Module-22-resource-estimation-exercises.pdf

MICROMINE Training v10.1

Module 22– Resource Estimation

MODULE 22 TABLE OF CONTENTS Lesson 1 – Resource Estimation Concepts ............................................................................... ................................................ 1 CONCEPTS ............................................................................................................................................................................. 1 THE TRAINING PROJECT .......................................................................... ........................................................... ................... 2 NUMERIC EXCEPTIONS ............................................................................ ........................................................... ................... 2 Lesson 2 - Classical Statistics ........................................................... ........................................................... ............................. 4 CLASSICAL STATISTICAL ANALYSIS ............................................................................................... ....................................... 4 Tables................................................................................................................................................................................ 4 Lesson 3 – Generate Downhole Coordinates ........................................................ ........................................................... ....... 11 WIREFRAMING ........................................................... ............................................................ .............................................. 13 Lesson 4: Assay Data Flagging ................ ............................................................ ........................................................... ....... 15 FLAGGING/SELECTION ......................................................................................... ........................................................ 15 Flagging using Solid wireframes ........................................................ ........................................................... ................. 15 Flagging using DTMs ..................................................... ........................................................... ..................................... 15 Lesson 5 – Balancing Cut ......................................................................................................... .............................................. 18 Lesson 6 - Compositing .......................................................... ........................................................... ..................................... 21 Lesson 7 - Geostatistics .................................................................... ........................................................... ........................... 25 THEORY ........................................................... ............................................................ ........................................................ 25 Variography .......................................................... ............................................................ .............................................. 25 Anisotropy ................................................... ............................................................ ........................................................ 27 Semi variogram fo rmula ........................................................... ........................................................... ........................... 27 Semi variogram model formula .................. ............................................................ ........................................................ 27 PRACTICE ............................................................................................................................................................................ 30 Review......................................................... ............................................................ ........................................................ 30 Nugget......................................................... ............................................................ ........................................................ 30 Using Semi Variograms V ariograms .................................................. ........................................................... ..................................... 32 Optimum Lag ............................................................................ ........................................................... ........................... 33 Directional variogram, main: ................................................... ........................................................... ........................... 39 Direction of maximum continuity, primary variogram;................................................... ............................................... 40 Model the variograms:.............................................................. ........................................................... ........................... 41 Indicator variograms: ..................................................... ........................................................... ..................................... 44 Relative variograms: ................................................................ ........................................................... ........................... 48 Cross Validation: V alidation: .................................................. ............................................................ .............................................. 51 Lesson 8 – E mpty Cell Model .................................................................... ........................................................... ................. 55 BLOCK MODEL CELL SIZE ................................................... ........................................................... ..................................... 55 Flagging.......................................................................................................................................................................... 55 Checklist ................................................................................... ........................................................... ........................... 55 How to decide the block size ..................................................... ........................................................... ........................... 56 Subcelling ................................................................................. ........................................................... ........................... 57 Lesson 9 – Modelling Principles ............................................................................................................................................ 61 Declustering.......................................................... ............................................................ .............................................. 61 Specific gravity and block size .................................................. ........................................................... ........................... 62 Change of Support .......................................................... ........................................................... ..................................... 62 Proportional effect .......................................................... ........................................................... ..................................... 62 Interpolate parent blocks only .................................................. ........................................................... ........................... 63 Multiple runs ......................................................... ............................................................ .............................................. 63

Lesson 10 – Grade Interpolation ....................................................... ........................................................... ........................... 65 INTERPOLATION ................................................... ............................................................ .............................................. 65 Search ellipse ........................................................ ............................................................ .............................................. 66

Copyright © Micromine Pty Ltd 2006



How to decide search ellipse size ................................................... ........................................................... ..................... 67 INVERSE DISTANCE WEIGHTING ......................................................................................................................................... 68 Recommended Values ........................................................... ........................................................... ............................... 69 ORDINARY KRIGING ............................................................................................................................................................ 70 Ordinary kriging formula:.................................................... ........................................................... ............................... 71 ORDINARY KRIGING , RELATIVE VARIOGRAMS ......................................................... ........................................................... . 72 MULTIPLE INDICATOR KRIGING .......................................................... ........................................................... ..................... 73 KRIGING VARIATIONS .......................................................................................................................................................... 78

Lesson 11 – Model validation ................................................................................................................................................ 81 Global validation: .......................................................................................................................................................... 82 Declustered global estimate ........................................................... ........................................................... ..................... 82 Local validation: ........................................................ ............................................................ ........................................ 83 Model validation: ....................................................... ............................................................ ........................................ 84

Lesson 12 - Block Model Display .......................................................... ........................................................... ..................... 86 Lesson 13 – Resource Classification ...................................................... ........................................................... ..................... 88 Kriging variance: ........................................................................................................................................................... 89

Lesson 14 – Resource reporting ............................................................................................................................................. 91 Lesson 15 – Cut-off grades and grade tonnage curves ................................................................................................ ........... 95 Lesson 16 - Example NVG data Ordinary kriging start to end ......................................................................... ..................... 98 Step 1: Classical statistics exhaustive population .......................................................................................................... 98 Step2: Generate do wnhole coordinates .................................................... ........................................................... ........... 99 Step 3: Assign the wireframe to the assay file .................................................... .......................................................... 100 Step 4: Classical statistics orezone .............................................................................................................................. 100 Step 5: Apply a balancing cut ......................................................... ........................................................... ................... 101 Step 6: Composite the data to equal intervals .............................................................................................................. 102 Step 7: Geostatistics ..................................................................................................................................................... 103 Step 8: Cross validation ............................................................................................................................................... 109 Step 9: Build blank model ............................................................................................................................................ 110 Step 10: Ordinary Kriging ........................................................................................................................................... 111 Model report ..................................................... ............................................................ ................................................ 114 Validation ..................................................................................................................................................................... 114




How to decide search ellipse size ................................................... ........................................................... ..................... 67 INVERSE DISTANCE WEIGHTING ......................................................................................................................................... 68 Recommended Values ........................................................... ........................................................... ............................... 69 ORDINARY KRIGING ............................................................................................................................................................ 70 Ordinary kriging formula:.................................................... ........................................................... ............................... 71 ORDINARY KRIGING , RELATIVE VARIOGRAMS ......................................................... ........................................................... . 72 MULTIPLE INDICATOR KRIGING .......................................................... ........................................................... ..................... 73 KRIGING VARIATIONS .......................................................................................................................................................... 78

Lesson 11 – Model validation ................................................................................................................................................ 81 Global validation: .......................................................................................................................................................... 82 Declustered global estimate ........................................................... ........................................................... ..................... 82 Local validation: ........................................................ ............................................................ ........................................ 83 Model validation: ....................................................... ............................................................ ........................................ 84

Lesson 12 - Block Model Display .......................................................... ........................................................... ..................... 86 Lesson 13 – Resource Classification ...................................................... ........................................................... ..................... 88 Kriging variance: ........................................................................................................................................................... 89

Lesson 14 – Resource reporting ............................................................................................................................................. 91 Lesson 15 – Cut-off grades and grade tonnage curves ................................................................................................ ........... 95 Lesson 16 - Example NVG data Ordinary kriging start to end ......................................................................... ..................... 98 Step 1: Classical statistics exhaustive population .......................................................................................................... 98 Step2: Generate do wnhole coordinates .................................................... ........................................................... ........... 99 Step 3: Assign the wireframe to the assay file .................................................... .......................................................... 100 Step 4: Classical statistics orezone .............................................................................................................................. 100 Step 5: Apply a balancing cut ......................................................... ........................................................... ................... 101 Step 6: Composite the data to equal intervals .............................................................................................................. 102 Step 7: Geostatistics ..................................................................................................................................................... 103 Step 8: Cross validation ............................................................................................................................................... 109 Step 9: Build blank model ............................................................................................................................................ 110 Step 10: Ordinary Kriging ........................................................................................................................................... 111 Model report ..................................................... ............................................................ ................................................ 114 Validation ..................................................................................................................................................................... 114




Table of Exercises Exercise 22.1 Classical Stats ........................................ ............................................................ ................................................ 6 Exercise 22.2 Generate Downhole Coordinates ............................... ........................................................... ........................... 11 Exercise 22.3 Flagging using solid wireframe ...................................................... ........................................................... ....... 16 Exercise 22.4 Balancing Cut ................................................... ........................................................... ..................................... 18 Exercise 22.5 Compositing ..................................................... ........................................................... ..................................... 21 Exercise 22.6 Nugget .................................................... ............................................................ .............................................. 30 Exercise 22.7 Omni Variogram ........................................................ ........................................................... ........................... 33 Exercise 22.8 Horizontal Fan Variogram ................................................................................. .............................................. 35 Exercise 22.9 Vertical Fan Variogram.................................................................................................................................... 37 Exercise 22.12 Creating a Blank Block Model ..................................................... ........................................................... ....... 58 Exercise 22.13 I nverse Distance Weighting ...................................................................................... ..................................... 70



Lesson 1 – Resource Estimation Concepts


Notes:

After this lesson you will understand: •

What are we attempting to model;

•

What decisions do we need to make;

•

What are the most important decisions affecting the modelling;

•

How do we check how good our model is.

Concepts Resource estimation can be conducted for 1D, 2D and 3D models. The resource sector generally requires 3D models except for gridding and other 2D techniques which use surface data to identify anomalous areas that are indicative of prospective subsurface mineralisation. Important considerations for 3D modelling are the search ellipse, compositing, domaining, the interpolation method, dealing with erratic high grades, anisotropy, block sizes and validation (1) The search ellipse includes sample grades relevant to the estimation of block grades and excludes redundant (not required) grades; (2) The compositing ensures the grades used for estimation are weight averaged back to the same length so the estimation process is not biased. (3) Domaining divides the deposit into separate areas such as lodes that have unique geological or grade characteristics that must be interpolated and modelled independently. (4) The interpolation method is the method selected for modelling. This may be Inverse distance weighting (which does not require variography); ordinary kriging; median indicator kriging or multiple indicator kriging. Classical statistics, in particular the shape of the histogram, the shape of the probability plot and the coefficient of variation are useful to select the most appropriate interpolation method. (5) Erratic high grades can be allowed allowed for by applying balancing cuts to grades or by using nonlinear methods such as multiple indicator kriging. An allowance must be made for high grades so that they do not bias the entire model and affect large areas of the model to bias the t he model higher. (6) Anisotropy is the preferred continuity of grade in one direction; isotropy means the grade is equal in all directions. All deposits should exhibit anisotropy and this reflects the nature of deposition and the style of mineralisation. Gold in particular is very changeable and more prone to continuity in one direction to another (7) The size of the blocks required can be directed by the engineers who indicate the SMU (Smallest Mining Unit) or by the drillhole spacing, and by the


Page 22.1



style of the deposit. The parent cells can be used for estimation and subcells can improve the definition to provide an accurate volume.

Notes:

(8) Validation can be both global and local. Global validation means the raw sample data and the wireframe envelope are compared to the block model tonnes and grade to ensure the model reflects the data that was used for the estimation.

The Training Project A good example for resource estimation is an iron project because the data is more regular and can produce strong models. models. For the resource estimation we we will use iron data – Files used in this training project: Collar: IRON COLLAR.DAT Survey: IRON SURVEY.DAT Assay: IRON ASSAY.DAT Assay: IRON ASSAY COMP2.DAT COMP2.DAT Wireframe: Iron.tdb

Numeric Exceptions Always have the numeric exceptions ticked on for all three categories: Ignore characters, characters, Ignore blanks and blanks and
Page 22.2




Notes:

Lesson Summary This lesson has introduced some of the fundamental concepts involved in Resource Estimation. In the following lessons, we will put these concepts into use: search ellipse compositing domaining the interpolation method dealing with erratic high grades anisotropy block sizes validation numeric exception

Good Practice Use Numeric Exceptions throughout.

Help Topics For information on:

See:

Numeric Exceptions

Numeric Exceptions


Page 22.3



Lesson 2 - Classical Statistics

Notes:

Classical Statistical Analysis The aim of the classical statistical analysis is: • To check for the mixing of grade populations and the necessity of separation of grade populations if there are more than one population, • To derive the top cut grades for grade interpolation process, • To determine the natural cut-off grades for interpretation of mineralisation, • To determine the distribution of grades, • To assess the validity of Kriging interpolation process, and • To obtain the statistical parameters for grades.

Classical statistics are used to examine initially the global population and then the mineralisation population. The histogram, the cumulative frequency plot and the probability plot all reveal information about the distribution of population grades.

Tables To obtain statistical parameters, run Stats | Descriptive | Normal/Ln for each element and for each potential domain separately. The statistical parameters are to be recorded in the output file, tabulated and included into the report. Histogram The histogram bin size is selected so the shape of the distribution is apparent, it must be small enough to show the shape and large enough to contain sufficient data. The histogram shows two populations, one at low background grades and another at higher grades; there is a discreet break between the two. The moments such as mean, medium etc are displayed on the right of the histogram. Probability plot The change in the angle of the line on the probability plot helps to indicate the grade at which mineralisation grades occur as opposed to background grades. In other words this is the cut-off grade separating country rock from the ore zone. This grade is used for the grade to design the outlines on the drillholes for the ore zone interpretation. This grade is applied for the delineation of mineralisation polygons.

Histograms, Log Histograms and (Log) Probability Plots should be generated for each element and potential domain using Stats | Distribution process. Use filters to separate domains if possible. All graphs should be plotted, studied and

Page 22.4



included into the report. The potential mixing of grade populations, grade cutoffs for interpretation and top-cuts for grade interpolations should be determined from the histograms and cumulative frequency plots. One of the most important questions is to identify the number of grade populations. The number of populations can be estimated using the Stats | Distribution, (select Probability Plot and Natural Log options). When the probability plot is displayed, run Model | Decompose from the top menu to obtain the statistical parameters of grade populations. This can subsequently be displayed on the drill traces and checked against the geological model and structures.


Notes:

Obtain the grade value that separates background grades from mineralization grades. This is the inflection point on the probability plot. Alternatively this figure may be stated by the Government. This grade value is the grade at which all mineralization polygons will be interpreted in section. This Stage can take from several hours to several days depending on number of elements and number of domains. Domains are separated data for different block models, such as different lodes, or weathering or very high grade, high grade and low grade areas. Please note, it is a good practice to calculate a coefficient of variation that might indicate a potential quality of variograms (COV = St. Dev. / Mean) and the nature of interpolation method required. For a COV of 1 or less than 1.5 IDW may be appropriate, above 1.5 a kriging method should be applied.

Interpretation When interpreting the orebodies and wireframing the ore zones and mineralisation, a decision needs to be made whether a hard boundary or soft boundary is used. A typical hard boundary is where mineralisation is truncated by a fault; a typical soft boundary may be where a stockwork of veins extends from the main mineralisation zone into the country rock. A wireframe can be used for the hard boundary, however grades outside the wireframe or indeed no wireframe may be used when modelling the soft boundary. CHECKLIST •

Classical stats tables are generated for all elements and domains

•

COV values are calculated

•

Histograms, log histograms and probability plots are generated for all elements and domains

•

Tables and graphs are studied and modelling methodology is determined

•

Use numeric exceptions if there are any character values in numeric fields

•

The distributions are described


Page 22.5



Exercise 22.1 Classical Stats

Notes:

1.

Select Stats | Descriptive | Normal/Ln.

2.

Fill in the dialog box: Prompt

Setting

Input File

IRON ASSAY

Type

DATA

Fields | Field Name (1)

T FE

Fields | Field Name (2)

SIO2

Output File

DESCRIPT IRON

As discussed in Lesson 1, all Numeric Exceptions should be selected. 3. Click Run. You should see a window displaying the Normal and Logarithmic descriptive stats showing T Fe. Click Next to review the SIO2 stats. 4. Click Close to exit. You can right-click on the Output File: DESCRIP IRON.REP to view the file. Copy and paste to you Resource document and then Close. 5.

Select Stats | Distribution with: Prompt

Setting

File

IRON ASSAY

Type

DATA

Graph field

T FE

Graph type

HISTOGRAM

Values used 6.

Numeric Exceptions are all set.

7.

Graph Limits are set to:

NORMAL

Prompt

Setting

Graph min

0

Bin Size

1

Graph max

71

Graph increment

10

Page 22.6



8.

Graph Options are:

9.

Click OK and you should see the histogram graph below:


Notes:

The histogram of the exhaustive (total) iron population shows a very strong normal distribution suitable for Inverse Distance Weighting or Ordinary Kriging interpolation, which will be explored in later lessons.


Page 22.7



11. From the menu, select Mode to change the Graph type to CUM FREQUENCY to display the following cumulative frequency graph:

Notes:

The cumulative frequency curve shows the frequency of grades at varying grade cut-offs. 12.

Change the Graph type again to PROBABLITY PLOT for this graph:

With the probability plot set Normal, the major inflection point around a grade of 10% indicates the change from background to the mineralised grades.

Page 22.8



13.

Change the Values used to NATURAL LOG.


Notes:

When the probability plot is set to natural log and the population plots as a straight line then the distribution is normal, this is the case for the iron population.


Page 22.9



Notes:

Lesson Summary This lesson has introduced the concepts of: •

Descriptive stats

•

Histograms

•

Cumulative frequency graphs

•

Probability plots

•

Natural Log scales

To create these outputs and graphs, we used Stats | Descriptive and Stats | Distribution.

Good Practice Keep saving your outputs into a document or folder so that you can build a resource estimation report as you work through the process. This will be easier while you have the data fresh in your mind. Label everything carefully so that it will be easy to understand what it is in a months time or to a third party reader.


See:

Descriptive stats

Descriptive stats

Distribution stats

Distribution stats

Page 22.10



Lesson 3 – Generate Downhole Coordinates


Notes:

Prior to interpolation, the assay data file should be desurveyed (i.e. 3D coordinates should be calculated for the centroid of each sample interval) using Dhole | Generate | Downhole Coordinates. All geological domains and mineralised envelopes (or seams) can be interpreted interactively on screen using Strings or Outlines in VizEx. These Strings or Outlines are used to generate solid wireframes; Strings are generally used to create surfaces or DTM’s such as weathering surfaces. When all strings or outlines have been generated, they should be loaded into the 3D Viewer and checked for potential errors (missing interpreted sections or drillholes etc). Interpretation can take from 1 to 2 days, to several weeks depending on the complexity of deposit, necessity to interpret geology, domains and multiple elements.

Internal dilution can also be automatically integrated into the interpretation process. MICROMINE can use Dhole | Compositing | Grade to set a trigger grade at which compositing occurs in conjunction with various settings for internal dilution, minimum grade, minimum and maximum waste intervals. Once the settings are enabled then a new grade composite file incorporating the internal dilution can be produced. The new file is then set up in section and the interpretation is performed on this file instead of the assay file.

Exercise 22.2 Generate Downhole Coordinates 1.

From the menu, select Dhole | Generate | Downhole Coordinates and fill in as below:


Page 22.11


2.


The Collar/survey File More button needs to be filled in as follows: Notes:

3. You will need to write in new field names for the new Interval file fields (East, North and RL):

4.

Close the Collar/Survey Setup form and Run.

5.

Examine the modified Interval File by right-clicking on it.

If you are using the Drillhole Database option on the 3D Coordinates form, you can just tick the Create new Coordinate fields tickbox and the new fields in the Interval file will be generated in accordance with you Form Options under Options | Forms.

Page 22.12



Wireframing


Notes:

The geological model is an attempt to model by wireframing the mineralisation in the ground, there are no economic decisions at this stage, it is a mineralisation model.

Wireframing was covered in a previous section of the training. At the completion of the wireframing, assign the wireframe to the assay file to allow the wireframed mineralised data to be segregated from the unmineralised data outside of the wireframe. The finished wireframe is Iron.tdb. A tdb file is a database file and is short for triangle database. Use Iron.tdb for all work.


Page 22.13



Notes:

Lesson Summary This lesson has introduced the concepts of •

Wireframing

Good Practice Checklist •

All sections are interpreted for mineralised zones and/or geology;

•

All Strings/Outlines are snapped to drillhole intervals;

•

Interpretation is visually checked in 3D.


Page 22.14

See:



Lesson 4: Assay Data Flagging


Notes:

FLAGGING/SELECTION When all the wireframes are generated, they should be used to code the assay database in order to select the part of the database to be used for geostatistics and grade interpolation. If the wireframing stage was not required, the interpreted outlines can also be used to flag the Assay database. Prior to flagging, additional fields should be generated in the files where the flags will be recorded. The Assay file should be de-surveyed. By this method the assay file will be coded for those intervals inside the wireframe and those outside of the wireframe, and then a filter can be applied to consider grades only in the wireframe.

Flagging using Solid wireframes If solid wireframes are used to flag assays (or points), this should be done using Modelling | Assign |Wireframes. There is no necessity to use the sub blocking option; because we are flagging assay data and sub blocking factors are not required. If there are several overlapping solid wireframes, the flagging process should be run several times using the solids in the required order or recording flags into different fields.

Flagging using DTMs DTMs can be used to flag any point data (e.g. samples) for their location below, above and outside of the DTM using process Strings | DTM | Assign. It is always a good practice to avoid absent values in numeric fields if possible. Therefore, it is recommended to use all three options in the process and to generate flag values for the point data below, above and outside of the DTM (e.g. generate values 1, 2 and 3). Make sure the DTM covers all samples. A grid file can be generated and converted to a DTM to get sufficient coverage.


Page 22.15



Exercise 22.3 Flagging using solid wireframe

Notes:

1. Select Modelling | Assign |Wireframes and fill in as below:

2. We will not use a Block Model here so select Point Data. 3. The Input file will be IRON ASSAY.DAT. Once it is selected, you will need to right-click on the filename and select Modify. You will need to insert a new Field after SiO2. The Field Name is WFcode, Type C (Character) and Width 5. You do not need to give decimals as this will default to 0. 4. The Wiretrame will be selected from the IRON.TDB for the Type and the orebody Name is also iron. 5. Under Attributes to Assign you can tick Clear target field and Overwrite target field. Click the More button.and fill in (by double-clicking) as below:

6. Create a Report file name, eg WFassign. 7. Click Run. 8. Right-click the Report file to show 3933 records were updated.

9. Check the Input file (IRON ASSAY) by right-clicking.

Page 22.16




Notes:

Lesson Summary This lesson has introduced the concepts of

Good Practice •

The flagged samples are displayed in 3D and c hecked visually

•

It is a good practice not to have absent values (blanks) for flags

•

DTMs cover all samples involved in the resource estimation

Help Topics For information on: M

See:

M


Page 22.17



Lesson 5 – Balancing Cut

Notes:

The balancing cut is used for block models produced using a linear estimation method such as Inverse distance weighting or ordinary kriging. A balancing cut is the use of a more conservative grade instead of a few higher grades in the interpolation process. A balancing cut is determined by using a Cumulative Frequency plot for the mineralized grades only, mineralised grades are those inside the wireframe. At the grade where 97.5 percent of the grades occur, read from the cumulative frequency curve, is the grade to be used for the balancing cut.

Exercise 22.4 Balancing Cut 1.

Select Stats | Distribution with: Prompt

Setting

File

ASSAY IRON

Type

DATA

Filter

Selected WFCODE = 1 numeric

Graph field

T FE

Graph type

HISTOGRAM

Values used

NORMAL

2.

Save your Filter as Mineralisation.

3.

In the plotted example the grade at which 97.5 percent of the population occurs is 65.

Page 22.18



4.


Select Fields | Calculate with the following settings: Prompt

Setting

File

IRON ASSAY

Type

DATA

Filter

Selected

Notes:

Mineralisation 5.

Right-click on IRON ASSAY and select Modify.

6.

Insert a new row before T Fe: Prompt

Setting

Field

TFe Cut 65

Type

N

Width

5

Decimals

2

7.

Close and Save.

8.

Fill the table on the right hand side: Prompt

Setting

Input

T Fe

Function

Cut highs to

Input

65

Result

TFe Cut 65

9.

Click Run.

10.

Right-click IRON ASSAY and you can check that rows 11 and 12 have been reduced down to 65.

The cut grades will be used for the interpolation process and for reports on the resource and the reserve.

The 97.5 percent figure is a western standard discovered from the modelling of many projects, generally the grade at 97.5 percent is sufficient to reduce the influence of the high grades in the estimation avoiding sections of the model biasing the overall result too high. A balancing cut is required because if another hole was drilled down beside the high grade samples it is quite likely that a much lower sample grade will be returned, this is the nugget effect, and as such the model must be more conservative for very high grades.


Page 22.19



Notes:

Lesson Summary This lesson covered the concepts of : • •

Cumulative Frequencies for just the minerealisation. Applying a cut to the data

Good Practice

Help Topics For information on: Fields Calculate

Page 22.20

See:

Core | Files | Fields | Calculate




Lesson 6 - Compositing

Notes:

When deciding upon interval length, if the composite length is not obvious, a histogram should be produced for the Assay file, it should be calculated in File | Fields | Calculate by subtracting FROM from TO. Then run the process Stats | Distribution for the INTERVAL field, study the obtained histogram and make a decision on the composite length. Produce the balancing cut prior to compositing. Samples can be composite using Dhole | Compositing | Downhole process. Set up sample composite length equal to average sample interval length. Avoid mixing of samples from different populations or geological domains when composites are generated. It can be achieved by applying filters and then appending the result files together. Generally wireframes will be assigned to the assay file before compositing, if so then use the wireframe assign field as the Constant field in the compositing routine. CHECKLIST: •

Run stats and make sure there are no strange composite interval lengths

•

Use weighted average method for compositing

•

Run compositing separately for each domain

•

Set up minimum composite length (usually equal to the half of the composite length)

•

Use numeric exceptions if there are any character values in numeric fields

Exercise 22.5 Compositing 1.

We already have interval lengths in the field INT.

Determine the most

frequent sample interval to composite to for the iron data. Plot the interval size on the histogram as follows –


Page 22.21


2.


The image below should be produced. Notes:

Clearly a two metre interval should be used for the composite length, by using two metres most assays will remain unaltered for the estimation but ultimately all will be of equal length. 3. Select Dhole | Compositing | Downhole and fill in as below (Output File is IRON ASSAY COMP1):

There will be errors reported but they are only where there are breaks in mineralisation within a given drillhole. The first item in the report file rep will be in hole CK2 at a depth of 326.3m. This is the start of the second mineralisation in that drillhole. Where the last interval of the mineralisation is less than 2 metres, the same TO value is used and the actual interval is written in.

Page 22.22




4. Intervals (INT) in uncomposited IRON ASSAY.DAT file vary. The image below has less relevant fields hidden.

Notes:

5. Recalculate your intervals by right-clicking on IRON ASSAY COMP1 and selecting Edit, then select Calculations from the toolbar. Prompt

Setting

Input

TO

Function

Minus

Input

From

Result

INT

Clear result field

Selected

Overwrite result field

Selected

6. The composited IRON ASSAY COMP1.DAT file has 2 metres intervals and values for the mineralisation only (again some fields are hidden). This file is then used for all further interpolation.


Page 22.23



Notes:

Lesson Summary This lesson has introduced the concepts of •

Compositing the mineralisation

Good Practice


Page 22.24

See:



Lesson 7 - Geostatistics


Notes:

Theory Classical statistical analysis should be repeated using the same procedures described in the Lesson 2. However, this classical stats analysis will have the following differences: •

The analysis will be run for sample composites

•

Only flagged composites will be used for the analysis, those assays inside the wireframe

•

Stats will be run for each geological / lithological / structural / mineralogical domain separately

Final decisions will be made regarding the method of grade interpolation, variography, mixing of population and top cuts. Mixing of grade populations within each domain should be carefully considered using the same process described in Lesson 2. If it is not possible to separate grade populations using domaining, then the MIK grade interpolation method should be used.

Variography Variography will be run for each element and domain separately. For every domain there are three variograms, each at right angles to each other. For example, if we have three elements and five domains, the task will be to generate 45 directional variograms. If MIK is applied, then the number of final variograms will be 450 (if 10 thresholds are used). The first step would be to generate omni variograms. Omni variograms will indicate the general ranges and variances of grade populations and whether the chances of getting good directional variograms are good or bad. They also assist with the lag sizes. Variograms are to be generated using the process Stats | Semi Variograms. The second step would be to identify the main axis of directional anisotropy if any. If omni variograms are reasonable, a rosette of horizontal directional variograms should be generated. Ideally, a variogram map should also be generated which will clearly show the minimum and maximum ranges and directional anisotropy of grade distribution. A direction of maximum continuity should be identified from the horizontal variogram rosette. That will be the azimuth of the main axis. Then a rosette of vertical variograms should be generated with the azimuth of dipping equal to the azimuth of the longest continuity of horizontal variograms. A variogram with longest ranges will show the angle of dipping of the main axis of directional anisotropy. Downhole variograms are to be used to model nugget effect. Once the azimuth and dipping of the main axis of directional anisotropy is identified, three variograms are to be generated and modelled. The first variogram will be in the direction of the main axis, the second one –


Page 22.25



perpendicular to the first variogram, and the third one – perpendicular to the first two variograms.

Notes:

If geology and mineralisation are well studied and interpreted, sometimes the main directions of directional anisotropy are obvious and the above steps could be simplified or skipped. The directional variograms are to be displayed and modelled in Stats | Semi Variograms. It would be a good practice to generate Direct, Log and Relative semi variograms to obtain the main features. When experimental variograms are displayed on screen, they can be modelled using the Model menu. Select the variogram type (e.g. Model | New | Spherical). Then you will be prompted for the number of structures. Select the number of structures (for example 2). Then you will be able to model the nugget effect and sills of every structure using the mouse. When you specify the model parameters with the mouse, MICROMINE will display the modelled variogram parameters. Please note that Sill parameters there are actually C values. Sill values will have to be calculated (Sill = C partial sill + Nugget). All modelled variogram parameters should be saved to a Form. Variography can take from several days to several weeks depending on the number of elements, domains and selected interpolation method. If MIK is selected, the exercise could be very time consuming due to the large number of variograms to be modelled. CHECKLIST: •

Note the “Sill” in MICROMINE is actually, the partial sill.

•

Use downhole variograms (or vertical) to estimate nugget effect

•

Make sure all 3 variograms have the same nugget, C value and total sill

•

Handle zonal anisotropy (if any) by adding another structure to variograms

Sill co + c

Nugget co gamma

Range m

Variogram properties

Page 22.26




Anisotropy

Notes:

There are two types of anisotropy: 1. Zonal. The sills and the ranges are different in the three directions. If this occurs two structures must be used with a second long range to get the sills back to the same level at a very large range, such as 10,000 metres 2. Geometric. The sills are the same but the ranges are different for the three directions

Semi variogram formula The semi variogram and variogram is basically the same thing, technically they are different as the semi variogram is divided by two. Algorithm as follows, semi variogram –

γ (h ) =

1 2 N (h )

N ( h )

2

∑ ( xi − y i ) i =1

Once the semi variogram has been displayed then a model must be fitted to the gamma values.

Semi variogram model formula The rule of thumb is that a spherical model is fitted to most gold data; in some circumstances an exponential model may be used. 3D modelling requires three variograms orthogonal to each other, the nuggets should be the same and the partial sills must be the same, the ranges can differ for each direction The nugget is best determined from the downhole data as the data is the most closely spaced, the lag can be determined from the omnidirectional variogram which is an average of the lag spacing. Fitting the variogram model is done interactively, the Noel Cressie statistic can show the quality of the fit using a least squares regression, however the best guide is a visual fit of the line to the gamma values. The smaller the Noel Cressie statistic then the better the variogram model fit, theoretically. However in practice the first and second gamma values greatly influence this result. Use the visual fit in conjunction with the test button. Ordinary kriging requires the data to resemble a normal population. If there are mixed populations (this is apparent on the histogram and probability plot) then a method such as multiple indicator kriging must be used. The variograms are saved together in a form that retains the model parameters and the attitude; azimuth and plunge of the variogram. The weighting is then performed automatically within the ordinary kriging routine.


Page 22.27



Search ellipsoid parameters

Notes:

The parameters entered here define a search ellipse used to select samples for modelling. That is, the samples that will be used to calculate the estimated value. Radius - Enter the primary radius of the search ellipse. This value is a length or distance that becomes the base value by which the three factors below are multiplied to determine the dimensions of the search ellipse. Azim (deg) Enter the Azimuth (bearing in degrees) of the long axis of the search ellipse. This has a range of values 0 - 360 measured clockwise from north = zero. It corresponds to geological strike, or the trend of the long axis of a plunging body. Azim factor Enter the factor for the length of the long axis of the ellipse. This will be multiplied by the Radius to determine the actual length of the Azimuth axis. The Azimuth factor is generally the longest dimension of the search ellipse. Commonly the Radius is set equal to the along-strike search, typically 1.25 to 1.5 times the average section spacing, and the azimuth factor set to 1. The other two factors would then be defined as decimal values between zero and one. Plunge (deg) The plunge is the downward inclination of the orebody along the strike. It must be positive; plunge values are always in the range 0 - 90. For example, a tabular structure with Azimuth 30 degrees and a Dip of 60 degrees to the South-East will have Azim = 30 and Dip = -60. If the plane Page 22.28



contains a tubular or elliptical structure that plunges downward at 30 degrees to the north-east then the value required is Plunge = 30.


Notes:

If the same tabular structure has a tubular or elliptical structure that plunges downward at 30 degrees to the south-west then the value required is still Plunge = 30 but the Azim must be 210 and the Dip required i s 60. Thick factor This describes the thickness component of the search ellipsoid. Enter a factor for the length of the search axis perpendicular to the plane of the Azimuth and Dip values. This value is multiplied by the Radius value to determine the actual length of the Thickness search axis. The thickness factor usually describes the short axis of the search ellipsoid. Dip (+/- deg) Dip is an angle, with range -90 to 90 measured from the horizontal, perpendicular to the azimuth axis. It corresponds to geological dip. The convention used throughout MICROMINE is that clockwise rotation, looking in the Azimuth direction, has negative dip values and counterclockwise rotation has positive dips. Thus a bed striking at zero degrees and dipping 60 degrees east will have a Dip angle of -60. Dip factor Enter a factor for the length of the dip axis of the ellipse. This value is multiplied by the Radius value to determine the actual length of the Dip axis search. This is the down dip search dimension of the search ellipsoid.


Page 22.29



Practice

Notes:

Iron deposit (3d): The iron deposit will be modelled using ordinary kriging and median indicator kriging. The iron deposit is a good example because it has an excellent linear population and produced strong variogram models with geometric anisotropy. Files: Collar: IRON COLLAR.DAT Survey: IRON SURVEY.DAT Assay: IRON ASSAY.DAT Assay: IRON ASSAY COMP.DAT COMP.DAT Wireframe: IRON.TDB

Review After the 3d coordinates were created for the assay file, the wireframe was assigned to the assay file. This ensures we know which grades are the relevant mineralised grades and which grades fall outside the wireframe and are redundant. This file has already been produced and is IRON ASSAY.DAT; ASSAY.DAT; the code field should be WFCODE with WFCODE with the code iron. iron. A 2 metre composite file was also created called IRON ASSAY COMP.DAT. COMP.DAT.

Nugget The IRON ASSAY.DAT file ASSAY.DAT file is now used to calculate the variograms and so, to create the variogram models. Do not use the composite file at this stage because it may inadvertently display zonal anisotropy because compositing smooths the data in the file and will change the variance to a greater degree in one direction than another. The composite file is only used for the interpolation.

Exercise 22.6 Nugget 1.

Open the semi variogram form by selecting Stats | Semi variograms. variograms .

Page 22.30



2.


The Semi Variogram Type will Type will be Downhole in this instance. Notes:

3. Select the IRON ASSAY.DAT ASSAY.DAT as the Raw Data file. We initially have to create a semi variogram file from a raw data file. file. Later these semi variogram files can be re-used by clicking the second option. 4. Apply a filter to the Raw Data file in order to only use data inside the wireframe. Make sure you save this filter using Forms | Save as. as.

5. Select Show Variance Variance under Data Values Values and complete the form as shown below:

6. Under Semi Variograms, select Show semi variograms and Write semi variograms to file. file. The name of the File will be Vario be Vario DH of DH of Type DATA. DATA. 7. You will of course have set Numeric Exceptions Exceptions and then saved the form using the Forms button (third from left on the Toolbar) as Downhole Nugget. Nugget. 8.

This will display a Downhole variogram to variogram to determine the nugget size.

9. This gives a Nugget Coefficient established at about 50.


Page 22.31



Using Semi Variograms

Notes:

When you create a semi variogram, you will notice that the menu at the top of the screen changes. The two menu items of particular interest are Display and Display and Variogram. Variogram.

The menu items with icons are available from the Semi Variogram Toolbar. The ones that are probably used the most under Display menu Display menu are: Form – Form – you should always save your forms so that you can easily reproduce a result. Dump Dump – creates a screen shot of the active window that you can paste into a report that you might be generating as you work. Zoom + Area – Area – allows you to zoom into the area of interest. The relevant scale scale will also be shown automatically. Display Mode Mode – takes you to the variogram parameters page where you can make changes or check entry details. Show Together – Together – if you have more than one set of parameters, all of them can be shown on the screen at the same time. This is an alternative to leafing throught them by using the Page Up and Page Down keys on your keyboard. When using Show Together, you might want to use Display Mode first and change the Display Mode of some of the less likely candidates to None. This will simplify the display and let you focus on the more likely candidates. All of the items under the Variogram the Variogram menu menu are of use: Previous – Previous – lets you leaf back through individual variograms. This can be done more easily by using the Page Up key Up key on the keyboard. Next Next – lets you leaf forward through individual variograms. variograms. This can be done more easily by using the Page Down key Down key on the keyboard. Model Model – lets you model a curve through the points to represent a best fit representation of an ideal curve.

Page 22.32



Optimum Lag


Notes:

Determine the optimum lag size by using an omni directional variogram with various lag sizes. The omni directional variogram displays the average of lags. We will use the composited data found in the IRON ASSAY COMP.DAT file. For this iron example, lags of 50 to 60 metres produce well behaved variogram results. This distance can now be used to find the direction of maximum continuity.

Exercise 22.7 Omni Variogram 1. Open the semi variogram form by selecting Stats | Semi variograms.

2. The Semi Variogram Type will be Omnidirectional. 3. Select the IRON ASSAY COMP.DAT as the Raw Data File. Keep the filter WFcode = iron and select TFECUT65 as the Semi variogram field.


Page 22.33



4. Select Show Variance under Data Values, click the Omnidirectional Semi Variograms button and complete the form as shown below:

Notes:

5. Save the above form as Optimal Lag – Omni Comp. 6. Under Semi Variograms, select Show semi variograms and Write semi variograms to file. The name of the File will be Var Omni Comp of Type DATA. 7. Save the main form also as Optimal Lag – Omni Comp. 8. These will display Semi Variograms. To leaf through the displays, use the Page Up and Page Down buttons on your keyboard. It is advisable to zoom in to the left-hand side using the magnifying glass with the square inside it from the toolbar 9. The screenshot below shows the orange 60_50 values. reasonable fit.

Page 22.34

This gives a




10. We should now repeat the search with a tighter range of say 35 to 75, or even tighter.

Notes:

We will use a value of 50° for our lag or interval for now. The next step is to discover the direction of maximum continuity. This will have the longest total range. Set the variogram fan for 30 degree increments for 180 degrees, it is not necessary to do 360 degrees as one half is the mirror of the other. Set the tolerance to 15 degrees so they do not overlap and apply a conical search. The geology is often a very good guide to the direction of maximum continuity. The Mode button applies the value in the first row to all other valid rows. The display modes you can choose from are: None: The data for that azimuth will not be displayed. Useful when you want to switch a direction off temporarily to simplify the display. Line: Data for the azimuth will be plotted as a simple line graph. You can enter a symbol number when LINE is selected. The corresponding symbol will appear at each interval distance. Its size will vary proportionally to the number of pairs in that interval. Graph: The data will be displayed as a graph with two lines. The area between the lines can be hatched. To generate the lines, alternate values from interval one to the maximum calculated distance interval are connected. The intervening values are then connected back to the first interval value. This displays the difference between values in adjacent intervals (but loses information on the number of pairs in each interval). Symbol: The interval semi-variogram value for the azimuth will appear as a symbol. The symbol size is relative to the number of pairs in the interval. Pairs: The Pairs display option displays a fixed size symbol with the number of pairs written beside the symbol.

Exercise 22.8 Horizontal Fan Variogram 1. We will use the Semi Variograms form again. Type will be Directional.


The Semi Variogram

Page 22.35



2. Change the Semi Variograms File to Var Horiz Comp. The rest of the form stays unchanged.

Notes:

3. You will notice that the button under Data Values has changed to Semi Variograms Directions. Click this and complete the form as shown below:

4. Save the above form as Horizontal Fan. 5. Notice that we are using the Display Mode of Lines. Lines are easier to read than Pairs but there is much more information in Pairs. 6. Save the main form also as Horizontal Fan. 7. Again leaf through the displays using the Page Up and Page Down buttons on your keyboard. It is advisable to zoom in to the left-hand side using the magnifying glass with the square inside it from the toolbar. 8. Your graphs should show that the best fit will be between 35 and 55 degrees for the Azimuth. The screenshot below comes from a second run from 35° to 55°. Blue 39° was slightly better than Pink 41°. We will use 40°.

Page 22.36



Exercise 22.9 Vertical Fan Variogram


Notes:

1. Open the semi variogram form by selecting Stats | Semi variograms. 2. The Semi Variogram Type will again be Directional. 3. Change the Semi Variograms File to Var Vert Comp. The rest of the form stays unchanged. 4. Click the Semi Variograms Directions button and complete the form as shown below:

5. Save the above form as Vertical Fan. 6. Save the main form also as Vertical Fan. 7. Click Run to display the variograms as shown below.


Page 22.37



8. Again leaf through the displays using the Page Up and Page Down buttons on your keyboard. It is still advisable to zoom in to the left-hand side using the magnifying glass with the square inside it from the toolbar.

Notes:

9. Your graphs should show that the best fit will be -6°.

Omni directional variogram to determine optimum lag:

The setup for defining the Omni directional variogram;

Page 22.38




Notes:

The display of the Omni directional variograms;

Directional variogram, main: Step 4: Finding the principle direction;

Defining the settings to narrow down the principle direction variogram;


Page 22.39



Notes:

Display of the variograms for the principle direction;

Direction of maximum continuity, primary variogram; Step 5: Directions of maximum continuity; once the approximate direction of maximum continuity is known from step 4, then the lag can be experimented with and then the exact direction of maximum continuity in terms of azimuth and plunge can be investigated and modelled.

Page 22.40




Notes:

Clearly the lag of 50 or 60 metres and a zero degree plunge produce the best behaved semi variograms. The variograms have a good regular pattern, sill out close to the variance and do not have a saw toothed appearance.

Model the variograms: Step 6: The azimuth is 142 degrees, the lag 50 and the plunge zero. Then fit a spherical model to the gamma values on the variogram. Retain the same nugget as the downhole, vertical variogram with the same partial sills, one and two retained for all three orthogonal variograms.

Principal direction: 142 degrees azimuth;


Page 22.41



Notes:

Principal direction: 142 degrees azimuth, fitted model;

Intermediate: Directional variogram 232 degrees azimuth;

Page 22.42




Notes:

Intermediate: Directional variogram 232 degrees azimuth, fitted model;

3rd direction: 180 degrees azimuth, 90 degrees dip;


Page 22.43



Notes:

3rd direction: 180 degrees azimuth, 90 degrees dip, fitted model; These are the required variogram models for Ordinary kriging or to establish the search ellipse dimensions for Inverse distance weighting.

Indicator variograms: Only use indicator variograms if you intend to interpolate using a non linear model method such as median or multiple indicator kriging. For median indicator kriging determine the median of the grades inside the wireframe by plotting the cumulative frequency curve of the data. The also find the grade ranges at 10 percent, 20 percent etc up to 90 percent. These grade cut-offs will be used for the bins in the median indicator kriging routine and the 50th percentile or median will also be used at the cut-off grade in the indicator variogram modelling procedure.

Page 22.44




Notes:

Grade thresholds for the bins: % of data

Fe %

10

26.3

20

29

30

33.8

40

37.6

50

42

60

46.8

70

51.8

80

56.9

90

61.1


Page 22.45



Notes:

The median occurs at a cut-off grade of 42% Fe; use this cut-off grade for the indicator variograms. Use the same procedure as for semi variograms to find the direction of maximum continuity, the intermediate variogram and the third direction. Note the nugget and partial sill must still be the same for all three indicator variograms.

Page 22.46




Notes:

Once the three indicator variograms are modelled then go to Modelling | 3d block estimate | multiple indicator kriging and save the indicator variogram form, this should be used in the cut-off box.


Page 22.47



Notes:

Relative variograms: If the model area exhibits proportional effect, where the mean and variance change in proportion to each other across the model area then a relative variogram must be used to ensure the sills of the lags are all at the same level so the variograms appear sensible and will allow the fitting of a variogram model. A test for proportional effect can be conducted by using modelling | 3d block estimate | statistical.

Display the result in Stats | scattergrams | simple linear, plotting the mean on one axis versus the standard deviation on the other axis. If the result plots as close to a straight line then a proportional effect is present, if the cloud is wide as is the case with the iron example then no proportional effect exists and relative variograms are not required.

Page 22.48




Notes:

To produce an ordinary kriged model using relative variograms the variograms will be modelled using the relative gamma values. When the normal direct variograms are modelled the relative gamma value is also stored inside the file. Use the relative gamma values for the 142, 232 and vertical normal direct variogram output files by selecting display relative variogram from file. Then fit a model and save the form for all three relative variograms.


Page 22.49



Notes:

Page 22.50




Notes:

Cross Validation: Cross validation is conducted by removing a raw data value and using the surrounding raw data values to estimate the removed value. The value is then compared to the estimate and is repeated throughout the dataset. The total average estimates are compared to the actual estimates; if the variogram model is robust the figures should be very close.


Page 22.51



Notes:

The average error statistic should be close to zero and the standard deviation of the error statistic close to one. The results of the iron estimation by cross validation were 8.0575 for the standard error and -0.005125 for the error statistic. The standard error is a little high and could be improved but the error statistic is close to zero and is a good result.

Page 22.52



Actual versus estimated values can be plotted on a scattergram to see how well the kriging process reproduces the sample data. Actual value versus the error statistic demonstrates the conditional bias


Notes:

The Means are very close so the global cross validation is good, the precision is 19%, and the result was influenced by some low grades that did not produce a low estimate because of the amount of data found by the search ellipse. The cross validation is reasonable for the direct variograms to be used for ordinary kriging.


Page 22.53



Notes:


Good Practice Keep


See:

M

Page 22.54



Lesson 8 – Empty Cell Model


Notes:

Grades can also be interpolated into an empty cell block model generated (or imported) earlier.

Block Model Cell Size Model cell size should be selected depending on the following parameters: Drilling density Variability of grade Smallest mining unit SMU Final model size The cell size should be sized to be small enough to produce a grade map for grade distribution and big enough that it reflects available data. When block model cell size is selected and the extent of model is calculated, an empty cell block model should be generated using Modelling| 3D Block | blank block model. The block sizes should be saved in a form and restrict to wireframe option should be used to save only the blocks in the mineralisation.

Flagging If wireframe solids and DTMs are modelled, then they should be used to flag the block model. Basically, flagging should be carried out the same way as described in the section 6. The only difference would be if sub blocking is required. When the block model is flagged for all possible domains / zones / ore bodies etc, all other cells (unflagged) should be deleted from the model to reduce the size of the file and number of records (File | Filter | Subset). That will also help to control the interpolation process. The process should not take more than several hours.

Checklist 

The model should not have too many cells. An average block model has several hundreds of thousands cells.



Generate sub blocks to represent volume more accurately.



Generate as many new fields for flags, as there are wireframe/string/outline types that are used for flagging. It is easy to combine them later, if necessary.


Page 22.55



How to decide the block size

Notes:

The block size of 10 metres east, 20 metres north and 5 metres in rl is displayed at the centre of the search ellipse. This block size for iron estimation is appropriate given the sample spacing of 100 metres. The blocks must reflect grade distribution, showing a local map of grade occurrence, so the block cannot be too big as the grade change will not be shown and cannot be too small because the file will be unnecessarily large and the grade estimate will become less reliable. The block size for an iron deposit will be bigger than the block size for more densely spaced shear hosted gold deposits or VMS hosted base metals because the samples are more closely spaced and the geology is far more variable.

Page 22.56



If the sample spacing for a gold deposit was 25m between sections, 10m between the drillholes along the section and 1m sample intervals then an appropriate block size would be 5m by 2m by 2.5m in rl.


Notes:

Subcelling Subcelling is the creation of smaller blocks on the edge of the wireframe when the parent cell is not fully inside the wireframe. The numbers entered into the sub block boxes are how much the parent cell is divided by to define the subcell in metres. If the parent cell is 10 metres in east and the sub blocks east is entered as 5 then the subcells will be 2 metres in the easterly direction. A sub block factor is different, the cells are not subblocked, rather a number between zero and one is defined for the percentage of the block inside the wireframe.


Page 22.57



Exercise 22.12 Creating a Blank Block Model

Notes:

1. We know we are going to be asked for the extents of the Iron wireframe, so firstly open that in Vizex in Plan view. Jot down values to encompass the wireframe in Eastings and Northings. You should have jotted down something like: 19000mE-19900mE and 35000mN-36500mN. 2. Switch to Looking North view and note the RLs as well. You should have something like -650mRL-0mRL. 3. Select Modelling | 3D Block Estimate | Blank Block Model and fill in as below. The Output entries are all typed in as we are creating a new file.

4. Under Restrictions | Wireframes click on the More button. Fill in the details as shown below. Again, the Block Factor Field is going to be created and the Sub-block values represent the number of times you want the block sub-divided not the size to which you want it sub-divided.

5.

Close the Restrict With Wireframes form.

Page 22.58




Notes: 6. Click on the Block Definitions button. This is where we use the values that we jotted down earlier. 7. Enter the values but notice that you are asked for the Block Centre. To accommodate this and keep our blocks on round number co-ordinates, add half the relevant block size (Spacing) to each Origin Block Centre and subtract half the relevant block size from each End Block Centre.

8.

Select Forms | Save As and save the form for later use as Iron OBM.

9.

Close the Block Defintions form.

10. Select Forms | Save As and save the form as Iron OBM. 11. Click the Run button. This may take a couple of minutes and progress is shown in the bottom left of the screen. 12. Right-click on Iron OBM in the Output File box and firstly select Min/Max. You will notice various details including the creation of about 110,000 records and that the data does not start until 19025mE. 13. Right-click on Iron OBM a second time and select view. Notice that the BF field stores a value between 0 and 1 representing the number of virtual subblocks inside the wireframe as a fraction.


Page 22.59



Notes:


Good Practice Keep


See:

M

Page 22.60



Lesson 9 – Modelling Principles


Notes:

Declustering If samples are clustered then the samples must be declustered to allow a fair estimation of the unknown value in the search ellipse. Declustering is required to minimise interpolation bias from high-density assay areas, which often occur in high-grade zones. If data is not declustered the clustered data has an undue overwhelming influence in the grade interpolation on the surrounding area. If a large number of raw data values are picked up by the search ellipse from one area then these points will preferentially ensure that this area weights the interpolation of the point of estimation more so than the scattered data points. Sectors or cells can be employed to decluster the data, a maximum and minimum number of points can be stipulated for each sector. MICROMINE currently employs a sector method in some model modules to subdivide the search ellipse and allow the thinning of the number of points to be interpolated by specifying the maximum number of points allowed within the sector.


Page 22.61



Notes:

Specific gravity and block size Block size is determined by drill spacing and the smallest mining unit required for the resource, the SMU. The SMU is often suggested by the mining engineer, even at the stage of resource estimation. Specific gravity can be defaulted or interpolated for the blocks.

Change of Support Discretisation is where ordinary kriging in this case estimates point grades inside the block which are then averaged to produce the block grade. Block kriging was designed to combat the change of support, where the grade of a truck load of ore is more even and reliable compared to the grade of a far smaller often more variable sample. Block kriging is now not considered the best method of dealing with the change of support but there are few practical alternatives.

Proportional effect When the local variability of data changes across the model area this is known as heteroscedasticity, the proportional effect is a form of this. For the positively skewed distributions the local variance increases with the local mean. The proportional effect is detected from a scatterplot of the local mean versus the variance-calculated from moving window statistics. The proportional effect can be calculated in Micromine by using the Modelling | 3d | statistical, defining a block size and writing a file containing the local mean and variance. Proportional effect will render the sample semi variogram uninterpretable. Clustering combined with the proportional effect results in the high clustered values contributing to the lower lags. The corresponding lag mean is large and because of the proportional effect the lag variance is also large. As distance (h) increases the data that contributes to the lag becomes more representative, the

Page 22.62



lag mean and variance decrease. The trend results in the lag variance results in overestimation of the semi variogram value at short range and also the relative nugget effect. An inaccurate variogram model generates inaccurate weighting. A relative variogram is required as opposed to a traditional variogram. The relative variogram is standardised by the gamma value divided by the lag variance.


Notes:

Interpolate parent blocks only If parent cells only are interpolated then it means that if the parent block was divided into subcells because it was on the wireframe border, then the subcells will be assigned the grade estimate for the parent cell and they will not be independently estimated.

Multiple runs Often when the blank model is created and interpolated into the first search ellipse size is not sufficiently large enough to populate all blocks. When the name of the block model file is the same for ‘Define blocks from file’ and the output block model file then the grades will be written into the blank model. Note that the input field name, width, type and decimals must be the same in the input file as the define blocks from file data file. If they are different then a result will not be written. After the first run with the first search ellipse then increase the search ellipse size for run 2 and possibly run 3 until all blocks been populated with an estimated value.


Page 22.63



Notes:


Good Practice Keep


See:

M

Page 22.64



Lesson 10 – Grade Interpolation


Notes:

INTERPOLATION The block model is generated simultaneously with interpolation of grades. This means the following routine should be used: 1. Specify all input parameters, files, search ellipse and variogram parameters in the processes Modelling | 3D Block Estimate | IDW or Modelling | Kriging. Run the interpolation. 2. The generated block model should now be flagged for domains etc. using wireframes or outlines (as specified in chapter 8). All unflagged cells (above the surface, outside of wireframes etc.) should be filtered out (File | Filter | Subset). 3。

All the assigned cells should be checked for whether all cells are populated with grades or not (Stats or Min/Max). There should be no missing grades in the block model. 4。

5. If not all cells are all populated, then repeat the steps from 1 to 4 with altered (increased) search parameters until all cells are informed with grades. Generate a Run Number field in the output model file. You will have several saved search ellipses that increase in size, each one represents a grade interpolation run. 6. Add all the generated models together in such a way, that earlier models would update the later ones (File | Merge | MM). 7.

Repeat all steps from 1 to 6 for all domains and for all elements.

The interpolation process can take from several hours to several days. CHECKLIST: • Apply top cuts if necessary • Interpolate grades using several methods for validation • All cells in the Resource model should be informed with grades. • If Kriging was used, run cross validation to check if the variogram is

appropriate. Run the variogram model that produces a good estimate with the lowest kriging error. • If MIK is used, make sure the search parameters are the same for all bins

for a particular interpolation run • The number of interpolation runs should be equal minimum to: No of

elements x No of domains x No of interpolation volumes x No of interpolation methods. It is a good idea to save all these runs in a macro.


Page 22.65



Search ellipse

Notes:

When defining the search ellipse in MICROMINE – • Define the number of sectors, if there is a large amount of data, then eight or sixteen is an appropriate number of sectors to employ. • Define the maximum and minimum number of points to be used in the search ellipse, a maximum of 6 means a total number of six per sector will be used, the six nearest samples to the point of estimation will be used for the estimate; the other points in the sector will not be used. With the minimum set to 2, if only one sample is found then the point of estimation will not be estimated. This is data declustering performed on the fly by MICROMINE in the estimation process. • Define the attitude of the search ellipse, the azimuth, dip and plunge. This should be defined to include the relevant samples for estimation and to exclude the redundant points. Several different search ellipses of increasing sizes may be required to interpolate all of the blocks in the model. Often three runs may be needed to populate the model, the search ellipse will be increased in radius with each run, the blocks that have already been estimated will not be overwritten, only empty blocks will be populated by each new run.

Page 22.66



How to decide search ellipse size


Notes:

The search ellipse radius is determined from the variogram parameters or from the sample spacing. For the iron example the radius of 250 metres above contains around seven drillholes in the ellipse each with two metre interval samples, so there are abundant samples for the first pass estimation because the drillholes are spaced around 100 metres apart. Save the search ellipse forms. Load the search ellipse in 3D Viewer, bring up the grid and look from different angles with the ellipse transparency on to observe that the attitude of the search ellipse is the required design. This is an important validation step.


Page 22.67



Notes:

Inverse Distance Weighting Inverse distance weighting uses the inverse of the distance to the value of a selected power as the mechanism whereby the samples are preferentially weighted. For the simple example the unknown grade at the point of estimation is ?, results are tabulated Grade

Distance

1/d2 (m)

As fraction

% weight

Weighted grade

0.31

110

1/12100

0.000083

0.162

0.05

0.25

80

1/6400

0.000156

0.304

0.076

0.18

120

1/14400

0.000069

0.135

0.024

0.21

70

1/4900

0.000204

0.399

0.084

1

0.234 kg/m3

?

Using a power of 2 for the inverse distance weighted calculation the point of estimation equates to 0.234 kg/m3.

Algorithm =

n

vi

i =1

i

∑ d w n

∑ d i

w

i =1

Where the algorithm result = point of estimation; d = distance; v = data value; n = number of points to calculate cell node value; w = size of the power

Page 22.68




For the different powers, 1, 2 and 5, as the power is increased then the weighting on the nearest sample to the point of estimation increases, the higher the power then the greater this weighting to the nearest samples.

Notes:

With an inverse power of 1 the grade weights are more evenly spread amongst the samples, based evenly on the distance from the point of estimation. As the power increases to a power of 5 then the samples closest to the point of estimation at 6 and 7 metres respectively receive nearly all the weighting. The grade estimate increases to 4.28 since the 2 nearest grades are also the highest grades; most of the other samples have very little influence on the grade estimate. POWER = 1 GRADE

DISTANCE

DISTANCE to POWER

3.1 4.5 2.1 1.2 4.2

10 6 12 18 7

10 6 12 18 7

1/ DIST

GRADE WEIGHTING

GRADE * WEIGHT

0.100000 0.166667 0.083333 0.055556 0.142857

0.1823 0.3039 0.1520 0.1013 0.2605

0.5653 1.3676 0.3191 0.1216 1.0941

0.548413

1.0000

3.4676

GRADE WEIGHTING

GRADE * WEIGHT

0.010000 0.027778 0.006944 0.003086 0.020408

0.1466 0.4072 0.1018 0.0452 0.2992

0.4544 1.8324 0.2138 0.0543 1.2565

0.068217

1.0000

3.8114

GRADE WEIGHTING

GRADE * WEIGHT

0.000010 0.000129 0.000004 0.000001 0.000060

0.04935 0.63460 0.01983 0.00261 0.29361

0.15297 2.85571 0.04165 0.00313 1.23315

0.000203

1.00000

4.28662

POWER = 2 GRADE

DISTANCE

DISTANCE to POWER

3.1 4.5 2.1 1.2 4.2

10 6 12 18 7

100 36 144 324 49

1/ DIST POW

POWER = 5 GRADE

DISTANCE

DISTANCE to POWER

3.1 4.5 2.1 1.2 4.2

10 6 12 18 7

100000 7776 248832 1889568 16807

1/ DIST POW

Recommended Values When using MICROMINE for inverse distance weighting for iron, interpolate both the cut and uncut fields in the composite file. For gold set a power of 2 or 3; 3 is most commonly used for gold. For iron a power of 2 is appropriate. Interpolate only the grades in the wireframe and define the blocks from the file to update the block model. The composite file must be used; if grade intervals are not of equal length then the model will be biased and will be a less accurate estimate.


Page 22.69



Exercise 22.13 Inverse Distance Weighting

Notes:

Ordinary kriging Kriging is an interpolation method, which uses the measured anisotropy of the deposit to preferentially weight the samples to varying extents in the three defining directions within the deposit. Anisotropy may or may not be present dependent upon the nature of the deposit. Anisotropy is the uneven distribution of grade within the deposit. If the deposit is isotropic and the variogram range does not change with direction then an omnidirectional variogram may be fitted. The omnidirectional variogram will have a tolerance of 90 to look in all directions and will weight the samples as an average of all variogram models. The weighting mechanism is determined by the variograms that are modelled. The variogram model is then applied to the kriging algorithm to estimate block values. The variogram models are a geostatistical measure of variation in grade with distance along a spatially defined direction. Three variogram models will be Page 22.70




produced for 3d modelling in x,y,z and two in x and y for 2d modelling. Kriging is dependent upon being able to model variogram models, variography will confirm or disprove the geologist’s intuition and assumptions relating to the deposit.

Notes:

Ordinary kriging formula:

n(u )

Z (u ) = ∑ λ α (u ) Z (uα ) with OK

* OK

α =1

n(u )

∑ λ α

OK

(u ) =1

α =1

Where gamma is the mean squared differences between pairs such that the mean squared error in the matrix column sums to zero. The matrix is the K matrix. A second matrix is set up which calculates the mean squared differences between the data points and the cell nodes. This is accomplished using gamma values from the variogram models. This matrix is the M2 matrix. The K matrix is divided by the M2 matrix to obtain the kriging weight, which is λ. Any left overs are accounted for by the Lagrange coefficient,µ. The Lagrange parameter is a condition in the equation that requires that the total kriging weights sum to one. The ordinary kriging dialog boxes are the same as the Inverse distance weighting with the exception that the routine uses not only distance but also the variogram models in the three orthogonal directions to weight the estimation. The three variogram models are setup as saved forms in the semi variogram parameters box. The longest range is the main direction, followed by the intermediate and the third is the shortest direction. Some rules apply to the form for saving the variograms – • The nugget must be the same for all three variograms • The partial sills must be the same for all variograms • The three variograms must be orthogonal to each other, note that if the main direction is 180 degrees with zero dip, then the second direction must be 270, not 90 degrees, the angle must be bigger. • If these parameters are wrong then the kriging variogram form will not be saved, this is a validation step in MICROMINE.

Block kriging can be used which will then enable the discretisation, this means that several points can be estimated into the block and are then averaged for the block estimate. Iron example;


Page 22.71



Notes:

Run the routine to generate the block model output file.

Ordinary kriging, relative variograms Once the relative variograms are saved together in a form then the ordinary kriging routine can be run. There is no need for a back transformation for the grade estimate, this is performed automatically; the kriging variance and kriging standard error however are rendered useless and should not be used.

Page 22.72




Notes:

Run the routine and report the tonnes and grade.

Multiple Indicator Kriging MIK takes account of different anisotropy at different grade levels. MIK is particularly useful where there are mixed populations present as it is a non linear method. It is also better at handling the higher grade values and avoids the need to cut top grades. MIK uses different grade levels by asking for a cut-off value above which all raw data values are transformed by an indicator to a value of one. All values below this cut-off are assigned an indicator of zero. This is an indicator transformation. The various grade bins are selected by using a cumulative distribution frequency curve to group grades into percentile bins. The indicators are then modelled with semi variograms and these variograms are applied to each cut-off. By applying a more suitable variogram model to the various grade levels the anisotropy is correctly honoured. This type of modelling method is attempting to model higher grades that are mixed in with the main population and cannot be domained out to be modelled independently. Micromine has an MIK option. The indicator variography and the indicator kriging both exist with multiple cut-offs with associated multiple variogram models. Up to ten cut-offs are available, so that deciles can be defined.


Page 22.73



Example: Notes: Apply a cut-off; model the variography, run the model with the indicator cut-off and associated variogram model. Do this for each cut-off; you will then have five kriged models. Indicator bins, Au, more importantly the range will vary for each indicator. 0.2 g/t everything above 0.2 has a primary variogram atti tude of 260 degrees 0.5 g/t everything above 0.5 has a primary variogram atti tude of 260 degrees 0.9 g/t everything above 0.9 has a primary variogram atti tude of 265 degrees 1.5 g/t everything above 1.5 has a primary variogram atti tude of 275 degrees 6.5 g/t everything above 6.5 has a primary variogram atti tude of 290 degrees Do for the three directions and save the formset, repeat for each grade cut-off; alternatively do once at the 50th percentile, this is called median indicator kriging and is much faster as only three variograms and one variogram formset is needed. Also the variograms at the 50th percentile are always the easiest to do. Create a blank block. Setup the cut-offs and enter the corresponding variogram formset. Then run to generate an MIK model that applies different weights from the different variograms at various grade levels. An e type estimate for each block will be produced. MIK is more likely to be employed in resource estimation than grade control; however some of the large nickel operations do use this method in addition to others. Production pressures restrict its use. How the estimate is produced: Indicator kriging is a non-parametric estimation method, in that it does not assume that the population conforms to some type of distribution. Indicators below certain cut-offs flag the grades, the indicators are modelled with variograms which produce probability maps of the indicators. The mean bin grade is applied as the E type estimator to the probability maps to produce a grade estimate. Once the kriged probabilities are adjusted in the Micromine file then an estimate can be obtained. If the grade thresholds are as follows – 1, 2, 3, 5, and 10 and the probabilities are 1, 0.82, 0.61, 0.46 and 0.12 respectively then this means that = 0% below 1 g/t 1 – 0.82 = 18% of the material is between 1 and 2g/t @ 1.45 0.82 – 0.61 = 21% of the material is between 2 and 32g/t @ 2.47 0.61 – 0.46 = 15% of the material is between 3 and 5g/t @ 3.81 0.46 – 0.12 = 34% of the material is between 5 and 10g/t @ 7.1 = 12% of the material is above 10g/t @17.31

You need to obtain the mean of each bin; ex the mean of the raw data below 1 g/t is 0.5

Page 22.74



The mean of the data between 1 and 2 g/t is 1.45 g/t. MICROMINE calculates the mean value automatically and then produces the e type estimate.


Notes:

For the block estimate, the weighted equation bec omes (0.18 x 1.45) + (0.18 x 1.45) + (0.18 x 1.45) + (0.18 x 1.45) + (0.18 x 1.45) = 5.42 g/t. This is a single block estimate, called an E type estimate. This is then repeated to estimate all of the blocks in the blank model.

MICROMINE example; Median indicator kriging; Median indicator kriging uses one Indicator variogram modelled for the median at the 50th percentile defaulted to all of the grade thresholds, instead of different variograms for each threshold. Median indicator kriging is not as accurate but it is faster and is still a non linear technique that can deal with any high grades and dual populations. Once the three indicator variograms are modelled then go to Modelling | 3d block estimate | multiple indicator kriging and save the indicator variogram form, this should be used in the cut-off box. All of the remaining dialogs are the normal setup as for inverse distance weighting, interpolate the Composite file but interpolate the TFE field, do not model the cut field.

The cut-off box allows a form to be saved that should include all of the cut-offs for the grade bins and the indicator variogram form containing the 3 orthogonal indicator variograms; in this case the form is as saved as IND and is defaulted to


Page 22.75



each grade cut-off. The grade estimate is the method to calculate the bin grade to be used for the weighted estimate, if mean is selected it is the average of grades between 0 and 10%, 10 and 20% etc; if median is selected then it is the median grade for grades between 0 and 10%, 10 and 20% etc.

Notes:

The deciles do not have to be used for all bins, the top bin may be at 95% or some other figure that in the opinion of the modeller appropriately deals with the influence of a few high grades.

The IND formset includes the 3 saved indicator variograms for the median.

Page 22.76




Notes:

The first indicator variogram is the direction of maximum continuity.

The second indicator variogram is the intermediate variogram with the second longest range.


Page 22.77



Notes:

The third indicator variogram is the variogram with the shortest range.

Again use block kriging and interpolate the parent cells only. Run the MIK routine.

Kriging variations Page 22.78



The kriging calculations are not straightforward and only the advent of fast processing of millions of multiple equations by computers has allowed their application as an alternative to simpler methods. Numerous variations exist upon the listed algorithm with ordinary, universal, disjunctive, indicator kriging etc also available.


Notes:

Note that – Simple kriging differs from ordinary kriging in that simple kriging interpolates from a constant or known mean whilst ordinary kriging applies a local mean which varies across the sample area. MICROMINE supports ordinary kriging. Cokriging requires an inverse relationship of one element to another in the model area such as nickel and magnesium. Disjunctive kriging is another name for co indicator kriging.


Page 22.79



Notes:


Good Practice Keep


See:

M

Page 22.80



Lesson 11 – Model validation


Notes:

There are three types of validation local (section grades compared to model grades), global (tonnes and grade of data compared to model) and reconciliation (predicted versus modelled reconciled at the time of mining) the model result was validated globally and locally Globally means the total raw data grade was compared to the model result for each lode and totally Locally means that the drillholes were overlain on the block model and the raw grades were compared to the interpolated grades The generated grade model should be carefully validated for potential errors. There are several steps Check the block model tonnes and grade against the Modelling | Polygonal wireframe estimate | grade tonnage report. This is your best reference tonnes and grade check; if there are big differences then look for the reason why. Visual display of block model. Comparison of block grades against drill hole sample grades. Plotting sections and plans together with geological outlines. General trends and continuity of interpolated mineralisation should not be too different from the expected trends. Interpolation of grades using different methods. Comparing the global figures between different interpolation methods (e.g. Kriging vs IWD or IWD2 vs IWD3) Comparison of resource figures with previous reports. Compare wireframe volume with the block model volume.


Page 22.81



Notes:

Global validation: The model and wireframe volume should correlate very closely, less than one percent variation. The grade comparison between the model and the wireframe can vary because often the raw samples are clustered and not evenly spaced whilst the model blocks are evenly distributed. The grade can vary up to 5 percent.

Declustered global estimate Analysis of the declustered assay data is required to validate the interpolated grades in the block model. The main method to generate a declustered global estimate is Averaging. Averaging

Steps to generate a declusted assay database using Averaging option: Run the Modelling | 3D Block | Statistical process. It will assign unique Index values for all samples. Index field will have the same values for assays within the same blocks. The cell size should be estimated on the basis of the sample density or “incremental test” (see below). Generate weighted average grade values for each Index from the assay data using Dhole | Calculations | Extraction. Constant field should be Index, Extraction Type = Weighted Av.

Page 22.82




Notes:

Place the grade against the block dimensions for x and y, repeat the process until a grid of cells versus estimates is created that can then be contoured, the lowest value is the optimum cell size and the declustered estimate. Repeat from x and z for three-dimensional data. Then use the cell dimensions to obtain a single estimate which is the declustered global estimate. Note if your data is very regular and spaced on a grid then the data is not clustered and this process is not required, the global estimate will be immediately apparent.

Local validation: The drillhole grades and the block model grades can be compared to look for any aberrations.


Page 22.83



Model validation:

Notes:

Use the modelling | subblocking | validate block model routine to check for any overlapping, duplicate blocks, or any blocks beyond the permitted project boundary. The report file should not report any errors. This is an important validation step.

Page 22.84




Notes:


Good Practice Keep


See:

M


Page 22.85



Lesson 12 - Block Model Display

Notes:

The block model can be displayed in Vizex. Do not load the block model in the 3d viewer; the speed of a large block model is such that it should be viewed in Vizex.

Page 22.86




Notes:


Good Practice Keep


See:

M


Page 22.87



Lesson 13 – Resource Classification

Notes:

There are no standard rules and procedures how to classify resources however the JORC code provides guidelines. Different users have different approaches. The most common approaches follow: Variogram ranges can be used to classify resources. For example, if cells are populated using search radii equal to 2/3 of short ranges of variograms and more than 3 samples were used; those cells can be classified as Measured. If cells are informed using search radii equal to long ranges of variograms and more than 1 sample was used, those cells can be classified as Indicated. All other cells are classified as Inferred or not coded at all. If this approach is accepted, interpolation run will actually be the Resource Class. Drilling density or exploration grid can also be used for resource classification. Some users accept an idea that for the given commodity and given complexity of deposit, resources can be classified on the basis of the generally accepted exploration grid (e.g. 50 x 50 m for Measured Resources, 100 x 100 m for Indicated and 250 x 250 m for Inferred). If this approach is accepted, the user should digitise outlines in plans or sections for each resource category and use the Assign processes to classify model cells. If necessary, wireframe solids for Resource Classes can be generated and used to assign Categories to block model cells. If Kriging was applied for grade interpolation, estimated kriging variance can also be used for resource classification. Obviously, the estimated variance reflects the reliability of grade estimate in each cell. If this approach is accepted, then user should define a range of variances for each Resource Category and run File | Fields | Generate to assign Categories. When classifying resources, the user should remember that other parameters should also be considered, such as: 

Reliability of rock density values



Reliability of sampling and analytical data



Methods and reliability downhole survey data

 Accuracy of drill hole collar locations  Availability and potential effect of topographic terrain 

Results of site visit



Other parameters



Confidence from interpretation of the continuity and design of the mineralisation.



Confidence in the total input data from above.

CHECKLIST:  No Measured resource class will be declared without a site visit.  No Measured resource class will be declared if any bias or precision

issues are identified in the sample quality analysis  Display/plot the classified model and compare with the expectations

Page 22.88



Kriging variance: There is no single kriging variance value; rather kriging variance and kriging standard error are calculated for every interpolated block in the block model. The size of the error can be used as a guideline to the confidence of the accuracy for block estimation, and can be used to categorise the deposit. Categorisation divides the deposit into different confidence levels for the accuracy of estimation; often it will reflect the need for further drilling or more accurate data, if poor recovery, sampling and analysis contribute to doubt. The model can be colour coded on the kriging variance, reflecting regions of lower sample density and estimation reliability.



Notes:

Page 22.89



Notes:


Good Practice Keep


See:

M

Page 22.90



Lesson 14 – Resource reporting


Notes:

Use Modelling | Block model report;

The block model is comprised of underscore fields which is the block size, a subblock factor, an sg and an estimated grade.


Page 22.91



Notes:

For each block tonnes = _EAST X _NORTH X _RL X sbf X sg; if subcelling is used and not the sub block factor then block tonnes = __EAST X _NORTH X _RL X sg Grade = estimated grade, seen in the coloured field. The model can be queried on a bench by bench level by using the multi types thickness field with the _RL field.

Page 22.92




Notes:

By using the thickness field and the material field the model can be reported for different levels with different grade cutoffs, this is useful for planning purposes.


Page 22.93



Notes:


Good Practice Keep


See:

M

Page 22.94



Lesson 15 – Cut-off grades and grade tonnage curves


Notes:

When Resource Class is assigned to each block model cell, resources can be calculated and reported in terms of tonnes, volumes and average grades (Modelling | Block Reserves | Report). Usually Resources are reported for various cut-off grades and for each resource category separately (other subdivisions could be areas, zones, lenses, ore bodies etc.) as well as the total figures. When resource figures are generated for various cut-off grades, they can be imported into Excel and used to generate grade-tonnage curves or use Display | graphs | general.

By using a cut-off set with the largest value down a report file on resources can be produced that can be used to display a grade tonnage curve. Run the model report to create the report file.

Then go to display | graphs | general to display the report file results


Page 22.95



Notes:

Move the cursor along the grade tonnage curve to see the active reading of available tonnage with a change in cut-off grade of mining.

Page 22.96




Notes:


Good Practice Keep


See:

M


Page 22.97



Lesson 16 - Example NVG data Ordinary kriging start to end

Notes:

Step 1: Classical statistics exhaustive population Go to Stats | distribution; use the assay file and the Auave field to examine the distribution of the population in the histogram and the probability plot modes.

Page 22.98



Clearly the distribution is neither normal nor lognormal, the background versus mineralisation cutoff evident on the probability plot is around 0.6 g/t.


Notes:

Step2: Generate downhole coordinates Go to Dhole | generate | downhole coordinates; and setup the collar, survey and assay files to create three d coordinates in the assay file, the fields must be added to the assay file and will be populated when the routine is run. Ensure the azimuth correction is set to 40 degrees for the NVG data as seen in the previous exercises.

Run the routine; the east, north and rl fields are now populated with the coordinates of the interval centroid.


Page 22.99



Step 3: Assign the wireframe to the assay file

Notes:

Go to Modelling | assign | wireframes and write a code into a field in the assay file to identify which intervals are inside the wireframe and which intervals are outside the wireframe. Use the All ore wireframe set. Note this exercise assumes that the interpretation and wireframing has been completed from a previous exercise.

Step 4: Classical statistics orezone Let us use the ore lode MV1S as an example. Set a filter to look at MV1S lode only.

The histogram of the mineralised grades shows a population that approximates a normal distribution, the probability plot in lognormal model also shows almost a straight line, indicative of a normal distribution.

Page 22.100




Notes:

As such a linear method of interpolation such as ordinary kriging or inverse distance weighting is accepatable for modelling.

Step 5: Apply a balancing cut The cumulative frequency curve of the mineralised grades indicates that at 97.5% the grade is equals 27 g/t, this becomes the upper or balancing cut.

Use file | fields | replace to replace the grades in a new field auavecut with those above 27 g/t back to 27 g/t.


Page 22.101



Notes:

Step 6: Composite the data to equal intervals Go to dhole | compositing | downhole and run the routine with the constant field set to the wireframe flag field, this means the composite will start and finish as the drillhole enters and then leaves the wireframe. Composite the data to 1 metre, this is the most common interval, composite the cut field.

If the coordinates are now badly affected then rerun the generate downhole coordinates to fix the problem in the composite file. Check that intervals are to one metre and grades are recalculated by opening the file to have a look.

Page 22.102



Step 7: Geostatistics


Notes:

Begin the search for the three orthogonal semi variogram models to be used for the ordinary kriging model by following the same procedures for the NVG data as used for the iron data. The first step is to investigate the size of the nugget from the downhole semi variogram. The best lag size to use is obviously the most common sample spacing. For the NVG data if a choice is made to use the composite file and the data has been composite to 1 metre then clearly three lag should be set to 1 metre for the downhole variogram. Set the mode to calculate from raw data and the semi variogram type to Downhole, fill out the remaining dialog boxes using the lag of 1 metre, and ensure a filter is used so that only the data in the MV1S ore envelope is used for the analysis.


Page 22.103



Notes:

The downhole variogram indicates the size of the nugget, the nugget is set at 2.

Page 22.104




Notes:

The omnidirectional variogram indicates an optimum 7 metre distance for the lag size.

After producing a fan of directional variograms the principal direction is found to have an azimuth of 15 degrees. A lag of 10 metres produces the best behaved semi variogram. The variogram has a dip of 5 degrees.


Page 22.105



Notes:

A two component semi variogram model is fitted to the gamma values, nugget 2, partial sill 9 and second partial sill 8. The first range is 22 metres and the second range is 40 metres.

Tick on, ‘let MICROMINE calculate the angles for the second and third directions’, MICROMINE will then create the orthogonal directions so the gamma values can be modelled. Save all variogram models as formsets and ensure the nugget and partial sills are the same for each variogram.

Page 22.106




Notes:

The second variogram is modelled, orthogonal to the direction of the first variogram. Only the ranges are different for the second variogram model from the first.


Page 22.107



Notes:

Finally the third variogram is modelled, again only the ranges vary between the three variograms.

Page 22.108



Step 8: Cross validation


Notes:

The three variograms are combined into a single formset, enter a name in model parameters and click right to create the new combined formset. Then double click to select the saved forms for the first, second and third directions, the ranges should decrease from the principle direction to the third direction. Save the combined form, if a mistake was made then MICROMINE will not allow the form to be saved. Once the form is saved run the cross validation routine, the actual versus estimated mean should be similar, 7.69 versus 7.74, an acceptable result. The results of the iron estimation by cross validation were 2.28 for the standard error and -0.0122 for the error statistic. The standard error is acceptable and the


Page 22.109



error statistic is close to zero, a good result. The variograms are accepted for interpolation.

Notes:

Step 9: Build blank model

Create a blank model that can be populated during the interpolation process, select on the fly optimise to reduce the number of blocks in the model. Subcell the parent blocks, in this case by a factor of 10 for each direction. Name the model appropriately and run the routine.

Page 22.110



Step 10: Ordinary Kriging


Notes:

Setup the ordinary kriging dialogs boxes selecting define blocks from file on the main dialog. This file is the unpopulated block model produced in the blank block model. Set a filter to interpolate only the blocks in the wireframe, these are the only grades relevant for the interpolation. Ensure the auavecut field in the assay composite file is the sale as in the blank block model file. Select block kriging, this will estimate to points regularly spaced in the block which are then averaged to produce a block grade.

The disretisation sets the spacing of the points to be estimated within the blocks.

Add an extra field called run, several runs with different search ellipses may be required to populate the entire model, set the value to 1 for the first run, this will be written to the file.


Page 22.111



Notes:

Tick on all numeric exceptions.

Select the saved form name by double clicking on parameters to call on the combined variogram form.

Complete the search ellipse first run form and view in 3d to ensure that it accomplishes the aim of including relevant samples and excluding redundant

Page 22.112



samples for the first run. Increase the search radius and save the forms for second and third runs.


Notes:

Second run form.

Third run form. When the model is run for the second run ensure the value in extra fields is changed to 2 and changed to 3 for the third run.


Page 22.113



Notes:

Model report

Go to modelling | model report | block report. Use the block factor field to ensure the correct tonnage and select usecutoff set with largest value down. Enter a report file and run the report.

Ordinary kriging model report.

Validation

Page 22.114




Notes:

Go to Modelling | polygonal wireframe estimate | grade tonnage report. Ensure ignore missing intervals is ticked on, select the assay file and the wireframe and run the routine.

Wireframe grade tonnage report

The ordinary kriging result is 952,750 t @ 7.2 g/t The wireframe estimate is 952,942 t @ 7.5 g/t.

The validation is acceptable and there can be high confidence in the result of the ordinary kriging estimation. Complete local validation and display the model result.

The NVG resource estimation lessons are complete.


Page 22.115

Module-22-resource-estimation-exercises.pdf

Recommend Documents