland estimates. 	 16 ................................................................................Figure 4: Project development timeframe for a geothermal power plant. 	 17 ...................................................Figure 5: Geothermal exploration and development: Cost per development stage.	 18 ...................................Figure 6: Worldwide geothermal generation capacity (2015), per country, in MWe total. 	 20 ........................Figure 7: Classification and basic stats for current and projected geothermal production levels. 	 21 ..............Figure 8: Geo-producers by size, number of  countries vs. total generating capacity (2015).	 22 ......................Figure 9: Percent change in geothermal production by country producer geothermal electricity. 	 23 ............Figure 10: Percent change in geothermal production for new market entries. 	 24 ...........................................Figure 11: Typical duration per development stage of  a mining project. 	 25 ....................................................Figure 12: Typical costs associated with mining development stages. 	 26 ..........................................................Figure 13: Potential geothermal direct use in the mining sector (modelled after Lindal).	 29 ...........................Figure 14: Steps to addressing research objectives.	 48 .........................................................................................Figure 15: Steps, methods and software used to address research objectives.	 49 ..............................................Figure 16: List of  GPPs on OpenEI.org.	 57 .........................................................................................................Figure 17: a) HTML with embedded GPP data; and b) HTML for a single OpenEI row. 	 58 ......................Figure 18: Drill-down information for each GPP from OpenEI.org.	 59 ...........................................................Figure 19: Batch renaming harvested files using Automator. 	 60 ........................................................................Figure 20: Data scrubbing in Trifacta.	 61 .............................................................................................................Figure 21: Table Pivots in MS Excel. 	 61 ..............................................................................................................Figure 22: Effects of  normalization on data trends. Plots in 2-D.	 63 .................................................................Figure 23: Effects of  normalization on data trends. Plots in 2-D and 3-D.	 64 ..................................................Figure 24: Hierarchy of  decision criteria and alternatives for example AHP problem. 	 67 .............................Figure 25: Generalized fuzzy sets to characterized normalized values that sum up to 1.	 70 ...........................Figure 26: Equations for the completion fuzzy set.	 71 .........................................................................................Figure 27: Heat maps for city comparisons in the AHP example. 	 76 ...............................................................Figure 28: Worldwide distribution of  power generation capacity, per country/GPP type.	 77 ........................Figure 29: Current (2015) and projected (2020) contribution of  small geothermal producers.	 79 ..................Figure 30: Current (2015) and projected world geothermal production levels (2020).	 80 ................................Figure 31: Degrees of  separation between the GPP and ΔT.	 81 ........................................................................Figure 32: The geo-production dataset compilation process. 	 83 .......................................................................Figure 33: Screenshot of  the ThinkGeoEnergy plant map.	 84 ..........................................................................Figure 34: HTML source of  the ThinkGeoEnergy plant map page containing scraped data.	 85 .................Figure 35: Incorrectly positioned GPP in the thinkgeoenergy.com plant map.	 86 ...........................................xFigure 36: Older version of  the thinkgeoenergy.com plant map data, on Google Maps. 	 86 ..........................Figure 37: Screenshot of  geothermal listing of  the Global Energy Observatory site.	 87 .................................Figure 38: Power plant information page.	 88 .......................................................................................................Figure 39: Geothermal power plant listing on OpenEI. 	 89 ...............................................................................Figure 40: Map of  geothermal power plants on OpenEI.	 90 ............................................................................Figure 41: Screenshot of  the Wikipedia page on geothermal power stations.	 91 .............................................Figure 42: Relative accuracy calculation for Latitude values for the 4 source GPP datasets.	 93 ......................Figure 43: Map of  geothermal installations in Iceland. 	 94 ................................................................................Figure 44: Classification of  GPP types based on temperature.	 96 .....................................................................Figure 45: Generalized fuzzy set for completion. 	 96 ...........................................................................................Figure 46: Equations for the completion fuzzy set.	 96 .........................................................................................Figure 47: Total generation capacity per GPP type vs. count of  installation per GPP type. 	 100 ...................Figure 48: Normalized capacity and count per GPP type. “Capacity density” ρ in insert.	 101 ......................Figure 49: Cumulative capacity vs. cumulative count of  geothermal installations by country.	 102 ................Figure 50: Worldwide geothermal generation capacity. 	 104 ..............................................................................Figure 51: Geothermal production in New Zealand. 	 105 .................................................................................Figure 52: Geothermal production in Japan.	 106 ................................................................................................Figure 53: Geothermal production in Iceland.	 107 .............................................................................................Figure 54: Geothermal production in the Philippines and Indonesia.	 108 .......................................................Figure 55: Geothermal production in the US State of  Alaska, and in Russia.	 109 ..........................................Figure 56: Geothermal production in Hawaii, USA.	 109 ..................................................................................Figure 57: Geothermal production in Kenya and Ethiopia.	 110 .......................................................................Figure 58: Geothermal production in Turkey.	 110 ..............................................................................................Figure 59: Geothermal production in Central America.	 111 .............................................................................Figure 60: Geothermal production in the continental USA.	 112 ......................................................................Figure 61: Geo-production in the TVZ - individual GPP units and multi-type GPP complexes. 	 115 ..........Figure 62: GPP distribution by type and generation capacity.	 116 ....................................................................Figure 63: Degree of  separation analysis for the indicators related to geothermal potential.	 120 ...................Figure 64: Simple visual overlap between volcanoes and GPPs worldwide.	 122 ..............................................Figure 65: Volcanoes and GPPs around the world.	 123 ......................................................................................Figure 66: Overlap between volcanoes and flash-steam GPPs.	 125 ...................................................................Figure 67: Overlap between volcanoes and dry-steam GPPs.	 126 .....................................................................Figure 68: Overlap between volcanoes and unconventional GPPs.	 126 ............................................................Figure 69: Overlap between volcanoes and binary GPPs.	 127 ...........................................................................Figure 70: Physical proximity between dry-steam plants and their closest volcano. 	 129 .................................Figure 71: Σ(Capacity) vs. min(GPP-Volcano distance), using single closest volcano to each GPP.	 130 .........Figure 72: Σ(Capacity) vs. min(GPP-Volcano distance), country totals per GPP type.	 131 .............................Figure 73: Volcanoes with combination attributes best matching GPP distribution.	 134 ................................`iFigure 74: Classification of  volcanoes into primary and secondary indicators.	 135 .........................................Figure 75: Combination of  volcano attributes best matching the geographic of  GPPs.	 136 ..........................Figure 76: Nested buffers of  increasing radius around GPPs and their overlap with volcanoes.	 138 .............Figure 77a: Distribution of  earthquakes around GPPs and primary volcanoes (5.0≤M≤5.4).	 139 ................Figure 77b: Distribution of  earthquakes around GPPs and primary volcanoes (5.5≤M≤7.9).	 140 ...............Figure 77c: Distribution of  earthquakes around GPPs and primary volcanoes (8.0≤M≤8.9).	 141 ................Figure 78: Proximity analysis setup between GPPs and volcanoes.	 142 ............................................................Figure 79: Proximity between GPP types and volcano types.	 145 .....................................................................Figure 80: Proximity between GPP types and volcano types, with normalized band widths.	 146 ..................Figure 81: Proximity between geo-production and volcano attributes.	 147 ......................................................Figure 82: Country-based distribution of  GPP types.	 151 ..................................................................................Figure 83: Country-based distribution of  volcano types.	 152 .............................................................................Figure 84: Histogram of  non-numerical values in the original thermal springs dataset.	 153 ..........................Figure 85: Distribution of  numerical and non-numerical values in the original springs data.	 154 .................Figure 86: Histogram of  numerical values in the original thermal springs dataset.	 155 ..................................Figure 87: Data transformation process of  the Thermal Springs dataset.	 157 .................................................Figure 88: The pressure-enthalpy (P-h) diagram for water.	 157 .........................................................................Figure 89: Histogram of  numerical values in classified thermal spring dataset.	 158 ........................................Figure 90: Reclassification of  the thermal springs dataset (continental US subset).	 160 ..................................Figure 91: Actual and reported locations of  the Makushin Volcano Fumaroles. 	 161 .....................................Figure 92: Reclassification of  the thermal springs dataset (continental US subset).	 162 ..................................Figure 93: Simple overlap between quakes, boundaries and GPPs.	 164 ...........................................................Figure 94: Tectonic plate boundaries, earthquakes and GPPs in New Zealand. 	 166 ......................................Figure 95: Tectonic plate boundaries, earthquakes and GPPs in Japan.	 167 ....................................................Figure 96: Tectonic plate boundaries, earthquakes and GPPs in Guadeloupe.	 168 .........................................Figure 97: Plate boundaries, earthquakes and GPPs in Central America.	 169 .................................................Figure 98: Plate boundaries, tectonic settings, and GPPs in the Philippines and Indonesia.	 170 ....................Figure 99: Tectonic plate boundaries, earthquakes and geothermal production in Iceland.	 171 ....................Figure 100: Tectonic boundaries, earthquakes and GPPs in the Portuguese Azores.	 172 ...............................Figure 101: Tectonic boundaries, earthquakes and GPPs in Hawaii, USA.	 172 ..............................................Figure 102: Tectonic boundaries, earthquakes and GPPs in in the East-African Rift Valley.	 173 ...................Figure 103: Tectonic boundaries, earthquakes and GPPs in Western USA.	 174 .............................................Figure 104: Tectonic boundaries, earthquakes and GPPs in Mexico.	 175 ........................................................Figure 105: Worldwide tectonic plate boundaries and GPPs.	 176 .....................................................................Figure 106: Proximity of  different GPP types to different earthquake magnitudes.	 178 ..................................Figure 107: Numerical relationship between proximity and earthquake magnitude.	 179 ...............................Figure 108: Histogram of  heat flow values.	 180 ..................................................................................................Figure 109: Surface heat flow layer; data range 0-100 [mWm-2].	 181 ...............................................................`iiFigure 110: Surface heat flow layer; data range 100-200 [mWm-2].	 182 ..........................................................Figure 111: Surface heat flow layer; data range 200-300 [mWm-2].	 182 ..........................................................Figure 112: Surface heat flow layer; data range 300-400 [mWm-2].	 183 ..........................................................Figure 113: Buffer analysis results for the heat flow dataset, based on the GPP buffers..	 183 ..........................Figure 114: Proximity analysis results for surface heat flow (in mWm-2) and GPPs.	 185 .................................Figure 115: Entity-Relationship diagram for the MRDS dataset. 	 187 .............................................................Figure 116: Transformation of  a multi-value row into single-value rows for the MRDS dataset.	 190 ...........Figure 117: Transforming multi-value fields into a single value fields, and filtering.	 190 .................................Figure 118: Statistical analysis of  the MRDS data based on size/stage combo.	 193 .......................................Figure 119: Filtered subset of  MRDS data contributing to the mineral potential data layer.	 194 ..................Figure 120: a) Loose; b) strict; and c) compromise filters for the MRDS data.	 195 ..........................................Figure 121: The IntelligenceMine front-end interface, showing the generate/download buttons.	 197 ..........Figure 122: Entity-relationship diagram for the fields of  greatest interest in the InfoMine dataset.	 198 ........Figure 123: Common development stage naming scheme for InfoMine & MRDS.	 199 ................................Figure 124: Row count comparison per development stage for InfoMine & MRDS. 	 199 .............................Figure 125a: Project counts through the mineral development life cycle stages. 	 201 ......................................Figure 125b: Project counts through the mineral development life cycle stages.	 202 .......................................Figure 126: Cumulative counts of  projects progressing through the various life cycle stages.	 203 ..................Figure 127: Size distribution for commodity and economic class for the InfoMine Dataset.	 209 ...................Figure 128: Sizing the InfoMine properties - base metals.	 210 ...........................................................................Figure 129: Sizing the InfoMine properties - industrial minerals, non-metals, or metalloids.	 211 ..................Figure 130: Sizing the InfoMine properties - gemstones & precious metals.	 212 ..............................................Figure 130: Sizing the InfoMine properties - gemstones & precious metals.	 212 ..............................................Figure 131: Sizing the InfoMine properties - energy metals & critical rare earth metals.	 213 .........................Figure 132: Statistical analysis of  the InfoMine data based on size/stage/state combo.	 214 ..........................Figure 133: InfoMine filtered subset contributing to the mineral potential data layer.	 215 ..............................Figure 134: Resulting coverage of  a: a) strict; b) loose; and c) compromise filter for InfoMine.	 216 ...............Figure 135: AHP matrix for the a) MRDS; and b) InfoMine datasets.	 218 ......................................................Figure 136: AHP matrix for the a) MRDS; and b) InfoMine datasets.	 219 ......................................................Figure 137: Reclassification/calculation of  combined wstate/stage. 	 220 ...............................................................Figure 138: Index value distribution and listing order weights for MRDS and InfoMine.	 220 .......................Figure 139: Pairwise comparison matrices for the relative importance of  mineral indicators. 	 222 ................Figure 140: Mineral potential weights (wmineral) per commodity and economic mineral class.	 222 .................Figure 141: Relative geographical scopes and point overlap of  the MRDS & InfoMine datasets. 	 223 .........Figure 144: Volcano types most closely associated to geo-production. 	 229 ......................................................Figure 145: Percent of  GPPs closest to a volcano, per buffer size.	 230 ..............................................................Figure 146: Buffer analysis for earthquakes (summary) — counts. 	 232 ............................................................Figure 147: Selecting appropriate buffer size for tectonic boundaries analysis.	 233 .........................................`iiQFigure 148: Selecting appropriate buffer size for tectonic boundaries analysis.	 234 .........................................Figure 149a: Quake count for tectonic plate boundaries — for OSR, CCB, CTF, and OCB. 	 235 ..............Figure 149b: Quake count for tectonic plate boundaries — for OTF, SUB, CRB and ALL.	 236 .................Figure 150: Mineral lease shape versus actual extent of  Kennecott/Bingham Canyon mine	 239 .................Figure 151: Surface footprint as a function of  total production, sample gold mines.	 241 ................................Figure 152: Tracing the pit surface areas of  sample US gold mines (part 1).	 241 .............................................Figure 153: Tracing the pit surface areas of  sample US gold mines (part 2).	 242 .............................................Figure 154: Histogram of  gold production totals for the InfoMine dataset.	 244 ..............................................Figure 155: Histogram of  traced surface areas for sample gold properties.	 244 ...............................................Figure 156: Filtering based maximum search radius - buffer distance.	 246 .......................................................Figure 157: Near analysis table (for volcanoes) with extended indicator attributes. 	 247 ..................................Figure 158: Geomine weights (wgeomine) per commodity and economic mineral class.	 249 .............................Figure 159: Geomine indicator potential (wgeomine) for precious metals.	 250 .....................................................Figure 160: Buffer creation around all geothermal and mineral indicator data points. 	 252 ...........................Figure 161: Indicator buffers and fishnet grid used in the calculation of  geomine potential.	 253 ...................Figure 162: Clipping the extent of  the rectangular grid.	 254 .............................................................................Figure 163: Combining potential into grid-based layers using individual geomine points.	 255 ......................Figure 164: Geographic extent of  geomine potential for all indicators.	 256 .....................................................Figure 165: Geographic extent of  geomine potential for a) base; and b) precious metals. 	 257 ......................Figure 166: Geographic extent of  geomine potential for a) energy; and b) industrial minerals. 	 258 .............Figure 167: Geographic extent of  geomine potential for a) gold; and b) copper.	 259 ......................................Figure 168: Geographic extent of  geomine potential for a) lead; and b) zinc.	 260 ...........................................Figure 169: Geographic extent of  geomine potential for silver.	 261 ..................................................................Figure 170: World map of  wgeomine, redrawn using a 1/2-std. dev. class break interval. 	 262 ..........................Figure 171: wgeomine for the Western US, redrawn using a 1/2-std. dev. class break interval. 	 263 ..................Figure 172: wgeomine for South America, redrawn using a 1/2-std. dev. class break interval. 	 264 ...................Figure 173: Commodity contributions to wgeomine in a) NV, CA, and AZ, & b) in Chile.	 265 .........................Figure 174: Mineral indicators for NV, CA, and AZ for a) gold; and b) copper, sized by wmine.	 267 ..............Figure 175: a) Geo-indicators in NV, USA; and b) corresponding buffers, by type. 	 268 .................................Figure 176: a) Mineral indicators (gold) in NV, USA; and b) corresponding buffers. 	 269 ...............................Figure 177: 5 × 5 km wgeomine grid in NV, CA & AZ, a) for gold; and b) for copper.	 270 .................................Figure 178: 25 km2 geomine wgeomine grids in Chile, for a) wgeo; b) wmine; & c) wgeomine.	 272 ............................Figure 179: wgeomine recommendation map for gold and copper in CA, NV, AZ.	 274 ....................................Figure 180: Recommendation map for Chile, based on the 25 km2 geomine wgeomine grid.	 275 ....................Figure 181: Approach overview for the calculation of  geomine potential. 	 279 ...............................................`QvAcknowledgements My deepest gratitude goes out to my research supervisor, Dr. Dirk Van Zyl, for his interminable patience, unwavering support, and boundless compassion. For stepping in when needed, standing by me through the darkest of  times, and consistently treating me like a human being. For providing a safe space where I was allowed to confide in and disagree with him, converse and debate, and ultimately stay the course. Not only for believing in me, but more importantly, for being a light in the darkness, someone I aspire to emulate. None of  this would have been possible without him.  I would especially like to express my gratitude to my co-supervisor, Dr. Sadiq Zarrouk, who has very graciously and very patiently stuck with me through this journey. For actually reviewing the material and for providing invaluable feedback, advice, and direction. And for being the wonderful and inspirational teacher (“the tough one I’ll remember”) that instilled in me a love for geothermal during my short exchange tenure at the University of  Auckland, in New Zealand. What I know about geothermal energy, I learned from him.  Many thanks are also due to Dr. Bern Klein, Dr. Marek Pawlik, and Dr. Scott Dunbar for seeing me through this long and arduous process. I am also indebted to Leslie Nichols, Maria Lui, Mac McLachlan, and Sally Finora for their friendship and support throughout these years.  Lastly, I would like to acknowledge InfoMine Inc. for very graciously granting access to and allowing for the use of  their mineral production data.   xvDedication Στον πατέρα μου, που έφυγε.  Στη μάνα μου, που μ’αντέχει.  Στον αδελφό μου, που αγαπάει και παιδεύει.  Στην Άντα, και στον Γιώργο.  Και στους λίγους άλλους, τους εκλεκτούς, που ως εκ θαύματος, δηλώνουν ακόμη παρόντες.   «…τοὺς δὲ φίλους ἐν ταῖς ἀτυχίαις διαγιγνώσκομεν…»  — Ισοκράτης Οι υπόλοιποι: Άγαμοι Θύται.  xviChapter 1: Introduction Mines are heavy energy consumers. They require energy to locate, extract, process and transport mineral resources to market. They require energy to sustain operations, provide safety and security, mediate environmental risks and close a project at the end of  its life cycle (Newfoundland and Labrador 2012). The amount and type of  energy a mine ultimately consumes depends on multiple factors, such as its location, the specific mineral being produced, the life cycle stage it is in, as well as the availability and cost of  energy sources at its disposal. Ideally, energy is purchased from the grid, but when that is not possible (e.g. in the case of  remote mines), alternative solutions must be sought. Energy is typically sourced from traditional fossil fuels but lately mines have began incorporating alternative and renewable sources to their energy consumption mix. Geothermal energy is produced by converting part of  the thermal energy contained in hot fluids (harvested via drilled wells from deep underground reservoirs), by passing the steam portion of  these fluids through a steam turbine. Geothermal is clean, locally produced and renewable. It has very high capacity factors (85-90%), very low environmental footprint and greenhouse gas (GHG) emissions, and it can generate base load power, which is of  particular importance to the typically round-the-clock mining operations (Kagel and Gawell 2005, Li et al. 2015). Although both geothermal and mining projects share similar development cycles, the mining industry has been a slow adopter of  geothermal energy in mining operations. Part of  the reason for the slow uptake is geothermal’s relatively longer development life cycles and scarcer availability compared to other alternatives. Another reason is a lack of  understanding for the resource by mineral developers, as for the most part the two industries do not overlap operationally, even in geographical regions of  co-occurrence. For decision-makers tasked with the responsibility of  selecting appropriate energy sources for mineral projects, it is imperative to have a clear understanding of  the availability, characteristics, strengths and weakness of  an energy source to be able to confidently  1decide upon its merits on whether or not it would make a suitable alternative for a given project; a decision which is, generally speaking, site-specific.  This research aims to provide a decision-making methodology for an initial assessment of  the combined geothermal and mineral potential of  a given area, at a lower cost and risk than traditional exploration and resource assessment methods. The developed methodology provides an estimate for the existence/coexistence of  geothermal and/or mineral resources, even in areas of  zero geothermal and/or mineral production information. This is achieved using a combination of  geographical analysis, statistical analysis, and decision-making techniques to map, analyze, and calculate the coincidence of  features known to be associated to either geothermal or mineral potential.  This estimate is based on publicly available, non-specialist data, that are more accessible to decision makers without sector-specific specialist knowledge in, for example, geochemistry or geophysics. The intent is to demonstrate that decision-makers charged with the task of  energy-source selection for a mining project could make a reasonable assessment on the viability of  integrating geothermal in mining, without having to resort to the analysis of  more complex, specialist data. This document is structured as follows: Chapter 2 introduces the concepts underlying the problem under investigation; Chapter 3 introduces the research questions, lists assumptions, and describes of  the overall research approach; Chapter 4 describes the process used in creating the geothermal production layer used in the geomine model; Chapter 5 focuses on the evaluation of  geothermal potential and the identification and mapping of  high-geo-potential areas from around the world; Chapter 6 presents the mineral development development indicator layers, which formed the basis of  the mineral potential calculations; Chapter 7 focuses on the steps taken to finalize the mineral and geothermal potential datasets; Chapter 8 presents the process of  calculating and mapping of  the resulting values of  geomine potential; and finally, Chapter 9 summarizes outcomes, lists contributions to knowledge, discusses limitations, and gives recommendations for further work.  2Chapter 2: Problem Background & Literature Review This chapter introduces the concepts underlying the problem under investigation, and provides information necessary for the subsequent development of  the problem statement and research questions. Topics covered include an overview of  geothermal energy and mineral production, a description of  energy needs of  mining projects and a discussion of  the intricacies involved in selecting an energy source for a particular site. Unless otherwise attribute, figures and diagrams are original, specifically produced by the author as part of  this research.  Geothermal energy production can be defined as the process through which economically valuable hot geothermal fluids are identified, located, extracted, and processed for their subsequent use in electricity generation or in direct non-electric applications (Dickson and Fanelli 2013). Electrical energy can be generated from hot fluids found in geothermal reservoirs deep within the Earth’s crust. Drilled wells are constructed to tap into these reservoirs and provide a pathway to the surface, at which point the steam portion of  the produced fluid is harvested and passed on to steam turbines, thereby converting part of  the steam’s thermal energy content to electricity.  The amount of  thermal energy that can be converted to electricity is largely a function of  the mass flow rate of  the fluid passing through the wellhead and the fluIDs pressure, temperature and enthalpy content. There are 3 main types of  geothermal power stations: dry-steam units that pass harvested steam directly through a steam turbine; flash-steam units that use separated mixtures of  liquid and vapour brines; and binary units that utilize working fluids with boiling points lower than that of  water (Dickson and Fanelli 2013).  Constructing a geothermal power station is typically a lengthy and very expensive process, taking an average of  7 years and millions of  dollars in investment to complete (Flóvenz 2012). The actual cost depends on the type of  power station being constructed and the achievable mass flow rates and temperatures of  the harvested fluids (Flóvenz 2012). Geothermal electricity has been  3produced commercially since the mid-20th century and the technology is well known and proven. Currently, high-capacity geothermal electricity production is situated in countries that are adjacent to or contain areas of  high volcanism, with USA, Indonesia and the Philippines making the top of  the producers list (Bertani 2015). Unlike mineral development, where the extraction of  minerals eventually depletes the resource and at the end it has to be abandoned, geothermal reservoirs have the ability to return to their natural state (i.e. their state before production), if  left un-produced for a certain period of  time. Geothermal power stations are also typically very-low- footprint installations, hence they carry much smaller reclamation requirements at the time of  closure, compared to mineral extraction projects (Flóvenz 2012).  The life cycle of  a modern mine can be divided in 5 distinct stages : 1) Prospecting, which is the 1search for ores and other minerals; 2) Exploration, i.e. the process of  determining as accurately as possible the size and value of  a mineral deposit; 3) Development, which consists of  environmental permitting, designing, financing and opening a mineral deposit for exploitation, either by stripping the overburden or by underground mining; 4) Exploitation or Operations, i.e. the actual recovery of  minerals from the earth in quantity; and finally 5) Closure and Reclamation, which includes closing a mine and re-contouring, revegetating, and restoring its water and land values. Reclamation is a relatively modern (1970 and onwards) addition to the mine life cycle, brought about by the increasing demands of  society for a cleaner environment and the stricter laws regulating future uses of  mine lands (Hartman and Mutmansky 2002). It follows an overall approach around the concept of  achieving “development that meets the needs of  the present without compromising the ability of  future generations to meet their own needs” (Brundtland Commission 1987).  The effort, time and investment required for mining project development are comparable to that of  geothermal. By and large, cost is constrained significantly though the first two phases of   Part of  this chapter has been published in (Patsa, Van Zyl, Zarrouk, & Arianpoo, 2015). 1 4development (Exploration and Advanced Exploration), until the projects has shown enough promise to pass onto Development and Construction (Flóvenz 2012). In fact, only “2 out of  every 200 projects that reaches the discovery stage moves to development. This is equivalent to about 1 out of  every 10,000 grassroots exploration projects” (Canada 2007). In general though, mineral development, just like geothermal, is heavily case-specific.  2.1 Energy use in mining Mine energy use is very site-specific, as it depends on the mine’s location, the type of  mineral resource in production and the extraction and mineral processes involved. Energy is purchased directly as electricity from the grid (if  possible), as fuel to run combustion-based equipment (e.g. transport vehicles and ore processing equipment), and for environmental control through the heating and cooling of  water and air (e.g. for mine ventilation, ore treatment, and space heating/cooling).  Different mine life cycle stages require different types of  energy input but generally, development and operations are the most energy-demanding life cycle stages. Canadian mines have in the past used diesel, gasoline, natural gas, explosives, light fuel oil, and Bunker C fuel oil to conduct operations, including (but not limited to) drilling, blasting, mucking, loading, excavating, underground and over-ground ore transport, crushing, grinding, hoisting, ventilation, dewatering, separation, floatation, space heating and lighting, and tailings disposal. This is on top of  the more general heating, lighting and other needs of  facilities and camps (Canada 2005a; Canada 2005b). Off-grid mines that are not connected to an electric power transmission and distribution network rely heavily on conventional sources of  energy, such as fossil fuels to meet their needs. Northern mines are particularly susceptible, as they are typically remote and cannot rely on future transmission grid expansions. Their total dependence on fossil fuels often comes with a high cost and risk. For example, the Diavik Diamond Mine lies about 300 km NE of  Yellowknife, in the Canadian  5Northern Territories, and it is home to 1,165 workers. Diavik’s remoteness and operating temperatures at arctic conditions (-30°C) make power security a critical safety and economic priority. Up to September 2012, the mine relied exclusively on 70 million litres of  diesel, shipped over ice roads during a 6-week winter road season. Energy costs constituted more than 25% of  the cost of  mine operations, which were conducted under the increasing threat of  climate change (directly affecting the short refuelling time window), very high fuel pricing, and fuel price volatility (Van Wyk 2013). Cost and risk reduction are top priorities for any mining operation and mining companies are taking a long hard look at alternative, renewable energy options, in an effort to shift away from a total dependence on fossil fuels. For Diavik Mine, the alternative solution most fitting to their needs and local resources was in the form of  a wind farm. Operational since September 2012, the four 2.3 MWe wind turbines are expected to provide 17 GWh of  renewable energy per year, corresponding to a 10% reduction in diesel consumption and a 6% reduction in GHG emissions (Van Wyk 2013). Much further to the south, Codelco, the largest copper producer in Chile, has a concentrated demand for process heat. The Gabriela Mistral Division mine relies exclusively on diesel fuel to heat water for its copper electro-winning process (in the 50-60°C range), which yields a final product that is 99.999% pure. By replacing 85% of  diesel consumption with 51,800 GWh of  thermosolar, Codelco is expecting annual energy savings equivalent to the cost of  almost two months of  fuel (Judd 2013b; Judd 2013a). Newmont’s projects in Ghana are almost exclusively hydro-powered and are proving more cost-effective than gas-fired or diesel plants. The same developer uses biodiesel at its Peruvian mines, and to run trucks and cut costs in Nevada (CleanEnergyBC 2013). The emerging shift towards alternative energy sources is not surprising; most are clean, locally produced and renewable. Geothermal energy offers the additional advantages of  very high capacity factors (85-90%), very low environmental footprint (in terms of  land use and GHG emissions), and the  6ability to generate base load power, a characteristic of  particular importance to the, typically, round-the-clock mining operations (Kagel and Gawell 2005; Li et al. 2015). 2.2 Mines, energy and decision-making In order to review how geothermal energy relates to mining operations, it is important to understand how mining companies make decisions on energy issues.  2.2.1 Operating costs Mines operate to make profit, which (in unchanging commodity pricing conditions) can be achieved either by reducing expenses or by improving production efficiency (Treadgold 2015). With sustained commodity price drops, international competition and low profit margins, operating costs can have a significant effect on a mine’s ability to survive market volatility. Sudden, steep drops in commodity prices have led to mine closures, as further production is deemed uneconomical. Mining operators are therefore constantly on the outlook for even marginal efficiency increases and cost reductions.  2.2.2 Access, safety, and power supply security Remoteness, physical ease and safety of  site access, connectivity to the electricity transmission and distribution network, availability and price of  alternative fuel options, and fuel transportation costs are some additional risk factors affecting fuel source selection. Fossil fuels derived from regions of  political turmoil and strongly controlled pricing by oil cartels can result in supply chain interruptions. In a world of  political and environmental uncertainty, short-term energy-supply security and long-term energy independence are of  paramount importance, which cannot be achieved without access to locally-produced, predictable and reliable energy (KPMG 2001). 
 72.2.3 Environmental requirements Modern legislation dictates that resource extraction must be undertaken in a manner that does not interfere with the integrity of  the environment, in terms of  pollution control, remediation and resource conservation. The potential environmental impact from mining activity includes erosion, soil, groundwater and surface water contamination, loss of  biodiversity, destruction/disturbance of  ecosystems and habitats, GHG emissions, and large-scale land stripping (Azcue 2012). Adherence to environmental laws and regulations is seldom optional in developed countries. For example, Taseko's Prosperity open-pit gold and copper mine project near Canada’s Fish Lake, B.C., has been rejected twice by the Federal Ministry of  Environment. An independent environmental review panel, fully supported by the local Tsilhqot'in Nation leadership, deduced that the project was likely to cause irreversible environmental damage to the Fish Lake water supply – a failed attempt that cost Taseko $110 million in prefeasibility costs. This kind of  political pressure from environmental advocates and aboriginal groups helps to further underline the mining industry's need to control pollution and shift towards cleaner operations (Klein 2011; CBC News 2013; CBC News 2014). 2.3 Geothermal resources Geothermal energy is generated from hot fluids harvested from deep within the earth’s crust. It is energy that is primarily carried to the surface by water (the carrier), which is then either used directly in heating applications, or put through a steam cycle to produce electricity. A geothermal resource is defined as the combination of  specific conditions that are present in a particular underground location, namely, heat and water in levels/amounts adequate to support production, and permeable enough (broken up or porous) rocks that allow the hot water to flow to the surface in quantities adequate for production (Dickson and Fanelli 2013). These three requirements of  water,  8heat, and permeability are absolute. The hot rocks must also be at shallow enough depths that can be reachable with currently drilling technology. This is typically the case in areas of  tectonic boundaries and active volcanism, specifically in zones of  subduction or around transform faults, or above mid-plate plumes (Sanyal 2010). The most prominent of  these is the circum-Pacific Ring of  Fire, which hosts the majority of  geothermal resources; about 10,000 MWe of  geothermal power capacity has been installed almost exclusively in this region (Sanyal 2010).  Geothermal, petroleum, and mineral resources are to an extent similar in that they are all extracted from different depths below the Earth’s surface. Petroleum and geothermal resources are both mostly handled as liquid carriers of  energy (as well as dispersed solids or entrained steam) and they are typically accessed through the drilling of  wells that are designed to tap into particular target formations, termed reservoirs. Their differences dictate the manner in which they are extracted, transported and utilized in the production of  energy or mineral/petroleum (by-) products. The main difference between geothermal and mineral/petroleum resources is that geothermal resources in convective hydrothermal systems are renewable through recharge (although typically at a slower rate than the rate in which energy is extracted in commercial installations); this rate actually varies significantly from site to site (Lawless 2010). Heat, the direct energy product of  geothermal production, must also be put to use in-situ and unlike fossil fuels, it is generally not transportable. The largest part of  geothermal heat is mostly derived through the decay of  radioactive elements in the crust (O’Sullivan 2014). A smaller part is sourced through the convection of  magma within the mantle, while an even smaller contribution is the heat carried from the core through conduction. The heat is eventually emitted as an upward flux through the entire planetary surface. The value of  this flux averages to about 65 mW/m2 worldwide; this corresponds to a temperature gradient of  30℃/km. In areas that host geothermal resources — for example around the tectonic plate boundaries — heat flow (and corresponding temperature gradient) values are much larger (O’Sullivan 2014).  9The term geothermal reservoir denotes a hydrothermal system that comprises of  a zone of  permeable rock that is shallow enough to be reached by drilling, and contains water that is hot enough and in adequate quantities to be suitable for production. Geothermal reservoirs are classified based on the phase of  the geothermal fluids at depth, the amount of  heat flow they contain, the manner in which this heat flows to the surface, and the geological setting they are associated with. On the lower end of  the scale, geothermal reservoirs are warm water systems with very little fluid movement and no boiling, while on the higher end of  the scale, geothermal reservoirs are convective two-phase systems that boil at depth. 2.4 Mineral resources Mining or mineral resource extraction is the process through which economically valuable mineral resources, namely: metals (including: ferrous metals, such as iron, manganese and tungsten; base metals, such as copper, lead and zinc; and precious metals, such as gold, silver and platinum), non-metallic/industrial minerals (i.e. non-fuel mineral ores that are not associated with the production of  metals, such as phosphate, limestone, and sulphur); or energy minerals (e.g. fossil fuels such as coal, petroleum, and natural gas, and uranium), are identified, located, extracted, and processed, for their subsequent use in consumer products (Hartman and Mutmansky 2002).  Most rocks and mineral deposits are formed in one of  three different ways, namely through igneous activity (i.e. the solidification of  magma), through metamorphism (i.e. the transformation of  one type of  rock into another), or through sedimentary processes (i.e. the accumulation of  loose particulate matter that is compacted and cemented into rock in the presence of  heat, pressure, and chemical agents). Some minerals can also be deposited as veins (formed when fluids that contain them in suspension migrate through fissures in rocks and deposit minerals onto the their walls) (Bangert 2016).
 102.4.1 Metals and minerals related to volcanism Generally speaking, high-temperature geothermal reservoirs are associated with high concentrations of  hydrothermal alteration minerals such as gold, copper and silver. This is due to the tendency of  precious metals to precipitate and deposit in response to boiling and mixing of  deep geothermal fluids (Simmons and Browne 2000). Due to the variety in the formation process of  different minerals, rocks, and metals, some are expected to be more closely associated with high-temperature geothermal production than others. Of  particular interest to this study are minerals that form through the process of  hydrothermal alteration, namely hydrothermal deposits such as gold, silver, copper, zinc, lead, molybdenum, tin, tungsten, and mercury. 2.5 Geothermal production Geothermal energy production can be defined as the process through which economically valuable hot geothermal fluids are identified, located, extracted and processed, for their subsequent use in electricity generation, or in direct, non-electric applications (Dickson and Fanelli 2013). In a way, the extraction of  geothermal fluid can be viewed simply as another mining project, with heat replacing mineral fuels as the economically valuable resource. In geothermal production, it is the fluIDs heat content (expressed as enthalpy), accessibility (in terms of  practically drill-able depths), and extractability (in terms of  achievable mass flow rates from a given well) that determine whether a geothermal resource is economically valuable and warrants production (DiPippo 2012). 2.5.1 Power production basics Electrical energy can be generated from hot fluids found in geothermal reservoirs deep within the Earth’s crust. Drilled wells are constructed to tap into these reservoirs and provide a pathway to the surface, at which point the steam portion of  the produced fluid is harvested and passed on to  11steam turbines, thereby converting part of  the steam’s thermal energy content to electricity. The amount of  thermal energy that can be converted to electricity is largely a function of  the quality and quantity of  the produced fluid. The term quality encapsulates three fluid characteristics, namely pressure, temperature and enthalpy. Quantity refers to the mass flow rate of  the fluid passing through the wellhead. As power potential is directly proportional to enthalpy and mass flow rate (Eq. 1) (Zarrouk and Moon 2014), more energy can be extracted from the same amount of  a higher enthalpy fluid; in fact, producing higher thermal energy fluids results in higher and financially more attractive power outputs from a given well. Assuming an ideal separator and turbine (i.e. with respective efficiencies η=1.0), the power produced from a given well, measured in MWe, can be roughly calculated as:  	 	 	 	 	 	 (1) Here,               is the power (i.e. the rate at which energy is being produced) output of  the turbine (in kWe),           is the steam fraction of  the fluid leaving the well and entering the turbine,            is the mass flow rate of  the geothermal fluids flowing through the wellhead (in kg/s), and Δh is the difference in enthalpy between the fluid entering the turbine and the cooling reservoir on the other side of  the cycle (e.g. ambient conditions, or the temperature of  the condenser, depending on system design). Enthalpy is the measure of  energy contained in water or steam (kJ/kg). 2.5.2 Geothermal power stations A typical geothermal power station (e.g. one that taps into a hydrothermal, liquid-dominated reservoir) consists of  a production well that allows the hot fluids to be pumped to the surface; a separator that takes in the geothermal brine and separates the steam from the liquid content; a powerhouse that includes the turbine used to generate electricity; a condenser that completes the cycle; and a cooling tower that serves as the cold reservoir needed for the completion of  the  12pelectrical = xsteam × !msteam × Δhpelectricalxsteam!msteamthermodynamic cycle. In most cases, an injection well that allows for the geothermal fluids to be rejected at suitable locations back into the reservoir is also present (Figure 1). The main difference between dry-steam and flash-steam plants is that the fluids coming out of  the well in dry-steam conditions are in a vapour phase and therefore do not require separation, so they are fed into the turbine more or less directly. In flash-steam plants, separation is necessary. Electricity from lower temperature resources can be produced by passing lower-enthalpy geo-fluids through a binary cycle that uses binary fluids with boiling points lower than that of  water. Figure 1: Example schematic of  a geothermal (flash-steam) power station.  132.5.3 The geothermal development cycle The construction of  a geothermal power station is not a small endeavour, requiring a considerable investment in terms of  time, money, and expertise (Flóvenz 2012). Florenz (2012) lists the typical development stages of  the geothermal power station as follows: Step 1: 	 Gathering and evaluation of  existing data  Step 2A:	 Surface exploration Step 2B:	 Exploration drilling Step 3:	 Pre-feasibility report Step 4:	 Environmental assessment Step 5:	 Drilling and testing of  confirmation wells  Step 6:	 Feasibility report Step 7:	 Design, construction, production drilling.  Step 8:	 Testing, commissioning, training Step 9:	 Operation In this example, development starts with data gathering and evaluation, and includes a preliminary resource assessment. It is followed by surface exploration and exploration drilling which allows for sampling and for attaining a better understanding of  the reservoir. At this stage, the question of  whether a particular reservoir can sustain production is explored. If  the answer is not a definite no, a project moves on to the next stage, the preliminary economic assessment or the prefeasibility study and the environmental assessment. These are further supported by additional drilling and testing of  confirmation wells. Finally at the feasibility study stage, a more detailed investigation is done to determine the probability that a geothermal installation can be successful. This is in terms of  both technical challenges (i.e. whether the system can be built and be sustained by  14the underground resource) but also in terms of  economics, environmental concerns, and legislation. Following the production of  a feasibility report, a project will typically enter the construction and development stage, which includes well-drilling, and building the steam pipe network and the power house itself. Further testing is needed once the system has been constructed and before it is commissioned. The final step that typically precedes the commencement of  operations is personnel training (Figure 2). Figure 2: Project development timeframe for a geothermal power plant. Adapted from (Flóvenz, 2012). 2.5.4 The renewability of  geothermal resources Up to this point in the process, the power station has not produced a single megawatt of  electricity. This happens when the project reaches the operation stage, which for geothermal depends on the resource and it can be as long as 50 years, depending on the resource. Unlike mineral development, where the extraction of  the mineral consequently depletes the resource that at the end of  operations has to be abandoned, geothermal has the ability to return to its natural state (i.e. its state before production), if  left un-produced for certain period of  time. This is thought to be 1.5-4 times the duration of  the production period (Rybach 2007). A geothermal power station also has a very low footprint hence much smaller reclamation requirements at closure, which may or may not be temporary (Flóvenz 2012). 
 152.5.5 Geothermal electricity costs Constructing a geothermal power station is therefore a lengthy and very expensive process. The cost will depend on the type of  power station being constructed; for example, binary systems are on the lowest end of  the enthalpy/production-capacity scale and actually have the highest cost per megawatt to install (Figure 3). Overall, project costs are typically kept at a minimum for the first few years of  exploration and feasibility, when the risk of  not completing a project is at its highest (Figure 4) (Flóvenz 2012). Table 1 lists the predicted costs per development stage for a sample geothermalproduction project — this one in Kenya. The most costly stage by far is Construction and Production Drilling. Figure 5 charts the cost analysis data listed in Table 1. The US Department of  Energy indicates that power is sold between $0.030 and $0.035 per kWh at the Geysers Power Plant in CA, USA, and that “a power plant built today would probably require about $0.05 per kWh” with “some plants” charging “more during peak demand periods” (US Department of  Energy 2017). Figure 3: Total cost per energy unit, Iceland estimates.  As given by (Flóvenz 2012).
 16Table 1: Sample project cost estimate for a 70 MWe geothermal power plant in Kenya. Adapted from (Flóvenz, 2012). Figure 4: Project development timeframe for a geothermal power plant. Adapted from (Flóvenz 2012).  17Figure 5: Geothermal exploration and development: Cost per development stage. Based on data provided by (Flóvenz 2012). The US Department of  Energy further states that development costs for a geothermal power plant are “heavily weighted toward early expenses, rather than fuel to keep them running”. This makes sense, as the fuel running the turbines is provided from the wells once they are built and connected to the surface-side steam pipe network (US Department of  Energy 2017).  More specific cost estimates can be complex, as geothermal power costs (both operating and capital) are highly case-specific, and are affected by numerous variables including but not limited to: the temperature and achievable flow rates from the wells; the chemical composition of  the brine and whatever requirements this may impose on the surface-side brine treatment, both before and after the turbine stage; the allowable condensate/waste brine disposal methods; the existence or need for an electricity transmission network, the current market price of  electricity the eventual electricity  18selling price per kWh; the availability and cost of  alternative energy sources; the availability of  local expertise/workforce; and the accessibility and cost of  building materials. The US Department of  Energy estimates that “the initial cost for the field and power plant is around $2,500 per installed kWe in the U.S., probably $3,000 to $5,000/kWe for a small (<1MWe) power plant. Operating and maintenance costs range from $0.01 to $0.03 per kWh. Most geothermal power plants can run at greater than 90% availability (i.e., producing more than 90% of  the time), but running at 97% or 98% can increase maintenance costs. Higher-priced electricity justifies running the plant 98% of  the time because the resulting higher maintenance costs are recovered” (US Department of  Energy 2017). 2.5.6 Geothermal production around the world Geothermal electricity has been produced commercially since the mid twentieth century, and the technology used to produce it is well known and proven. Figure 6 shows the amount of  geothermal produced around the world, summarized on a country level. On first inspection, most countries with high generation capacity are unsurprisingly adjacent to or contain areas of  high volcanism. According to Bertani (2015), official production figures place the USA, Indonesia and the Philippines at the top of  the producers list — this according to . The production levels in Bertani (2015) are reported on a country level, not on a project basis. They have been classified based on the total amount of  generation capacity into either small (S), medium (M), or large (L). This was done using the percentile approach: if  a country produced more than the 70% of  the largest reported production, it was classified as large; if  it produced less than the 30% of  the largest reported production, it was was classified as small; in-between values were classified as medium (Figure 7). Based on his classification, current production levels (2015) show that 27% of  counties are small producers (holding 0.21% of  worldwide capacity), 31% are large (holding 85.92% of  worldwide capacity) and 42% are medium (holding 13.87% of  worldwide capacity). 
 19
20Figure 6: Worldwide geothermal generation capacity (2015), per country, in MWe total. Based on data presented in (Bertani 2015).Figure 7: Classification and basic stats for current and projected geothermal production levels. Based on data presented in (Bertani 2015). The trend is very interesting. Even though the small, medium, and large country-producer groups are more or less equal in number, the large group generates the vast majority of  megawatts: almost 86% of  the total worldwide capacity. By contrast, the small group generates only a tiny fraction of  the world’s generating capacity, namely 0.21%. Bertani (2015) also reported in the same paper a projection of  geothermal production for the year 2020. Based on this projection, 25 countries that did not produce any geothermal electricity in 2015 but that are known to contain geothermal resources, will enter the market. Most new additions are small in size with the exception of  Chile, which is expected to enter the medium-sized production group (Figures 8 and 9).  As new countries are added to the list, or as existing countries increase their production levels by 2020, their rank with respect to their generation capacity changes. The overall trend (i.e. accounting for existing production as well as new development) is pretty constant. Obviously, when viewing the new development in isolation (Figure 10), the trend will be positive by definition. Within the context of  this analysis, it was important to identify current and future electricity producers (on a country level), as a first step to the analysis of  the potential for integration between geothermal and mining. Of  special interest are countries who are projected to rise in the rankings, as these are expected to take the most favourable stance towards development.  small medium large30% 70% 2122Figure 8: Geo-producers by size, number of  countries vs. total generating capacity (2015). Based on data presented in (Bertani 2015).Figure 9: Percent change in geothermal production by country producer geothermal electricity.  Difference between 2015 (current, confirmed) and 2020 (projected) levels.  The chart was constructed using the data reported by (Bertani 2015).
 23Figure 10: Percent change in geothermal production for new market entries.  Difference between 2015 (current, confirmed) and 2020 (projected) levels.  The chart was constructed using the data reported by (Bertani 2015).
 242.6 Mineral production The development chronology and investment requirements of  mining resource production projects are comparable to those of  geothermal resource production. Operating mines are developed in steps or stages, generally divided into 5 groups, namely:  1. Prospecting, Exploration and Advanced Exploration2. Preliminary Economic Assessment, Prefeasibility, and Feasibility3. Planning and Construction4. Operation5. Reclamation, Rehabilitation and ClosureThe time and monetary investment required in each of  these 5 stages also varies. The schematic in Figure 11 visually compares the relative duration of  each stage for a typical mineral development project. Figure 11: Typical duration per development stage of  a mining project. Based on data presented in (Canada 2007).   25Figure 12 also provides an indication of  the kinds of  costs associated with each development phase (Canada 2007). By and large, cost is constrained significantly though the first two phases of  development (Exploration and Advanced Exploration), until the projects has shown enough promise to pass to the Development and Construction stage. This seems logical considering the probability any given project has to make it all the way to Development and Operation (Canada 2016). It is important to note that this is only indicative; mineral development, just like geothermal development is heavily case-specific. Figure 12: Typical costs associated with mining development stages. Based on data in (Canada 2016). 
& Operation$M 262.7 Geothermal energy use in mineral production Central to the argument that geothermal is a highly attractive energy option for mining operators are the many characteristics, resources and processes that are common in both industries.  2.7.1 Mining, heat, and water use  Heat and water are cardinal elements of  geothermal systems. They also play a key role in mining and mineral extraction. As a transport medium, water mixes with crushed ore to produce ore slurry that can be piped through the processing plant. As an excess by-product in pits and underground tunnels however, it can disrupt access to the mine workings, and must therefore be removed (ICMM 2012). Heat addition to specific mineral processes can significantly enhance yields and production efficiencies. Conversely, excessive heat flow in underground mine galleries located in areas of  adverse temperature gradients constitutes a safety hazard for mine workers, and must be continuously cooled and ventilated at a correspondingly high operational cost.  Heat recovery is only valuable if  the recovered heat can be reused. Through careful whole-system analysis and design, heat losses and water use can be monitored in order to balance loads, improve performance, decrease emissions and minimize waste. For a mine with access to low-temperature resources, the addition of  waste heat recovery either from exothermic mineral processes or high-load ventilation and cooling systems, can further improve performance, efficiency, and cost-reduction. The captured heat can be stored on-site or used directly as process heat. For example, the Finish smelter operator Boliden Harjavalta Oy recovers 20 MWth of  heat from its sulphuric acid plant; half  of  this heat energy is used in the company’s adjacent copper and nickel plants, and the other half  is sold to a local district heating network (Laval 2011). Alternatively, the recovered heat can be supplied to an Organic Rankin Cycle (ORC) heat engine to generate supplementary electricity;  27water with temperatures as low as 78°C can be used as a heat source in this manner (Zarrouk and Moon 2014).  2.7.2 Geothermal applications akin to mining Generally, it is fluid enthalpy (h) that determines (or limits) how the extracted geo-fluids can be used, and what their expected economic value would be. Resources on the highest end of  the scale are typically employed in power generation: in dry-steam power plants for dry or superheated steam sourced from dry-steam or vapour dominated reservoirs; in single- or double-flash-steam power plants, for high- to medium- enthalpy fluids sourced from liquid-dominated hydrothermal reservoirs; and in binary cycle power plants, used with low-enthalpy resources, or as a supplementary stage to existing flash-steam plants to recover power from hot, waste brine. With the exception of  a number of  industrial applications, low to very-low temperature fluids are generally reserved for direct use. The very end of  the enthalpy scale contains geoexchange applications that use heat pumps to boost the amount of  energy harvested from shallow depths (< 300 m on average) (Curtis et al. 2005).  Figure 13 is the mining-specific version of  the Lindal diagram, which illustrates a variety of  mining-specify application for geothermal fluids, depending on specific resource temperatures (Gudmundsson and Lund 1985; Patsa, Van Zyl, and Zarrouk 2015). Most geothermal applications make use of  water in a range of  temperatures, rather than at a specific temperature point. Those on the higher end of  the temperature scale use steam, while the applications that are located the lower part of  the scale typically make use of  hot, liquid geothermal waters. Direct use of  geothermal water spans the entire temperature scale, which can be extended below 22°C to include geo-exchange/heat pump applications. Industrial geothermal use occupies the mid-to-high section of  the diagram, leaving the mid-low end of  the scale for less energy-demanding applications such as space heating and agriculture (Gudmundsson and Lund 1985; Dickson and Fanelli 2013).   28Figure 13: Potential geothermal direct use in the mining sector (modelled after Lindal) .
2 [oC] Power Minerals370270260 ZincExtraction250210200 (Saga) SaltCrystalization190 DiotomaceousEarth Drying180170150140130 Fresh Waterby Distilation120110 LithiumExtraction1009080 Intense De-icingOperations70 Enhanced CuHeap Leaching605040302010Concentration of Saline Solutions District Heating with GSHP'sSpace Heatingwith GSHP'sDe-IcingGeneral Direct UseMining-Specific Lindal DiagramSeawater Desalinationby ThermalDistilationAbsorptionRefrigerationWarm Water for Year-RoundMining in Cold ClimatesDirect Use for Op'sDistrictHeatingPowerGenerationfromDry/FlashSteamBinaryPowerGenerationPreheatingIron-OreConcentrate SlurryEnhanced AuHeap LeachingProcess HeatEvaporation of HighlyConcentrated Solutions (Gudmundsson and Lund 1985; Patsa, Van Zyl, and Zarrouk 2015)2 292.7.2.1 Power production/supplementation for remote mines  It is not only mines in the far North that worry about fuel supply. Veladero gold mine is located more that 4 km high up the Andes Mountains, in the San Juan Province of  Argentina. In production since 2005, Veladero has 5.1 million oz. of  estimated gold reserves. The mine consumed about 30 million liters of  diesel fuel to generate the 12.5 MWe it required to operate. Seventeen tanker trucks traveled 500 km every week to transport this fuel to site. Gold – a hydrothermal (epithermal) alteration mineral – is closely associated with geothermal energy. Medium-low geothermal resources have been found on-site, with surface thermal springs up-flows measuring between 76-78°C. The operator was looking at an estimated 8-14 MWe binary plant installation that would run on geofluids sourced from 1-1.5 km depths. The proposed plant would cover 66-100% of  Veladero’s operational needs, and it was expected to generate annual savings of  19-30 million litters in fuel consumption and 53,000-93,000 tons in GHG emissions (Borders 2013; BarrickGold 2014).  2.7.2.2 Minerals extraction  The chemical composition of  geothermal fluids varies greatly between reservoirs. Rock composition, temperature and pressure at depth all affect the eventual fluid mineral composition. Rich brines, though more difficult to handle in the power production process, can extend the economic value of  a resource through the exploitation of  the primary power generation by-products, such as various water-soluble minerals and precious metals. Actually, the higher the chemical concentration, the more minerals could be potentially extracted from a given brine. In some cases, extracting a mineral from geothermal brine may be more economically attractive than mining it from rock (Bakane 2013).  Silica, as the most abundant, dissolved, and precipitating mineral in geothermal brine, typically interferes with electricity generation by precipitating from solution and adhering to pipe and equipment walls. It therefore needs to be removed. The extraction process employed for this purpose  30will determine the particle size of  the extracted silica, and hence the quality of  the end marketable product. At present, silica extraction purity is close to 99% (Bourcier et al. 2006). In Mammoth Lakes, California, marketable amounts of  quality silica were extracted from the geothermal brine that is supplied to the Mammoth Pacific power generation plant. Mammoth used reverse osmosis to produce very high quality silica, mainly due to the field’s very low salinity, very low calcium, negligible iron, and other heavy metals content. Bloomquist (2006) evaluated the annual yields at 
$11 million, based on the typical market price $0.75 per pound for precipitated silica used in rubber manufacturing, and a silica recovery of  7,200 tons per year. The Mammoth brine also contained extractable lithium, tungsten, caesium and rubidium (Bloomquist 2006) (Bourcier et al. 2006). Lithium (Li) is used in a broad spectrum of  products and industrial applications, including but not limited to batteries, ceramics, glass, rubber, lubricating greases, pharmaceuticals, and in primary aluminum production. It is more stable when processed into compounds such as lithium carbonate (Li2CO3) and lithium hydroxide (LiOH), and it is primarily produced in Chile, Australia, China, and Argentina, either from hard-rock, open pit or underground mines, or through solar brine evaporation. Although demand levels (2013) – at 160,000 tonnes per annum – were lower than the estimated word production levels – at 186,000 tonnes per annum – the US Department of  Energy (DOE) lists lithium as a strategic mineral and predicted a 60% market growth by 2017, driven by the high tech and automotive industries (US Department of  Energy 2011; Kaufmann 2014).  Simbol Materials demonstration plant in Calipatria, CA has been developing an alternative, high quality, lithium extraction process to use with Salton Sea’s geothermal brine, one of  the most concentrated mineral brines in the world (with a mineral content ranging between 200,000 – 250,000 ppm). The lithium-bearing brine is sourced from EnergySource’s Featherstone 50 MWe geothermal power plant, at a flow rate of  6 gal/min and at an outflow temperature of  110°C. Extraction is completed in stages, first by removing the silica (SiO2), then iron (Fe), and finally lithium  31carbonate (Li2CO3), which is the primary product in this process. Three additional to silica and iron by-product minerals will also be extracted at full-scale production, namely manganese (Mn), zinc (Zn) and potassium (K). When operational, the full-size plant will be able to process 1,000 times the demonstration plant flow rate, generating 16,500 tonnes of  high-grade lithium per year – currently valued at $6,000 per tonne – from an average brine input of  6,000 gal/min (Duyvesteyn 1992). In fact, the economic value of  the Salton Sea minerals was estimated at $1.5 billion dollars. This is higher than the economic value of  the net combined 327 MWe produced by the 10 power plants operating in the Salton Sea Known Geothermal Resource Area (KGRA). Full-scale production yields from the Featherstone plant were estimated as: 16,000 tons of  lithium carbonate equivalent, 24,000 tons of  electrolytic manganese dioxide, and 8,000 pounds of  zinc metal (Harrison 2010). According to the DOE, Simbol's business model separates the geothermal operator from the business of  mineral extraction, thus reducing risks and costs (Klein and Gaines 2011; CalEnergy 2014).  2.7.2.3 Enhanced heap leaching  (Hartman and Mutmansky 2002) define heap leaching as the process of  recovering minerals from typically low-grade metal ores with copper, gold, or uranium content; this is achieved through the application of  an aqueous leachate solution on piles of  broken ore stacked on impermeable pads. The leachate solution impregnated with the extracted metals is collected as it percolates out from the base of  the heap and further processed to produce doré (a semi-pure alloy of  gold and silver), which is then transported to a refinery for further purification (Kappes 2002). Adding heat to the leaching solution in small-scale experiments accelerated the chemical reaction behind extraction by improving the kinetics of  the leaching process (Trexler, Flynn, and Hendrix 1990). Also known as enhanced heap leaching, this process can increase gold extraction rates by 5-17% and copper extraction rates by 1.2% per degree Centigrade change in the heap solution temperature. Enhanced heap leaching also allows for year-round operation of  a mine site, independent of  weather conditions, which is  32quite advantageous as typically heap leaching stops when temperatures fall below 4°C (Bloomquist 2006; Sigmundsson 2012).  In Nevada, a total of  10 producing gold, silver, or gold/silver mines have geothermal resources on-site or in close proximity to leaching facilities. A number of  them are already using geothermal brine in enhanced heap leaching. At the Round Mountain Gold mine, geothermal fluid at 82°C is used to heat the cyanide leach solution by flowing fed through counter-flow heat exchangers at an average flow rate of  70 L/s. The system has an installed capacity of  14.1 MWth and uses the equivalent of  42 TJ of  thermal energy per year (Lund 2003). Annual production levels at the Florida Canyon mine are 905 kg for gold and almost 800 kg for silver (Driesner and Coyner 2008). The heat is transferred from geothermal fluids at 99°C to the barren cyanide solution using a shell and tube heat exchanger (Trexler, Flynn, and Hendrix 1990). The 1.4 MWth system uses an estimated 42 TJ per year (Lund 2003). In the past, the gold- producing Freeport Jerritt Canyon Mine and the silver-producing Gooseberry Mine also used thermally-enhanced cyanide heap- leaching process (Flynn, Trexler, and Hendrix 1986; Bakane 2013).  (Sigmundsson 2012) investigated the potential advantages of  using low-temperature geothermal fluids in heap leaching for copper at Chile's Collahuasi Copper Mine. In 2012, the mine was extracting about 40% of  the copper contained in its ore heaps. An acidic mixture of  water and sulphuric acid (termed raffinate) was used as the copper- bearing heap leaching solution. The extraction process comprised of  heating the raffinate from 15°C to 35°C using diesel oil and propane, prior to re-circulating it to the heap pads. The study indicated that using 70°C geofluid as the primary heat source in a geothermally-enhanced heap-leaching alternative would increase production levels by an average of  1.2% per degree Centigrade change in the raffinate temperature. The resulting fuel-cost savings for the proposed system upgrade corresponded to a 12-month projected payback period.   332.7.2.4 Desalination  The number of  people working on a mine site varies from project to project but it can be in the thousands. The water used in some operations can potentially be of  too low a quality to be suitable for human consumption. Nevertheless, for sites accommodating human workers, access to a potable water source is absolutely necessary but not always easy to secure. A number of  operators are resorting to desalination to meet this need. In western Australia, CITIC Pacific Mining required a full-scale desalination plant – complete with water transmission lines – for transporting iron ore slurry at their Sino ore project mine site. 140 ML of  water was generated every day, enough to fill 56 Olympic-sized swimming pools. With fresh water demand for copper production on the rise (expected to increase by 38% by 2021), Chile is pushing hard to make desalination in mining processes mandatory, for mines consuming more than 150 L/s. A number of  companies have already introduced desalination plants into their operations, for example BHP at Escondida, Freeport-McMoRan at El Abra, and Coldeco at their Radomiro Tomic and Chuquicamata divisions. Salt removal in Chile has become increasingly costlier in recent years (amounting to $5/m3 versus 
$2.8/m3 in Mexico), and a mandate to desalinate will directly impact operational costs and profit margins (Jamasmie 2014).  Mines with access to geothermal resources may have the option to use it for desalination. A demonstration/research project on the Greek island of  Milos used geothermal brine to run a seawater desalination unit. The developed device, successfully separated seawater into two streams of  potable, low-salt-concentration water, and highly concentrated salt brine. Four production wells providing 300 m3/hr of  geothermal fluid at 55-99°C, can be used to thermally distil, in stages and under pressure, seawater at 80,000 ppm salinity. This system has a production capacity of  
75-80 m3/hr of  drinking water, enough to cover the needs of  the entire island at an estimated cost of  €1.5/m3. The harvested heat would also be enough to generate a supplementary 470 kWe of  power  34using an ORC unit (K4RES-H 2010). Unfortunately, due to a breakdown in stakeholder relationships (primarily between the State and the local communities of  Milos) and bureaucratic restrictions, the desalination plant was never hooked up to the local drinking water distribution system and the desalination unit itself  was abandoned.  2.7.2.5 Mining areas of  high geothermal potential  In certain cases, high geothermal potential can actually act as a hindrance to mining activity, particularly in underground mining. For example, the Enterprise Mine in the Ota Banda district in Western Australia is that country’s deepest and hottest underground operation. A high grade copper mine, it extends from about 1,000 m to almost 2,000 m below surface. Temperature increases with depth and at Enterprise Mine, extreme depths combined with a high geothermal gradient of  200°C/km, a surface rock temperature of  28°C and high surface ambient temperatures during summer, result in extreme heat stresses. Additional heat sources such as surface climate, auto-compression, plant machinery and equipment, oxidation, explosives, broken rock, lighting, personnel and service water compound the problem. Left untreated, extreme heat stresses can be detrimental to worker safety. Extreme heat has also been linked to significant decreases in productivity, and higher accident rates (Brake and Fulker 2000). Typically, underground conditions are managed by ventilation in a process that floods the mine with air to remove heat (which in this context is regarded as an unwanted air-borne contaminant). But for mines that are deeper than 1,000 m, ventilation cannot provide adequate fresh air to the workings to remove heat, blasting gases and diesel fumes. In cases such as these, refrigeration is unavoidable; in fact, Enterprise Mine refrigeration needs exceed 40 MW(R) (Brake and Fulker 2000).  Most operating mines have additional onsite space heating and/or cooling needs, e.g. within administrative buildings or live-in camps. Very low temperature ground source heat pump systems (GSHPs) generally operate within a 5-25°C temperature envelope and can potentially provide for this  35kind of  need. GSHPs use a binary fluid that has a very low boiling point (e.g. -26.3°C for tetrafluoroethane (R134a)) to transfer heat from source (e.g. the ground) to target (e.g. mine camp). The overall system comprises of  3 separate components: 1) an open or closed loop installed in the ground or submerged in a large water mass, at depths of  relatively constant temperature; 2) a heap pump unit that supplements and boosts the heat extracted from the ground with electricity; and 3) a distribution system that uses air or water to transfer the heat from source to the target environment. GSHPs operate at very high efficiencies (300-600% or 3 < COP < 6) and vary in size from individual modular units used for single rooms to district-sized units that can generate enough capacity to serve large communities – one such example is installed in Dalian, China, where 68 MWth of  total heating and 76 MWth of  total cooling loads are generated by a district cooling and heating seawater system, from source temperatures between 2°C (winter) to 21°C (summer) (Patsa 2009). According to Koufos (2012), Canadian mines stand to gain considerable energy savings by switching their space heating and cooling systems to GSHPs. Based on a study involving 12 mines across Manitoba, Ontario and Quebec, such a switch would result in total annual heat savings of  20,915 kWth, equivalent to CAN$1.5 million/year in cost savings and 18,850 tonnes in CO2 emission reductions. For the systems examined in the study, heat was sourced from water mined from depths between 800 and 3,100 m, extracted at flow rates between 7-63 L/s, at source temperatures between 10°C and 22°C.  2.7.2.6 District heating & cooling from abandoned mines  More than 1 million abandoned mines are thought to exist around the world, some of  them in close proximity to densely populated areas (Preene and Younger 2014). Although past their operational life, they still hold economic value as sources of  heat mining. Flooded underground mines are essentially large heat storage units. Depending on the depth of  the mine and the local geothermal gradient, enough thermal energy could be extracted from the mine-works to supply a district-sized heating and cooling system, optionally coupled with GSHPs. The technological  36feasibility of  this concept has proved successful: overall, more than 15 deployed projects in Canada, the US, Germany, Norway, the UK, the Netherlands, Russia and Spain are utilizing mine water for heating and cooling (Preene and Younger 2014). The Heerlen Minewater project in the Netherlands for example, was built as part of  a regeneration scheme for an area that was devastated by the closure of  coal mines. The system taps into heat stored in four flooded underground coal mines, and services 33,000 m2 of  residential space, 3,800 m2 of  commercial/cultural space, and 13,700 m2 in health care and educational institutions. Mine water is harvested from a depth of  700 m, at 22 L/s, and passes through heat exchangers operating at a ΔT of  5°C (Koufos 2012). Further to the North, the British Geological Society is teaming up with Glasgow City Council, to look into the potential of  heat mining, as part of  the Clyde Urban (regeneration) Super Project. A heavy-industry boom in the 19th century and an abundance of  coal and iron in the general Lanarkshire region, led to Glasgow’s ascent as the world’s preeminent shipbuilding centre. Now mostly defunct, the majority of  these mines lay beneath the city. The project is looking at the potential heat stored within mine waters, superficial deposits and bedrock aquifers, and the feasibility of  using it as a source for district heating and cooling. Geological modelling has identified a number of  drilling targets, primarily within mine shafts that are most likely to be structurally preserved (British Geological Survey 2014; Macnab 2011; Ramos, Breede, and Falcone 2015). 2.7.2.7 Mines, carbon taxes and the Kyoto Protocol Clean Development Mechanism  Emissions savings are fast becoming a prominent concern of  the mining industry, primarily due to an anticipated need to mitigate some of  their emissions-related fees. Climate legislation, such as President Obama’s Climate Action Plan in the US and Australia’s Clean Energy Act 2011, is expected to have a considerable impact on operating costs across the entire mining industry. Operators that fail to reduce their greenhouse gas emissions will be required to pay carbon taxes and fines, and are therefore actively looking at clean energy for offsetting some of  their operation-related  37emissions. Newmont Mining Corporation, concerned about potential carbon fees for its US and Australian projects, is taking advantage of  the Kyoto Protocol Clean Development Mechanism (CDM), under which, “climate-friendly, sustainable development projects in developing countries are eligible for Certified Emission Reductions (CERs)” (UNESCO 2014). CERs define ways for governments and corporations to meet their compliance obligations, under cap-and-trade schemes pending in the U.S. and Australia. The Newmont is positive that geothermal energy provides them with an opportunity to “not only reduce emissions and increase energy efficiency, but also to earn CERs to use later in cap-and-trade schemes” (Newmont Mining Corporation 2010). CERs developed prior to the deployment of  a cap-and-trade scheme are expected to cost less than half  of  the same CERs purchased post-deployment (Hannam 2014; The White House 2014). 2.7.2.8 The case of  Lihir  The Lihir Gold Mine in Papua New Guinea (PNG) presents a fine example of  geothermal and mining activity integration. In production since 1997, it employs more than 2,200 people and has an annual production of  649,340 gold ounces (2013 figures). This mine is of  particular interest, as it is on an active hydrothermal environment, has proven high-temperature geothermal potential onsite, and is without access or in proximity to a power distribution network (Bertani 2015). Current production uses geo-fluids primarily extracted from a depth of  1,000 m at 240-250°C, but temperatures greater than 300°C have been measured. Four wells located on the property, initially drilled for the purpose of  releasing ground pressures while open pit mining, supplied steam to a 6 MWe non-condensing power plant, installed in 2003 (Bixley 2003). Two subsequent expansions – a 30 MWe single flash, condensing geothermal unit 2005, and 20 MWe extension in 2007 – raised the total combined generation capacity to 56 MWe. This power is generated for exclusive consumption at Lihir Gold and covers about 75% of  the mine’s power needs. 30 MW are used by the oxygen  38plant, less that 3 MWe are directed offsite for used in local villages, while the remaining capacity powers the mine, onsite camps and offices.  Although secondary to the gold mining activity, the geothermal system has generated significant annual profit: US$40 million in savings from offsetting heavy fuel oil consumption (which corresponds to >50% of  the mine’s energy cost), and US$4.5 million from sales of  carbon credits on the global market. The plant’s emissions savings actually correspond to approximately 280,000 tonnes per year and allow it to trade carbon credits under the CDM. The operators also contribute to the broader PNG economy and local communities, through “taxation and royalties to national, provincial and local governments, salaries and wages, landowner contracts, investments in public infrastructure and services, and support of  Lihirian and PNG suppliers, […] including access to health services, the provision of  electrical power and water to local villages” (Limited 2017; Melaku 2005). Finally, it is interesting to note that the incremental transition from diesel fuel to geothermal power (from 5 MWe in 2003, to 35 MWe in 2005, and finally to 56 MWe in 2007), allowed the company to gradually build its on understanding of  the geothermal reservoir, and eventual trust in its potential. 2.8 The Analytic Hierarchy Process (AHP) The Analytic Hierarchy Process, which was used in the development of  the geomine model in this research, was first introduced by Thomas L. Saaty in the 1970s (Saaty 1980). It is a multi-criteria approach to decision- making that places an strong emphasis on the fact that “decisions are ultimately dependent on the creative process by which the decision problem is formulated” (Golden, Wasil, and Harker 1989). It is an “intuitive and relatively easy method for formulating and analyzing decisions” and its overall philosophy is to provide “a solid, scientific method (the analytic part) to aid  39in the creative, artistic formulation and analysis of  a decision problem” (Golden, Wasil, and Harker 1989). More specifically:  — 	 It is analytic, in the sense that it is a method that uses mathematical/logical reasoning to make a selection out of  multiple alternatives (as opposed to holistic decision making, which consists of  simply choosing the alternative that is most desired).  —	 It is hierarchical, because it structures decision problems using levels that combine to construct the premise and setting of  the decision-making problem at hand, namely: goals, criteria, sub-criteria, and alternatives. By breaking decision-making down into levels and by focusing on smaller decision subsets, decision-makers can tackle problems of  considerable complexity in a variety of  application areas (Saaty 2008; Vaidya and Kumar 2006; Saaty and Vargas 2013; Thomas 1988; Zahedi 1986; Thomas, Kevin, and Luis 1991).  —	 And it is a process that forces decision-makers to adopt a more careful, detailed, and structured approach to decision-making, compared to relying on ambiguous assertions of  expertise or “gut-feelings” (Golden, Wasil, and Harker 1989). The AHP process breaks decision-making into 4 steps (Saaty 2008): 1.	 Defining both the problem at hand and the decision that needs to be made. 2.	 Structuring the decision hierarchy by defining the decision’s goal, objectives (from a general perspective), to the intermediate level criteria upon which subsequent elements depend, to the low-level set of  alternatives the decision-maker must choose from.  3.	 Constructing a set of  pairwise comparison matrices, by comparing upper-level objectives to intermediate- and lower-level criteria and alternatives.   404. 	 Weighing objectives priorities by sequentially adding up component and alternative importance weights into global importance weights, normalizing to allow for comparisons, and scaling the final weights based on a predefined set of  relative importance weights for each top-level objective.  Weights are assigned using a numerical scale (termed the fundamental scale of  absolute numbers — see Table 5) that indicates “how many times more important or dominant one element is over another element with respect to the criterion or property to which they are compared” (Saaty 2008).  AHP continues to be successfully employed in a wide range of  real-word applications — examples include (but are not limited to): construction management (Esa, Halog, and Rigamonti 2017; Erdogan, Šaparauskas, and Turskis 2017; Bitarafan et al. 2012); inventory management (Lolli, Ishizaka, and Gamberini 2014; Balaji and Kumar 2014); solid waste management (Milutinović et al. 2017; Vučijak, Kurtagić, and Silajdžić 2016; Soltani et al. 2015); healthcare (Agapova et al. 2017; Lee, Vaccari, and Tudor 2016; Yuen 2014); general sustainability (Garcia et al. 2016); air quality (Martenies, Wilkins, and Batterman 2015); and climate change (Hendrickson, Nikolic, and Rakas 2016; Berrittella et al. 2008).  Numerous examples exist in the literature on combining Geographical Information Systems (GIS) with AHP, primarily for site selection that requires multi-criteria decision making. For example: Mishra, Deep, and Choudhary (2015) applied AHP within GIS to identify suitable sites for organic farming in rural areas with inadequate transportation services in an effort to boost rural economies and promote rural tourism to make self-sustainable villages in Uttarakhand, India. Janke (2010) used a multi-criteria GIS model to identify areas that are suitable for wind and solar farms. Uyan (2013) used AHP and GIS to conduct a site selection analysis for solar farms in the Karapinar region, Konya/Turkey. Sánchez-Lozano et al. (2013) used GIS and AHP to evaluate numerous solar farm  41locations in SE Spain. Roig-Tierno et al. (2013) combined AHP and GIS to determine the best location for new development in the retail sector in Murcia, Spain. Şener et al. (2010) used AHP and GIS to locate a site of  a landfill in the Lake Beyşehir catchment area in Konya, Turkey. Anane et al. (2012) applied GIS and AHP to ranking suitable sites for irrigation with reclaimed water in the Nabeul-Hammamet region, in Tunisia. Aydin, Kentel, and Duzgun (2013) applied AHP and GIS to identify the more feasible locations from a list of  priority sites for the development of  hybrid wind solar–PV renewable energy systems, in western Turkey. Zhang et al. (2015) used AHP and GIS to conduct a land suitability assessment for tobacco production, in the Shandong province of  China. Abudeif, Moneim, and Farrag (2015) used AHP and GIS to select the most suitable of  4 candidate sites located in the NW Coast of  the Red Sea, in Egypt, for a suitable location for siting a nuclear power plant. There are also some limited examples of  GIS and AHP use within the geothermal literature, most notably in exploration targeting and resource assessment. For example: Sadeghi and Khalajmasoumi (2015) used AHP and fuzzy login in ArcGIS to integrate petrological, volcanism, hot spring and fault data to derive a preliminary assessment of  the geothermal potential of  the East Azarbayejan Province, in NW Iran. Moghaddam et al. (2014) used spatial analysis and multi-criteria decision making to create a regional-scale geothermal favourability map, by studying the spatial distribution and association between known geothermal resources and publicly available regional geoscience data. Moghaddam et al. (2013) also conducted spatial data analysis for exploration of  regional scale geothermal resources in Akita and Iwate provinces within the Tohoku volcanic arc, in northern Japan. Kimball (2010) used publicly available data to build a geothermal favourability map for British Columbia, Canada, by combining evidence layers created using expert knowledge, weighted summation, and AHP methods within ArcGIS. Malczewski and Rinner (2015), Malczewski  42(2006; 1999), and Ferretti (2011) give a comprehensive overview of  GIS-based MCDA and provide a recent coverage of  the state of  the art on problems and solution approaches in the literature.  In mining, AHP has been applied in problem solving relating to resource assessment, operation, mine closure, and reclamation. Liang et al. (2017) used AHP and GIS in assessing regional selenium resources in soils in the Boshan region, in Kazakhstan. Sui et al. (2016) derived an AHP-based quantification method for evaluating the fracability of  shale in oil & gas operations, by optimizing hydraulic fracturing of  shale gas reservoirs, and enhancing shale gas recovery. Azadeh, Osanloo, and Ataei (2010) improved the Nicholas technique, a well-known mining method selection technique, by using AHP to address some of  the problems it had with dealing with unsteady and uncertain characteristics of  mineral resources. Jianqing (2011) applied AHP in the development of  a mine gas prevention and control system. Soltanmohammadi, Osanloo, and Bazzazi (2010) developed an analytical AHP-based approach for ranking alternative options in post-mining land-use determination. The application of  the AHP process is described in further detail in Section 3.3.2.  No example was found in the literature of  the approach adopted for this study, namely using GIS and AHP to identify locations with high combined geothermal and mining resource potential. This study is in these terms, novel.
 43Chapter 3: Methods and Research Design This chapter introduces the research questions, and presents the assumptions from which they derive. In addition, it provides a description of  the overall research approach, including the statistical, computational, and decision-making methods employed in addressing the research questions.  3.1 Assumptions and research questions 3.1.1 Assumptions The main assumptions that apply to this work are as follows: —	 Due to the time, effort and monetary investment involved in bringing a geothermal or a mineral production project to the operation stage, it is reasonable to assume that the presence of  an operating production facility, be it a power station or a mine, indicates the known and proven existence of  resources, in economically viable quantities. It is therefore reasonable to regard current geothermal/mineral production as the strongest possible indicator of  geothermal/mineral potential. — 	 Geothermal and mineral resources do not exist at random; they are rather the result of  specific physical and chemical processes, which are in turn known to directly or indirectly relate to specific geological, hydrological, or tectonic conditions. Different types of  processes and conditions give rise to different types of  mineral or geothermal resources. Different types of  processes and conditions are themselves directly or indirectly associated with specific (surface) features that are both identifiable and mappable.  —	 This association between resources, and processes/conditions, and between resources and (surface) features can be quantified using statistical and/or geographical analysis techniques.  44— 	 It is therefore assumed reasonable to extend the association between resources and processes/conditions, and between resources and (surface) features, to define an association between processes/conditions and (surface) features that can be identified, mapped and quantified using statistical and/or geographical techniques. —	 If  it is assumed reasonable to use the presence of  specific processes/conditions to make inferences about the presence of  geothermal or mineral resources; also if  it assumed reasonable to use the presence of  specific (surface) features to make inferences about the presence of  specific processes/conditions; it is by extension assumed reasonable to use the presence of  specific (surface) conditions to make inferences about the presence of  geothermal or mineral potential, and to be able to quantify this relationship using statistical and/or geographical techniques.  —	 Using geographical, statistical, and decision-making techniques to map, analyze, and calculate the coincidence of  features known to be associated to either geothermal or mineral potential, it is possible to estimate (at different levels of  confidence), the existence (and by extension coexistence) of  geothermal and mineral resources. —	 It is assumed that it is possible to use data to represent features that are considered to indicate the presence of  geothermal or mineral potential. Such features are termed indicators (or indicator features) in this text, because they are used to indicate (directly or indirectly) geothermal or mineral potential. A strong or primary indicator is related to high potential, while a weak or secondary indicator is related to low potential.  Based on the above-stated assumption, the following research questions can be posed.  453.1.2 Research questions “Which indicator features can be used to indicate high geothermal resource potential, even in areas of  zero geothermal production information?” “Which indicator features can be used to indicate high mineral resource potential, even in areas of  zero mineral production information?” “Which of  the identified indicator features known to coincide with high mineral and geothermal potential can be used to predict the coexistence of  mineral and geothermal potential, even in areas of  zero mineral and/or geothermal production information”? 3.1.3 Limitations on scope and data complexity  This project does not use specialized data to calculate mineral and geothermal potential. The aim is instead to use publicly available data of  a simpler nature, which would be more accessible to decision makers without specialist knowledge in, for example, geochemistry or geophysics. This is generally in contrast to the type of  data typically collected during preliminary exploration, where the contribution of  multiple experts to data analysis is required to identify and potentially size a resource.  3.2 Research design The overall approach taken to address the stated research questions and objectives was a complex one, with multiple steps requiring the use of  a large suite of  software, and calculation, and graphical representation methods. In general, the process entailed locating and acquiring data that was cleaned up and processed into a format that was suitable for analysis.  The analysis consisted of  the following steps:  —	 Identifying appropriate indicator features for both mineral and geothermal resources; —	 Ranking indicator features in terms of  their importance to the research questions;  46—	 Combining relative indicator contribution into a composite metric termed geomine potential; —	 Mapping geomine potential in GIS, in order to identify the areas most likely to host mineral development that can successfully integrate geothermal energy in mining operations.  3.2.1 Process overview This multi-step process is shown as flow charts in Figures 14 and 15. Figure 14 shows the general flow of  the calculation process, starting with data harvesting and ending with the mapping of  geomine potential — and assumes that the data requirements identification stage has already been completed. Figure 15 extends this process description by including a list of  methods and software packages used in each step, a short description of  which is provided below: Data location	 Online sources of  suitable data were located through an extensive literature and online search, following the identification of  data requirements, which were derived directly from the research questions. Retrieval	 Ready-made data retrieved through downloading — typically in CSV (comma-separated values), XLS (MS spreadsheet), or SHP (geospatial vector data) format — was converted into CSV, which was universally read by all applications in this study. Harvesting	 Data not directly downloadable as a ready-made dataset had to be harvested, for example from client-side web pages and embedded maps displaying data generally stored on a server-side database. Harvesting was done iteratively, using AppleScript automation scripts that queried and downloaded the data by evoking the wget command in the command line (OSX Terminal). The harvested files were stored locally in HTML format, and batch converted to TXT using Automator.
 47
48Figure 14: Steps to addressing research objectives. Figure 15: Steps, methods and software used to address research objectives.
 49Scraping	 The harvested pages contained the desired data, embedded in the HTML source that was used to display them; typically each located feature in a set was saved as a separate page. An AppleScript script iteratively scanned each harvested file, located the data points of  interest (e.g. the Lat/Lon coordinates of  a particular feature), extracted all data pertaining to said feature, and temporarily store them in an in-memory array. At the end of  the run, the array was written to disk, in either SQL or CSV format, depending on whether the scraped data would be stored in a local database or on disk.  Scrubbing	 Scraping data often results in duplicate entries or encoding issues, which subsequently show up as corrupt or illegible characters in the set. Certain characters, such as quotes and forward slashes can also be problematic when saved in a database or read from a CSV file. To avoid problems, all sourced data was cleaned/scrubbed using a combination of  methods and programs, such as simple “Find & Replace All” in a text editor, column splits and character substitutions in Trifacta, and more elaborate logic-based editing in MS Excel and R Studio. Data stored locally in a database also needed to be organized and formatted thematically, based on the corresponding ER-diagrams drawn up before setting up the database tables that would eventually contained them.  Transformation 	Trifacta, pivot tables and VLOOKUPs in MS Excel were extensively used to transform datasets as needed. This included imposing custom classification schemes depending on the needs of  the analysis. For example, certain volcano types that were essentially synonymous (e.g. compound and complex volcanoes) were merged using a common classifier attribute (here compound was replaced by complex).   50Ranking	 Expert opinion applied through pairwise comparisons to the Analytic Hierarchy Process (AHP) allowed for the calculation of  the relative importance of  the various indicators and factors used in the model, e.g. the relative importance of  the different types of  power generation technology. When appropriate, fuzzy logic was also employed to convert the calculated importance ranks into classification attributes that were subsequently used to update the datasets. For example, the percentage of  NULL cells in the data was used to assign a classifier attribute to the dataset Completion metric, defined as either Very Low, Somewhat Low, Low, Mostly Low, Medium, Mostly Medium, High, Mostly High, and Very High.  Analysis 	 This research project included mathematical, statistical, and geographical analysis, as well as visual analytics. Extensive mathematical and statistic analysis conducted in R Studio and MS Excel aggregated the stored data, and helped to better understand data and to combine individual data layers into more complex information layers that formed the basis of  the geomine model. Geographical analysis provided the tools for quantifying the physical relationships between features and for calculating the importance of  specific classes or subclasses of  features based on, primarily, proximity to high geothermal production and/or potential. Interactive visualizations allowed for data exploration, which in turn resulted in a better understanding of  data trends. Visualization 	 A wide range of  visualization methods and graphics were created to supplement analysis and support argumentation. The type of  visualization was selected based on the argument that was made in each case, accounting for any applicable data constraints. Atypical visualization types (i.e. other than the classic bar/column/line/scatter graphs) are described in more detail in section 3.3.5.   51Analytics 	 A number of  visualizations were originally created as interactive, and snapshots were exported for inclusion in this thesis. The original interactive versions are published online and are publicly accessible. They have been included because they allow for a more immersive interaction with and exploration of  the data. Mapping	 As a final step to the analysis and visualization, maps were created in ArcGIS and Tableau showing the geographic extend of  all identified geothermal, mineral, and geo-mine potential. Where applicable, QR codes are provided to the reader as links to the online versions of  these maps.  3.2.2 Identification of  data requirements 3.2.2.1 Assumptions on data representation and format Two basic assumptions apply to the concept of  data retrieval: a) sourced data can represent either physical entities or phenomena (e.g. a geothermal power station or a heat flow measurement at a given location), or conceptual elements or characteristics (e.g. the administrative borders denoting a specific country and its political system); and b) data (whether drawn as points, lines, or shapes) must be geo-referenced, i.e. always attached to a particular geographical location defined as set of  
Lat/Lon coordinate pairs (to allow for the location and/or colocation of  features).  References are made in this text to such concepts as geothermal potential, geothermal production, mineral potential, mineral production, and indicator features, as well as to the concepts of  coexistence and of  ranking (production and potential) with regards to size. Reference is also made to the case of  a complete absence of  production information. These terms can be translated to specific requirements that subsequently inform data selection and retrieval. The following requirements were identified from an analysis of  the research questions.  523.2.2.2 Data requirements deriving from the research questions —	 Geothermal production refers to currently operating geothermal power plants. In order to accommodate for the concepts of  size and importance and to allow for classification and more in-depth analysis, the geothermal power plant type and the level of  installed generating capacity were also included. Geothermal production plays a central role in the geomine model; it is used as the benchmark which other geothermal indicator features are compared to and ranked against. An absence of  data in the case of  geothermal production would therefore necessitate that such data must be developed by manually locating geo-production installations around the world. —	 Geothermal potential refers to the potential presence of  geothermal resources at a specific location. As these resources are located at depth, any indication of  potential carries a level of  uncertainty, which in the industry is mitigated through extensive exploratory drilling, geophysical and geochemical analysis, and geological modelling. As one of  the objectives of  this research is to investigate whether a reasonable approximation for geothermal potential can be made in the absence of  “specialist” information and expertise, more generalized data were used to represent resource potential. Data selection was based primarily on the definitions of  geothermal potential and of  the processes, conditions and phenomena that are known to be associated with its origin; they include volcanism, tectonics, seismic activity, surface heat flow, surface features, and known geothermal resource areas (KGRA’s). They also include information on geothermal development activity preceding the actual construction of  a geothermal power station (GPP), i.e. exploration areas and projects in the (pre-) feasibility stage. Classifying, qualifying, and/or quantifying data was used for analysis, classification, and ranking.  53—	 Mineral production refers to mines that are either currently operating, have suspended, or terminated production. A variety of  reasons can be behind the decision to suspend or terminate mine operations, therefore indicators of  mineral production should not only be restricted to currently operating mines. For example, cessation of  activity at a mine site may quite plausibly be the result of  a drop in commodity pricing, rather than the result of  a depleted mineral resource. To allow for in-depth analysis, classification, and ranking, data on the type of  facility, and the type and size of  the mined commodities was also included.  —	 Mineral potential refers to the potential presence of  mineral resources at a specific location. Mineral resources are also located primarily underground, which, similarly to the case of  geothermal resources, adds a level of  uncertainty to the process of  locating them. A mineral development project goes through a set number of  different stages prior to reaching production. As shown in Chapter 2, only a fraction of  currently active projects will progress through the stages preceding production (e.g. from prospecting to exploration, and from exploration to feasibility). This means that earlier stages carry greater uncertainty than later ones, and although areas of  exploration can not be used to represent mineral potential, they can be used to represent mineral potential at a (much) lower confidence than that would have been afforded by the presence of  an operating mine at the same location. Data selection was therefore focused on identifying mineral development activity at various stages, known mineral resources, and identified deposits that may or may not be currently exploited. In each case, classifying, qualifying and/or quantifying fields were included to allow for data analysis, classification, and ranking.  543.2.3 The definition of  indicator features Indicators features, known to be closely associated to geothermal and/or mineral resources, can be mapped and subsequently used to model geothermal and/or mineral potential. Used in this context to indicate the presence of  a resource, they are thusly termed indicators or indicator features. Not all features are equally important indicators of  potential; for example, it can be shown that volcanoes are more closely clustered around geothermal resources than earthquakes of  a certain magnitude. The confidence that expresses the level of  association or coincidence of  a feature and the resource potential whose presence it indicates is termed importance in this text. An indicator that is more closely affiliated with a resource is more important than an indicator that is not as closely associated with the same resource — e.g. volcanoes are more important indicators of  geothermal potential than earthquakes.  Another equally important confidence measure is the one which relates to the quality, completion, and applicability of  the datasets (created/sourced and subsequently) used in this analysis. Particular effort has been made to assess this data confidence measure — termed significance in this text — for each dataset used. 3.2.3.1 Primary indicators of  geothermal and mineral potential It is assumed that in either case, production is the strongest indicator of  potential: operating mines in the case of  mineral production, and GPPs in the case of  geothermal. The confidence level assumed for these two indicators is 100%, due to the amount of  effort, time, due diligence, financial investment, and expertise that go into bringing a power station or a mine into the production stage (see background section for more detail). All other applicable indicators are compared to production indicators to assess their corresponding level of  confidence.  553.2.3.2 Secondary indicators of  geothermal and mineral potential In reality, mineral and/or geothermal production is not present everywhere where there exists mineral and/or geothermal potential. Other factors, such as political stability, access to market and infrastructure, and environmental restrictions can be barriers to or delay the installation of  production units/projects. In the case of  geothermal for example, although both Chile and Canada are known to hold high-temperature geothermal resources (identified through extensive exploration), there is no commercial electricity production from these resources to-date (Farias et al. 2015; Lahsen, Munoz, and Parada 2010; Kerr Wood Leidal Associates 2015; Grasby et al. 2011). If  production were the only indicator used to identify areas of  potential, resources not currently commercially produced would be left unidentified. It was actually the recognition that the need to find alternative indicators of  potential (even at a lower level of  confidence) — of  primary interest to this work — that initially led to the formation of  the Assumptions, and the Research Questions. 3.3 Methods used to answer the research questions 3.3.1 Data harvesting methods The bulk of  the analysis work for this study was conducted on an Apple iMac computer. The geographical mapping and analysis done in ArcGIS, which required a Windows environment to run. This was also done on an iMac through a virtual machine. The iMac OS X system is Unix-based, and it allows for the installation and use of  a sleuth of  command line applications; those were all run through the Terminal, which is the native OS X command line interface.  3.3.1.1 Command Line and wget  One command line tool that was used extensively for harvesting is GNU wget. It allows for the download of  HTML pages from a website. It can retrieve results recursively, i.e. by downloading  56pages and images that are linked to any given target page. Because it is run through the Terminal, it can be called iteratively, either by a “for loop”, or by passing a list of  addresses to it through a TXT file. The GPP data from OpenEI.org were harvested in this manner (see Chapter 4).  The list of  power stations on OpenEI was organized in a table on http://en.openei.org/wiki/Geothermal/Power_Plant#tab=List_of_Plants (Figure 16); it was actually embedded in the HTML source of  that page (Figure 17). The <tr></tr> tags open and close table rows, and contain information about individual power stations. The data was further split into table cells, denoted by the <td> </td> tags. The opening <td> tag also contained attributes, specifying the class of  the tag (a .css property that defines how the cell should be displayed). Nested in each <td> tag was specific GPP information, such as the GPP title, owner, facility type, operation start date (year), geothermal area, and geothermal region. Figure 16: List of  GPPs on OpenEI.org.  57Figure 17: a) HTML with embedded GPP data; and b) HTML for a single OpenEI row. 
(a)(b) 58Further information was harvested by asking wget to follow the links embedded in the GPP <a> tags. For example, each GPP page provided additional information on the location of  the GPP(as a lat/long pair, and its power generation capacity in MWe — Figure 18). This data was also captured in a similar manner, and saved to disk.  Figure 18: Drill-down information for each GPP from OpenEI.org.  593.3.1.2 Automator and AppleScript The harvested files were saved in a predetermined location and renamed to TXT in batch-mode using Automator (another Mac-specific utility that is part of  OS X — Figure 19). Figure 19: Batch renaming harvested files using Automator.  Once the harvested files were in TXT format, they could be read by ApplesScript. A script was written to scan through the harvested HTML pages, locate the tags containing the data, and store them in an array. At the end of  the iteration, after each file was scanned, the script wrote the array to a CSV file on disk, which could be further processed in Trifacta, MS Excel, and Sublime 2 (a text editor). Data transformations were also done in Trifacta (Figure 20), which is specifically designed to deal with typical data scrubbing needs of  data scientists, and handles character replacements and column splits with greater ease compared to both MS Excel and AppleScript. Even so, at times, the inherent flexibility of  scripting through AppleScript, or the ease and speed of  table pivoting in Excel (Figure 21), proved equally useful. All of  these tools were used interchangeably  60throughout the data processing stage, as needed. No single one was adequate in meeting all data harvesting, scraping, and scrubbing needs, but they worked very efficiently when combined.  Figure 20: Data scrubbing in Trifacta. Figure 21: Table pivots in MS Excel. 
 613.3.2 Data transformation methods 3.3.2.1 Normalization The vast majority of  data was normalized before mapping. This is especially true for the data that was transformed through statistical and mathematical calculations, and for those resulting from decision-making calculations. For example, the GPP dataset compiled in this study (see Chapter 5 for more details) was assumed to represent the population of  all GPP installations around the world. The Type field stores the type of  power generation technology installed onsite; there are 4 possible types populating that particular column, namely: Binary, Flash-steam, Dry-steam and Unconventional. These 4 subclasses are members of  the superclass Type, which is considered the primary qualifier field for this dataset. 
A second column stores the power generation capacity of  each GPP (in MWe), and its values range between 0.05 MWe and 303 MWe — this was considered the dataset’s main quantifier field. The individual capacity entries of  the GPP dataset were actually ranked in two ways, once to account for the importance of  their type (using the qualifier field weight, calculated separately), and a second time to account for their individual generating capacity (using the quantifier). The resultant scaled weights had to be normalized to fall within the interval [0,1], as per the requirements of  AHP, and in order to combine them into more complex metrics at subsequent stage in the analysis.  Four different normalization methods were used in this analysis, namely Max, Min-Max, Sum, and Bound; they are presented below. Figure 22 shows the original data plotted in 3D (capacity vs. Lat-Long), along with their normalized values. Figure 23 compares the cumulative results in both 3-D (capacity vs. Lat-Long) and 2-D (capacity vs. GPP).  The Max method has the specific advantage of  not equating maximum values to 1 and minimum values to 0. It is defined as:  62
63Figure 22: Effects of  normalization on data trends. Plots in 2-D. With and without original values.
64Figure 23: Effects of  normalization on data trends. Plots in 2-D and 3-D. With and without original values.		 	 	 (2) This can be problematic if  the normalized values are used as multiplication factors at a later stage in the process; their inclusion will result in data loss, which is best avoided. Data loss does occur with the more traditional normalization method of  min-max, whereby: 	 (3) Data loss may be acceptable in certain cases, if, for example, the goal is to reduce data noise by removing the smallest of  values and retaining only those who are of  most significance — provided the reduction on information resolution resulting as a consequence from the data loss can be tolerated. In Excel, the Sum method would be defined as: 	 	 	 	 (4) The Sum method is used when the normalized data can be conceptually considered to be parts of  a whole. For example, it was used to calculate the relative importance of  the different types of  GPPs. The resulting weights were applied to each row, scaling capacity to account for the importance of  GPP type.  At the end of  the data transformation, the normalization of  all GPPs was done using Max instead. This was also done because GPP importance was applied as a factor in the transformation to scale the data, in an attempt to retain as much of  the original data trends as possible. Another alternative to the min-max that deals with the issue of  0’s, under- and over saturation is the Bound method. Here, the max and min values in the data are set to a pre-defined min and max limit (or bound). This results in the elimination of  0’s and also of  over saturation of  the maximum value in the data, which is set to 1 in other methods. This would be preferable for example, in the case where value_normalizedmax =value_originalmax(all _ values)value_normalizedmin−max =value_original −min(all _ values)max(all _ values)−min(all _ values)value_normalizedsum =value_originalsum(all _ values) 65the max value in the sample is not the max value in the population, or as a way to limit the reduction in information resolution for the normalized dataset. It may thus be imprudent to set it equal to 1 (or 100%). In excel, the Bound method would be defined as: (5) One clear difference between the normalizations is how they affect data trends. With Sum, the normalized values all sum up to 1, and the normalized curve look squashed as a result, while the other methods retain the relative difference between the largest and smallest values better (Table 2). Table 2: Summation of  original and normalized values of  GPP dataset (Capacity field).  3.3.2.2 AHP & Pairwise Comparisons The Analytic Hierarchy Process (introduced in Chapter 2) was used in the decision-making part of  this analysis. The following example is used to demonstrate how AHP is applied in decision-making. It is taken and adapted from (Golden, Wasil, and Harker 1989). A recent graduate has to choose 1 of  4 job offers, in 4 different cities: Boston, Los Angeles, St. Louis, and Houston. She is originally from Philadelphia. The jobs in each of  the four cities are basically equal, so the graduate must decide which job to take on the basis of  the overall quality of  life in each city. She defines quality of  life based on the following criteria: 1) distance from her home town; 2) cost of  living; 3) climate; 4) ease of  commuting to and from work; and 5) arts and recreational facilities. If  distance was the sole criterion, Boston would be her most preferred city, followed by St. Louis, Houston, and Los Angeles. The decision criteria and the alternative options were organized in a hierarchy in Figure 24.  value_normalizedbound =upper _bound,  if (value_original ==  (max(all values))lower _bound,  if (value_original ==  (min(all values))value_original − min(all _ values)max(all _ values)− min(all _ values) ,  otherwise⎧⎨⎪⎪⎩⎪⎪ 66Figure 24: Hierarchy of  decision criteria and alternatives for example AHP problem. Adapted from (Golden, Wasil, and Harker 1989).  The decision making involves making a judgement of  how well each of  the alternative cities meets each criterion. For criteria which can be represented by data that is measurable (e.g. distances between cities), AHP uses these data directly. For criteria for which direct measurements or other values are not available, pairwise comparisons are required to determine the relative priority (or importance) of  each alternative. Table 3 shows how the distance between Philadelphia and the 4 alternative cities was used to address criterion (1). Cost of  Living was also measurable; it was represented by the statistical measure of  Cost of  Living Index (COL). The corresponding ranking for the 4 cities with respect to the cost of  living is shown in Table 4. For the remaining three criteria ((3) to (5)), the graduate will need to apply judgement calls to rank each alternative. Judgements were made using a predetermined preference scale that is traditionally used in AHP decision-making problems (Table 5).
 67Table 3: Inverse distance to Philadelphia data. Table 4: Comparison of  cost of  living between alternatives Table 5: The AHP preference scale. In each case, cell values were populated based on how each row compared to each column intersecting the cell. For example, the climate in Boston was assessed to be strongly more preferable (=5) than the climate in Houston, which in turn was considered moderately less preferable (=1/3) than the climate in Los Angeles. Table 6 shows the result of  the pairwise comparison for the climate criterion. The chosen preference values indicate the decision maker likes sunnier weather — clearly, a decision maker who instead prefers more rainy weather would produce a very different comparison matrix. In the case of  the graduate above, example, the cities rank based on climate were calculated in order as Los Angeles, Boston, St. Louis, and Houston.  C.R. stands for consistency rate; it denotes how likely it is that the values in a pairwisecomparison matrix were selected at random. For the matrix to be considered consistent, C.R. must be less than 0.1.  68Table 6: City comparison w.r.t. climate. Table 7 shows the pairwise comparisons for the last 2 criteria (commute and arts & recreation), their corresponding C.R. values, and their normalized relative priorities. Not all criteria are equally important (Table 8). Their relative significance also needed to be calculated prior to deriving a final combination ranking (Table 9). In this example, Boston was slightly more preferable an option to Los Angeles, both of  which topped the candidate list.  Table 7: City comparison w.r.t. commuting and arts & recreation.Table 8: Comparison of  criteria w.r.t. goal. Table 9: Composite priorities for the cities. 
 69Fuzzy logic There is an inherent amount of  uncertainty in this research. The datasets used in the analysis are representations of  physical phenomena, not the phenomena themselves. The data was collected from multiple sources, and put together by different people. Hence, it is hard to be absolutely confident about their accuracy. On the other hand, the results of  most of  this analysis are floating points; they may imply a level of  accuracy that does not correspond to the confidence held in the data. To account for this, and in order to translate the end calculations of  combined weights into a more human-sensible scale, 5 fuzzy sets were developed and applied to most normalized variables with value ranges in the interval [0,1]. The fuzzy sets and their associated equations are shown in Figures 25 and 26 below.  Figure 25: Generalized fuzzy sets to characterized normalized values that sum up to 1. The fuzzy sets above represent the concepts of  Very Low, Low, Medium, High, and Very High degree of  belief  (or membership) in the generalized function, and they were defined by the equations corresponding to the trapezoidal function in Figure 25. One representative example where these fuzzy classes were applied was in the assessment of  completion for the harvested datasets.   70Figure 26: Equations for the completion fuzzy set. Completion, which is described in detail in the next section (3.3.3), is assessed based on the number of  NULL (or empty) values contained in a dataset. The completion rates of  the harvested datasets for geothermal production is presented in the Table 10 below, which is a copy of  Table 18, presented in detail in the next Chapter.  Based on the fuzzy equations defined above and the percentages of  completion calculated for each dataset, the following fuzzy classifications apply: Wikipedia (41% Medium, 59% High); Global Energy Observatory (100% High), ThinkGeoEnergy (100% High); OpenEI (91% Very High, 9% High); and the custom dataset created in this project (100% Very High). Table 10: Fuzzy classification for completion.  
 713.3.3 Data quality assessment methods  Data Quality Metrics The decision of  whether a candidate data set should be included or excluded from the model was based on the combination of  three metrics: completion [C], accuracy [A], and suitability [S].  Accuracy	 Accuracy [A] refers to the geo-referenced part of  the data, more specifically it is an assessment of  how close the location of  a specific feature in the data is, compared to the actual location of  said feature. The measure is not exact, as for most datasets the data provider indicates the level of  accuracy in general terms, for example as very high or unreliable. The assessment of  accuracy for the datasets used by the model was done following a thorough overview of  the data and was largely based on expert opinion.  Completion 	 Completion [C] in the measure of  the number (or percentage) of  NULL, empty or zero values contained in the data. Datasets with numerous columns or fields are not always consistently populated across all columns. For example, a dataset might have longitude and latitude noted on every row but only a small part of  the rows filled in for elevation. The completion of  the lat and long columns will be very high in this case and very low for elevation. Completion can be used to decide whether to include or exclude either specific columns within a set, or entire datasets depending on the severity of  incompletion on critically important columns. Typically the value of  completion is based on simple statistics. Suitability 	 Suitability [S] is the most subjective of  the three metrics, and it was defined based on expert opinion, which assessed how well a specific dataset fits the role it has been selected to play in the model. For example, the suitability of  the volcanoes layer is very high due to the fact that it is used to represent volcanism.  72All three metrics use a common index metric scale (Table 11). Table 11: Common index metric scale for accuracy, completion and suitability.  Calculating Confidence [ f  ] These three metrics eventually combine into the metric of  confidence, denoted by f  and calculated using AHP and pairwise comparisons. It is defined as:  (6) Confidence was calculated using the analytic hierarchy process (AHP) and pairwise comparisons, by evaluating the relative importance of  the completion, accuracy, and suitability metrics in the calculation of  the confidence metric. The pairwise comparisons, and hence the evaluation of  each metric’s importance, was largely based on expert opinion. Consistent with AHP, α, β, and γ were calculated using pairwise comparisons. They represent the relative importance of  A, C, and S and their values add up to 1 (Tables 12). They are calculated for each data layer incorporated into the model. For α=β=0.143 and γ=0.714, Eq. 7 becomes: (7) The confidence metric f  is simply a measure of  data quality, and it is used to characterize the datasets. It has no bearing on how important the data layers and the information layers derived from each dataset will be for the geomine model. It could very well be that an information layer that has only a limited bearing on the decision-making process on the geo-model was derived from datasets with very high levels of  confidence. f =αA + βC + γ Sf = 0.143A + 0.143C + 0.714S 73Table 12: Pairwise comparison matrix & results for completion, accuracy and suitability. 3.3.4 Geographical analysis methods Proximity analysis The term Proximity Analysis denotes the process by which ArcGIS calculates the distances between each feature in one layer to each feature in another layer. The result was used to calculate the (proximity-based) importance weights for the layers that were included in the GeoMine model. The importance of  volcano types for example, was calculated from the proximity analysis with GPPs as follows: (8) Here, CapacityGPP is the production capacity of  the GPP in MWe. The result matrix for the GPP-Volcano comparison returned more than 500K rows, as the distance from each of  the 314 GPPs to each of  the 1532 volcanoes had to be calculated. A similar process was used to rank individual properties with respect to their proximity to GPPs. The values of  proximity analysis were calculated as:  (9)	wvolcano_ type =CapacityGPPdistanceVolcano−GPP⎛⎝⎜⎞⎠⎟∑ volcano_ typeproximitydistance_interval =productionmineral ×CapacityGPPdistanceMine−GPP⎛⎝⎜⎞⎠⎟∑ distance_interval 743.3.5 Visualization methods Visualizing the data was essential to aid in the understanding and the discovery process as the research was being conducted. This strong focus on visualization is also evident by the more than 181 images that are included in this thesis, 170 of  which were created from harvested, downloaded, or calculated data, specifically for presenting the results of  the analysis. Different visualization tools were used, including as MS Excel, Tableau, ArchGIS, and circos.ca. Additional touch ups where were done in Adobe Photoshop and Adobe Illustrator as needed.  Most of  the visualizations presented in this thesis are snapshots/static exports of  interactive visualization tools, which were instrumental in discovering data trends and guiding overall data analysis.  3.3.5.1 Heat maps A heat map displays data graphically on a matrix using cell colour to represent the strength of  association. They can provide concise visual summaries for large datasets (Cairo 2016). Figure 27 redraws Tables 6 (Climate), 7 (Commute and Art & Recreation), 8 (w.r.t. Goal), and 9 (composite priorities) as heat maps using a common scale. Compared to the 4 tables, the heat map comparison below depicts differences and similarities between the tables in a more perceptible manner.  3.3.5.2 Treemaps A treemap is a special type of  heat map — where heat maps use colour to represent the strength of  association between two variables, treemaps use colour to distinguish between data groups, and area size to represent magnitudes. Primarily used to display hierarchical information, treemaps comprise of  nested rectangles that are sized proportionally to the corresponding data points they represent. Treemaps may be best suited for more generic judgements on data, such as providing an overview of  the general production trends for geothermal energy. The one in Figure 28 looks at production from just small country producers. 
 75
76Figure 27: Heat maps for city comparisons in the AHP example. HEATMAPS FOR AHP  PAIRWISE COMPARISONS 
77Figure 28: Worldwide distribution of  power generation capacity, per country and GPP type.3.3.5.3 Chord Diagrams Chord diagrams are an effective method of  visualizing interrelationships and reciprocal flow between entities, by comparing differences or similarities within a dataset or between different groups of  data. These interrelationships can be quantitative (e.g. similarity) or binary (e.g. is/isn't connected). Chord diagrams work best with tabulated information and when flow is primarily bidirectional. Nodes, which represent table rows and columns, are arranged around a circle and they are connected to one another using arcs. Arc size is drawn proportional to the cell values corresponding to the row-column intersection that is being visualized by each connection/node pair. Colour is typically used to distinguish different data categories and improve the readability of  the chart (The Data Visualisation Catalogue 2016; Brath and Jonker 2015; Krzywinski 2016). The chord diagrams included in this thesis (see example of  Figure 29) were created with the use of  the online interface of  Circos, a software package designed to produces circular visualizations. Circos diagrams are created by defining 2 text files: a configuration file that stores user preferences on the visual display; and a data file that contains information about the nodes and arc sizes, along with a small number of  optional display settings, such as colour, order, and default node size (Krzywinski 2016).  3.3.5.4 Slopegraphs Slopegraphs were first introduced by Edward Tufte in his 1983 book “The Visual Display of  Quantitative Information” (Tufte 1983). Originally termed table-charts, this new type of  graphic was useful for displaying information insights in 5 different ways: 1) the hierarchical ordering on items on the two axes; 2) the specific values associated with each point on each axis; 3) the manner in which values change (e.g. over time) for each point pair, through their slope; and 4) the way each point’s rate of  change compares to the rate of  change of  other points and to the average trend, again through their slopes. Figures 9, 10 and 30 are examples of  slopegraphs that depict change in geothermal production levels over time.
 78Figure 29: Current (2015) and projected (2020) contribution of small geothermal producers.CIRCOS DIAGRAM OF CURRENT VS. PROJECTED GEOTHERMAL PRODUCTION 79
80Figure 30: Current (2015) and projected world geothermal production levels (2020).
Only small producers are shown. SLOPEGRAPH OF CURRENT VS. PROJECTED WORLDWIDE GEOTHERMAL PRODUCTION3.3.5.5 Collaboration Graphs Collaboration graphs are used in mathematics and social sciences to model node networks (also known as social networks.) Facebook and Twitter are a type of  social network and its members and interactions can be modelled using collaboration graphs. The term social network is not equivalent to social media though. UBC’s campus computer network (intranet) is also a type of  social network. Collaboration graphs consist of  nodes, representing concepts such as users or computer terminals, and edges, which represent associations between two different nodes. They are typically used to measure the closeness of  collaborative relationships between the participants of  a network. The distance between any two nodes in a collaboration graph is called the collaboration distance, which is equal to the smallest number of  edges in the edge-path connecting them. For this analysis, collaboration graphs were created using OmniGraffle. The example of  Figure 31 depicts the conceptual degrees of  separation between geothermal power stations and the geothermal temperature gradient. A collaboration graph was created to depict the results of  the degrees-of-separation analysis for the geothermal potential indicators (Figure 63). Figure 31: Degrees of  separation between the GPP and ΔT. 
d(ΔT)=(3,3,3,4)DEGREES OF SEPARATION BETWEEN GPPS AND ΔΤ 81Chapter 4: Geothermal Production This chapter describes the process of  creating the geothermal production layer used in the geomine model, from data harvesting, to processing, and mapping. 4.1 Indicators of  geothermal production The previous chapter established that geothermal power stations would be used as primary indicators of  both geothermal production and geothermal potential. As such, the first step of  the analysis is compiling a list of  geothermal power stations from around the world, and using this list to create the GPP dataset that was subsequently used in the model. Due to just how critically important the GPP layer is to this analysis (it is used as a comparison benchmark for most other indicators), it was deemed necessary to derive as much of  a complete and accurate dataset as possible. This in turn required assessing the quality of  the resultant dataset, as well as taking any necessary steps to maximize it. This process is depicted in the flowchart in Figure 32 and described in detail herein.  4.2 Compiling the GPP data layer 4.2.1 Data sources for the GPP data layer  A number of  online sources were queried for geothermal production data. Four main providers were identified: 1) thinkgeoenergy.com; 2) the Global Energy Observatory; 3) OpenEI.org; and 4) Wikipedia. They are presented below, along with details on how data was harvested and processed in each case. An initial overview of  the data providers listed above indicated that there were distinct differences between the datasets, both in terms coverage (i.e. number of  rows/points stored in each set, and number of  fields held), and location accuracy. 
 82
83Figure 32: The geo-production dataset compilation process.thinkgeoenergy.com ThinkGeoEnergy is a leading provider of  news and research on geothermal energy. It was originally launched as a geothermal blog in 2008 and operates out of  Iceland. The site provides a “Plant Map”, which is based on Google maps and shows geothermal power stations around the world, complete with production information and type (Figure 33). Figure 33: Screenshot of  the ThinkGeoEnergy plant map. The data underlying this map is not directly accessible through download, for example as a KML/KMZ, XLS, or an XML file. The map was actually created as a Google map, hosted on Google servers, and was embedded within the thinkgeoenergy page shown above. Embedded map data can be acquired from the page, but that requires both scraping and scrubbing to turn them into a format that is actually usable by other applications. A screenshot of  the HTML source containing parts of  the data is shown in Figure 34 below. The Javascript code that defines one single point highlighted in blue (in this case the US-based Aidlin geothermal power station). The data was originally contained in a series of  HTML files that were captured and saved on disk. An AppleScript script was written to extract each power plant location, separate its attributes (e.g. latitude, longitude, and installed capacity), and export it in CSV format. MS Excel was then used to conduct a simple  84statistical analysis (primarily to assess the completion of  the dataset) before mapping (Table 13). Although this dataset was the most complete of  all the sources considered, it still had issues of  both completion and accuracy (Figure 35), and most significantly, it is not accessible to a researcher lacking web scraping and data scrubbing skills. This makes the site an excellent resource for the generalist, but for the purposes of  research, the inability to access the data directly through, for example, a download, is problematic. Figure 34: HTML source of  the ThinkGeoEnergy plant map page containing scraped data. As a final note, an older version of  this particular dataset can also be found on a separate Google map site that is titled Geothermal Power Plants and is identified to have been created by the thinkgeoenergy.com team (Figure 36). The downloaded data was actually much less accurate than the ones underlying the Plant Map on thinkgeoenergy.com, and it contains no information on either facility type or power production capacity. The statistical analysis and data mapping presented below was conducted on the version of  the data hosted on the ThinkGeoEnergy site proper.  85Figure 35: Incorrectly positioned GPP in the thinkgeoenergy.com plant map. Figure 36: Older version of  the thinkgeoenergy.com plant map data, on Google Maps.  Table 13: Fields, sample row and descriptive statistics for the thinkgeoenergy.com data. 
✔ ✘✔https://www.google.com/maps/d/viewer?mid=1kVn_quMo8uI18noPxbFEqK219gY&hl=en_US  86Global Energy Observatory The Global Energy Observatory is a project developed by and hosted at the Los Alamos National Laboratory in the US. The site hosts a collection of  global data on energy production, transmission and consumption. The data is open and accessible through a web interface; no direct access to the database is given. The site provides information on 95 geothermal power stations as tabulated links, each of  which is pointing to a page with more extensive information on each power plant (Figures 37 and 38). A script similar to the one used for the thinkgeoenergy.com plant map was written to save each page linked in the table, and subsequently scrape power station information, which was saved in CSV form. Descriptive statistics on the resulting dataset are given in Table 14 below.  Table 14: Fields and sample row of  the scraped Global Energy Observatory data.Figure 37: Screenshot of  geothermal listing of  the Global Energy Observatory site.  87  Figure 38: Power plant information page. One of  the scraped fields ([UnitNumber]) indicates the number of  installed units in each power plant. It returned equal to 1 for all scraped rows. This was obviously problematic, as unit capacity did not match the plant design capacity across all rows. A following calculated value for [Unit Capacity] was used instead, calculated as: ! 	 	 (10) [Number of Units] = CEILING [Plant Capacity][Unit Capacity]⎛⎝⎜⎞⎠⎟ 88OpenEI  OpenEI is an open wiki site containing crowdsourced data in a variety of  energy-related topics. It Geothermal Power Plant listing (http://en.openei.org/wiki/Geothermal/Power_Plant#tab=List_of_Plants) contains 252 entries, and includes data on the facility type, capacity, date of  operation and ownership (Figures 39 and 40). The data is open and crowdsourced, which means that its accuracy can be very problematic.  For the purposes of  this analysis though, as each row in the resulting dataset was eventually manually checked and verified, OpenEI was regarded as an acceptable data source. Lat/long locations were missing from the list but were imbedded in the map source. They were scraped by downloading each each power plant hyperlink page in the table using wget, scraping using AppleScript, scrubbing in Trifacta, and exporting to CSV, which was finally added to the dataset through the use of  the VLOOKUP( ) function in Excel. The updated sample row and descriptive statistics table is shown below (Table 15). Figure 39: Geothermal power plant listing on OpenEI (Information, 2016b).   89Figure 40: Map of  geothermal power plants on OpenEI (Information, 2016a). Table 15: Fields, sample row, and descriptive statistics for the supplemented OpenEI data. OpenEI is the only data provider of  the ones considered for this project that provided (some) information on the Geothermal Area and the Geothermal Region each power station belonged to. List of  geothermal power stations - Wikipedia Wikipedia also provided a limited listing of  a geothermal power stations with a generating capacity greater than 50 MWe. The wiki page title “List of  geothermal power stations” (https://en.wikipedia.org/wiki/List_of_geothermal_power_stations) contained a total of  68 power stations (as  90of  June 2016) complete with their production capacity and location coordinates. Figure 41 is a screenshot of  the wikipage. Table 16 provides a listing of  fields, a sample row, and basic NULL statistics for this dataset. The data itself  was embedded in HTML and required some text processing in order to extract the information from every row, convert coordinates in a lat/long format, and export all columns in an appropriate file format that could be read by the spreadsheet (i.e. CSV). !Figure 41: Screenshot of  the Wikipedia page on geothermal power stations. Table 16: Fields, sample row, and descriptive statistics for the Wikipedia data. 
 914.3 Assessment of  data quality for the GPP dataset  4.3.1 Accuracy Assessments of  accuracy for the source and resultant GPP datasets Accuracy was calculated for each dataset (denoted by the subscript “source”) by comparing the locations of  the GPP location to the corresponding location of  the same GPP in the resulting GPP dataset (denoted by the subscript “custom” (Eq. 11). The locations contained in the resulting set were assumed to be absolutely correct for the purposes of  this analysis, due to the fact that they were verified manually by visual inspection using Google Earth when the dataset was put together. Because at this stage in the analysis the data had not yet been inserted into ArcGIS, accuracy was calculated as the arithmetic difference between the corresponding latitude and longitude values (in decimal degrees): (11)  Calculations for Latitude and Longitude were conducted separately. The larger the [Δ], the greater the deviation from the exact position, the lower the assessment of  accuracy. Figure 42 shows the resulting calculations for ΔLat and ΔLong. Lat has the greatest amount of  deviation across all all 4 source datasets; the Global Energy Observatory data have the greatest number of  inaccurate Latitude locations and the OpenEI data have the greatest number of  inaccurate Longitude locations. For the final calculation of  accuracy for each dataset as a whole, the sum, mean and median values of  the Δ’s were also calculated. Figure 43, mapped in Tableau, shows just how differently some of  the same features are mapped by these 4 sources. The differences are more pronounced when viewing a map with a smaller rather than a larger scale. Iceland is used as the example in Figure 43.  ΔLat=Latsource−LatcustomΔLong=Longsource−Longcustom 92Figure 42: Relative accuracy calculation for Latitude values for the 4 source GPP datasets.  93
94Figure 43: Map of  geothermal installations in Iceland. Includes data 
sourced from thinkgeoenergy.com, the Global Energy Observatory, OpenEI, and Wikipedia.4.3.2 Completion Each of  the presented datasets was assessed in terms of  its accuracy, completion, and suitability, and a value for confidence was calculated (Table 17). The ThinkGeoEnergy dataset had the highest amount of  confidence due to a very high level of  completion. It did have some accuracy issues that had to be addressed before any data was passed on to model. The final GPP dataset is made up of  269 data points, of  4 different power station types that have, in all, a total of  11,010.31 MWe (or about 11 GWe) of  generation capacity.  The geothermal power plants were reclassified into 4 general types that were in turn grouped into 2 main categories of  power generation; namely: a high-temperature production category that included GPP types consistent with flash and dry-steam production, and a low-temperature production category that included GPP technology consistent with binary installations and with other, unconventional types of  geothermal electricity production. Figure 44 below illustrates this classification scheme.  As a final step, the accuracy, completion and suitability of  the source and resulting datasets was evaluated, so as to calculate their measures of  confidence. Completion was defined as 
1-NULL(%) for each dataset, where NULL(%) denotes the % of  NULL entries contained in thedataset fields selected as most pertinent to this analysis to make up each data layer (Table 17). Figure 45 shows the generalized fuzzy set, originally defined in Chapter 3, applied to the concept of  completion. Figure 46 lists the associated equations for the fuzzy competition set.  Table 17: Assessment of  completion for the four sources of  GPP data, compared to custom dataset.  95!Figure 44: Classification of  GPP types based on temperature. Figure 45: Generalized fuzzy set for completion.  Figure 46: Equations for the completion fuzzy set. High-T ProductionLow-T ProductionBinaryDry SteamFlash SteamTriple FlashDouble FlashUnconventionalBackpressure SteamDirect Steam TurbineDry SteamEGSHybridSingle FlashSteam with entrained water separated at wellheadBinaryWet Steam 96To simplify matters, intermediate classes between Very High, High, Medium, Low, and Very Low were defined using the classifiers Mostly and Somewhat, for completion values falling outside the 100% membership of  a particular class. For example, at 89.1% completion, the OpenEI dataset was calculated as 0.09 High and 0.91 Very High, based on the fuzzy class definition and its corresponding equation set (Figures 45 and 46). The corresponding fuzzy classification for the remaining dataset sources are listed in Table 18. With the simplified approach of  using the modifiers, OpenEI was classified as Mostly High instead; Table 19 lists all dataset sources, sorted order of  decreasing completion, along with their simplified completion classifications.  Table 18: Fuzzy classification for completion.  Table 19: Assessment of completion, sorted in order of decreasing completion.  4.3.3 Coverage Coverage denotes the size of  a dataset, in terms of  the number of  non-null rows it contains. The Global Energy Observatory dataset is actually the second largest in terms of  coverage. The most complete was the ThinkGeoEnergy set, indicated by the longest line forming on the x-axis — this assessment also agreed with the previous assessment of  completion above. The Wikipedia dataset seemed to be the most accurate of  the 4, based on the % of  its points that laid on the x-axis, but it was also the smallest of  them, providing only about 1/5 of  the coverage of  the largest one (i.e. the ThinkGeoEnergy set). Table 20 below tabulates the calculated coverage for each dataset examined.  97Table 20: Coverage calculations for the GPP source and resultant datasets. 4.3.4 Suitability As per its definition in Chapter 4, suitability is a subjective metric that is defined on expert opinion. In this case, geothermal power stations have been used to represent geothermal electricity production. Unsurprisingly, most of  the datasets were determined to be of  Very High suitability in this context. The one exception to the rule is the Wikipedia dataset, due to its very low coverage/completion. The suitability assessment for each dataset is shown in Table 21 below.  Table 21: Assessment of  suitability for the source and resultant GPP datasets. 4.3.5 Confidence As stated above, confidence has been defined as a combination metric of  accuracy, completion and suitability, and it is given by Eq. 7. Using the calculations from the previous sections, confidence was assessed for all source datasets and for the resulting data layer (Table 22). Table 23 provides a list of  fields, a sample row and descriptive statistics for the GPP layer. Table 22: Assessment of  confidence for the four sources of  GPP data, compared to custom dataset.  98Table 23: Fields, sample row, and descriptive statistics for the resultant GPP data layer. 4.4 Analysis of  the resulting GPP Dataset The GPP dataset that resulted from the process described above is described in more detail in this section. Descriptive statistics were drawn to help understand data trends. Table 24 and Figure 47 show an initial assessment of  the collected data. Table 24: Summary statistics for resulting GPP dataset. The initial analysis indicated that flash stream power stations are the most prevalent in the industry; they also hold the greatest cumulative amount of  power generation capacity. Binary plants are the second most installed unit type worldwide, accounting for 34% of  the total count. Dry-steam plants are third in the list, representing only 19% of  installations. Unsurprisingly though, the 59 dry-steam plants hold a greater total capacity than the 103 binary plants — by definition, the dry-steam powering the dry-steam installations hold much higher enthalpy than the hot water typically utilizing binary plant technology. Dry-steam resources have a higher enthalpy content than flash-steam. Flash- 99steam installations still dominate because they are more prevalent in the industry. In fact, dry-steam resources require very special conditions to occur, and as a result, they can be found only in limited and specific areas around the world (more specifically in Italy, Indonesia, the US, and Japan). Figure 47: Total generation capacity per GPP type vs. count of  installation per GPP type.  Online GPPs only in this graph.  4.4.1 Scaling the GPP data for GPP type importance and mapping An importance weight for each GPP type, denoted wgpp, was also calculated and subsequently used to transform the data by scaling capacities to account for GPP type importance. Defining the total capacity over the total count per GPP type, gives an indication of  which GPP type is most “potent” capacity-wise. A metric was defined for this purpose by taking the ratio of  average capacity and average count, for each type of  GPP. The metric was equivalent to a capacity “density” for each type, and it was denoted wgpp in the model.  100	 (12) The result of  this calculation is shown in Table 25 and Figure 48 below.  Table 25: Average count and average capacity of  geothermal power stations per installation type.  Flash-steam plants, as a GPP type, generate the most amount of  power capacity with a capacity “density” = 1.41. Dry-steam plant are close behind, with a ρ of  1.29, or 91% as important as that of  dry-steam plants. Binary plants have a capacity density of  0.37, close to 1/4 of  the dry-steam ρ. Figure 48: Normalized capacity and count per GPP type. “Capacity density” ρ in insert. ρcapacity =(c)typeΣ(c)(#)typeΣ(#)⎛⎝⎜⎜⎜⎞⎠⎟⎟⎟ρ capacity 101wgpp = ρcapacity =(c)typeΣ(c)(#)typeΣ(#)⎛⎝⎜⎜⎜⎞⎠⎟⎟⎟Unconventional plants (here denoting hybrid bio-geothermal and EGS) are last at ρ=0.13. They are also plotted in Figure 48. The recorded capacities and counts were summarized by country producer. The percent difference between the normalized capacity and normalized counts are plotted in Figure 49. Producers that are higher of  the graph have installations that generally are of  a higher capacity density, i.e. they are more “potent”. The opposite also applies. Surprisingly, Italy was closer to the middle of  the chart, even though its power is generated almost exclusively from dry-steam resources. This was attributed to the fact that most dry-steam installations in Italy are smaller in relative capacity compared to the large production-levels attained by country-producers with primarily liquid-dominated resources. The calculated values for wgpp were incorporated into the data and used to scale production for each GPP.  Figure 50 summarizes production by country and type. Figures 51 to 60 are maps of  geothermal production from around the world. They are based on the unscaled GPP data, and were created using Tableau and Adobe Illustrator. Figure 49: Cumulative capacity vs. cumulative count of  geothermal installations by country.  102It is important to note that geothermal production maps of  printable quality that are accurate, complete, and with a worldwide coverage are not currently available in the literature. These maps were created to directly fill this gap. Table 26: Average count and average capacity of  geothermal power stations per country.  103
104803Figure 50: Worldwide geothermal generation capacity.  
105Figure 51: Geothermal production in New Zealand.  
106Figure 52: Geothermal production in Japan. 
107Figure 53: Geothermal production in Iceland. 
108Figure 54: Geothermal production in the Philippines and Indonesia. Figure 55: Geothermal production in the US State of  Alaska, and in Russia. Figure 56: Geothermal production in Hawaii, USA.  109Figure 57: Geothermal production in Kenya and Ethiopia. Figure 58: Geothermal production in Turkey.  110
111Figure 59: Geothermal production in Central America. 
112Figure 60: Geothermal production in the Continental USA. Chapter 5: Indicators of  Geothermal Potential This chapter focuses on the evaluation of  geothermal potential and uses numerous different indicator layers to identify and eventually map high-geo-potential areas from around the world. The analysis begins with an assessment of  the geo-potential from production plants using the GPP data layer that was presented in the previous chapter. It then extends this assessment by adding secondary indicators to the map, and assess them in turn for their relative geo-potential indicator strength.  5.1 Using geothermal production to indicate geothermal potential The strongest indicator of  geothermal potential (or geo-potential) is the presence of  an existing geothermal power plant (or GPP), mainly due to the amount of  time, effort, and expenditure that is required to build operating geothermal power plants. Therefore, GPPs were assigned an indicator strength of  100% (or 1.0). The complete and up-to-date list of  geothermal power plants (GPPs) presented in the previous chapter was used in this section to represent geo-potential with the highest level of  confidence.  In addition, production levels were assumed to be relative to the levels of  geothermal potential held by the resources they tap into. This analogy is not exact for a number of  reasons; for example, resources are typically utilized locally and as such, local demand will play a great role in determining production levels. Achievable production rates also depend on the methods employed in harvesting the geothermal fluids, on whether the project has a re-injection program in place, or on which stage of  its development the overall project is (among other factors). But in general, it is assumed (admittedly somewhat simplistically) that a larger geothermal power station would indicate the presence of  a larger resource/reservoir. In ranking resource potential, demand-side effects were generally ignored and GPP MWe production capacities were used to rank geo-potential based on production levels.  113Unconventional plants were excluded because, by their definition, they do not depend on the type of  the reservoir they are taping into. EGS installations are defined on the premise that drilling deep enough will eventually yield adequate temperature levels suitable for production even in areas with the typically unsuitable for conventional production average geothermal gradients of  30℃/km. In EGS systems, the other two required elements of  a geothermal system, namely water and permeability, are created at the target depth using hydrothermal fracturing.  The question was posed on whether binary plants should also be excluded from this analysis. Some binary plants are indeed installed at the end of  a flash-steam cycle as either bottoming or topping units, to increase the overall output efficiency of  the power plant. In this case, the choice to install a power plant was not directly related to the nature of  the reservoir produced and would therefore not be very informative in an analysis that uses the type (and production level) to provide context for the local geo-potential. To answer this question, the distribution of  binary plants (in terms of  total counts) was calculated, to identify how many of  the binary plant installations are independent units and how many are installed alongside a unit of  different type (i.e. flash-steam or dry-steam). The GPPs in the production data layer were grouped into complexes, based on their locale (and ownership); for example (Figure 61) the 15 geothermal installations within the Taupo Volcanic Zone (TVZ) in New Zealand includes 2 complexes: Kawerau, which has 1 flash-steam and 2 binary units, and Wairakei, which has 3 flash-steam and 1 binary unit.  The complexes were then analyzed for heterogeneity (Table 27), with data indicating that the majority of  complexes are actually of  a unique type, and that only 12% of  the binary plants are installed within a complex that contains other types of  GPPs; in fact, 88% of  binary plants are unique installations. Excluding them from the analysis was therefore deemed imprudent, as their presence is actually indicative of  favourable geothermal potential. The GPP type breakdown is also charted visually in Figure 62.   114Figure 61: Geo-production in the TVZ - individual GPP units and multi-type GPP complexes.  Table 27: Constitution of  GPP complexes, by type.   115
116Figure 62: GPP distribution by type and generation capacity.Before transforming the GPP data points in the geo-production layer into geo-potential data points in the geo-potential layer, the following parameters had to be accounted for: Installation Type. The type of  GPP that can be installed in a particular location depends on the enthalpy content and achievable mass flow rates. The GPP layer created in Chapter 5 grouped GPP installations into 4 specific types based on a pre-defined classification scheme, namely: dry-steam, flash-steam, binary, and unconventional. An importance weight for each of  these types was also defined (as capacity density) and calculated. The capacity density for each of  the 4 types was based on the number of  GPPs installed per type and the total amount of  capacity they generate. The final ranking of  GPP types was based on this capacity density, which expresses the relative strength of  GPP generation technologies (Table 28).  Table 28: Ranking of  GPP types based on each type’s calculated capacity density value.  Individual GPP (MWe-) Capacity. Based on the assumption that GPP capacity levels can be used to indicate the size of  the resource from which they produce, i.e. that the achievable production levels (in MWe) correspond to the size of  the resource they tap into, the capacity of  each GPP must be taken into consideration when ranking the GPP layer entries and converting them to geo-potential. Capacity density was used to scale production levels for each row in the geo-production dataset, to account for the importance of  their particular GPP type. A sample of  this calculation is shown in Table 29. The resultant scaled capacities was subsequently normalized (/max) and used as the final geo-indicator weights. The overall effect of  scaling based on capacity densities was boosting the original capacity of  the flash and dry plants and depressing those of  binary and unconventional plants. Dry-Steam Flash-Steam Binary Unconv’lCapacity density 
ratio: ρc = Σ(rc)/Σ(r#) 1.293 1.411 0.372 0.13 117Table 29: Sample of  rows showing calculation of  geo-potential for the GPP layer. 5.2 Identifying and classifying additional geo-indicators Using just GPPs to indicate geothermal potential is problematic; it is relatively straightforward to presume geothermal potential exists everywhere where there is geothermal production, but the opposite does not hold true: not all areas of  geothermal potential are currently being actively produced. For example, although Chile hosts high-enthalpy geothermal resources, identified through extensive exploration activity, the country has no production. Thus a model that uses GPPs as the only geo-indicator would severely underestimate Chilean geothermal potential. For this reason, there is a need to identify alternative strong geo-potential indicators other than GPPs.  The initial data needs assessment listed 5 different datasets that were deemed necessary to represent geothermal potential; they were: (1) planned, under-construction, and currently operating power stations representing geothermal production; (2) volcanos and (3) earthquakes representing geological setting conditions; and (4) surface heat flow; and (5) thermal springs, representing surface manifestations of  the geothermal resource itself. The aim of  this chapter is to rank geo-indicators with respect to their relative importance. The aim of  this section is to simplify their ranking by first separating geo-indicators into two groups: primary and secondary. Primary geo-indicators can be used to indicate the presence of  geothermal potential with the highest level of  confidence, because they are known to be directly associated with geothermal resources. Secondary geo-indicators can also be used to indicate the presence of  geothermal potential, albeit at a lower confidence level, due to the fact that they are known to be only indirectly associated with geothermal resources.  1185.2.1 Degrees of  separation In order to classify the distinction between primary and secondary, geo-indicator association was mapped and assessed based on the concept of  “degrees of  separation”, also known as “collaborator distance” in the field of  Mathematics (Girvan and Newman 2002). Figure 63 shows a collaboration graph that maps the manner in which the identified geo-indicators relate to geothermal potential. A “Degrees of  Separation” analysis was conducted to calculate the relative importance of  each of  these indicators with respect to their capacity to indicate geo-potential, or their relative strength/importance as geo-indicators. Each node on the graph represents a geo-indicator. The links connecting the nodes represent conceptual associations between indicator features. The distance between two nodes in a collaboration graph is called the collaboration distance. The collaboration distance between two distinct nodes is equal to the smallest number of  edges connecting them (Yegnanarayanan and Umamaheswari 2011). The maximum distance between any pair of  nodes in the diagram is called the “diameter” (Easley and Kleinberg 2010).  The resulting distances were inverted to compute the importance (ranking) weights for each layer (Table 30). The most important layer is GPP, with an importance w=1; the least important layer is thermal conductivity (k) with a importance w=1/6. Table 30: Relative importance scales for all geo-indicator layers.  119!Figure 63: Degree of  separation analysis for the indicators related to geothermal potential.
 1205.3 Using volcanoes to indicate geothermal potential Due to how closely associated as phenomena volcanism and geothermal potential are, volcanoes were examined in detail to determine their relative importance as geo-indicators. The objective here was to assess whether specific types of  volcanoes could be used as strong indicators of  geo-potential, ideally as strong as GPPs. For this, three types of  analysis were utilized: visual comparison, buffer analysis, and proximity analysis.  5.3.1 Comparison between GPPs and volcanoes The initial comparison was visual, aimed at identifying apparent patterns in the geographic distribution of  the two types of  features. The relationship or association between the two feature types was classified as either Tight (or Strong), Loose (or Weak), or Mixed (or Undeterminable).  5.3.1.1 GPPs in general vs. volcanoes in general Figure 64 superimposes the association of  volcanoes and GPPs. Specific subtypes were marked using different markers and colours. The visual comparison is not 100% clear; although there seems to be considerable amount of  co-occurrence in many areas (with installation closely following volcanic arc “directional trends”), some GPPs have been installed away from volcanic centres. Conversely, many volcanoes are not located near a production facility. To assist in the visual differentiation, information about subtypes were removed from the map. Using a single marker for volcanoes (in general) and a separate single marker for GPPs (in general), both features were mapped as overlapping layers to identify areas of  co-occurrence (Figure 65). Even with a cursive look, it is easy to see there is considerable overlap between GPPs and volcanoes. This is clearest to see in the case of  the Philippines, Indonesia, Japan, and New Zealand, where installations are located very close to volcanoes. There is also excellent overlap in the East African Rift Valley, as well as in Central America and in Papua New Guiana.  121There is a good agreement in Mexico with the exception of  a single site. Hawaii and Iceland, both hotspots, also show good agreement; so does Italy. Overlap in the USA is not perfectly matched, with seemingly more dispersed production in relation to the location of  volcanoes. State-wise, there seems to be a weak association between volcanoes and GPPs in Idaho, Utah, and Oregon; California fairs better in this regard, but installations in Nevada are not clearly associated to the presence of  volcanoes. Other areas do not display the same tightness between volcanoes and GPPs, for example in Germany and Turkey, where installation are not adjacent to a volcano, at least not as close as in the examples seen above. The installations in Australia also appear to be independent of  the presence of  volcanoes. Figure 64: Simple visual overlap between volcanoes and GPPs worldwide.
Volcanoes & GPPsVolcano GPP 122
123GPP TypeFlash SteamBinaryDry SteamUnconventionalKindGPPVolcanoComplexCrater rowsStratovolcanoCalderaExplosion craterFissure ventLava domeMaarShieldSubglacialSubmarineTuff RingUnknownVolcanic coneVolcanic fieldVolcano TypeVolcanoes & GPPsFigure 65: Volcanoes and GPPs around the world. 5.3.1.2 GPPs by type vs. volcanoes in general In order to investigate the underlying differences between areas of  strong/tight and weak/loose association between GPPs and volcanoes, GPPs of  different types were separately mapped alongside volcanoes in general, and the general occurrence trends were visually contrasted and compared: flash-steam GPPs (marked in blue) versus volcanoes (Figure 66), dry-steam GPPs (marked in red) versus volcanoes (Figure 67), unconventional GPPs (marked in purple) versus volcanoes (Figures 68), and binary GPPs (marked in yellow and orange) versus volcanoes (Figure 69). All volcanoes, regardless of  volcano type, are marked in brown.  A very strong/tight visual association is evident in both the flash- and dry-steam maps. Unconventional are by definition unrelated to resource type, as they are primarily EGS installations (with the exception of  a single hybrid/biomass plant). The associations between volcanoes and binary plants was more interesting; in some areas (e.g. in New Zealand, Japan, and the Philippines) the visual association was generally tight, while in others (e.g. in Nevada, USA and in Germany, Austria, and France) it was far more loose.  Mapping independent binary installations (marked in yellow) separately from the plants that are part of  a complex (marked in orange) indicated that the loosest association was between volcanoes and independent binary GPPs; complex binary plants were tightly associated with volcanoes. The latter is unsurprising; complex binary plants are installed as secondary units, in order to augment the production from a primary flash- or dry-steam plant. As both flash- and dry-steam plants are tightly associated with volcanoes, and since the primary installation would be dominant, their associated binary counter-plants are also tightly associated with GPPs. The independent GPPs in Figure 69 are of  mixed association; for example there is a good overlap in Japan, the East African Rift Valley, California, the Azores, and Central America, while installations in Nevada (marked [A] on the map), Germany/Austria ([B]), China ([C]), and Australia  124([D]) seem to have very little overlap with volcanoes. The reason for this may lie in the geological setting within these areas, data for which have been excluded from this analysis, primarily for practical reasons: geological data is typically 3-D (and this analysis in limited to 2-D datasets), they are not freely available for all the areas included in the analysis (which has a global scope), and typically require specialized knowledge to analyze (which is counter to one of  the main objectives of  this project: to work with primarily non-expert, freely available datasets). Regardless of  the underlying reasons, it is clear that in terms of  using GPPs to indicate geo-potential, flash-steam, dry-steam, and binary systems that are part of  complexes are very strong indicators of  geo-potential; independent GPPs, although they should not be ignored by the analysis, were considered as weaker indicators of  geo-potential. In any case, all GPPs were scaled by a factor of  1.0 in the final geo-potential dataset, based on the importance weights calculated as part of  the degrees of  separation analysis.  Figure 66: Overlap between volcanoes and flash-steam GPPs.  125Figure 67: Overlap between volcanoes and dry-steam GPPs. Figure 68: Overlap between volcanoes and unconventional GPPs.  126Figure 69: Overlap between volcanoes and binary GPPs. 5.3.1.4 Areas of  high production Geothermal production is often reported by country producer in the literature. In order to investigate the connection between production levels and physical proximity to volcanoes, the country of  location was generated using the lat-long coordinate pairs in the production dataset (using custom automation coding and Google’s inverse geolocation API) and added to the data. Production capacity sums were therefore calculated for each country producer, and the minimum distance between volcanoes and the GPPs located in each country were extracted from the proximity results and plotted (Figure 70). The log-log plot in Figure 71 also shows the power trend line that was fitted to the data. The trend had a distinct downward slope indicating that for the country-summarized data, production is inversely proportional to the min distance between GPPs and volcanoes for the same country; in fact, the largest country producers (in total MWe) have the smallest minimum distances from volcanoes. This is a logical conclusion, as it was expected that the highest production  127would be physically associated with areas of  high volcanism, and it was another indication that volcanoes are indeed very strong indicators of  geothermal potential. The chart in Figure 72 was drawn using the same data as Figure 71. The difference in this case is that the data was grouped first by GPP type and then summarized by country of  locale. This resulted in more plot points; for a country that has GPPs of  all 4 types, 4 points were generated for the chart. The GPP-type breakdown was done so as to investigate whether all types of  GPPs follow the general country-based trend with distance from volcanoes. The downward trend holds true for flash-steam GPPs, flash-steam + dry-steam, and for independent binary plants. The trend for complex binary plants and for dry-steam GPPs is also horizontal, indicating that there is little correlation between GPPs production levels and min distances between GPPs and volcanoes. The flash-steam and the flash-steam + dry-steam trend lines match the overall trend best. The reasons behind the horizontal slope of  the dry-steam GPPs is thought to be 2-fold; on the one hand, dry-steam GPPs are all located very close to another and in all cases, in very close physical proximity to a volcano (Figure 70). This is true for Larderello complex, whose plants are installed within 14 km of  the caldera of  Larderello volcano (and as close as 1 km from its centre), the Geysers, Japan and Indonesia. In fact, the areas covered by these GPPs are a very small fraction of  the total area of  the country. In addition, flash-steam systems dominate the data, as they represent 63.5% of  the total MWe produced (compared to 24.1% for dry-steam). The almost horizontal slope of  the complex binary was attributed to the fact that the decision to supplement flash- or dry-steam GPPs with a topping of  bottoming binary unit is primarily dependent on surface-side conditions (such as the technical characteristics and turbine efficiency of  the main installation unit, atmospheric conditions affecting the thermodynamic cycle of  the cold reservoir side, and economics). Complex binary plants also represent a very small portion of  the data (3% of  production), so they would not affect the overall trend much. They were therefore not expected to directly relate to reservoir conditions. 
 128The slope of  the independent binary trend line is a bit more steep, indicating a slightly higher sensitivity; this makes sense considering the fact that most of  these installations are of  lower temperature and at shallower depths than flash. The overall trend is also strongly downward, and it is not greatly affected by the dry and independent binary plants because they account for a rather small portion of  the overall data (Table 31). Table 31: GPP types per MWe-production.  !!Figure 70: Physical proximity between dry-steam plants and their closest volcano. 
 129
130Loose/Weak association 
between volcanoes and GPPsStrong/Tight association 
between volcanoes and GPPsFigure 71: Σ(Capacity) vs. min(GPP-Volcano distance), using single closest volcano to each GPP.
131Figure 72: Σ(Capacity) vs. min GPP-Volcano distance, country totals per GPP type.5.3.1.5 GPPs by type vs. volcanoes by type In order to investigate if  specific GPP types are more closely associated to geo-production (and by inference geo-potential), volcanoes and GPPs were superimposed one more time, including information about volcano type (Figure 73). In this example, GPP type is indicated by colour, as is volcano type. Not all volcanoes are included; only the types that best match GPPs and reduce the amount of  noise (i.e. non-matching volcanoes) are shown. The map was created in ArcGIS and the selection of  volcano types best matching the geographic distribution of  GPPs was selected visually, by running consecutive queries on the data and examining the overlap of  each type. In this comparison, the auxiliary attributes contained in the volcano dataset that give rock type, tectonic setting, and crust type were used as filters in the queries to further reduce result returns.  The combination of  volcanoes best matching GPPs while reducing non-matching returns (noise) is shown graphically in Figure 74. This chart was created by listing the query conditions used to filter out volcanoes. Those returned as best-matched were classified as Primary; the rest were marked as Secondary.  Figure 75 was created in Tableau; it is an attempt to graphically represent 5-dimensional information; the dots represent a particular combination of  volcano type and rock type. Only those types that made it to the Primary volcanoes query is shown in the graphic. Each dot was then replaced with a pie chart, which show the crust type and tectonic setting that combined and was included in the query for each particular volcano-rock type combination. For example, andesitic, basaltic, rhyolitic, trachybasaltic, and trachytic maars of  all tectonic settings and all crust types are returned in the primary results (as indicated by row 4 in the graphic). In contrast, andesitic and basaltic shields are included only if  they are associated with intraplate continental or intraplate oceanic crust (first and second pie dot in row 5 of  the graphic).   132This classification scheme ended up splitting the volcano dataset almost precisely in half; 765/1531 volcanoes were classified as Primary and 766/1531 volcanoes were classified as Secondary; a 50/50 split. Half  of  the volcanoes were subsequently scaled using a higher importance weight, supplementing the sparse geo-indicator layers created using the relative small number of  GPP-derived indicator point, which were essentially tripled in number with the addition of  765 primary volcanoes to the 314 GPPs making the primary indicators dataset.  The importance of  all volcano types was calculated using proximity analysis (see next section), and subsequently scaled based on an adjusted relative importance weight derived from the degrees of  separation analysis. The adjusted weights are shown in Table 32; the main difference with the ranking dered from the degrees of  separation analysis is that volcano types marked as primary were assigned a relative importance value of  1.0 (same as GPPs), while the rest were scaled by 0.5. Table 32: Updated list of  relative importance weights.   
 133
134Figure 73: Volcanoes with combination attributes best matching GPP distribution.
135Figure 74: Classification of  volcanoes into primary and secondary indicators. 
136Figure 75: Combination of  volcano attributes best matching the distribution of  GPPs. 5.3.1.6 Validating the primary/secondary volcano classification An attempt was made to further validate this design choice with supporting evidence. Buffer analysis was used to derive the distribution of  earthquakes around areas of  geo-potential, once using a set of  nested ellipsoid buffers created around GPPs, and a second time using nested ellipsoid buffers created around volcanoes marked as primary. A model was created in GIS to generate the buffers and another to iterate through each buffer, select the data and export it (Figure 76). Further code was written in Automator to go through and merge the returned files, process them accordingly and generate the final list. Returned quakes were contained within the doughnut-shaped buffers of  a particular radius, for each buffer used in the model. The returned data included the buffer radius and the quake IDs, which were used in turn to populate earthquake magnitudes corresponding to each return quake, from the original data. At this particular point, the number of  earthquakes at each magnitude was used to create the plot matrix shown in Figures 77a, 77b, and 77c. The trends of  the data returned by the two buffer analyses were very close and supported the design decision to use primary volcanoes with the same importance as GPPs as geo-indicators, and validated the use of  primary volcanoes as primary geo-indicators. 5.3.2 Ranking volcano geo-indicator strength Proximity analysis was used to rank the different volcano types. Their ranking was based on a calculation of  importance weights for each volcano type and took into account both the capacity of  the GPP and the distance between GPPs and volcanoes (Eq. 13). A proximity matrix was generated by calculating the distances (in km) between each GPP and each volcano. The schematic of  this is shown in Figure 78, for one example volcano; distances from each volcano in the analysis to every GPP in the production data set was returned, corresponding to 480,734 distances between 314 GPPs and 1531 volcanoes.
 137
138VolcanoFigure 76: Nested buffers of  increasing radius around GPPs and their overlap with volcanoes. Figure 77a: Distribution of earthquakes around GPPs and primary volcanoes (5.0≤M≤5.4). 139Figure 77b: Distribution of earthquakes around GPPs and primary volcanoes (5.5≤M≤7.9).  140Figure 77c: Distribution of earthquakes around GPPs and primary volcanoes (8.0≤M≤8.9).  141
142Figure 78: Proximity analysis setup between GPPs and volcanoes. 	 (13) Tables 33 and 34 list the data that was used to create the next three Circos diagrams, which visualize the proximity between GPP and volcano types. The first chart (Figure 79) is based on the first 4 (numerical) columns of  Table 33. The second chart (Figure 80) was based on the same data; it was drawn with normalized band widths for all data rows and columns, in order to better visualize the contribution breakdown for the smaller bands. The third chart (Figure 81) expands on the first by including the results calculated for the auxiliary volcano attributes, given in the first 4 numerical columns of  Table 34.  Basic versions of  these graphs were created using the tableviewer tool on circos.ca, exported in SVG format, and subsequently edited (in terms of  its colours and fonts) in Adobe Illustrator for readability and data visibility. The analysis showed that stratovolcanoes was the volcano type most closely associated to high capacity production, and particularly flash-steam production; this is indicated by the widest band on Figure 80. Flash-steam production is dominant for the proximity to most volcanoes types; other than explosion craters, which are exclusively associated to dry-steam production, the widest band indicating GPP type on all other volcano types in Figure 81 is the blue band (representing flash). When including the auxiliary attributes in the analysis, the following attributes are indicated as prominent: for flash-steam, continental crust, andesitic and basaltic rocks, and subduction zones; for dry-steam, continental crust, andesitic and basaltic rocks, and rift zones; and for binary, continental crust, andesitic and basaltic rocks, and subduction and rift zones. The bands leading to crust type, volcano type, rock type, and tectonic setting are of  equal width, because their analysis is based on the same GPP-volcano pairs. 
wvolcano_type =wGPP ×CapacityGPPdistanceVolcano-GPP⎛⎝⎜⎞⎠⎟ volcano_ type∑ 143Table 33: Summary proximity matrix for volcano-GPP comparisons. Table 34: Summary proximity matrix for volcano-GPP comparisons — attribute columns.  
 144Figure 79: Proximity between GPP types and volcano types.  145Figure 80: Proximity between GPP types and volcano types, with normalized band widths.  146Figure 81: Proximity between geo-production and volcano attributes. 
 147The final importance weights for volcano types were calculated using the proximity sums of  all three GPP types, inverted and normalized. They are shown as attribute pairs (corresponding to the attribute combos used to reclassify volcanoes into primary and secondary indicators) in Table 35.  Table 35: Final importance indicator weights for volcano types. 5.4 Using thermal springs to indicate geo-potential Geothermal surface features are surface manifestations of  sub-surface geothermal systems. They are direct by-products of  the existence of  geothermal systems in their locality and can be used to infer information about the characteristics of  the sub-surface systems and conditions. Using fundamental techniques, such as mapping feature location, heat output, and chemistry, they can support management decision-making by providing reliable information about how a system behaves, how it responds to production/utilization, and what its natural state may be.  Table 36 is a proposed classification system for different types of  geothermal surface features (Scott and Robson 2012; Scott 2012).   148Table 36: Proposed classification for geothermal surface features.  Adapted from (Scott and Robson 2012). At the time of  writing, only limited geo-referenced information was available on world-wide geothermal features. The most complete source was a dataset that lists thermal springs located in the contiguous USA, Alaska, and Hawaii, provided by NOAA (NOAA 2016). For this reason, surface features in this model are only represented by thermal springs. The original dataset had a number of  issues that required dealing with before using the data in the model. More specifically:  Limited geographic coverage. The scope of  the dataset was restricted to the USA alone. This was a problem due to the fact that the rest of  the data used in the model were global in scope. On the one hand, patterns of  geographic distribution and proximity evaluations are affected by scale, therefore analysis conducted on a limited dataset cannot automatically be used to draw conclusions on feature distributions and inter-relationships on a world-wide basis (Mitchell 2005). On the other hand, surface features are very closely associated with the two primary indicators of  geothermal potential, namely volcanoes and GPPs, based on the degrees-of-separation analysis results presented earlier in this chapter. Such data therefore, if  available, should be included in the model. Looking again at the relative distribution of  geothermal installations from around the world (Figure 82), out of  the 25 countries with geothermal production represented in the diagram, the US distribution in indicated to be one that most closely matches the worldwide average. Similarly, the distribution of  volcanoes also follows this trend Surface Geothermal Feature TypesFeature Type:Geyser Mud Geysers FumarolePrimary 
Flowing SpringsMixed Flowing SpringsMud Pots 00 Steaming GroundPrimary 
Non-Flowing PoolsMixed Non-Flowing PoolsMud Pool Heated GroundReservoir Type:Primary 
Geothermal FluidMixed/Diluted 
Geothermal FluidMixed Diluted Fluid and/or Steam HeatedSteam FedPlus landscape features like explosion craters, singer terraces, and others.  149(Figure 83). Just like the the distribution of  GPPs, the distribution of  volcanoes in the USA is the closest one to that calculated for the worldwide average. As volcanoes and GPPs have been determined to be both primary indicators of  geothermal energy, and because of  the similarities in the distribution of  these two features between the USA and the world-wide average, it was deemed appropriate to study the proximity of  thermal springs using the USA-limited dataset, and use the results on the limited scope analysis in the global model. A modified approach was therefore employed, whereby subsets of  the volcano and GPP layers were taken and used in conjunction with the USA-base thermal springs layer. The proximity weights derived from this analysis were only applied to the USA, as the only available data for the model was the USA dataset. It is assumed nevertheless, that should additional data become available in the future, that these weights would be applicable with a relatively acceptable degree of  confidence. Data consistency. Extensive cleaning and analysis was conducted to calculate the confidence value of  this dataset, and to decide whether or not to incorporate it in the model. The data in the NOAA dataset date back to 1980, with numerous researchers and geologists contributing to its compilation. In all, 3,771 thermal springs are listed. Each row contains information across 11 fields. Of  particular interest are the spring’s location, given as a lat/long coordinate pair, as well as its Name and Temperature (reported in both ℃ and ℉). The main issue with data consistency concerns the temperature field, and it was of  critical importance. The column actually contained 2 types of  entries: numerical values ranging from 20 to 133, corresponding to temperatures in degrees Celsius; and the 3 classifier values of  “W”, “H”, and “B”, corresponding to a classification of  Warm, Hot, and Boiling respectively. The rows had either one of  the two types, i.e. either a temperature or a class. More specifically, out of  1,661 springs, 1,480 had a numerical value for temperature, and 181 were classified as either boiling, hot, or warm (Figure 85). The classified rows represented about 11% of  the dataset. 
 150
151Figure 82: Country-based distribution of  GPP types. Figure 83: Country-based distribution of  volcano types. Ideally, any post-processing should retain as much information as possible and use as accurate a measure as possible — in this case, the numerical values of  temperature. Converting the non-numerical values in the data into a numerical value corresponding to one of  the three classifications would introduce too much uncertainty in the data. Only one of  the originally classified values was indicated to be Boiling, so the uncertainty associated with assigning its temperature value equal to the corresponding Tsat for its elevation would not be significant.
 152Table 37: Completion of  the non-numerical values of  the original thermal springs dataset. !!Figure 84: Histogram of  non-numerical values in the original thermal springs dataset.  Figures 84 and 86 are histograms of  the non-numerical and numerical values in the original thermal springs dataset respectively. Figure 85 shows their combined geographical distribution.
Boiling Hot Warm NULL 153
154Figure 85: Distribution of  numerical and non-numerical values in the original springs data. Table 38: Assessment of  completion for the numerical values of  the original thermal springs dataset. !!Figure 86: Histogram of  numerical values in the original thermal springs dataset. The rows indicated as Warm and Hot on the other hand represented 38% and 60% of  all of  the originally classified rows (or 4% and 7% of  the entire dataset respectively — Table 39). In addition, the classification Hot spans 44% of  the temperature range the numerical data lies in, between 59℃ and 96℃ (Table 40). Selecting a single value for the 38% of  the classified data marked as Hot would lower the accuracy measure of  this part of  the dataset considerably. The decision was  155made to transform the numerical values to classifiers instead, in an effort to maximize data coverage and completion rates. Relative high accuracy would also be retained, as for the purposes of  this rather higher-level analysis, a temperature class would be accurate enough to use in the GeoMine model. The steps taken to transform the data are illustrated in Figure 87 below.  Table 39: NULL statistics for the non-numerical values of  T in the thermal springs dataset.  Table 40: Temperature range for the classified values of the post-processed thermal springs dataset.  Initially, the Google Elevation API was used to retrieve the elevation values h corresponding to each lat/long coordinate pair. This was fed to Eq. 14, which returned the saturated vapour pressure Psat (for pure water) corresponding to that h. Psat in this case was assumed to be equal to the ambient pressure (in absolute bar) at elevation h.  (14) The saturated vapour temperature Tsat corresponding to that Psat was then retrieved using the Excel Macro add-on X-Steam (X-Steam, 2016). The final step was comparing Tsat to the temperature indicated for the spring, Tspring. If  Tspring>= Tsat, then the phase of  the fluid would lie above the saturation dome on the P-h diagram (Figure 88) in the superheated steam region, and the spring would be classified as Boiling. Otherwise, the spring would be classified as Hot. patm@h =101,325 × (1− 2.25577 ×10−5 × h)5.25588106 156"Figure 87: Data transformation process of  the thermal springs dataset. Figure 88: The pressure-enthalpy (P-h) diagram for water (Patsa, 2010).
 157Table 41 lists Completion values for the transformed data. The three fields of  particular interest (Lat, Long and Temperature Class) are all of  Very High completion. Figure 89 is the histogram of  the transformed and classified Thermal Springs Dataset. There was no clear-cut off  point for the classification boundary between Hot and Boiling (as it was a function of  elevation), so the two classes overlap in the bins that lie on the higher end of  the temperature scale. The final assessment of  confidence in the transformed Thermal Spring dataset is shown in Table 42; it was assessed as Very High. Figure 89: Histogram of  numerical values in classified thermal springs dataset.
20-29 30-36 37-49 50-59 60-69 70-79 80-89 90-99 100-109 130-139 150-160WarmHotBoilingSpring CountTemperature [℃]Histogram of Numerical Values in Classified Thermal Springs Dataset 158Table 41: Assessment of  completion for the transformed thermal springs dataset.  Table 42: Assessment of confidence for the thermal springs dataset.Dealing with outliers: Two outliers were identified in the transformed and mapped data points, indicating a temperature much higher than 100℃ (Figure 90). One of  the points pointed to Makushin volcano fumaroles in AK (53.982,-165.93), at a temperature of  154℃. Google maps of  the area put the given location off-shore (Figure 91). The actual location of  the caldera (centre) is given as ( 53.885312°, -165.929812°) by Google Earth, indicated at 1,720 m ASL, that is 11.20 km (ground distance) from the location in the dataset, at a 358 degree heading. The caldera was measured to be 2.14 km cross to 1.44 km wide at its widest points. Motyka and Ruscetta (1982) indicated that “temperatures of  fumaroles and steam vents were at or near the boiling point” and that “fumaroles and hot springs occur within the 2.5 km diameter of  the Makushin caldera”. A twitter image, posted on Sep 11, 2014 by photographer Roberto Carlos Lopez showing the steaming fumaroles within the caldera diameter was overlaid on the caldera location in Google Earth 
(Figure 91). 
 159
160Figure 90: Reclassification of  the thermal springs dataset (continental US subset).Figure 91: Actual and reported locations of  the Makushin volcano fumaroles .  3The temperature of  this point was therefore adjusted to the boiling point temperature at that elevation, namely to 94.25℃, which corresponds to the saturated vapour temperature at the saturated vapour pressure of  0.822 bar. Its exact position was also updated to reflect that location described in the source. The second outlier was the Sherman Crater fumaroles in WA, USA (48.77, -121.813) given at 130℃. Its location was accurate, and its elevation was given by (SummitPost 2016)at 3,098 m. The fumarole temperature was adjusted to the more accurate temperature of  89.64℃, which corresponds to the Tsat at that elevation for a Psat of  0.692 bar. The final reclassified version of  the thermal springs dataset is mapped in Figure 92. original locationcorrected location https://pbs.twimg.com/media/BxSQHleIAAAbCCI.jpg)3 161"Figure 92: Reclassification of  the thermal springs dataset (continental US subset). 5.4.1 Ranking the surface features layer The layer importance for the thermal springs layers was given as 0.50 by the degrees-of-separation analysis. Proximity calculations were conducted using GPPs that were located in the USA alone, as only US-based thermal springs data was available. Table 43 lists the results of  the proximity analysis and the final importance weights for the thermal springs temperature classes; it indicates warm thermal springs and dry-steam as the dominant attribute pair in this analysis. Table 43: Proximity analysis results and final weights for the thermal springs layer. !! 1625.5 Quakes and tectonic boundaries as geo-indicators The analysis of  the earthquakes and tectonic boundaries layers used the same approach for calculating proximity weights as the one used for volcano types. The geographic distributions of  earthquakes and tectonic boundaries were first examined visually, prior to calculating the importance weights of  each layer using proximity analysis. The two indicators were also contrasted visually together, because the phenomena they represented as strongly inter-related: the vast majority of  earthquakes are a result of  tectonic plate motion, the consequences of  which are in turn felt more frequently along tectonic plate boundaries. GPPs were the comparison basis for the proximity analysis of  both layers.  5.5.1 Tectonic plate boundaries, earthquakes and geothermal production Figure 93 is a map of  superimposed tectonic plate boundaries, earthquakes, and geothermal power stations. A casual inspection of  Figure 93 indicates that majority of  earthquakes can be found along tectonic plate boundaries, in an overlap trend similar to the one between volcanoes and tectonic plate boundaries. A number of  tectonic plate boundaries, primarily between oceanic plates, have no associated earthquakes; this is not due to a lack of  seismic activity in the ocean, but rather a lack of  monitoring equipment in those areas.  This lack of  data is not particularly limiting for this project, as the geographic scope of  this analysis excludes areas that are not in-land. Earthquakes are marked on the map using circles whose diameters are proportional to quake magnitude (a logarithmic index that expressed the level of  energy released by a quake). Boundary types are distinguished on the map by colour. 7 different types of  boundaries are included, namely continental convergent boundary (CCB), continental transform fault (CTF), continental rift boundary (CRB), oceanic convergent boundary (OCB), oceanic spreading ridge (OSR), oceanic transform fault (OTF), and subduction zones (SUB).  163
164Figure 93: Simple overlap between quakes, boundaries and GPPs. Earthquakes, tectonic plate boundaries and GPPsQuakePlate BoundaryGPPGPP BufferDry-SteamFlash-SteamBinaryThe next 10 figures show areas of  high geo-production, tectonic and seismic activity 
(Figures 94-107). Figure 94 shows the tectonic boundaries, earthquakes and geothermal power stations in New Zealand. Geothermal production is limited to the TVZ zone in the North Island. Both islands are seismically active, but the tectonic settings are different; in the North, the majority of  geo-production is situated along a continental rift boundary running parallel to a subduction zone. A small number of  earthquakes have been recorded within the area of  geothermal production, but the majority of  seismic activity in the North Island is concentrated between the subduction zone and the fault/boundary line stretching out between the TVZ and Wellington. The seismic activity in the South Island is associated with a continental transform fault in the Christchurch area, and the extension of  the subduction zone further south to the Fiordland National Park.  A similar situation in seen in Japan, one of  the most seismically active countries in the world (Figure 95). Both earthquakes and geo-production in Japan are recorded running mostly parallel to a subduction zone. This trend is also evident in the case of  Guadeloupe (Figure 96), where the only geothermal power station is located in an area of  concentrated seismic activity that runs parallel to a subduction zone. This is also the case for the adjacent Central American countries of  Costa Rica, Nicaragua, El Salvador, and Guatemala (Figure 97). Geothermal production and seismic activity in Sumatra, Java, and the lesser Sunda Islands in Indonesia, as well as production and seismic activity in the Philippines, also run parallel to a subduction zone (Figure 98). Other areas of  geothermal production portray a different profile. In Iceland for example, which is a combination of  an intra-plate hotspot and divergent boundary, the rather limited seismic activity is mostly concentrated along an oceanic transform fault running along the southern part of  the country. Both areas of  concentrated geo-production on the other hand are located along oceanic spreading ridges, in both north and south (Figure 99).  165Figure 94: Tectonic plate boundaries, earthquakes and GPPs in New Zealand.  Similar conditions to the Iceland are also seen in the Azores, with production being adjacent to an Oceanic Spreading Ridge and very limited seismic activity (Figure 100). Seismic activity on the Hawaiian hotspot is also limited; located mid-plate (and hence no-where near a tectonic boundary), Hawaii has geothermal production on the island of  Hawaii, in close proximity to the Kilauea caldera (Figure 101).  166Dry-SteamFlash-SteamBinaryQuakeGPPsFigure 95: Tectonic plate boundaries, earthquakes and GPPs in Japan.  167Dry-SteamFlash-SteamBinaryGPPsFigure 96: Tectonic plate boundaries, earthquakes and GPPs in Guadeloupe.  168Flash-SteamGPPsFigure 97: Plate boundaries, earthquakes and GPPs in Central America.  169Flash-SteamBinaryGPPs
170Plate Boundaries, Earthquakes, and Tectonic  Settings in the Philippines and IndonesiaFigure 98: Plate boundaries, tectonic settings, and GPPs in the Philippines and Indonesia.
171Figure 99: Tectonic boundaries, earthquakes and GPPs in Iceland. Flash-SteamBinaryGPPsFigure 100: Tectonic boundaries, earthquakes and GPPs in the Portuguese Azores. Figure 101: Tectonic boundaries, earthquakes and GPPs in Hawaii, USA.  172Flash-SteamBinaryGPPsGPPsFigure 102: Tectonic boundaries, earthquakes and GPPs in in the East-African Rift Valley. In the western United States (Figure 103), three distinct earthquake distribution patterns emerge. Dry-steam production at the Geysers is in close proximity to a continental transform fault and average levels of  seismicity. Further to the south, highly active seismic areas can be found along a continental convergent boundary. Geothermal production coincides with seismic activity close to the Southern border with Mexico.  173Flash-SteamBinaryGPPsFigure 103: Tectonic boundaries, earthquakes and GPPs in Western USA. Very limited seismic activity is present in Nevada where there is a lot production far away from a tectonic plate boundary. Production here is from blind resources. The main trends of  interest in this area are flash-steam production with seismicity over a CTF boundary, and dry-steam production with limited seismicity along a CTF boundary. Plate Boundaries, Earthquakes and Geothermal Production in Western USA 174Dry-SteamFlash-SteamBinaryGPPsFigure 104: Tectonic boundaries, earthquakes and GPPs in Mexico. Production in South Mexico is found along as seduction zone and is accompanied by seismicity; the subduction zone here is somewhat removed from the occurrences of  geothermal production (Figure 104). Further south, the trend of  high seismicity and high production levels alongside a subduction boundary is also evident. Some plants and earthquakes can be found along a continental convergent boundary, as well as a continental rift boundary.  The relative importance of  earthquake magnitudes and tectonic plate boundaries were subsequently assessed using proximity analysis.
Plate Boundaries, Earthquakes and Geothermal Production in Mexico 175Flash-SteamBinaryGPPs5.5.2 Ranking tectonic boundaries and quakes using proximity analysis The same process employed for volcanoes was used to rank the importance of  the different tectonic plate boundary types. Proximity was calculated using the distances between GPPs and the points generated through the conversion of  the line-basses tectonic boundary dataset. !Figure 105: Worldwide tectonic plate boundaries and GPPs.  176Proximity units were the same as the ones calculated for volcanoes and normalized 
(/max) [MW/km]. They are shown in Table 44, along with the subsequent importance weights calculated for each type by taking their respective inverse values. A Circos diagram corresponding to the tabulated values is also drawn (Figure 105). The Oceanic Transform Faults, Oceanic Spreading Ridge, and Subduction Zones emerged as the tectonic boundary types with the closest proximity to flash-steam power production.  Table 44: Proximity values and final importance weights for tectonic boundary type. 5.5.3 Ranking earthquake data using proximity analysis The proximity analysis was repeated to calculate the distance between each earthquake in the significant seismic event dataset and each GPP. The results were pivoted/summarized by earthquake magnitude, which is the main quantifier field of  that dataset.  The chord diagram in Figure 106 shows the result of  the proximity analysis for different earthquake magnitude values; the closest proximity between earthquakes and high-capacity GPPs is indicated for quakes of  magnitude 7.2. Figure 106 plots earthquake magnitudes against their corresponding proximity weights, in attempt to treat magnitudes as numerical values and to identify a quantifiable potential trend between magnitude and proximity (expressed as a percentage).  The trend equation was used to estimate a proximity value given an earthquake magnitude; its usefulness was assessed at the time of  mapping the contribution to geothermal potential from the quakes layer.  177Figure 106: Proximity of  different GPP types to different earthquake magnitudes.  178Figure 107: Numerical relationship between proximity and earthquake magnitude. 5.6 Using heat flow to indicate geo-potential The Global Heat Flow Database of  the International Heat Flow Commission is a collection of  point surface heat flow data, provided as a grid of  64,801 nodes covering the surface of  the Earth (IHFC 2016). It is a complex dataset, providing three attributes per geographic coordinate pair, namely surface heat flow, thermal conductivity, and temperature gradient. Due to quality issues, only the heat flow data layer was initially considered for inclusion in the model. The map in Figure 109 is a subset of  the data, covering points that range between 0 and 100 mWm-2. The range for the full dataset runs from -126 to 489,000 entire data range. The majority of  the data range between 0 and ~500 mWm-2 (Figure 108). The presence of  the outliers (both the minima and maxima values) are problematic when mapping with a colour scale, especially due to the fact that the high limit of   179majority of  the data range (500 mWm-2) is 3 orders of  magnitude smaller than the overall dataset maxima (489,000 mWm-2). For this reason, the map data were filtered to exclude outliers. Figure 108: Histogram of  heat flow values. Figures 110, 111, and 112 were drawn for data values ranging between 100-200 mWm-2, 200-300 mWm-2, and 300-400 mWm-2 respectively. As the maps focus on higher heat flow values,their geographic distribution gets closer and closer to the distributions of  GPPs and volcanoes, and that of  tectonic plate boundaries. This is both encouraging and expected, as lava flows and other geothermal surface manifestations are expected to increase the amount of  heat flow through the crystal surface. Figure 113 shows the result of  buffer analysis on the heat flow data, using the GPPs buffers. The overall trend of  the data is distinctly downwards, indicating that surface heat flow diminishes with increasing distance from GPPs. The trends for specific GPP types mostly agree with the general trend; the one exception of  unconventional plants (which has a positive slope) is of  little consequence, for the reasons given elsewhere in this chapter.
 180
181Figure 109: Surface heat flow layer; data range 0-100 [mWm-2]. Figure 110: Surface heat flow layer; data range 100-200 [mWm-2]. Figure 111: Surface heat flow layer; data range 200-300 [mWm-2].  182Figure 112: Surface heat flow layer; data range 300-400 [mWm-2]. Figure 113: Buffer analysis results for the heat flow dataset, based on the GPP buffers..  1835.6.1 Ranking the heat flow layer Proximity analysis was conducted for the heat flow data set using GPPs. The same approach was used in this case as for all other geo-indicators. Although the heat flow values were numerical, they were grouped into increments of  10 mWm-2 and then treated as classifiers. With 62,226 rows, the heat flow layer was the most numerous geo-indicator dataset in the study. A full comparison with the 316 GPPs proved to be too computationally intense for the hardware that was available, leading to the decision to limit the scope of  the comparison to data that lay within 300 km of  each GPP point. Proximity comparisons between GPPs and the resulting subset (of  52,176 points, instead of  the full 19,601,190) was conducted successfully; proximity results are shown in Figure 114.  Tabulated data were too lengthy to include with the text. The exclusion of  heat flow points that were located more than 300 km away from GPPs was not considered to be particularly problematic, as their relatively larger distances in the denominator would diminish the contribution of  these points in the overall (summed) calculation of  proximity. 5.7 Limitations in this chapter The geology of  area of  geothermal potential was not included in the model; the reason for this was 2-fold: 1) the model was 2-dimensional and geologic settings are typically examined in 3-D, with increasing depths; and 2) as the model is global in scope, data requirements of  value would be too prohibitive to procure, even if  data was freely available for all areas of  interest (which, sadly, is not). Such analysis was therefore deemed to be outside the scope of  this research project and was left for others to explore.  184"Figure 114: Proximity analysis results for surface heat flow (in mWm-2) and GPPs.  185Chapter 6: Indicators of  Mineral Potential This chapter presents the mineral development datasets used to create and rank the mineral development indicator layers, which formed the basis of  the mineral potential calculations. Also presented are detailed descriptions of  the processes employed is scrubbing and processing, prior to merging it into a single unified dataset of  mineral development, as well as those used to classify and transform the data by defining and applying different indicator importance weights to specific field attributes. 6.1 Mineral development data 6.1.1 The USGS MRDS dataset The first dataset used in this analysis was the Mineral Resources Data System (MRDS) of  the U.S. Geological Survey (USGS). MRDS is “a collection of  reports describing metallic and non-metallic mineral resources throughout the world” that is freely available online (https://mrdata.usgs.gov/mrds/) (USGS-MRDS, 2017). It provides detailed information on specific worldwide deposits, such as location, commodity, production, reserves, and resources. USGS explicitly states that MRDS “large, complex, and somewhat problematic”; provided data is not of  consistent quality or completion, having been accumulated over a long period of  time by multiple investigators. It contains a large amount of  empty/NULL fields, is provided “as-is”, and gives no quality or accuracy guarantees. It is nevertheless the most complete, extensive, open-source dataset on mineral development available, and was thus the starting point for this analysis. The data used was supplied as a collection of  tables that were connected through a single foreign key (Deposit ID [Dep_ID]); an entity-relationship ER diagram was drawn up as a first step to visualizing and understanding the interconnections between them (Figure 115). A detailed preliminary metadata analysis was also conducted to help identify specific fields of  interest (see Appendix A).    186
187Figure 115: Entity-Relationship diagram for the MRDS dataset.  Here Occurrence is defined as “ore mineralization in outcrop, shallow pit or pits, or isolated drill hole(s)”, for which “grade, tonnage, and extent of  mineralization” is “essentially unknown”, and for which “no production has taken place and there has been no or little activity since discovery with the possible exception of  routine claim maintenance”; Prospect is defined as “a deposit that has gone beyond the occurrence stage”, and includes “subsequent work such as surface trenching, adits, or shafts, drill holes, extensive geophysics, geochemistry, and/or geologic mapping” to enough of  an extent that makes it possible to, at the very least, “estimate grade and tonnage”, and potentially includes the undertaking of  a feasibility study “that would lead to a decision on going into production”; Producer is defined as “a mine in production at the time the data was entered”, and includes “intermittent producer(s) that produce on demand or seasonally with variable lengths of  inactivity”; Plant is defined as “a processing plant (smelter, refiner, beneficiation, etc.) that may or may not be currently producing at the time of  data entry” and for which no associated geological information is provided; Past Producer is defined as “a mine formerly operating that has closed, where the equipment or structures may have been removed or abandoned”; and Unknown is used to classify rows for which “either the development status was unknown or the data source this record came from did not specify this value, at the time of  data entry” (USGS 2015). Commodity information in the original MRDS data was given by three separate multi-value fields, namely COMMOD1 (defined as “primary commodities”, which “have a strong effect on the economics of  the project, and might be economically viable as the only commodity”); COMMOD2 (defined as “secondary commodities”, which “can be economically recovered but have little effect on the economic viability of  the project”; and COMMOD3 (defined as “tertiary commodities”, which “are economically interesting but not economically recoverable as of  the date of  the source information”). This classification directly informed the importance weight allocations for the various commodities.  The three multi-value fields had to be further separated into single-value rows (using Trifacta and Excel), each providing information for a single (unique) commodity reported for a particular (uniquely- 188defined) deposit. For example, deposit #10001948 is associated with 5 primary commodities, 2 secondary commodities, and 0 tertiary commodities. The single multi-value row in this case was replaced by 7 rows associating this same deposit with 7 different commodities. Value, assigned as [1] for commodities listed in the COMMOD1 field, [2] for commodities listed under COMMOD2, and [3] for those listed under COMMOD3, was also added to new rows, along with the order in which each commodity was listed in the original multi-value row — this was of  consequence, as entries listed further to the left on the multi-value fields were considered to be more important than commodities listed further to the right. In this previous example, deposit #10001948 is primarily a zinc mine but does produce some barite of  high economic value. Figure 116 illustrates this example transformation. The commodity transformation split the original 277,806 multi-value rows (representing 100% of  the dataset prior to the split) to 466,007 rows, a 168.75% increase in row count (Figure 117).  As a final step at this stage, data rows not associated with a particular commodity were removed from the working dataset. This transformation was necessary to allow for subsequent analysis of  individual commodities, and in order to comply with the uniqueness and elimination of  duplication design requirements of  relational database systems. To enhance context and reduce visualization complexity, commodity names were re-classified using a more simplified naming scheme. For example, the rows with a commodity entry of  either Titanium, Titanium-Heavy Minerals, Titanium-Ilmenite, Titanium-Metal, Titanium-Pigment, Titanium-Rutile were given a common Generalized Commodity name of  Titanium, and were assigned to the Commodity Grouping of  Titanium (Table 46). In all, the MRDS data provided information for 115 distinct commodities.  Commodity Groups were further assigned into 7 economic classes, namely: Industrial Minerals, Non-Metals, or Metalloids; Base Metals; Critical Rare Metals or Metalloids; Energy Metals or Minerals or Fuels; Precious Metals; Rare Earths; and Gemstones. Table 47 provides a full listing of  Economic Classes, Commodity Groups, and Commodity names used to re-classify the MRDS dataset.
 189Table 46: Re-classification of  the MRDS commodities using a simplified naming scheme. !!Figure 116: Transformation of  a multi-value row into single-value rows for the MRDS dataset. Figure 117: Transforming multi-value fields into a single value fields, and filtering.
 190Table 47: Commodity groupings (including economic classes) for the MRDS commodity field.  
 1916.1.1.1 Size-based filters The MRDS dataset uses the following classes to denote size: (L: Large; M: Medium; S: Small; Y: Some (unspecified) production; N: No production; U: Unknown; NULL: Empty cell value) 
(Figure 118). Size information is provided for 82,956 rows that have a defined development stage (corresponding to 16.8% of  the 466,007 rows derived from the previous step). Denoting size as a necessary attribute, without which rows could not be passed on to the mineral potential data layer, would exclude 92.2% of  the MRDS data, thereby reducing the layer’s scope, extent, and resolution (Figure 120b). A looser filter would include all data rows regardless of  their size attribute 
(Figure 120a); this includes currently active production properties indicated as having no production, which is unrealistic as, by definition, actively producing properties must have production levels greater than zero. On the other hand, non-production development stages, such as Occurrence, are not practically easy to size with a great level of  confidence, due to the large number of  unknowns at such early stages in the mineral development life cycle. Size values provided for rows that are in non-production development stages could therefore be safely ignored in the filtering. This would allow for the definition of  a compromise filter that excludes rows with unspecified production size — i.e. Plant, Producer, or Past Producer — where size is important, and include rows regardless of  size definition for rows in non-production development stages — i.e. Occurrence or Prospect — where size is not important. Figure 121 shows how such a compromise filter was defined to include non-production rows regardless of  reported size and to exclude production properties for which sizing is missing or indicated as Unknown — thereby improving the quality of  the data and minimizing data exclusions (Figure 120c). The final number of  rows passing though the compromise size/stage combo filter was 268,963 (or 56.72% of  the original data), reducing exclusions from 82.2% to 42.28% or from 383,051 to 197,044 rows. The treemap of  Figure 119 shows the breakdown of  included and excluded data, by Development Stage, Size, and commodity Economic Class. 
 192Figure 118: Statistical analysis of  the MRDS data based on size/stage combo.  193
194Figure 119: Treemap of  the filtered subset of  MRDS data contributing to the mineral potential data layer. Figure 120: a) Loose; b) strict; and c) compromise filters for the MRDS data. 
(a)(b)(c) 1956.1.2 The InfoMine dataset The second dataset used to create the mineral potential data layer was InfoMine’s IntelligenceMine, a comprehensive and up-to-date proprietary mining database that is hosted at infomine.com (InfoMine, 2015). Access is usually restricted to subscribed members, but following negotiations with the data providers, temporary, short-term access was granted for the purposes of  this research project. The level of  access provided was similar to that of  a subscriber to the service. Clients typically log onto the IntelligenceMine front-end interface and query specific views of  the server-side database, using predefined queries or keys (Figure 121). Direct access to the data (for example through phpMyAdmin) for the purpose of  directly querying database tables using SQL was not granted. The data query and retrieval process consisted instead of  iteratively running a long series of  predefined queries by clicking on links on the front end of  the IntelligenceMine website, generating individual XLS files containing the results of  each query sent to the server by clicking on a button, and pressing yet another button at the end of  the file generation to download it to the local machine. This was a lengthy and cumbersome process that resulted in the generation of  hundreds of  XLS files, and which was assisted in part by an automation script written to alleviate some of  the interaction burden on the client-side. The downloaded files required a substantial amount of  further processing to filter, clean up, and export in a format that would allow the data to be imported into MySQL. An Entity-Relationship diagram (Figure 122) was created to illustrate the relationships between the data, and to help identify the most important (to this analysis) fields of  the InfoMine dataset, which were Latitude, Longitude, Activity Status, and Current Activity. Three separate fields providing commodity information were also added to the analysis, namely Commodity Exposure, Commodity in the Production table, and Commodity in the Reserves table. Size was not explicitly defined, but some limited information was given in terms of  reserves approximations and historical production data for a small subset of  the data.   196Figure 121: The IntelligenceMine front-end interface, showing the generate/download buttons. In terms of  NULL values, 48.3% of  the original 33,510 rows in the InfoMine dataset (i.e. 16,196 rows) had an undefined lat/long pair value and were therefore excluded from the analysis. Unknown values in the Development Stage column were also removed from the dataset, further reducing its working size by 83 rows to a total of  17,231 rows (Table 48).  Table 48: Row counts per stage for the InfoMine data (NULL & non-NULL lat/long values).  197Figure 122: Entity-relationship diagram for the fields of  greatest interest in the InfoMine dataset. 6.1.2.1 Development Stage Development Stage is defined in a higher resolution for the InfoMine data than the MRDS data: the MRDS dataset uses 5 distinct development stage (namely Occurrence, Prospect, Production, Plant, and Past Production, while the InfoMine dataset uses 9 — namely Prospect, Exploration, Advanced Exploration, Preliminary Economic Assessment, Prefeasibility, Feasibility, Development, Production, and Closed (Table 48). Although no direct correlation between the defined stages was provided with either dataset, if  viewed in a chronological perspective, they could be matched to the mineral development lifecycle stages mining projects typically adhere to (Figure  198123). Comparing the relative distributions of  project counts by development stage between the two datasets, and reducing the resolution of  the InfoMine classification scheme to match that used in the MRDS dataset, did not reveal a similarity in the two trends. It was therefore difficult to justify reclassifying the InfoMine data (using the regroupings shown in Figure 124) and using the same approach in calculating a common development-stage-based importance weight for the two sets; these weights were calculated separately for each dataset instead. Figure 123: Common development stage naming scheme for InfoMine & MRDS. Figure 124: Row count comparison per development stage for InfoMine & MRDS.  As discussed in Chapter 3, for a typical life cycle, mineral development projects start as mere acknowledgement of  a likelihood of  the presence (or occurrence) of  a particular mineral that is high enough so as to warrant further investigation into its economic potential. This typically takes the Common:InfoMine:MRDS:%-Count-per-Stage 199form of  prospecting and exploration activity, which if  successful, can gradually progress through the more analytical stages of  prefeasibility, feasibility, and eventually, actual development. Production of  an operating mine is typically active while the produced ore’s characteristics and overall market conditions combine to make for a profitable venture. When that ceases to be the case, or when a resource is depleted, the project closes. With each stage, work goes into better informing reserve estimates through modelling, physical exploration, sampling, and analysis. Projects can be dropped at any stage in their lifecycle, so long as projections (or actual production yields) indicate the venture is not profitable. Consequently, the further a project progresses through the mineral life cycle stages, the higher its likelihood of  success. A mineral indicator point derived from a project in a later life cycle stage (e.g. Feasibility or Development) would therefore carry a higher mineral potential indicator weight than one derived from a project in an earlier stage (eg. Prospect or Exploration).  In practical terms, only a small fraction of  all identified prospects progress from prospecting all the way to production and eventual closure. The exact number was not available in public domain literature at the time of  this study, but the Government of  Canada does indicate that “about 1 out of  every 200 projects that reaches the discovery stage moves to development. This is equivalent to about 1 out of  every 10,000 grassroots exploration projects.” (INAC, 2016). The Development Stage field of  the InfoMine data was of  high enough a resolution (with its 7 values each indicating a distinct life cycle stage) to allow for an assessment of  such a metric from the data, using cumulative counts of  projects making it through each progressive stage. Figure 125 shows how a flowchart of  the mineral development life cycle of  projects marked as currently Active was constructed, and Figure 126 illustrates these figures using a stacked bar chart. Only about 6% of  all identified Prospects recorded in the InfoMine dataset make it to the end of  their life cycle and close; about a quarter of  Prospects make it to Production.  200
201InfoMine Properties: Mines, Processing Facilities & ProjectsProspectAbandoned 4,800Suspension 54,805Active 3,657Current 8,462Past/Complete 24,669All 33,131ExplorationAbandoned 6,015C&M 13Suspension 206,048Active 6,315Current 12,363Past/Complete 12,306All 24,669Advanced ExplorationAbandoned 243C&M 18Suspension 18279Active 3206Current 3485Past/Complete 9100All 12306Prelim. Econ. AssessmentAbandoned 0C&M 0Suspension 00Active 116Current 116Past/Complete ?+ All 116PrefeasibilityAbandoned 7C&M 1Suspension 311Active 380Current 391Past/Complete 8593All 8984A project is termed prospect before exploration commences. Advanced exploration is typically always preceded by exploration.This is problematic here; not all prefesibility studies have been preceded by a preliminary economic assessment.Cannot know % of prefeasibility studies that have been preceded by a preliminary economic assessment.A preliminary economic assessment is always preceded by advanced exploration. A preliminary economic asessment will always be followed by a prefeasibility study.  Assumption: a project will always have a prefeasilibity study and sometimes a preliminary economic assessment.Figure 125a: Project counts through the mineral development life cycle stages. 
202Figure 125b: Cumulative counts of  projects through the mineral development life cycle stages. InfoMine Properties: Mines, Processing Facilities & ProjectsFeasibilityAbandoned 8C&M 1Suspension 1019Active 299Current 318Past/Complete 8275All 8593DevelopmentAbandoned 13C&M 23Suspension 2056Active 518Current 574Past/Complete 7701All 8275ProductionAbandoned 124C&M 327Suspension 263714Active 4948Current 5662Past/Complete 2039All 7701ClosedAbandoned 902C&M 14Suspension 4920Active 1119Current 2039Past/Complete 0All 2039A feability is typically always preceded by a prefeasibility.Development requires the completion of a feasibility study - always. Closed is and stays closed. No production w/out Development.Hence: Development = Abandoned + Care & Maintanance + Suspension + PAST.PAST = Production (All).Will add to current tally because this is a snapshot view; all 166 are assumed to eventually proceed to the prefeasibility stage.Each project can only be in one Stage and have one Status at any give time.Suspension: Temperary SuspensionC & M: Care & Maintenance
203Figure 126: Cumulative counts of  projects progressing through the various life cycle stages.6.1.2.2 Activity State The InfoMine dataset also included definitions of  Activity State for a subset of  the data; properties were classified as either being Active, Abandoned, in Temporary Suspension, or under Care & Maintenance. The row counts for each of  these are shown in Table 49. The majority of  the data were classified either as Active or Abandoned.  Table 49: Row counts per activity state for the InfoMine data. 6.1.2.3 Commodity Data rows used in the creation of  the mineral potential data layer required commodity information; for the InfoMine dataset, this was drawn from three separate fields, namely:  —	 Commodities in the CommodityExposure table; a multi-value field. It was transformed into a single-valued column using a similar process as the one used for the same transformation in the MRDS dataset. The multi-value version of  the column consisted of  32,710 rows, including 4 blank rows. The transformed version was enlarged by 175%, to 57,282 rows that link properties to a single commodity, and noting its order in the listing. —	 Commodity in the Production table; it contained a total of  34,883 rows corresponding to 33,509 unique property IDs. Out of  these, only 4,450 rows were not empty, corresponding to 3,123 unique IDs. That meant that there was commodity information and production information for only 9.3% of  the original data.  204—	 Commodity in the Reserves table; its original version had 61,888 rows, 33,510 of  which were on unique PIDs. Of  those, 3,999 rows had a non-NULL value in the included metal field/column, and 5,089 rows had a non-NULL value in the Tonnage column/field, corresponding to 9.3% and 15.18% of  the original data respectively.  One additional commodity-related transformation was applied to the data; values listed in the commodities field were reviewed and regrouped to match the commodity classification used in the MRDS dataset, to allow for eventually merging the potential weight rows derived from each set. This was again a complex process, as multiple values were used in the data to indicate what this analysis regarded as a single commodity; for example, Table 50 shows the 50 different names used to indicate Coal. The final reclassification of  commodities and their groupings matching those used for the MRDS data analysis is shown in Table 51.  Table 50: Coal classifications in the reserves table.   205Table 51: Commodities, commodity groups and economic classes defined for the InfoMine dataset.   
 2066.1.2.4 Size The InfoMine dataset does not use predefined classes to denote size, providing instead historical production information for a small subset of  the listed data; more specifically, 12.5% of  the rows had a development stage that was neither NULL or Unknown, a State that was not Unknown, and a size that was greater than zero. Production levels were given as numerical values accompanied with a corresponding production unit; for example gold production was given in [kozt], while copper was reported in [kt]. At the time when the InfoMine data was accessed, production levels had only been reported for the period between 1996 and 2014 inclusive, in the form of  annual production totals. No later access to the dataset has been granted since, therefore 2014 was the latest year for which production data was available for analysis. Using data rows that have a positive indicated production and known development stage and activity state would exclude 86.5% of  the InfoMine rows — such filtering would be too strict. If  size were to be used in filtering the InfoMine data, the supplied production levels had to be converted into a more generalized size class, similar to one used to size the properties listed in the MRDS dataset. Commodity production values could not be ranked within the overall set, as commodities have different economic values per production unit and production data was reported in a variety of  different units. For example, the maximum value of  total historical (1996-2014) production reported for bitumen was 1,678,500,000 m3 while the corresponding maximum for gold was 496,923 kozt. Production values of  these 2 commodities are not comparable, which makes the process of  classifying production totals more complex than simply comparing values to one another. To overcome this problem, the following approach was adopted:  —	 Historical property production totals were split into commodity-based subsets, which had a single unit of  measurement (e.g. kt, or kozt); within each commodity/subset, the maximum individual reported production was noted and used as the basis of  classification, by assigning  207properties with a reported historical total less than 30% of  the noted maximum a class of  Small (or S), and by assigning properties with a reported production greater than 70% of  the reported maximum a class of  Large (or L). Values falling within those 2 limits were assigned a class of  Medium (or M).  —	 Reserves and Contained Metal values were also sized using the 30% and 70% of  maximum reported per commodity bounds.  —	 Size classes were assigned to individual properties for specific commodities. Some properties reported production, reserves, or metal content for multiple commodities. Consequently, for some property-commodity pairs, multiple (up to three) size-based classes were assigned. In the case of  disagreement (e.g. for a specific pair, size indicated by total production was S but size indicated by reserves was L), production-based classes were given precedence. In the case of  a disagreement between metal content and reserves, the larger value was used.  —	 In all, size classes were returned for 7,832 unique property-commodity pairs, corresponding to information on 4,917 unique properties and 73 unique commodities.  The treemap of  Figure 127 illustrates the breakdown in property count per assigned size w.r.t. commodity and economic class. The vast majority of  properties, across all classes, were assigned a Small size classifier. Figures 128-131 show size assignments for each property that was classified based on production figures, taking into account commodity maxima and thereby retaining some of  the detail even after the transformation was applied. Figure 132 shows how a compromise filter (similar to the one defined for the MRDS data) was applied to the dataset to enhance the quality of  the data and retain much of  its coverage/extent. The treemap of  Figure 133 shows the breakdown of  the data filtered through using the compromise filter, per economic class, commodity, and size. Figures 134a, 134b, and 134c map the coverage of  the strict, loose, and compromise filters respectively.
 208209Figure 127: Size distribution for Commodity and Economic Class for the InfoMine Dataset.
210Figure 128: Sizing the InfoMine properties - base metals.
211Figure 129: Sizing the InfoMine properties - industrial mineral, non-metal, or metalloid.
212Figure 130: Sizing the InfoMine properties - gemstones & precious metals.
213Figure 131: Sizing the InfoMine properties - energy metals & critical rare earth metals.Figure 132: Statistical analysis of  the InfoMine data based on size/stage/state combo.  214
215Figure 133: InfoMine filtered subset contributing to the mineral potential data layer. Figure 134: Resulting coverage of a: a) strict; b) loose; and c) compromise filter for InfoMine. 
(a)(b)(c) 2166.2 Mineral potential The term mineral potential is used to denote a quantitative measure for the “propensity” of  a specific location to yield high-value mineral production. It is calculated using AHP and comparison matrix calculations, using the following criteria: a) development stage — the more advanced the stage the greater the likelihood a given project will be successful, with advanced stages carrying a greater importance weight than more preliminary ones; b) activity state — the current status of  a given project, i.e. whether or not it is currently active, completed or under temporary suspension will also affect its likelihood for success, with active projects carrying a greater importance weight than inactive ones; c) property size — in general terms, a larger size will be of  more value to a developer and therefore carry a higher mineral potential importance weight. Commodity was used as a filter/classifier rather than a ranking criterion in the analysis — it allowed for the production of  summary statistics per Commodity (and/or Commodity Economic Class), and for the reduction in visualization complexity. Component weights were calculated for the 3 above-listed criteria, which were subsequently used to define a simple measure of  mineral potential.  6.2.1 Mineral development indicator weights 6.2.1.1 Development Stage Figure 135a-b shows the AHP pairwise comparison matrices for the MRDS and InfoMine datasets respectively. For the MRDS data, a Plant was assumed to have the same importance as a Production site, which reduced the comparison matrix from a 5×5 to a 4×4. In the InfoMine data, Exploration was assumed to have the same importance as Advanced Exploration, and Preliminary Economic Assessment, Prefeasibility, Feasibility were assumed to have the same importance as Development.   217Figure 135: AHP matrix for the a) MRDS; and b) InfoMine datasets. Table 52 lists the resulting development-stage-based importance weights for both datasets, arranged in an approximate chronological (w.r.t. the mineral development life cycle) order.  Table 52: Development stage weight comparisons for the MRDS & InfoMine datasets. (a)(b) 2186.2.1.2 Activity State Activity State was not defined in the original MRDS data; to align the comparison with that of  the InfoMine set, a value for Activity State was defined and assigned to the MRDS data rows based on their corresponding value for Development Stage (Table 53). Data rows with a Development Stage other than Past Producer were assigned a value of  Active for the Activity State field. Past Producer rows were assigned an Activity State value of  Closed.  Table 53: Activity state assignments for the MRDS data based on development stage.  Figures 136 shows the common 4×4 AHP pairwise comparison matrix for the 2 datasets. Figure 137 combines the Activity State and Development Stage importance weights. A simplified combination weight was then calculated to replace the product weights based on a classification of  (Primary, Secondary, Tertiary, Quaternary), which were replaced with the corresponding values: (1.00, 0.50, 0.33. 0.25) (Figure 137). Figure 136: AHP matrix for the a) MRDS; and b) InfoMine datasets.  219!Figure 137: Reclassification/calculation of  combined wstate/stage.  6.2.1.3 Commodity The values of  the commodity listing order indices assigned through the transformation process ranged between 1 and 16 for both datasets, and were therefore used to create a common weight scheme that was in turn used to scale the mineral potential for listing order (Figure 138).  !Figure 138: Index value distribution and listing order weights for MRDS and InfoMine.  2206.2.1.4 Size No ranking weights were defined for Size, which was provided either directly as an attribute (e.g. in the case of  the MRDS dataset), or inferred from reported production levels (e.g. in the case of  the InfoMine data); its use was instead limited to filtering out low-confidence data rows (such as those reporting zero total historical production on active producers). 6.2.2 Relative importance of  component layers The weights calculated in the previous sections were eventually combined in an overall mineral potential indicator weight using Eq. 15. Prior to adding each individual contribution, weights were scaled with respect to their relative importance as indicators, which was also calculated using AHP and pairwise comparisons (Figure 139). Development Stage dominated the calculation of  mineral potential, with a relative weight (wrstage) of  59%, compared to 30% for Activity State (wrstate), and 11% for Commodity Order (wrorder). As the Development Stage and Activity State weights were combined at a previous step, a combined scale factor for wrstate/stage was also calculated by summing the two individual factors. Figure 140 summarizes the mineral potential weights per commodity and economic mineral class.  	 	 	 	 	 	 	 	 	 (15) In summation, this chapter presented the two mineral potential datasets used in this analysis and described the steps taken to clean, reclassify, and eventually merge the data into a single dataset of  mineral potential indicators. Filtering and classification schemes were described in detail separately for both datasets, as was the manner in which the AHP relative importance weights that were used in merging the data were calculated. The geographic coverage of  the combined dataset resulting dataset is shown in Figure 141.  221wmineral potential  = wstate/stager ×wstate/stage  + worderr ×worder"Figure 139: Pairwise comparison matrices for the relative importance of  mineral indicators. !Figure 140: Mineral potential weights (wmineral) per commodity and economic mineral class. 
 222Figure 141: Relative geographical scopes and point overlap of  the MRDS & InfoMine datasets. Both a) and b) show the same data; in a) the MRDS layer is on-top of  the InfoMine layer, 
while in b) the InfoMine layer is on top of  the MRDS layer.  (b)(a) 223Chapter 7: Suitable Application Areas This chapter focuses on the steps taken to finalize the mineral and geothermal potential datasets, and select suitable application areas for mineral and geothermal potential indicators, including mines, GPPs, volcanoes, quakes, boundaries, thermal springs, and heat flow. 7.1 Finalizing the mineral and geothermal potential datasets The geothermal indicators presented in this chapter will eventually combine to produce a different kind of  map, one that identifies particular geographic areas of  high combined geothermal and mineral potential, rather than one that provides an abstract representation of  geo-potential.  The first steps to achieving this was to create potential layers for both types of  indicators (i.e. mineral and geothermal) using the importance weights calculated in the previous two chapters. To this end, the geothermal potential datasets were scrubbed and updated in R Studio by adding a column that stores their corresponding importance weights (which are sub-type dependant). Following the calculation of  one additional parameter, namely the suitable application area radii, discussed below in section 7.2), the geothermal and mineral potential datasets were merged and exported in CSV form, and subsequently imported in ArcGIS, where they were used to create .shp files that were in turn used to calculate geomine potential (see Chapter 8 for further details).  7.2 Selecting suitable application areas for geothermal indicators One final variable was needed before proceeding with the mapping of  geomine weights, one that would dictate the manner in which weights would be applied to each indicator point when mapping. Each indicator was represented by a single 2D point defined by a (lat, lon) pair. A point has zero actual surface area but the effects of  the various mapped features are not limited to a single point location. For example, geothermal power plants are typically located within a geothermal field,  224in close proximity to wells (reducing the distance brines need to travel from wellhead to steam turbines reduces enthalpy loses and increases power yields). GPPs are used in this research to indicate geothermal potential. Potential is not physically restricted to the exact location of  the power plant, but it is instead related more to the extent of  the geothermal reservoir, as mapped during exploration and subsequent production.  A similar case can be made for other features: e.g. volcanoes used to indicate active volcanism, quakes used to indicate tectonic activity, etc. Therefore with each mapped point, it is important to have a rough estimate of  how much space around each point should be marked as an area of  higher potential (contributed by said point). For the sake of  simplification, these areas were drawn as circles centred around each point, and a distance or radius (termed suitable application area (SAA) radius within the context of  this document) was selected for each type of  indicator. This process is described in further in sections 7.2.1 and 7.2.2 for geothermal and mineral indicators respectively.  7.2.1 Suitable Application Areas: Geothermal power plants The original GPP layer was mapped in detail in Chapter 5 (see Figures 50 to 60 inclusive). In these maps, the drawn points represent (unscaled) capacity. The physical extent of  the points is by no means representative of  the physical extent of  the geothermal area on which these GPPs have been constructed. Their radii are rather abstract (and somewhat dependent on the select map scale) and are used to facilitate the comparison of  different production levels. Prior to producing the geomine map, it was necessary to gain an understanding of  the geographic extent (or scope) that each point in the map represents. These areas will still not be exact; the level of  accuracy inherent in the data used by this model does not allow for a strict delineation of  geo-potential area limits. These areas were instead represented, for the sake of  simplicity, by circular polygons created around each indicator point depending on the point type and (for the most part) final point importance weight. The final step was consolidating all polygons into a single geo- 225potential area, taking into account the relative importance of  different indicator layers (computed using degree of  separation analysis) and dealing with the amalgamation of  (and calculation of  combined importance weight for) any overlapping features.  The polygons, termed areas of  applicability, were created separately for each layer. For all layers apart from GPPs, the change of  occurrence of  each feature type with an increasing distance from GPPs was calculated to help select an appropriate value for the SAA radii. For GPPs, no such comparison was possible. Instead, the capacity of  the GPP was used to define a radius using the following equation: y = 0.0538 x1.3843 	 — (16) Here, y is area in km2 and x is the capacity in MWe. This equation was derived by taking a sample of  GPPs for which reservoir boundaries are considered known, and by superimposing published reservoir maps (as images) of  each GPP on the actual physical location of  the GPP using Google Earth, tracing their boundaries, and extracting a surface area value of  the projected (albeit approximate) surface area (in km2). Figure 142 shows an example of  the traced polygons for the reservoirs at Larderello-Nord, Larderello-South, and Mt. Amiata, in Italy.  The results of  the mapping/tracing are tabulated in Table 54; the sampled plants were for the most part dry-steam installations. Dry-steam plants were selected for this purpose because they are typically more condensed in a geographical sense, which makes for easier and more manageable information retrieval (e.g. published reservoir map or schematics from the literature) and subsequent tracing — all dry-steam plants are situated within 7 distinct geothermal reservoirs.  Figure 143 charts the tabulated data, which were fitted with the trend line that provided 
Eq. 16. Two flash steam plants were added to the chart and a separate trend line was derived; it was parallel to the line derived for the dry steam plants, indicating an agreement in the overall trend.   226Table 54: Traced and measured surface areas for selected geothermal fields. !!Figure 142: Tracing the resistivity boundaries of  the Larderello geothermal fields. [km2] [MWe] [km2] [MWe] [km2/MWe] 227Resistivity  boundariesGPP!Figure 143: Mapping GPP generation capacity and reservoir surface areas for sample GPPs.  7.2.2 Suitable Application Areas: Volcanoes The rate of  occurrence of  volcanoes with increasing distance from GPPs was used to determine SAA radii for the volcano layer. Buffer analysis was conducted using buffers similar to the GPP buffers in Figure 76, initially for all types of  volcanoes combined, and subsequently for combinations of  particular volcano and GPP types.  Using buffer analysis, it was also possible to identify which types of  volcanoes are typically found in areas of  geo-production. Figure 144 shows the cumulative counts of  volcanos found within specific incremental distances from GPPs. Figure 145 shows the results of  the buffer analysis for different volcano and GPP types. 
 228229Figure 144: Volcano types most closely associated to geo-production. Figure 145: Percent of  GPPs closest to a volcano, per buffer size.  230The main observation from Figures 144 and 145 is that (excluding unconventional installations) all GPPs are located within at most 100 km from a volcano. Some volcano types are generally found closer to volcanic centres than others — for example, if  a GPP is most closely located to a crater row, it will be at most 20 km away; the corresponding distance for calderas can be as long as 90 km. Table 55 summarizes the suitable application area radii corresponding to each volcano type. The maximum value indicated for either flash or dry-steam plants by the above analysis was the one ultimately used. Table 55: Suitable application area radii corresponding to volcano types.  !7.2.3 Suitable Application Areas: Earthquakes A similar approach was used to select suitable application areas for the Earthquakes layer. Once again, Magnitude was treated as a qualifier and used to filter through the buffer analysis results. The same GPP buffers used for volcanoes were used to calculate the occurrence distribution of  seismic events.  The results of  the buffer analysis are presented in Figure 146. Of  particular interest was the existence of  an inflection point in Figure 146; past the second peak count at 90 km, and a drop in the count at 100 km, the chart line curves steeply upwards, indicating a much higher occurrence of  earthquakes after 100 km from GPP. This was attributed primarily to interference from too-large buffer radii, whereby earthquakes considered to be unrelated to the presence of  geothermal potential were accounted for in the results. The individual Magnitude interval lines followed a more or less  231similar trend up to the inflection point, and as such, a distance of  90 km was selected as the suitable application area radius for all earthquakes irrespective of  M level.  Figure 146: Buffer analysis for earthquakes (summary) — counts.  7.2.4 Suitable Application Areas: Tectonic plate boundaries The tectonic plate boundary layer required a modified approach from the one employed up to this point, due to the fact that relative to the distances calculated between GPPs and other geo-indicator features, boundaries were situated much further away. This necessitated the creation of  a separate set of  buffers, which were centred around boundaries instead of  GPPs.  To accommodate the larger physical distances between GPPs and tectonic plate boundaries, buffer sizes were extended to 1,000 km. The Tectonic Boundaries dataset consisted of  line segments, which was problematic because the analysis, as it was defined in this project, used distances between  232points and point counts for the calculations of  proximity and occurrence distributions respectively. To overcome this difficulty and to retain consistency in the approach, line segments were converted to equidistant sets of  points, each of  which carried the corresponding attributes of  the segment upon which it was defined. Buffers of  increasing radius were then created around each tectonic boundary point. Figures 147 and 148 show a sample of  3 (out of  the 10 in total) buffer sets of  increasing radius created around boundaries in and around the geothermal region in New Zealand. It is easy to see for example that a 100 km buffer around tectonic plate boundaries — a distance that was more than adequate to cover the occurrence of  volcanoes around GPPs — was unsuitable a choice in the case of  tectonic plate boundaries, as associated GPPs would not be detected. In the case of  New Zealand a 400 km buffer would be required to capture GPP occurrence distributions using boundary-based buffers. In other areas, those distances were even larger. The results of  the buffer analysis for the tectonic plate boundaries is shown in Figures 149a and 149b. Each graph corresponds to a specific tectonic plate boundary type and the frequency counts used were for earthquakes (Table 56). !Figure 147: Selecting appropriate buffer size for tectonic boundaries analysis. Table 56: Select sizes (in km) for tectonic plate boundaries per boundary type. !400 km200 km 233Buffer
234Figure 148: Selecting appropriate buffer size for tectonic boundaries analysis.Flash-SteamBinaryGPPsBufferQuakeGPPBoundary!Figure 149a: Quake count for tectonic plate boundaries — for OSR, CCB, CTF, and OCB.   235!Figure 149b: Quake count for tectonic plate boundaries — for OTF, SUB, CRB and ALL.  2367.2.5 Suitable Application Areas: Heat flow and thermal springs Application areas for the heat flow and thermal springs data were not defined, as the decision was made to exclude both datasets from the analysis. This was done so as to deal with issues of  computational complexity: the heat flow dataset was too large to include in the model. It is actually the largest of  them all, accounting for more than 86% of  all data points (Table 57).  Table 57: Breakdown of  row counts for the geo-indicator datasets. !Computational times on the available equipment and within the applicable project constraints took too long to process. In addition, the heat flow layer was of  relatively low importance; the overall importance weight for heat flow assigned from the degrees of  separation analysis was 0.20 — the lowest of  them all. The decision was therefore made to drop the dataset completely. The springs data was also eventually excluded from the model due to the fact that the only available dataset was regional and the GIS process looked for patterns and distance analysis on a global scale. The discrepancy would lower the confidence in the results and therefore this dataset was also excluded. 7.3 Selecting suitable application areas for mineral indicators An approach similar to the one adopted for calculating SAAs for geothermal power stations was adopted for the calculation of  SAAs for mineral potential. The rows in the mineral development dataset represent locations geo-referenced by lat-long point pairs. No indication of  the surface area  237or the surface footprint of  each mine is provided. An ideal mineral potential map layer would be able to account for the physical extent of  mineral development projects, which can be substantial; for example, the Kennecott Copper Mine southeast of  Salt Lake City, Utah, covers 1,900 acres, twice as large as Downtown Vancouver, BC.  Mapping the mineral potential layer required some assessment of  surface footprint and a departure from mapping mines as points, particularly for the production of  higher-resolution/smaller-scale maps. The ideal would be to roughly correlate, if  possible, mine surface area with mine production levels (indicating deposit size). A mine’s surface footprint, shape, and size can vary significantly depending on a number of  factors, such as mineral production methods, location, budgetary or legislative constraints, and actual mineral product mined. Surface/open-pit mines have the largest surface footprint among mine types by definition (as opposed to, for example, underground mines that extend underground, in a vertical direction), and they grow in different directions, depending on topology and deposit characteristics. For example, even through the lease for the Kennecott Copper Mine is an accumulation of  mostly connected rectangles, the mine itself  is more teardrop in shape (Figure 150). It was therefore further decided to map polygons representing mine surface footprint as an equivalent circle centred on each lat-long-defined map point. Chapter 8 provides more details on mapping both mineral and geothermal potential. For the sake of  simplicity, all mines were assumed to be surface or open-pit mines. To arrive at an estimate for SAAs, a number of  mines would have to be selected and their visible surface footprint traced using Google Earth Pro, to get a measure (in km2) of  each area traced. Ideally, SAAs would be have been calculated for each of  the 143 commodities listed in the dataset, a prohibitive task for this project’s resources. Selected mines were therefore: a) surface or open-pit projects; b) gold-producers (as gold is typically found in active geothermal areas); c) with production levels (in kozt) provided in the dataset; with gold as the primary commodity listed for each property; d) located in  238the US State of  Nevada (to minimize the effects from differences in available technology, applicable legislation, and/or culturally-driven decision-making); e) listed as currently active producers in the dataset; and f) identifiable as an active mine site using up-to-date satellite photography. !Figure 150: Mineral lease shape versus actual extent of  Kennecott/Bingham Canyon mine (http://www.duluthnewstribune.com/news/4195276-kennecott-asks-expand-copper-search) The above filters reduced the list to 21 properties (Table 58). Each was subsequently traced to measure its surface area, which was then plotted against known total production (Figure 151). Figures 152 and 153 show a sample of  traced areas. Even if  this rough estimate of  surface area as a function of  production levels were to be applied to all types of  mines in the data, the fact remains that the vast majority of  the data from both datasets did not provide production information, which made the application of  the above estimate difficult to apply across the entire dataset. Statistical analysis on a subset of  the InfoMine data (namely gold mines with reported production, no geographic restrictions imposed) indicated that mean and median total production were 1,067 and 230 kozt respectively, with a largest bin size in the 100-250 kozt range (Figure 154), and a range (defined as max-min) of  43,883.3 kozt (Table 59a).
 239Table 58: Total production and measured surface footprint of  sample gold mines in NV, USA. !! 240sFigure 151: Surface footprint as a function of  total production, sample gold mines. "  Figure 152: Tracing the pit surface areas of  sample US gold mines (part 1).  241"Figure 153: Tracing the pit surface areas of  sample US gold mines (part 2).  242Area size range on the other hand was 59.5 km2, and mean and median values were 15.7 and 12.5 km2 respectively (Table 59b). The histogram indicated that most properties had a surface area ranging between 10-25 km2 (leaning slightly to the right within that range) (Figure 155). The decision was consequently made to assign a constant value of  20 km2 to all SAAs, regardless of  production levels or commodity; this translated to an effective radius for the circular polygons of  2.5 km. Table 59: Descriptive statistics for a) total production; and b) for traced areas. The analysis presented below used buffer areas around each point to measure distances between mines and geo-indicators. The larger the buffer, the greater the number of points included in the analysis, and the larger the number of data rows returned in the near table. It is possible to set this distance to infinity and create a near table that includes distances between all points. Similar to heat flow, this would have resulted in a computational load that would have been prohibitive within the technological and resource constraints of this project. It was therefore necessary to limit the buffers to reduce the number of points found within set distances from geo-indicators and, by extension, the rows returned in the distance matrix. The applied limits were set equal to the suitable application areas calculated for each indicator type and subtype; they are listed in Table 61. A considerable portion of the data were thus filtered out (Figure 156), changing both the overall number of points going into the analysis and their relative distribution.  243!Figure 154: Histogram of  gold production totals for the InfoMine dataset. !Figure 155: Histogram of  traced surface areas for sample gold properties.  244Table 60: Measured surface areas (in km2) for sample mines. !Table 61: Assigned limits for search radii for the proximity analysis per geo-indicator subtype. ! 245
246Figure 156: Filtering based maximum search radius - buffer distance.Chapter 8: GeoMine Potential This chapter presents the process of  calculating and subsequently mapping of  the resulting values of  geomine potential from around the world. 8.1 Calculating the geomine potential indicator weights The proximity analysis returned of  a set of  rows that comprised of  a pair of  IDs (one corresponding to a mineral indicator, and the second to a geothermal indicator), and the geodesic distance between them (in km). The table was exported from ArcGIS and imported into R Studio, where the remaining properties of  each indicator in the pair were added to complete the near table (Figure 157).  !Figure 157: Near analysis table (for volcanoes) with extended indicator attributes.   247Proximity p was calculated for all points in the dataset, i.e. for both mineral and geothermal potential indicators, using equation (Eq. 17), which takes into account the distance between indicators for each pair-row retuned in the near matrix of  the proximity analysis:  (17)  was given by Eq. 16 (see Chapter 6), and! was calculated separately >0.The circos diagram in Figure 158 shows the geomine potential weights per commodity, grouped by economic mineral class. The scatterplot in Figures 159 show how wgeomine values range with respect to mine count for individual commodities in the 7 economic mineral classes.  p = wgeoindicator ×wmineral indicatorpair distancewmineral potential wgeothermal potentialfor all types of geo-indicators, and was presented in Chapter 5. Individual p values were subsequently summed up either per mineID, mineral commodity, or geothermal indicator type/class, scaled by their corresponding layer importance weights (assigned a value of 1.0 for all mineral indicators or as listed in Table 30 for geothermal indicators), and eventually assigned to each type of  indicator as an importance weight, denoted wgeomine. The geomine weights were then merged with  both mineral and potential indicator layers, and were exported for mapping in ArcGIS. A layer of geo-potential marked the location and geomine values for all geothermal potential indicators and a layer of mineral potential marked the location and values for all mineral potential indicators. In addition, 148 layers were created for each of the 148 commodities of interest in the model, for which wgeomine 248Figure 158: Geomine weights (wgeomine) per commodity and economic mineral class.  249!Figure 159: Geomine indicator potential (wgeomine) for precious metals.  2508.2 Mapping geomine potential 8.2.1. Potential layers, buffers, and fishnet grids At the completion of  the above-described steps 149 .shp files were extracted and imported as geo-referenced layers into ArcGIS for the calculation of  geomine potential. The modelling process from this point onwards was common for all layers: —	 Geodesic buffers were created around each point in the layer, using the ArcGIS Buffer Tool. The values of  the suitable application areas radii presented in Chapter 7 were used to define the radius of  the circular buffer drawn around each point (Figure 160). — 	 A 50 km × 50 km rectangular grid was created using the ArcGIS Fishnet tool and it was used to aggregate layer geomine potential defined by individual data points. The grid was also clipped (using the ArcGIS Clip tool) to land areas, which was consistent with the scope of  this study. Figure 161 shows the clipped grid (in grey), overlaid by the geothermal (in red) and mineral (in black) indicator buffer layers.  —	 The fishnet grid was further reduced in size by merging all point potential layers together and using them to clip the land-based grid, thereby reducing its extent to only those cells that intersect any potential point (Figure 162). This resulted in an overall reduction in the number of  grid cells from 1,460,340 in the original (unclipped) layer, to 69,384 in the final (point-based) version, which in turn reduced the computational load by 2 orders of  magnitude, and corresponding calculation times from several days to several hours per pass. —	 A grid-based potential layer was then created for each indicator layer, by intersecting the clipped grid layer with the corresponding indicator buffers, and by summing up the potential contribution of  all intersected buffer slivers to each corresponding intersecting cell in the grid (Figure 163 — cells that intersect multiple points display a higher opacity that cells with fewer intersecting points). This was done by applying the ArcGIS Spatial Join and Dissolve tools to each layer pair in turn.
 251
252Figure 160: Buffer creation around all geothermal and mineral indicator data points. 
253Figure 161: Indicator buffers and fishnet grid used in the calculation of  geomine potential. 
254Figure 162: Clipping the extent of  the rectangular grid.Figure 163: Combining potential into grid-based layers using individual geomine points. 8.2.2 Overview geomine potential maps The 149 grid layers created in the previous step were ultimately combined into a single layer of  geomine potential, using the Merge tool (Figure 164). Commodities were also grouped by economic class and merged to produce 9 additional aggregate geomine potential maps for select economic mineral classes and commodities (Figures 165-169).  Separate colour scales were used for the aggregate maps, the class-based maps, and when mapping individual commodities, to facilitate inter-group comparisons. Although a representative Partially overlapping buffers around individual (GPP) pointsClipped, geodesic, 50 km × 50 km rectangular fishnet gridBuffer slivers intersecting each cell — higher opacity indicates larger degree of  sliver overlapAggregate potential of  overlaying slivers assigned to each cell 255sample of  the produced maps has been included in this document, the full list of  interactive maps for all commodities analyzed ca be found online, at: www.lenapatsa.com/thesis/w_geomine_map.html.  The following maps are presented: the overall geomine potential given by all mineral indicators combined (Figure 165); the potential of  base metals (Figure 166a), precious metals (Figure 166b), industrial minerals (Figure 167a), energy metals, minerals and fuels (Figure 167b); and the potential of  gold (Figure 168a), copper (Figure 168b), lead (Figure 169a), zinc (Figure 169b), and silver (Figure 170). Interactive versions of  all maps, plus additional ones corresponding to all listed commodities can be found online, at www.lenapatsa.com/thesis/. "Figure 164: Geographic extent of  geomine potential for all indicators.  256!Figure 165: Geographic extent of  geomine potential for a) base; and b) precious metals. 
 257!Figure 166: Geographic extent of  geomine potential for a) energy; and b) industrial minerals.   258!Figure 167: Geographic extent of  geomine potential for a) gold; and b) copper.  259Figure 168: Geographic extent of  geomine potential for a) lead; and b) zinc.
 260!Figure 169: Geographic extent of  geomine potential for silver. 8.3 Regions of  high geomine potential incidence The maps presented above indicated that high values of  wgeomine are not ubiquitous but are for the most part concentrated in the West Coast of  the USA, in Central America, and in Chile. To help illustrate this with a bit more clarity, the wgeomine maps  (World, Western USA, and South America) were redrawn using class break intervals that were proportional to one-half  of  the standard deviation for wgeomine  (Figures 170, 171, and 172 respectively). Based on these figures, two generalized areas were selected for more refined analysis, the first comprising of  the US states of  Nevada, California, and Arizona, and the second covering the northern part of  Chile. Based on the individual commodity contributions to the total wgeomine for each area of  interest (Figure 173), the remaining analysis was further limited to gold and copper in the Western USA, and copper in Chile.
 261
262Figure 170: World map of  wgeomine, redrawn using a 1/2-std. dev. class break interval. !Figure 171: wgeomine for the Western US, redrawn using a 1/2-std. dev. class break interval.   263!Figure 172: wgeomine for South America, redrawn using a 1/2-std. dev. class break interval.   264!Figure 173: Commodity contributions to wgeomine in a) NV, CA, and AZ, & b) in Chile.
 2658.3.1 Nevada, California, and Arizona Figures 174a and 174b map gold and copper indicator points in the US States of  Nevada (NV), California (CA), and Arizona (AZ). Mapped points were sized by their corresponding mineral potential weight (wmine) for each (lat, lon) location. Areas of  high mineral potential can be roughly inferred to by inspection: they correspond to areas of  high point overlap. The vast majority of  these points were not derived from currently active operating mines (Table 62). For gold, active properties in the stage of  extraction amounted to 10.69% of  all points. For copper, the percentage is slightly higher, at 11.17%.  Table 62: Gold and copper mines by stage and state in NV, CA, and AZ. !Each point on the map also has a value for wgeomine, which depends on the wmine point value and on the wgeo values of  all geo-indicator points in its vicinity. To derive the wgeomine map for the Western USA, gold- and copper-derived mineral indicators were mapped, buffered using a 5 km radius, spatially joined with a a 5 km × 5 km rectangular grid, and dissolved to calculate an aggregate wgeomine value for each grid cell. Figures 175a and 175b, 176a and 176b, and 177a illustrate this process for gold in Nevada. Figure 177b shows the corresponding resultant (i.e. dissolved) wgeomine grid for copper. Aggregate wgeomine cell values for both grids were locally normalized (/max), by equating in each case the highest wgeomine value to 1.0.
 266!!Figure 174: Mineral indicators for NV, CA, and AZ for a) gold; and b) copper, sized by wmine.
 267(a)(b)
268Figure 175: a) Geo-indicators in NV, USA; and b) corresponding buffers, by type.  (a) (b)
269Figure 176: a) Mineral indicators (gold) in NV, USA; and b) corresponding buffers.  (a) (b)!Figure 177: 5 km × 5 km wgeomine grid in NV, CA & AZ, a) for gold; and b) for copper.
 270(a)(b)8.3.2 Chile Figure 178 shows three 5 km×5 km grids for Chile: grid (a) aggregates the geothermal potential (wgeo) of  intersecting geo-indicator buffers, per grid cell; grid (b) aggregates the mineral potential values (wmine) of  intersecting mineral indicator buffers (represented in this case by copper), per grid cell; and grid (c) aggregates the geomine potential (wgeomine) for the same mineral indicator/copper buffers, again per grid cell. Most geomine potential is concentrated in and around the Los Flamengos National Reserve that lies to the east of  Antofagasta. By contrast, the vast majority of  copper reserves are found south of  Antofagasta, between Copiapó and San Antonio. While in the US geothermal and mineral indicators physically overlap in numerous locations, Chilean mineral and geothermal potential show little overlap and are for the most part colocated in and around the Los Flamengos National Reserve area. Copper was always  known to be closely associated with geothermal systems, and thus the model was expected to indicate a very close proximity between Chilean copper resources and local geothermal potential. Actual global model results for copper indicated that it has the 3rd highest value for wgeomine after gold and silver. Unlike the US, which, as the largest geothermal power producer in the world, accounts for 30% of  all installations and 28% of  total worldwide MWe production, at the time of  the analysis Chile had zero active geothermal production. Due to this absence of  input GPPs, the resultant wgeomine values for copper in Chile were considered to be somewhat under-estimated, even after accounting for the mitigating step of  treating select volcano types as primary geo-indicators.  Ideally, as future geothermal exploration data is released to the public domain, it will be possible to update the model and derive a more accurate assessment of  both geothermal and geomine potential, particularly if  exploration identifies high potential geo-resources in areas coinciding with those of  high wmine. Regrettably, at the time of  analysis no such data was available, with most prospectors understandably keeping such information in-house to remain competitive.  271!Figure 178: 25 km2 geomine wgeomine grids in Chile, for a) wgeo; b) wmine; & c) wgeomine.
 2728.4 Recommendation scale The final step in this research was to define a simple decision-making scale of  recommendation levels, based on the calculated and normalized value ranges of  geomine potential wgeomine (Table 63): Table 63: Geomine potential wgeomine based recommendation level scale.The geomine (wgeomine) potential scale maps could then be re-drawn. Figure 179 shows the combined (for gold and copper) recommendation map for CA, NV, and AZ. A similar map was produced for copper in Chile (Figure 180). Following the completion of  the research and prior to publication, a new geothermal power plant (Cerro Pabellón (-21.8576461, -68.1514643) ) came online in Pampa Apacheta, in Ollagüe, N. Chile (Dixon and Nakagawa 2016). Dixon and Nakagawa (2016) provide an overview of  the 48-MWe installation, along with an assessment of  the current state and potential for development of  geothermal energy in mining in Chile. With regards to the model presented herein, this GPP lies in an area indicated as “moderately recommended” for co-development (Figure 180). This helps to a) partially validate the model, as the GPP was not built in an area that was classified as “not recommended”; and b) to further support the assessment that the value of  wgeomine is most likely underestimated, as the new GPP is located in an area that was classified as “moderately recommended” rather than “highly recommended”. Including the new installation in the results presented herein would have required repeating the analysis from the start, beginning with a reassessment of  wGPP, which would actually affect the entire downstream assessment of  wgeo and wgeomine. This was deemed unaffordable within the budgetary and time constraints of  this research and was therefore deferred for the future and added to the list of  recommendations for further work. 
 273
274Figure 179: wgeomine recommendation map for gold and copper in CA, NV, AZ. !.QO]ZM ":MKWUUMVLI\QWVUIXNWZ+PQTMJI[MLWV\PMSUOMWUQVM_OMWUQVMOZQL
 275Chapter 9: Conclusion This chapter provides a review of  the research objectives and approach, summarizes outcomes, lists contributions to knowledge, discusses limitations, and gives a number of  recommendations for future work.  9.1 Overview of  research objectives, approach & outcomes Mines need energy to operate, and they procure it from a variety of  sources that includes renewables. The industry has been slow to adopt geothermal energy, despite the many advantages this particular resource offers to mineral processing. Geothermal resources are for the most part found in specific areas around the globe that host a combination of  conditions necessary for their existence. This along with the fact that as an energy source, geothermal is typically extracted, produced and consumed in situ, limits any potential for integration between mining and geothermal to specific areas of  co-occurrence.  The literature review established that there is good technical and operational overlap between geothermal and mining, and that geothermal has the ability to supply electricity, steam, and hot water to meet a variety of  needs within the scope of  extractive activity, water management, ore processing and general operations. Moreover, the overlap between the development life cycle of  mines and geothermal power stations also supports the argument for integration. The ability to quickly and inexpensively assess whether geothermal merits further consideration as a potential source of  energy for a given project could help improve the uptake of  geothermal energy by the mining industry. The development of  such a tool/framework was the primary focus of  this research, and the intent was to demonstrate that such an assessment could be made by non-experts, without having to resort to more complex specialist analysis that is typically part of  exploration.  This research was conducted under the assumption that active production (of  either geothermal or mining) is the strongest possible indicator of  potential, and that resources do not exist  276at random, but are rather related to various phenomena and conditions that combined to create them. It was thus theorized that it would be possible to statistically and geographically study the associations between known production and such phenomena, and that using geographical, statistical, and decision-making techniques it would be possible to estimate the existence (and by extension coexistence) of  geothermal and mineral resources.  The flow chart of  Figure 180 gives an overview of  the approach taken to calculate geomine potential. On the mineral side, projects in a life cycle stage other than production were used to indicate potential at a lower degree of  confidence to that attributed to production sites. Similarly, lower confidence geo-potential was calculated by studying the geographical association between various phenomena and conditions known to be directly or indirectly associated to geothermal resources (such as quakes and volcanoes) and currently active production.  The initial list of  geo-indicators included geothermal power stations, volcanoes, earthquakes, tectonic plate boundaries, surface heat flow, and thermal springs. The corresponding list of  mineral indicators included properties whose state/status spanned the entire mineral development life cycle, namely recoded occurrences of  particular commodities, sites of  exploration, projects under feasibility study, mines under construction, currently operating facilities, and projects that are temporarily closed, permanently shuttered, or currently undergoing reclamation.  Following the identification of  the indicators that would be used to assess potential on both sides, suitable data was procured to physically/geographically represent each indicator. A complete listing of  geothermal power stations was compiled by harvesting partial listings of  geothermal production sites from online sources and supplementing with potential candidate production sites identified in the literature. The final list of  GPPs was created through painstakingly verifying the physical location of  each plant manually, using publicly available remote sensing (Google Earth), and by supplementing any missing information (such as type or capacity) from the literature.   277Data was procured for all identified geo-indicators, either through direct download (when possible), or through the vastly most complex process of  harvesting, scraping, clean-up/scrubbing, transformation and reclassification (Figure 16). Procuring data for the mineral potential indicators followed a similar process, with the bulk of  the data being supplied by the USGS in the form of  the MRDS dataset. A smaller subset of  the data was provided by InfoMine, an industry vendor. In both cases, harvesting, scrapping, clean-up/scrubbing, transformation and reclassification were needed prior to merging the two sets into a single mineral potential indicator dataset. Extensive exploratory data analysis and strategic filtering helped improve data quality and reduce both sets to a more manageable size prior to merging.  An importance weight (wgpp ) was defined for each GPP type based on average production data, and it was correspondingly assigned to each GPP data point. A metric for geothermal potential (wgeo) was then calculated for all non-production geo-indicator subtypes using proximity analysis, as the sum of  the ratios of  each GPP’s wgeo and the distance between said GPP and every geo-indicator point. The resulting geo-indicator weights (wgeo’s) were scaled by a relative importance weight that was calculated for each geo-indicator using degrees-of-separation-analysis. On the mineral side, a mineral potential weight wmine was calculated by combining component weights for each point’s size, state, stage, order, and value, using AHP and pairwise comparisons. Proximity analysis supplied the distances between each geo-indicator point and each mineral indicator point. These were then combined with wgeo and wmine to derive a value for geomine potential (wgeomine) for each point in the mineral indicator dataset. Geomine potential maps were produced by creating buffers around each point that were sized depending on point type, and by assigning point wgeomine to the buffers, spatially joining with a fishnet grid, aggregating overlapping buffer values, and assigning the combined geomine weight to the corresponding intersecting grid cell.  278In summation, the research questions and objectives presented in Chapter 3 have been addressed (Table 64). The developed methodology used a wide range of  analytical and graphical techniques to derive an estimate for the existence and coexistence of  geothermal and mineral resources, even in areas of  zero geothermal or mineral production information. Given data of  acceptable quality and a well-defined, computationally-manageable geographical scope, the methodology can potentially be used by mining operators to make a reasonable initial assessment on the viability of  utilizing geothermal energy in a given mining project. !Figure 181: Approach overview for the calculation of  geomine potential. 
 279Table 64: Summarized research outcomes.  2809.2 Contributions to knowledge This research makes the following contributions to knowledge: 1. The compiled comprehensive listing of  geothermal power plants, verified for locationaccuracy, which will be released to the public domain following the publication of  this thesis.2. Associated maps of  geothermal production, produced for select areas of  interest.3. A ranked listing of  Primary & Secondary indicators that can be used to assess geothermalpotential, even in areas with no current geothermal production.4. The AHP-based model for calculating relative importance weights for geo-indicators that waspresented, and which can be extended to include additional indicators of  geothermalpotential to the ones used in this research.5. An updated listing of  Thermal Springs in the USA, partially corrected for location accuracy.6. A comprehensive map of  mineral potential, for 148 distinct commodities. An interactive,simplified version of  the mineral potential map will be released online (in Tableau Public andin a shapefile (shp) format) following the publication of  this Thesis.7. A ranked listing of  148 mineral commodities, based on their proximity to geo-potential.8. A comprehensive map of  geomine potential, for 148 distinct commodities. An interactive,simplified version of  the mineral potential map will be published online (in Tableau Publicand in a shapefile (shp) format), following the publication of  this Thesis. 2819.3 Limitations This research focused on the development of  an evaluation and decision-making framework for geomine potential using primarily open and freely available datasets.  Confidence determinations on model results were heavily reliant on the quality (defined in terms of  accuracy and completion) of  the input data. As the majority of  the data inputs to the model were derived from freely-available, open, crowd-sourced datasets, whose quality was known from the outset to be problematic (as discussed in Chapters 4 and 5), steps were taken during scrubbing and analysis to improve data quality through the application of  strategic clean-up, visual verification, reclassification, and filtering. These steps improved upon the quality of  the data ultimately passed to the model, but were not able to eliminate all quality-related concerns (e.g. when dealing with old data or any physical location inaccuracies). This should be taken into account when interpreting any results deriving from this research.  Research findings are limited to the specific data inputs and assumptions to approach that were employed in constructing the model, calculating all importance weights (as they are currently defined), and exporting the maps included in this document. The AHP model in its current form accounts for 6 separate geo-indicators, namely: GPPs, volcanoes, earthquakes, tectonic plates, heat flow, and thermal springs. Expanding the model by adding criteria/indicators (e.g. locations of  active geothermal exploration) will necessitate the recalculation of  all relative importance matrices, by adding the new indicators to the existing matrix and recalculating the relative importance weights for all indicators (old and new). This is a core requirement of  the AHP method and can not be avoided, therefore anyone wishing to extend this model will need to have a good grasp of  AHP/multi-criteria decision-making basics. Because the model was designed to be used by non-experts, it is not in its current form exhaustive: in fact researchers with expert knowledge can improve upon it by extending, and/or  282refining the model, and by adding higher-quality data and their own expert opinions (through the re-definition and re-calculation of  the AHP pairwise matrices) which the model can inherently accommodate. For example, the current evaluation of  geothermal potential can be further refined through the addition of  geological (e.g. ore body) or geochemical indicators. At the time of  the analysis, there was no global thermal springs data available for download. The dataset that was considered for the model was limited (in terms of  geographic scope) to the USA, and it was notably out-dated and contained, in parts, location inaccuracies. Although steps were taken to improve its quality, the decision was ultimately made to exclude it from the model, until a more accurate and complete thermal springs dataset becomes available through future research and/or exploration. The geomine calculations were global in scope, but due to limitations in resources (time and available hardware), exported global geomine potential maps were of  low-resolution. Higher-resolutions maps were exported, but they had to be limited in geographic scope to render them processable by the available system/setup.  The completed analysis contains calculations for a total of  148 commodities, but due to limitations on the final length of  this document, maps for only a select number of  commodities are presented herein. More specifically, maps have been exported for 5 commodities of  interest, namely for gold, copper, silver, lead, and zinc. A simplified version of  calculated geomine proximity values for all 168 commodities have been exported to Tableau Public and will be made available for general release following the publication of  this thesis. The complete results will also be released as a .shp file, but they will only be accessible to users with expertise in and access to GIS software.  2839.4 Recommendations for future work Further research on the integration potential between mineral and geothermal resources is recommended: 1.	 Update the model to account for the new GPP in Northern Chile and rerun the analysis. 2.	 Keep the GPP dataset up-to-date, by adding new installations of  GPPs as they come online and by removing older GPPs as they are decommissioned.  3.	 Add further contextual fields to the dataset (e.g., plant ownership) to augment its value and to allow for more detailed analysis (e.g., on the current state of  the geothermal industry).  4.	 Re-apply the developed approach to a smaller geographic area (e.g. a particular US state), using locally-procured, higher-quality, and higher-resolution data. For example, in collaboration with a mineral producer, the model can be applied to in-house, not-publicly available datasets, to assist internal decision-making. If  the data is of  a type already assessed and ranked in this research, the model can be applied as-is; if  new indicators are introduced (e.g. geo-thermometry, or ore body geology), all of  the AHP ranking weights will need to be recalculated, as per AHP requirements.  5.	 Extend the model by including more complex, expert data (such as geo-thermometry), to allow for a more refined assessment of  geomine potential.  6.	 Extend the model by including data layers related to Sustainable Development. Examine the effects on non-technical, not resource-specific factors to the integration of  mining and geothermal, especially in areas of  high geomine potential.  7.	 With future improvements of  computer power, produce higher-resolution maps of  dswq	`geomine potential, or include datasets currently excluded from the model due to their prohibitively expensive calculation costs (e.g. surface heat flow).
 284Bibliography Abudeif, AM, AA Abdel Moneim, and AF Farrag. 2015. “Multicriteria Decision Analysis Based on Analytic Hierarchy Process in GIS Environment for Siting Nuclear Power Plant in Egypt.” Annals of  nuclear energy 75 682–92. http://dx.doi.org/10.1016/Fj.anucene.2014.09.024  Agapova, Maria, Brian W Bresnahan, Ken F Linnau, Louis P Garrison, Mitchell Higashi, Larry Kessler, and Beth Devine. 2017. “Using the Analytic Hierarchy Process for Prioritizing Imaging Tests in Diagnosis of  Suspected Appendicitis.” Academic Radiology 24 (5): 530–37. 
https://doi.org/10.1016/j.acra.2017.01.001  Anane, Makram, Lamia Bouziri, Atef  Limam, and Salah Jellali. 2012. “Ranking Suitable Sites for Irrigation With Reclaimed Water in the Nabeul-Hammamet Region (Tunisia) Using GIS and AHP-Multicriteria Decision Analysis.” Resources, Conservation and Recycling 65 36–46. 
https://doi.org/10.1016/j.resconrec.2012.05.006  Aydin, Nazli Yonca, Elcin Kentel, and H Sebnem Duzgun. 2013. “GIS-Based Site Selection Methodology for Hybrid Renewable Energy Systems: A Case Study From Western Turkey.” Energy Conversion and Management 70 90–106. https://10.1016/j.enconman.2013.02.004  Azadeh, Amir, M Osanloo, and M Ataei. 2010. “A New Approach to Mining Method Selection Based on Modifying the Nicholas Technique.” Applied Soft Computing 10 (4): 1040–61. 
https://doi.org/10.1016/j.asoc.2009.09.002  Azcue, Jose M. 2012. Environmental Impacts of  Mining Activities: Emphasis on Mitigation and Remedial Measures. Springer Science & Business Media. Bakane, Piyush A. 2013. “Overview of  Extraction of  Mineral/metals With the Help of  Geothermal Fluid.” Proceedings of  The Thirty-Eighth Workshop on Geothermal Reservoir Engineering. California 11–13. Accessed October 11, 2017. 
http://large.stanford.edu/courses/2016/ph240/makalinao2/docs/bakane.pdf   Balaji, K, and VS Senthil Kumar. 2014. “Multicriteria Inventory Abc Classification in an Automobile Rubber Components Manufacturing Industry.” Procedia CIRP 17 463–68. 
https://doi.org/10.1016/j.procir.2014.02.044  Bangert, Andrea. 2016. The Petrogenesis of  Rocks and Minerals. Santa Cruz Institute for Particle Physics. Berrittella, Maria, Antonella Certa, Mario Enea, and Pietro Zito. 2008. “Transport Policy and Climate Change: How to Decide When Experts Disagree.” environmental science & policy 11 (4): 307–14. https://doi.org/10.1016/j.envsci.2008.01.008  Bertani, Ruggero. 2015. “Geothermal Power Generation in the World 2010–2014 Update Report.” Proceedings World Geothermal Congress 2015. Accessed October 11, 2017. 
https://pangea.stanford.edu/ERE/db/WGC/papers/WGC/2015/01001.pdf   Bitarafan, M, S Hashemkhani Zolfani, Sh Lale Arefi, and EK Zavadskas. 2012. “Evaluating the Construction Methods of  Cold-Formed Steel Structures in Reconstructing the Areas Damaged in Natural Crises, Using the Methods AHP and Copras-G.” Archives of  Civil and Mechanical Engineering 12 (3): 360–67. https://doi.org/10.1016/j.acme.2012.06.015   285Bixley, PF. 2003. “Geothermal Development At Lihir Island.” Papua New Guinea: Proceedings 25th NZ Geothermal Workshop. Accessed October 11, 2017. 
https://www.geothermal-energy.org/pdf/IGAstandard/WGC/2005/1343.pdf   Bloomquist, R.G. 2006. “Economic Benefits of  Mineral Extraction From Geothermal Brines.” Accessed October 11, 2017. 
https://pangea.stanford.edu/ERE/pdf/IGAstandard/Russia/MEGB-2006/20Bloomquist.pdf   Bourcier, W., Ralph W., Johnson M., Bruton C., and Gutierrez P. 2006. “Silica Extraction At the Mammoth Lakes Geothermal Site.” Accessed October 11, 2017. https://e-reports-ext.llnl.gov/pdf/334693.pdf   Brake, DJ, and R Fulker. 2000. “The Ventilation and Refrigeration Design for Australias Deepest and Hottest Underground Operation: The Enterprise Mine.” Proc MassMin 611–21. Accessed October 11, 2017. http://www.mvaust.com.au/assets/v01-the-ventilation-and-refrigeration-design-for-australias-dee.pdf   Brath, Richard, and David Jonker. 2015. Graph Analysis and Visualization: Discovering Business Opportunity in Linked Data. John Wiley & Sons. British Geological Survey. 2014. “Heat Energy Beneath Glasgow.” Accessed October 11, 2017. 
http://www.bgs.ac.uk/research/energy/geothermal/heatEnergyGlasgow.html  Brundtland Commission. 1987. “United Nation Report of  the World Commission on Environment and Development (Our Common Future).” Transmitted to the General Assembly as an Annex to document A/42/427–Development and International Cooperation: Environment. Accessed October 11, 2017. http://www.un-documents.net/our-common-future.pdf   Cairo, Alberto. 2016. The Truthful Art: Data, Charts, and Maps for Communication. New Riders. CalEnergy. 2014. “Projects: Imperial Valley, United States - Generation, Geothermal.” Accessed October 11, 2017. http://www.midamericanrenewablesllc.com/imperialvalley_geothermal.aspx  Canada, Environment and Climate Change. 2016. “Environmental Code of  Practice for Metal Mines.” Accessed October 11, 2017. 
https://www.ec.gc.ca/lcpe-cepa/default.asp?lang=En&n=CBE3CD59-1&offset=4  Canada, Indian and Northern Affairs. 2007. “Stages of  Mineral Exploration and Development in the Northwest Territories.” Ottawa: Canada. Accessed October 11, 2017. 
http://www.un-documents.net/our-common-future.pdf   Canada, National Resources. 2005a. “Benchmarking the Energy Consumption of  Canadian Open-Pit Mines.” Accessed October 11, 2017. http://www.nrcan.gc.ca/sites/oee.nrcan.gc.ca/files/pdf/publications/industrial/mining/open-pit/Open-Pit-Mines-1939B-Eng.pdf   Canada, National Resources. 2005b. “Benchmarking the Energy Consumption of  Canadian Underground Bulk Mines.” Accessed October 11, 2017. http://www.nrcan.gc.ca/sites/www.nrcan.gc.ca/files/oee/pdf/publications/industrial/mining/bulk-mine/BulkMines-1939A-Eng.pdf   CBC News. 2013. “First Nations Declare Victory Against Taseko’s B.c. Mine.” Accessed October 11, 2017. http://www.cbc.ca/news/business/first-nations-declare-victory-against-taseko-s-b-c-mine-1.2325102  CBC News. 2014. “Taseko New Prosperity Mine At Fish Lake Rejected Again.” Accessed October 11, 2017. http://www.cbc.ca/news/canada/british-columbia/taseko-new-prosperity-mine-at-fish-lake-rejected-again-1.2553002   286CleanEnergyBC. 2013. “Fact Sheet: Run-of-river.” Accessed October 11, 2017. 
http://www.cleanenergybc.org/media/CEBC_FS-RunOfRiver_web.pdf   Curtis, R, J Lund, B Sanner, L Rybach, and G Hellstrom. 2005. “Ground Source Heat Pumps & Geothermal Energy for Anyone, Anywhere: Current Worldwide Activity.” Proceedings World Geothermal Congress, Antalya, Turkey 24–29. Accessed October 11, 2017. http://www.sanner-geo.eu/media/1437.pdf   Dixon, Andrea P. and Masami Nakagawa. 2016. “Geothermal Energy for Mining in Chile: Current Situation and Opportunities”. Proceedings of  The Forty-First Workshop on Geothermal Reservoir Engineering. February 22-24, 2016, Stanford University, California. Accessed March 8, 2018. 
https://pangea.stanford.edu/ERE/db/GeoConf/papers/SGW/2016/Dixon.pdf   Dickson, Mary H, and Mario Fanelli. 2013. Geothermal Energy: Utilization and Technology. Routledge. DiPippo, Ronald. 2012. Geothermal Power Plants: Principles, Applications, Case Studies and Environmental Impact. Butterworth-Heinemann. Driesner, Doug, and Alan Coyner. 2008. “Major Mines of  Nevada 2004.” Nevada Bureau of  Mines and Geology (NBMG), Special Publication P-19. Accessed October 11, 2017. http://minerals.nv.gov/uploadedFiles/mineralsnvgov/content/Programs/Mining/Forms_Publications/mm2008.pdf   Duyvesteyn, Willem PC. 1992. “Recovery of  Base Metals From Geothermal Brines.” Geothermics 21 (5-6): 773–99. https://doi.org/10.1016/0375-6505(92)90030-D  Easley, D., and D. Kleinberg. 2010. “Chapter 2: Graphs.” In Networks, Crowds, and Markets: Reasoning about a Highly Connected World, Cambridge University Press. Erdogan, Seyit Ali, Jonas Šaparauskas, and Zenonas Turskis. 2017. “Decision Making in Construction Management: AHP and Expert Choice Approach.” Procedia Engineering 172 270–76. 
https://doi.org/10.1016/j.proeng.2017.02.111  Esa, Mohd Reza, Anthony Halog, and Lucia Rigamonti. 2017. “Strategies for Minimizing Construction and Demolition Wastes in Malaysia.” Resources, Conservation and Recycling 120 219–29. 
https://doi.org/10.1016/j.resconrec.2016.12.014  Farias, Daniel Almarza. 2000. “Geothermal Development in Chile.” Accessed October 11, 2017. 
http://os.is/gogn/unu-gtp-sc/UNU-GTP-SC-18-06.pdf   Farias, Daniel Almarza, Juan Rojas, Diego Morata, and Valerio Acocella. 2015. “Geothermal Exploration in Chile: Country Update.” Proceedings World Geothermal Congress 2015. Accessed October 11, 2017. https://pangea.stanford.edu/ERE/db/WGC/papers/WGC/2015/01033.pdf   Ferretti, Valentina. 2011. “Integrating Multicriteria Analysis and Geographic Information Systems: A Survey and Classification of  the Literature.” Accessed October 11, 2017. 
http://porto.polito.it/id/eprint/2460560  Flóvenz, Ólafur G. 2012. “Phases of  Geothermal Development.” Renewable energy training program. Accessed 2017. 
http://www.esmap.org/sites/esmap.org/files/Flovenz%20Day%201%20-WB-2-phases-final.pdf   Flynn, Thomas, DT Trexler, and JL Hendrix. 1986. “Geothermal Enhancement of  Mineral Processing in Nevada: Final Report”. April 25, 1985 - June 30, 1986. 
https://www.osti.gov/geothermal/servlets/purl/7011778/7011778.pdf    287Garcia, Solange, Yara Cintra, SR Rita de Cássia, and Fabiano Guasti Lima. 2016. “Corporate Sustainability Management: A Proposed Multi-Criteria Model to Support Balanced Decision-Making.” Journal of  Cleaner Production 136 181–96. https://doi.org/10.1016/j.jclepro.2016.01.110  Girvan, Michelle, and Mark EJ Newman. 2002. “Community Structure in Social and Biological Networks.” Proceedings of  the national academy of  sciences 99 (12): 7821–26. 
https://doi.org/10.1073/pnas.122653799  Golden, Bruce L, Edward A Wasil, and Patrick T Harker. 1989. The Analytic Hierarchy Process. New York: Springier-Verlag. Grasby, Stephen Edward, D M Allen, S Bell, Z Chen, G Ferguson, A Jessop, M Kelman, M Ko, J Majorowicz, M Moore, J Raymond, and R Therrien. 2011. “Geothermal Energy Resource Potential of  Canada.” Geological Survey of  Canada. Accessed October 11, 2017. 
http://publications.gc.ca/collections/collection_2013/rncan-nrcan/M183-2-6914-eng.pdf   Gudmundsson, JS, and JW Lund. 1985. “Direct Uses of  Earth Heat.” International Journal of  Energy Research 9 (3): 345–75. Accessed October 11, 2017. 
http://onlinelibrary.wiley.com/doi/10.1002/er.4440090311/full  Hannam, P. 2014. “Australia’s Emissions Cut Target Triples Overnight Thanks to Failure to Repeal Carbon Tax.” Accessed October 11, 2017. Accessed October 11, 2017. http://www.smh.com.au/environment/climate-change/australias-emissions-cut-target-triples-overnight-thanks-to-failure-to-repeal-carbon-tax-20140602-39esv.html  Harrison, S. 2010. “Geothermal Technologies Program 2010 Peer Review: Technologies for Extracting Valuable Minerals and Compounds.” Accessed October 11, 2017. 
http://energy.gov/sites/prod/files/2014/02/f7/specialized_harrison_extracting_metals.pdf   Hartman, Howard L, and Jan M Mutmansky. 2002. Introductory Mining Engineering. John Wiley & Sons. Hendrickson, Thomas P, Milos Nikolic, and Jasenka Rakas. 2016. “Selecting Climate Change Mitigation Strategies in Urban Areas Through Life Cycle Perspectives.” Journal of  Cleaner Production 135 1129–37. https://doi.org/10.1016/j.jclepro.2016.06.075  ICMM, International Council on Mining & Metals. 2012. Water Management in Mining: A Selection of  Case Studies. London, UK: International Council on Mining & Metals. Accessed October 11, 2017. 
https://www.icmm.com/website/publications/pdfs/water/water-management-in-mining_case-studies  IHFC. 2016. “Global Heat Flow Database.” Accessed October 11, 2017. 
http://www.heatflow.und.edu/data.html  Jamasmie, C. 2014. “Chile Set to Make Mining Desalination Mandatory.” Accessed October 11, 2017. http://www.mining.com/chile-set-to-make- mining-desalination-mandatory-57150 Janke, Jason R. 2010. “Multicriteria GIS Modeling of  Wind and Solar Farms in Colorado.” Renewable Energy 35 (10): 2228–34. https://doi.org/10.1016/j.renene.2010.03.014  Jianqing, Zhang. 2011. “The Application of  Analytic Hierarchy Process in Mine Gas Prevention System.” Procedia Engineering 26 1576–84. https://doi.org/10.1016/j.proeng.2011.11.2341  Judd, E. 2013a. “Why Solar is a Good Fit for Mines: A Codelco Case Study.” Accessed October 11, 2017. 
http://assets.fiercemarkets.net/public/sites/energy/reports/solarminesreport.pdf    288Judd, E. 2013b. “Solar Pv for Codelco and Collahausi.” Canadian Clean Energy Conferences, Renewables and Mining, 2013. Accessed October 11, 2017. 
http://www.renewablesandmining.com/wp-content/uploads/2013/08/solar-pv-in-chile.pdf   K4RES-H. 2010. “Innovative Applications Geothermal Utilization for Seawater Desalination - Key Issue 5: Innovative Applications.” Accessed October 11, 2017. Accessed October 11, 2017. 
http://www.erec.org/fileadmin/erec_docs/Projcet_Documents/K4_RES-H/K4RES-H_Geothermal_desalination.pdf   Kagel, Alyssa, and Karl Gawell. 2005. “Promoting Geothermal Energy: Air Emissions Comparison and Externality Analysis.” The Electricity Journal 18 (7): 90–99. https://doi.org/10.1016/j.tej.2005.07.004  Kappes, Daniel W. 2002. “Precious Metal Heap Leach Design and Practice.” Proceedings of  the Mineral Processing Plant Design, Practice, and Control 1 1606–30. Accessed October 11, 2017. http://www.ore-max.com/pdfs/resources/precious_metal_heap_leach_design_and_practice.pdf   Kaufmann, K. 2014. “Simbol Materials’ Lithium Extraction Could Help Salton Sea.” Accessed October 11, 2017. http://archive.desertsun.com/article/20140222/BUSINESS0302/302220055/Simbol-Materials-lithium-extraction-Salton-Sea  Kimball, Sarah. 2010. “Favourability Map of  British Columbia Geothermal Resources.” diss., University of  British Columbia. Accessed October 11, 2017. 
https://open.library.ubc.ca/collections/ubctheses/24/items/1.0071365  Klein, Greg. 2011. “Prosperity for Whom? Taseko’s Bc Mine Hangs in the Balance.” Accessed October 11, 2017. http://resourceclips.com/wp-content/remote/pub_pdf/prosperity_for_whom_TKO.pdf   Klein, Kerry, and Linda Gaines. 2011. “Reducing Foreign Lithium Dependence Through Co-Production of  Lithium From Geothermal Brine.” GRC Transactions 35 645–49. Accessed October 11, 2017. 
https://anl.app.box.com/s/8jrm6w3d0l1ar8ygc1q68ujirppovwyf   Koufos, Katarina. 2012. “Assessing the Potential to Implement Open Loop Geothermal Systems in Canadian Underground Mines.” Masters Abstracts International 51 (04). Accessed October 11, 2017. 
http://digitool.library.mcgill.ca/webclient/StreamGate?folder_id=0&dvs=1507803503280~253  KPMG. 2001. “Capitalizing on Sustainable Development in Mining.” Accessed October 11, 2017. http://www.kpmg.com/ZA/en/IssuesAndInsights/ArticlesPublications/General-Industries-Publications/Documents/CSD_Mining_FINAL.pdf   Krzywinski, Martin. 2016. “Circos - Data Visualization.” Accessed October 11, 2017. 
http://circos.ca/intro/data_visualization/ Lahsen, Alfredo, Nelson Munoz, and Miguel Angel Parada. 2010. “Geothermal Development in Chile.” Proceedings World Geothermal Congress 25. Accessed October 11, 2017. 
http://www.geothermal-energy.org/pdf/IGAstandard/WGC/2010/0118.pdf   Laval, Alfa. 2011. “Waste Heat Recovery — Optimizing Your Energy System.” Accessed October 11, 2017. http://local.alfalaval.com/en-gb/about-us/news/Documents/PPI00443EN%20Optimising%20your%20energy%20system.pdf   Lawless, Jim. 2010. “Geothermal Lexicon for Resources and Reserves Definition and Reporting.” Accessed October 11, 2017. 
http://www.pir.sa.gov.au/__data/assets/pdf_file/0006/147876/Geothermal_Lexicon_2010.pdf    289Lee, Steve, Mentore Vaccari, and Terry Tudor. 2016. “Considerations for Choosing Appropriate Healthcare Waste Management Treatment Technologies: A Case Study From an East Midlands Nhs Trust, in England.” Journal of  Cleaner Production 135 139–47. https://doi.org/10.1016/j.jclepro.2016.05.166  Li, Kewen, Huiyuan Bian, Changwei Liu, Danfeng Zhang, and Yanan Yang. 2015. “Comparison of  Geothermal With Solar and Wind Power Generation Systems.” Renewable and Sustainable Energy Reviews 42 1464–74. Accessed October 11, 2017. 
http://www.cugb.edu.cn/uploadCms/file/20600/papers_upload/20151104105047960712.pdf   Liang, Ruoyu, Shuai Song, Yajing Shi, Yajuan Shi, Yonglong Lu, Xiaoqi Zheng, Xiangbo Xu, Yurong Wang, and Xuesong Han. 2017. “Comprehensive Assessment of  Regional Selenium Resources in Soils Based on the Analytic Hierarchy Process: Assessment System Construction and Case Demonstration.” Science of  The Total Environment 605 618–25. https://doi.org/10.1016/j.scitotenv.2017.06.150  Limited, Newcrest Mining. 2017. “Lihir, Png.” Accessed October 11, 2017. 
http://www.newcrest.com.au/our-business/operations/lihir/  Lolli, F, Alessio Ishizaka, and R Gamberini. 2014. “New AHP-Based Approaches for Multi-Criteria Inventory Classification.” International Journal of  Production Economics 156 62–74. 
http://dx.doi.org/10.1016/j.ijpe.2014.05.015  Kerr Wood Leidal Associates 2015. “An Assessment of  the Economic Viability of  Selected Geothermal Resources in British Columbia.” Report 2015-11. Accessed October 11, 2017. 
http://www.geosciencebc.com/i/project_data/GBCReport2015-11/GBC2015-11_KWL_Geothermal%20Economics_Project_Report_Revised.pdf   Lund, John W. 2003. “The USA Geothermal Country Update.” Geothermics 32 (4): 409–18. 
https://doi.org/10.1016/S0375-6505(03)00053-1  Macnab, Jamie David. 2011. “A Review of  the Potential Thermal Resource in Glasgow’s Abandoned Coal Mine Workings.” diss., University of  Strathclyde. Accessed October 11, 2017. http://www.esru.strath.ac.uk/Documents/MSc_2011/Macnab.pdf   Malczewski, Jacek. 1999. GIS and Multicriteria Decision Analysis. John Wiley & Sons. Malczewski, Jacek. 2006. “GIS‐based Multicriteria Decision Analysis: A Survey of  the Literature.” International Journal of  Geographical Information Science 20 (7): 703–26. 
http://dx.doi.org/10.1080/13658816.2015.1086924  Malczewski, Jacek, and Claus Rinner. 2015. Multicriteria Decision Analysis in Geographic Information Science. Springer. Martenies, Sheena E, Donele Wilkins, and Stuart A Batterman. 2015. “Health Impact Metrics for Air Pollution Management Strategies.” Environment international 85 84–95. https://dx.doi.org/10.1016%2Fj.envint.2015.08.013  Melaku, Markos. 2005. “Geothermal Development At Lihir–an Overview.” Proceedings of  World Geothermal Congress 1–4. Accessed October 11, 2017.  Milutinović, Biljana, Gordana Stefanović, Petar S Đekić, Ivan Mijailović, and Mladen Tomić. 2017. “Environmental Assessment of  Waste Management Scenarios With Energy Recovery Using Life Cycle Assessment and Multi-Criteria Analysis.” Energy. Accessed October 11, 2017. 
https://www.geothermal-energy.org/pdf/IGAstandard/WGC/2005/1343.pdf    290Mishra, Ashutosh Kumar, Shikhar Deep, and Abhishek Choudhary. 2015. “Identification of  Suitable Sites for Organic Farming Using AHP & GIS.” The Egyptian Journal of  Remote Sensing and Space Science 18 (2): 181–93. https://doi.org/10.1016/j.ejrs.2015.06.005  Mitchell, A. 2005. The Esri Guide to GIS Analysis, Volume 2: Spatial Measurements and Statistics. Redlands. CA: Esri Press. Moghaddam, Majid Kiavarz, Younes Noorollahi, Farhad Samadzadegan, Mohammad Ali Sharifi, and Ryuichi Itoi. 2013. “Spatial Data Analysis for Exploration of  Regional Scale Geothermal Resources.” Journal of  Volcanology and Geothermal Research 266 69–83. 
https://doi.org/10.1016/j.jvolgeores.2013.10.003  Moghaddam, Majid Kiavarz, Farhad Samadzadegan, Younes Noorollahi, Mohammad Ali Sharifi, and Ryuichi Itoi. 2014. “Spatial Analysis and Multi-Criteria Decision Making for Regional-Scale Geothermal Favorability Map.” Geothermics 50 189–201. 
https://dpoi.org/10.1016/j.geothermics.2013.09.004  Newfoundland and Labrador. “Labrador Mining and Power: How Much and Where From?” (2012). Accessed October 11, 2017. http://www.powerinourhands.ca/pdf/Labrador_Mining.pdf   Newmont Mining Corporation. 2010. “Geothermal Energy Takes Center Stage.” Accessed October 11, 2017. http://www.newmont.com/features/our-environment-features/Geothermal-Energy-Takes-Center-Stage  NOAA, National Oceanic and Atmospheric Administration. 2016. “Thermal Springs List for the United States.”. US Department of  Commerce. Accessed October 11, 2017. http://www.ngdc.noaa.gov/nndc/struts/results?op_0=eq&v_0=AK&op_1=l&v_1=&op_2=l&v_2=&t=100006&s=1&d=1  Li, Kewen, Huiyuan Bian, Changwei Liu, Danfeng Zhang, and Yanan Yang. “Comparison of  Geothermal With Solar and Wind Power Generation Systems.” Renewable and Sustainable Energy Reviews 42 (2015): 1464–74. https://doi.org/10.1016/j.rser.2014.10.049  O’Sullivan, John. 2014. “Introduction to Geothermal Science and Engineering.” Technical Workshop on Geothermal Regulation Environmental Licensing and Reservoir Modelling – Andes. Accessed October 11, 2017. 
http://www.irena.org/DocumentDownloads/events/2014/June/TechnicalTraining/1_Sullivan.pdf   Patsa, Eleni. 2009. “Seawater Heat Recovery for the University of  British Columbia’s District Heating System.” Accessed October 11, 2017. 
http://lenapatsa.com/lenapatsa/wp-content/uploads/x-contest-submission.pdf   Patsa, Eleni, Dirk Van Zyl, Sadiq J Zarrouk, and Nastaran Arianpoo. 2015. “Geothermal Energy in Mining Developments: Synergies and Opportunities Throughout a Mine’s Operational Life Cycle.” Proceedings World Geothermal Congress 2015. Accessed October 11, 2017. 
https://pangea.stanford.edu/ERE/db/WGC/papers/WGC/2015/02060.pdf  Patsa, Eleni, Dirk Van Zyl, and Sadiq J Zarrouk. 2015. “The Lindal Diagram for Mining Engineering.” GRC Transactions 39. Accessed October 11, 2017. http://pubs.geothermal-library.org/lib/grc/1032144.pdf   Preene, M, and PL Younger. 2014. “Can You Take the Heat? Geothermal Energy in Mining.” Mining Technology 123 (2): 107–18. http://dx.doi.org/10.1179/1743286314Y.0000000058  Ramos, Esmeralda Peralta, Katrin Breede, and Gioia Falcone. 2015. “Geothermal Heat Recovery From Abandoned Mines: A Systematic Review of  Projects Implemented Worldwide and a Methodology for  291Screening New Projects.” Environmental Earth Sciences 73 (11): 6783–95. 
https://doi.org/10.1007/s12665-015-4285-y  Roig-Tierno, Norat, Amparo Baviera-Puig, Juan Buitrago-Vera, and Francisco Mas-Verdu. 2013. “The Retail Site Location Decision Process Using GIS and the Analytical Hierarchy Process.” Applied Geography 40 191–98. https://doi.org/10.1016/j.apgeog.2013.03.005  Rybach, Ladislaus. 2007. “Geothermal Sustainability.” GHC Bulletin. Accessed October 11, 2017. 
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.371.254&rep=rep1&type=pdf   Saaty, Thomas L. 1980. The Analytic Hierarchy Process: Planning. Priority Setting. Resource Allocation, MacGraw-Hill, New York International Book Company 287. Saaty, Thomas L. 2008. “Decision Making With the Analytic Hierarchy Process.” International Journal of  Services Sciences 1 83–98. Accessed October 11, 2017. 
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.409.3124&rep=rep1&type=pdf   Saaty, Thomas L, and Luis G Vargas. 2013. The Logic of  Priorities: Applications of  Business, Energy, Health and Transportation. Springer Science & Business Media. Sadeghi, Behnam, and Masoumeh Khalajmasoumi. 2015. “A Futuristic Review for Evaluation of  Geothermal Potentials Using Fuzzy Logic and Binary Index Overlay in GIS Environment.” Renewable and Sustainable Energy Reviews 43 818–31. https://doi.org/10.1016/j.rser.2014.11.079  Sánchez-Lozano, Juan M, Jerónimo Teruel-Solano, Pedro L Soto-Elvira, and M Socorro García-Cascales. 2013. “Geographical Information Systems (GIS) and Multi-Criteria Decision Making (MCDM) Methods for the Evaluation of  Solar Farms Locations: Case Study in South-Eastern Spain.” Renewable and Sustainable Energy Reviews 24 544–56. https://doi.org/10.1016/j.rser.2013.03.019  Sanyal, Subir K. 2010. “Future of  Geothermal Energy.” Proceedings, thirty-fifth workshop on geothermal reservoir engineering, SGP-TR-188, Stanford University, Stanford 1–3. Accessed October 11, 2017. https://gondwana.stanford.edu/ERE/pdf/IGAstandard/SGW/2010/sanyal.pdf   Scott, BJ. 2012. “Guideline for Mapping and Monitoring Geothermal Features.” Accessed October 11, 2017. https://www.boprc.govt.nz/media/285502/guideline-mapping-and-monitoring-surface-geothermal-features-final-pdf-adobe-acrobat-pro.pdf   Scott, Brad, and Bridget Robson. 2012. “Surface Geothermal Featues.” IEA-GIA Environmental Mitigation Workshop. Accessed October 11, 2017. 
http://iea-gia.org/wp-content/uploads/2012/10/Scott-Surface-Features-Taupo-June-2012.pdf   Şener, Şehnaz, Erhan Şener, Bilgehan Nas, and Remzi Karagüzel. 2010. “Combining AHP With GIS for Landfill Site Selection: A Case Study in the Lake Beyşehir Catchment Area (Konya, Turkey).” Waste management 30 (11): 2037–46. https://doi.org/10.1016/j.wasman.2010.05.024  Sigmundsson, Arnar Freyr. 2012. “Using Geothermal Energy for Raffine Heating in Copper Production.” Accessed October 11, 2017. 
http://skemman.is/stream/get/1946/12073/30488/1/Arnar_Freyr_Sigmundsson.pdf   Simmons, Stuart F, and Patrick RL Browne. 2000. “Hydrothermal Minerals and Precious Metals in the Broadlands-Ohaaki Geothermal System: Implications for Understanding Low-Sulfidation Epithermal Environments.” Economic Geology 95 (5): 971–99. https://doi.org/10.2113/gsecongeo.95.5.971   292Soltani, Atousa, Kasun Hewage, Bahareh Reza, and Rehan Sadiq. 2015. “Multiple Stakeholders in Multi-Criteria Decision-Making in the Context of  Municipal Solid Waste Management: A Review.” Waste Management 35 318–28. https://doi.org/10.1016/j.wasman.2014.09.010  Soltanmohammadi, Hossein, Morteza Osanloo, and Abbas Aghajani Bazzazi. 2010. “An Analytical Approach With a Reliable Logic and a Ranking Policy for Post-Mining Land-Use Determination.” Land Use Policy 27 (2): 364–72. https://doi.org/10.1016/j.landusepol.2009.05.001  Sui, Lili, Yang Ju, Yongming Yang, Yong Yang, and Aishan Li. 2016. “A Quantification Method for Shale Fracability Based on Analytic Hierarchy Process.” Energy 115 637–45. 
https://doi.org/10.1016/j.energy.2016.09.035  SummitPost. 2016. “Summitpost.org”. Accessed October 11, 2017. http://www.summitpost.org  The Data Visualisation Catalogue. 2016. “Chord Diagram.” Accessed October 11, 2017. 
http://www.datavizcatalogue.com/methods/chord_diagram.html  The White House. 2014. “Climate Change and President Obama’s Action Plan.” Accessed December 31, 2014. http://www.whitehouse.gov/climate-change  Thomas, L. Saaty. 1988. The Analytic Hierarchy Process. Thomas, L. Saaty, P. Kearns Kevin, and G. Vargas Luis. 1991. The Logic of  Priorities. Treadgold, Tim. 2015. "Widespread Mine Closures To Follow The Commodity Price Collapse". Forbes.com. https://www.forbes.com/sites/timtreadgold/2015/07/21/widespread-mine-closures-to-follow-the-commodity-price-collapse/. Accessed March 1, 2018. Trexler, Dennis T, Tomas Flynn, and James L Hendrix. 1990. “Heap Leaching.” Geo-Heat Center Quarterly Bulletin 12 (4): 1–4. 
http://data.nbmg.unr.edu/public/Geothermal/GreyLiterature/Trexler_HeapLeaching_1990.pdf   Tufte, Edward R. 1983. The Visual Display of  Quantitative Information. UNESCO. 2014. “Education for Sustainable Development.” Accessed October 11, 2017. 
http://unesco.ca/home-accueil/esd-edd  US Department of  Energy. 2011. “Critical Materials Strategy. U.S. Department of  Energy Report.” Accessed October 11, 2017. http://energy.gov/sites/prod/files/DOE_CMS2011_FINAL_Full.pdf   US Department of  Energy, Office of  Energy Efficiency and Renewable Energy. 2017. “Geothermal FAQs.” Accessed October 11, 2017. https://energy.gov/eere/geothermal/geothermal-faqs  USGS. 2015. “Mineral Resources Data System (MRDS).” Accessed October 11, 2017. 
http://mrdata.usgs.gov/mrds/  Uyan, Mevlut. 2013. “GIS-Based Solar Farms Site Selection Using Analytic Hierarchy Process (AHP) in Karapinar Region, Konya/turkey.” Renewable and Sustainable Energy Reviews 28 11–17. 
https://doi.org/10.1016/j.rser.2013.07.042  Vaidya, Omkarprasad S, and Sushil Kumar. 2006. “Analytic Hierarchy Process: An Overview of  Applications.” European Journal of  operational research 169 (1): 1–29. 
https://doi.org/10.1016/j.ejor.2004.04.028 Van Wyk, L. 2013. “Diavik Diamond Mines Inc. Wind Farm Project.” Accessed October 11, 2017. https://www.bullfrogpower.com/remotemicrogrids/presentations/session_6_diavik_diamond_mine.pdf    293Vučijak, Branko, Sanda Midžić Kurtagić, and Irem Silajdžić. 2016. “Multicriteria Decision Making in Selecting Best Solid Waste Management Scenario: A Municipal Case Study From Bosnia and Herzegovina.” Journal of  Cleaner Production 130 166–74. 
https://doi.org/10.1016/j.jclepro.2015.11.030  Yegnanarayanan, V, and GK Umamaheswari. 2011. “A Note on the Importance of  Collaboration Graphs.” International Journal of  Mathematical Sciences and Applications 1 (3): 1113–21. 
http://ijmsa.yolasite.com/resources/8--sep.pdf   Yuen, Kevin Kam Fung. 2014. “The Primitive Cognitive Network Process in Healthcare and Medical Decision Making: Comparisons With the Analytic Hierarchy Process.” Applied Soft Computing 14 109–19. https://doi.org/10.1016/j.asoc.2013.06.028  Zahedi, F. 1986. “The Analytic Hierarchy Process—a Survey of  the Method and Its Applications.” Interfaces 16 96–108. https://doi.org/10.1287/inte.16.4.96  Zarrouk, Sadiq J, and Hyungsul Moon. 2014. “Efficiency of  Geothermal Power Plants: A Worldwide Review.” Geothermics 51 142–53. https://doi.org/10.1016/j.geothermics.2013.11.001  Zhang, Jiuquan, Yirong Su, Jinshui Wu, and Hongbo Liang. 2015. “GIS Based Land Suitability Assessment for Tobacco Production Using AHP and Fuzzy Set in Shandong Province of  China.” Computers and Electronics in Agriculture 114 202–11. https:doi.org/10.1016/j.compag.2015.04.004 
 294Appendix A: MDRS Preliminary Metadata Analysis Notes: — All of  these tables share a common field ( Deposit ID [dep_id] ), which acts as the foreign key that connects them together. I have excluded it from all but the Deposits table in the list below.  — The metadata analysis at this stage primarily consisted of running consecutive queries of  the type: SELECT distinct Field FROM TABLE, for each TABLE/Field combo of  interest — USGS provides a dictionary for the entire dataset (i.e. descriptions of  all tables and fields): 
http://mrdata.usgs.gov/mrds/full/dd.php.   Template: TABLE 	 Field [field_name] (example_content) DEPOSITS 	 Deposit ID [dep_id]  	 {1,2,3…} 	 Development State [dev_st]  	 {Occurrence | Past Producer | Unknown | Producer | Prospect | Plant} 	 Deposit Type [dep_tp]  	 {1164 unique types, eg. Alluvial | Auriferous Sulfide Quartz Veins | ...etc.} 	 Plant Type [plant_tp]  	 {Pigment Plant | Metal Plant | Smelter | Leach | Beneficiation (Mill) | …etc.} 	 Operation Type [oper_tp] 	 {Underground | Surface-Underground | Surface | Placer | …etc.} 	 Mining Method [min_meth] 	 {56 unique methods, eg. Open Pit | Open Stope - Breast Stoping | …etc.} 	 Milling Method [mill_meth] 	 {40 unique methods, eg. Sizing | Flotation | Hydrometallurgy Unspecified | …etc.} 	 Record Type [rec_tp]  ☜  USGS metadata: "General characteristics of  the record" 	 {Site | Deposit | District | Model Type | Region} 	 Production Size [prod_size] 	 {N | S | Y | U | L | M} 	 Plant Identifier [plant_ident]  ☜ USGS metadata:  "Subtype of  processing plant such as gravity, flotation, crush."   	 {30 unique entries, e.g. Tio2 Pigment-Sulfide | Precipitation | Hydromet |…etc.}  295COORDS 	 Latitude [wgs84_lat] 	 Longitude [wgs_84_lon] 	 Elevation [elev] 	 Elevation Units [elev_u] 	 Relative Position [rel_pos] 	 	 {58783 unique entries, e.g. "Estimated Location, Probably Within 0.5 Miles."} LOCATION ☜ redundant for QGIS and possibly Tableau, may help form cell groupings for grid calculations 	 Country [country] 	 State/Province [state_prov] 	 County [county] NAMES 	 Property Name [name] 	 Status [status]  	 	 {Previous | Current | Contains | Included in}  ☜ potential use as filter LAND_STATUS 	 Land Status [land_st] 	 	 {21 unique entries, e.g. Private | National Forest | State Park |…etc.}  OWNERSHIP 	 Owner Name [owner_name] 	 Owner Type [owner_tp] 	 	 {Owner-Operator | Owner | Operator | Unknown | Lessee/Operator | …etc.} 	 Beginning Year (of  ownership) [beg_yr]  ☜ to help determine current owner  	 End Year (of  ownership) [end_yr]  ☜ to help determine current owner  	 Stake (% of  ownership) [pct]   ☜ to help determine multiple owners TECTONIC 	 Tectonic Setting [tect_set] 	 	 {114 unique entries, e.g. Superior Province | Midcontinent Rift |…etc.} MODEL 	 Mineral Deposit Model [model_name]   ☜ USGS metadata: "Name or short title of  the  mineral deposit model to which the site conforms" 	 	 {193 unique entries, e.g. Polymetallic veins | Porphyry Mo, low-F | Porphyry Cu |…etc.}  296ROCKS 	 1st-Order Name [first_ord_nm]  ☜ USGS metadata: "Broadest category term for lithology" 	 	 {Sedimentary Rock | Unconsolidated Deposit | Volcanic Rock (Aphanitic) | …etc.} 	 2nd-Order Name [second_ord_nm]  ☜ USGS metadata: "Second broadest term for lithology" 	 	 {52 unique entries, e.g. Carbonate | Gravel | Pyroclastic Rock |…etc.} 	 3rd-Order Name [third_ord_nm]  ☜ USGS metadata: "Third broadest rock type category" 	 	 {75 unique entries, e.g. Tuff  | Basalt | Granite | Limestone | Diorite | Sandstone |…etc.} 	 4th-Order Name [fourth_ord_nm]  ☜ USGS metadata: "Fourth broadest rock type term" 	 	 {27 unique entries, e.g. Greenstone | Dunite | Harzburgite | Welded Tuff  |…etc.} 	 Lithologic Term [low_name]  ☜ USGS metadata: "Most specific lithologic term" 	 	 {156 unique entries, e.g. Dolomite | Gravel | Tuff  | Basalt | Pegmatite | Granite |…etc.} ALTERATION 	 Alteration [alterat]   ☜ This will also require extensive scrubbing, but may be useful in  identifying hydrothermal alteration deposits   	 	 {236 unique entries, e.g. Carbonitization | Complete serpentinization of  host rock |…etc.} 	 Alteration Type [alt_type]  ☜ USGS metadata: "General extent of  the alteration described,  either L for local or R for regional." 	 	 {L | R | NULL} 	 Alteration Text [alterat_text]  ☜ This will also require extensive scrubbing, but may be useful  in identifying hydrothermal alteration deposits   	 	 {7709 uniques entries, e.g. Limonite, quartz | Silicification | Kaolinitization |…etc.} RESOURCES 	 Probable Resources [probable]  ☜ USGS metadata: "Probable - the part of  a mineral resource  defined as PROBABLE under the current international JORC, CRIRSCO, CIM, and SME standards." 	 Proved Resources [proved]  ☜ USGS metadata: "The part of  a mineral resource defined as  PROVED under the current international JORC, CRIRSCO, CIM, and SME standards." 	 Reserves [reserves]  ☜ USGS metadata: "Reserves - the sum of  proved plus probable reserves." 	 Measured Resources [meas]  ☜ USGS metadata: "The part of  a mineral resource defined with a high level of  confidence by closely spaced sampling [...]" 	 Indicated [indic]  ☜ USGS metadata: "The part of  a mineral resource defined with a reasonable level of  confidence by sampling spaced closely enough to assume but not prove continuity [...]" 	 Demonstrated Resources [demo]  ☜ USGS metadata: "Sum of  measured plus indicated resources." 	 Inferred Resources [infer]  ☜ USGS metadata: "The part of  a mineral resource defined with a low  level of  confidence by a limited amount of  sampling. Geological and grade continuity are not assured." 	 Total Resources [tot_resources]  ☜ USGS metadata: "Sum of  demonstrated plus inferred resources."  297	 Total Endowment [tot_endowment]  ☜ USGS metadata: "The total amount of  material including  production, reserves, and resources. The original pre-mining endowment or magnitude of  the deposit." 	 Resource Type [res_tp] 	 	 {In-situ | Milling | Leaching | Combined milling and leaching | Tailings}  	 Units [units]  ☜ USGS metadata: "Units for all the reserve and resource fields." 	 	 {mt ore | NULL | mt concentrate | kg commodity | mt commodity | mt |…etc.} 	  RESOURCE_DETAIL 	 Commodity Name [commod]   	 	 {181 unique entries, e.g. Zinc | Copper | Tin | Gold | Molybdenum | Chromium |…etc.} 	 Commodity Importance [import]   	 	 {Major | Primary | Secondary | Trace | Minor | Tertiary | NULL} 	 Item [item]  ☜ USGS metadata: "Description of  the material for which reserve or resource estimates are  provided, such as "Cu at 0.6% cutoff", "WO3", "CuOx", or "Copper oxides". 	 	 {181 unique entries, e.g. Zn | Cu | Sn | NULL | Cr203 | Au | Ag | Fe | Mo | …etc.} 	 Commodity Group [commod_group]  ☜ USGS metadata: "Group of  the commodity,  such as PGE for platinum-group elements" 	 	 {72 unique entries, e.g. Zinc | Copper | Tin | Gold | Molybdenum | Chromium |...etc. } 	 Grade [grd]  ☜ USGS metadata: "Grade of  commodity in the amount fields of  the parent resource table.  Expressed as weight-% for most commodities, grams/metric ton for gold, silver, and platinum group metals." 	 Grade Unit [grd_unit]  ☜ USGS metadata: "The units for the grade of  the commodity. Weight percent  for most commodities, grams per metric ton for gold, silver, and platinum group metals." 	 	 {wt-pct | g/mt} COMMODITY 	 Commodity Code [code]  ☜ USGS metadata: "Distinct abbreviated code for the commodity. These codes are not drawn from an external standard." 	 	 {185 unique entries, e.g. CU | AU | AG | FE | MN | STN | F | ZN | PB | …etc.} 	 Commodity Name [commod]  	 	 {185 unique entries, e.g. Copper | Gold | Silver | Iron | Manganese | Stone |…etc.} 	 Commodity Type [commod_tp]  ☜ USGS metadata: "General type of  commodity: metal,  nonmetal, both, or (rarely) energy." 	 	 {Metallic | Non-metallic | M | N | Energy | E} 	 Commodity Group [commod_group]  ☜ USGS metadata: "Name of  a group to which a  commodity belongs, for example PGE for the various platinum-group elements." 	 	 {100 unique entries, e.g. Copper | Gold | Silver | Iron | Manganese | Stone |…etc.} 	 Commodity Importance [import]  ☜ USGS metadata: "The relative economic importance  of  the commodity - Primary, Secondary, or Tertiary. (*)"  	 	 {185 unique entries, e.g. Primary | Secondary | Tertiary | …etc.}  298WORKINGS (at the site) 	 Workings Type [work_type]   ☜ USGS metadata: "The name of  the workings, such as "South Pit".  Can be the entire mine." 	 	 {Surface | Underground | Unknown | Surface/Underground | Water} MATERIALS 	 Ore/Gangue [ore_gangue]  ☜ USGS metadata: "Category term indicating the role of  the material(**)" 	 	 {Ore | Gangue | Unknown | Trace} 	 Material [material]  ☜ USGS metadata: "Name of  a material present at the site. Primarily International  Mineralogical Association (IMA) approved mineral names, but also includes ore-commodity terms such as marble." 	 	 {779 unique entries, e.g. Gravel | Silver | Sand and Gravel | Limestone | Gold |…etc.} CON_PROC 	 Concentration Process [conc_proc]  ☜ USGS metadata: "Category terms, non-controlled text field." 	 	 {93 unique entries, e.g. GRAVITY | METAMORPHISM | HYDROTHERMAL |…etc.} 	 Concentration Process Description [conc_proc_text]  ☜ USGS metadata: "Text describing processes  that concentrated or enriched the minerals in the deposit, such as oxidation, evaporation, etc., from old MRDS." 	 	 {3188 unique entries, e.g. Oxidation | Hydrothermal Solutions |…etc.}	  PRODUCTION 	 Mined/Processed Amount [mined]  ☜ USGS metadata: "The amount of  ore mined or material  processed during the specified time period." 	 Units [units]  ☜ USGS metadata: "Mined units, such as grams, metric tons, or cubic meters." 	 	 {NULL | mt | cy | g | dollars | cm} 	 Material Type [item]  ☜ USGS metadata: "Type of  material mined or processed, such as "ore" 	 	 {814 unique entries, e.g. NULL | ore | HG | ORE AG PB ACC |…etc.} 	 Accuracy [acc]  ☜ USGS metadata: "Accuracy of  the MINED number" 	 	 {NULL | Estimate | Accurate} PRODUCTION_DETAIL 	 Commodity Code [code]  ☜ USGS metadata: "Abbreviated form of  the commodity" 	 	 {86 unique entries, e.g. AU | PB | BI | AG | ZN | CR | NI | CU | S_A | CD | …etc.} 	 Commodity Name [commod]   	 	 {86 unique entries, e.g. Gold | Lead | Bismuth | Silver | Zinc | Chromium |…etc.} 	 Commodity Group [commod_group]  ☜ USGS metadata: "Group of  the commodity,  such as PGE for platinum-group elements" 	 	 {63 unique entries, e.g. Gold | Lead | Bismuth | Silver | Zinc | Chromium |…etc.} 	 Commodity Importance [import]   299	 	 {NULL | Minor | Major | Trace | Primary | Secondary | Tertiary} 	 Material Type [item]  ☜ USGS metadata: "Further description of  the material produced,  such as 'Ore', 'Concentrate', 'MetaL'", 'WO3', or 'Copper oxides'.' 	 	 {188 unique entries, e.g. NULL | Al2O3 | Conc. | Ore | Cr2O3 | recovered Cu |…etc.} 	 Mined/Produced Amount [amt]  ☜ USGS metadata: "The amount of  the commodity recovered."	 	Units [units]  ☜ USGS metadata: "The units for the amount of  the commodity recovered." 	 	 {mt | NULL | g | WO3 | FLASKS | W | CARLOAD} 	 Recovery [pct]  ☜ USGS metadata: "The recovery percentage for this commodity from  the ore amount specified in the parent table PROD, MINED field." 	 Grade [grd]  ☜ USGS metadata: "The grade of  the commodity in the ore specified in the parent  table PROD, MINED field. (***)" 	 Grade Unit [grd_unit]  ☜ USGS metadata: "The units for the grade of  the commodity.  Weight percent for most commodities, grams per metric ton for gold, silver, and platinum group metals." 	 	 {NULL | wt-pct | MT | g/mt | %null% | oz/st | WT% | wt/pct | g/m3} 	  (*) Extended USGS metadata comment on Commodity: Commodity Importance Primary commodities have a strong effect on the economics of  the project, may be viable as the only commodity.  Secondary commodities can be economically recovered but have little effect on the viability of  the project.  Tertiary commodities are economically interesting but not currently recoverable. They include contaminants, impurities, prospecting pathfinders, ore-genesis indicators, etc. (**) Extended USGS metadata comment on Materials: Ore Gangue  Ore contains the recovered commodity. Gangue is waste or contaminant. Trace is low concentration but geologically significant. (***) Extended USGS metadata comment on Production_detail: Grade  Expressed as weight percent for most commodities, grams per metric ton for gold, silver, and platinum group metals.  300

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0365825/manifest

Comment

Related Items