4.   MINERAL RESOURCE AND RESERVES

4.2.1  

Sishen Mine

The principal techniques used for outlining the orebody are surface mapping, geophysical surveys (gravity) and surface drilling. The surface outcrop of the entire property has been mapped, and gravity surveys have been performed at intervals over the entire mining area. Drilling is done in a phased manner, with decreasing grid spacing, or infilling, with each successive phase. Phase I to Phase IV are diamond drilled, while Phase V uses Reverse Circulation (“RC”) Drilling. Phase I is drilled on a 400m grid, Phase II is on a 200m grid, Phase III is on a 100m grid, and Phase IV is on a 50m grid. Phase V RC drilling is completed in geologically or structurally complex areas to provide additional definition of the orebody and is drilled at a grid spacing of between 20m and 25m.

To date more than 13,000 exploration boreholes have been drilled, approximately 1,500,000m of drill core and percussion chips have been recovered. Annual drilling averages at between 40,000m and 50,000m and this rate has been sufficient to maintain or increase the total Mineral Resource remaining after mining. The majority of the exploration drilling is NQ (54.8mm) sized, although in areas where ground conditions are poor, typically with cavities and poorly consolidated ground, the holes may be drilled to BQ (42.1mm) size. All the exploration drill hole collar positions are surveyed with a differential GPS.

The rock chips from the RC drilling and the core from the diamond drilling are logged by a geologist, and the data stored in a drill hole management database. All visibly ferruginized core or rock chips are sampled. In addition to assaying all ferruginous material, 1m composite waste samples above and below the ore intersection are assayed. Sample lengths are variable, but the minimum sample length is 0.4m, the maximum is 3.5m, and the standard length is 3m. Samples are restricted to one material type. The core is diamond sawed in half lengthways, and one-half is crushed with a jaw crusher before being sent for assay, while the other is retained for reference. RC samples are bagged for every 0.5m of drilling and are dried and split on site. Composites of the chips are made after geological logging to create sample lengths of 3m or reflecting the material type logged, with a minimum sample length of 0.4m. A core recovery of greater than 90% is required for a sample to be taken.

On receipt of the samples at the mine laboratory sample preparation area, all samples, whether originating from percussion or diamond drilling, are then reduced to –2mm by a gyroll crusher and then split to a 0.6–1.4kg sample for analytical procedures using a rotary splitter. Pellets are then made from the pulverised samples, which are then automatically analysed for FeO total, K₂O, P₂O₅, Al₂O₃, MnO, MgO and CaO by XRF methods. Results are automatically recorded by the analytical instruments, captured electronically and sent directly to the geology department as digital data files.

Material that falls outside the traditional definition of ore, but that will be processed as part of the SEP is sampled in the same manner, and where such material has been drilled in the past, and not sampled, a programme is in place to sample all of this material for areas that could still be mined. The drilling standards have been modified for drilling in BIF to ensure that at least 25m below the ore contact is drilled to intersect sufficient material that could contribute to the SEP.

In addition to sampling of the drilling core and rock chips, during the pre-feasibility and feasibility studies, bulk samples of particular material types that could be part of the SEP were taken from stockpiles, and from in pit benches, to test the potential for upgrading the material to produce a saleable product. During the pre-feasibility study, each sample was 3kt, which was fed through a primary and then secondary crusher. An incremental sampler, sampling at set variable interval rates, covering the total sample, produces a final sample of approximately 2.5t. During the feasibility study, the samples taken were inj aggregate 80t and the final sample produced was approximately 3t.

In both the pre-feasibility and the feasibility studies the final samples were dispatched to Kumba Research and Development for screening in several size fractions, and beneficiation tests performed at several density separations. The results of these tests were used to generate the beneficiation algorithms applicable to the individual SEP material types. For the drilling samples, density measurements are performed on all sample pulps utilising a density meter, which is calibrated on a daily basis. The density measurements are validated against an empirically derived Fe% versus relative density graph which gives the expected upper and lower limits of the relative density expected from the measured Fe%.

Standard samples are submitted randomly with the production geology drilling samples (Phase V), but not with the exploration drilling samples (Phases I to IV). A standard sample has been created from Sishen iron ore, and is used by the MRM Department. Differences of greater than one Standard deviation from the accepted value are flagged for re-assay. Duplicate samples are not routinely submitted; however over the past couple of years, over 200 duplicates have been submitted to the mine laboratory. A good correlation between the original and the duplicate samples has been observed for the major elements analysed. Blank samples are not submitted as the XRF machine is calibrated to analyse samples with Fe% greater than 35% and a blank sample would not give a meaningful result.

The Sishen mine laboratory is SANAS accredited (No: T0195) as from November 2002 and since 1994 has complied with ISO 9002 Standards. The Sishen laboratory participates in a round-robin programme with different laboratories of Kumba, Mittal and Assmang. The laboratory has its own programme of analysing standards and duplicates in addition to those submitted with the mine samples.

The major lithological units in and surrounding the mineralised lithologies are modelled with wireframes. Verification of the classification of the lithological codes assigned to the intervals in the drill holes is done both visually, as well as automatically by comparing the assayed value with the material type and correcting ore materials assigned as waste materials and vice versa. Vertical sections are drawn on regular intervals and rings representing continuous lithological or material units are created. These are linked up between the sections to create the material wireframes. These solids are verified for overlaps and include waste materials within the ore where the intersections are sufficiently continuous.

The major mining areas (North Mine, Middle Mine and South Mine) are divided up into structural domains representing areas with relatively homogenous structural, geological, or chemical characteristics. These various domains are used to create block models with a parent block size of 20m x 20m x 12.5m in X, Y and Z dimensions, respectively. The blocks are sub-celled to dimensions of 10m x 10m x 6.25m to more accurately represent the volume of the ore wireframes. Data of Fe, SiO₂, Al₂O₃ K₂O, P and Density of material types are extracted from the validated drillhole database, within each of the material type wireframes, for each ore type within each structural domain and composited to 3m composites. Zero values are removed to ensure that geostatistical evaluations are based only on analysed data.

Anisotropic variograms are calculated for each element, for each structural domain, with the shortest axis perpendicular to the plane of the orebody and tilted to take into account the average dip of the structural domain. The modelled semi-variograms are typically either single or dual structured spherical models.

Geostatistical estimations of Fe, SiO₂, Al₂O₃ K₂O, P and density (“RD”) are calculated per element per material type per structural domain by means of Ordinary Kriging (“OK”) for main ore, conglomeratic ore and Banded Iron Formation (“BIF”). Directional search ellipsoids are orientated in the dip plane of the specific structural domain. An initial estimation is done using a search with 1.3 times the semi-variogram range. A second estimation is then undertaken using and expanded search of 2 times the semi-variogram range to estimate any blocks that remain uninformed by the initial run. This only affects zones classed as Inferred. A minimum of three and a maximum of 50 samples are used in the estimation.

Geostatistical estimations of Fe, SiO₂, Al₂O₃ K₂O and P for the different waste materials are calculated per element per material type per structural domain by means of Inverse Distance Squared (“ID²”). RD for the different waste materials is calculated according to the Fe grade, based on an empirically derived correlation.

Subsequent to the estimation of the block models, they are sliced up into corridors, based on the easting and separately on the northing values. The average grade of the blocks in the model within each corridor is compared to the average grade of the composite samples within the corridor, to ensure the model honours the trends in the data. If there is an observed significant deviation over a couple of successive sections, a decision may be taken to modify the grade of the block model to match that of the input data more closely.

The classification of Mineral Resources at Sishen takes cognisance of three parameters. These are the relative density of drilling data, the interpreted structural complexity, and the kriging error of the Fe estimate of each block. Initially the density of drilling is assessed. If only Phase I or II has been completed then an Inferred Category is assigned. When Phase III drilling has been completed, an Indicated category is assigned, and after Phase IV or denser drilling completed, a Measured category is applied. In areas that are interpreted to be structurally complex, taking into account; variations in thickness, dip, and the degree of intercalation of waste zones in the mineralised material; the level of confidence is less and area is downgraded to a lower confidence category.

Kriging error associated with Fe estimates is used to assist in delineation of complex areas of increased risk. Fe variance is averaged in the Z direction to create a Fe variance map. The Fe variance is compared to the mean and standard deviation of the mining areas and the structural domains, and areas of higher variance are determined, which are considered to be higher risk areas. These areas are then, in conjunction with the structurally complex areas downgraded to a lower confidence Mineral Resource Category.

The Mineral Resource and Mineral Reserve Statement for Sishen that has been reviewed by SRK is tabulated in Table 4.1.

Mineral Resources defined by Kumba are based on an iron ore quality for Main Plant ore of at least 60%Fe and for SEP ore of at least 58.5% (beneficiated) Fe, appropriate mining and processing methods and a market for the product. Kumba has considered the likely economic potential of Main Plant and SEP Mineral Resource through the use of an “optimistic” open-pit shell (an open pit shell with higher economic criteria) beyond the open-pit shell used to define the Mineral Reserves. Only the high grade (=60%Fe) Main Plant material has been defined as Mineral Resource beyond this “optimistic” open-pit shell as this is considered to have potential for underground mining.

SRK has separated the Mineral Resources and Mineral Reserves in terms of three categories (op1, op2 and op3) to reflect the process used by Kumba:

• op1 represents the material within the Final Pit shell (optimised);
  
• op2 represents the material between the final (optimised) pit shell and the the optimistic pit shell (with higher economic assumptions);
  
• op3 represents the material outside the optimistic pit shell, but above 60%Fe that has potential for underground mining.
  

The TEPs presented as part of this CPR reflect some 1,148Mt of material as total headfeed to the Main Plant and the SEP plant. There is a difference of 115Mt between the total headfeed and the Mineral Reserves as defined by Kumba. This material is strictly additional to even the Mineral Resources that have been defined although for consistency in reporting SRK has defined this as being from Inferred Resources in the LoM Plan. The average grade of the Mineral Resource is 58.6%Fe. SRK has undertaken certain checks and calculations of the Mineral Resources and Mineral Reserves and the appropriateness of the modifying factors as well as certain economic checks and confirms the statement contained in Table 4.1.

4.2.2

Sishen South Project

The primary exploration for Iron ore at Sishen South took place in the 1950s when Iscor undertook regional gravity and magnetic geophysical surveys in the Northern Cape. Anomalies were followed up and drilled with percussion drilling techniques. In the 1990s detailed gravity surveys around Sishen south proved highly successful and individual deposits were outlined to be followed up with drilling. Surface mapping was completed in all areas where there is outcrop and the ore is not covered by calcrete.

The initial drilling done on Sishen South was all percussion drilling, but has later been replaced with diamond drilling. Currently holes are drilled with percussion drilling until just short of where the ore is expected, or to where the ground becomes competent enough, and from there the hole is completed with core drilling. The anomalies are usually initially drilled on a 100m spaced grid and off-grid holes are drilled to more closely define the margins of the orebodies. Portions of some of the deposits are currently drilled on a more widely spaces grid. Secondary infill drilling decreases the grid spacing to approximately 75m to 72m. Where geologically and structurally complicated areas are encountered a closer spaced drilling grid will be implemented. Drillhole collar positions are surveyed. A random selection of drillholes were downhole surveyed and the results all showed the holes to be within 2° of vertical. All holes are thus assumed to be vertical.

For the percussion drilling, 0.5m samples are collected, mixed and logged, and where mineralised, a 1m composite sample is taken. The RC chips and fine material are collected over 0.5m intervals and the hole is flushed after each 0.5m drilled. The RC holes are 165.1mm in diameter. The fines and chips are mixed by hand at the core shed and riffle split into two samples. A portion of each sample is then logged on a palmtop computer, and where mineralised, one of the samples is sent to the Sishen Laboratory for assay. The Sishen mine laboratory is SANAS accredited (No: T0195) as from November 2002, and since 1994 has complied with ISO 9002 Standards. The Sishen laboratory participates in a round-robin programme with different laboratories of Kumba, Ispat – Iscor and Assmang. The laboratory has its own programme of analysing standards and duplicates in addition to those submitted with the mine samples.

The electronic log is then transferred to the Sable Database where it is verified. Final depths of drilling are checked for the RC holes and verified when the core drilling begins. RC boreholes account for <5% of the total number of boreholes drilled.

The core drilling is predominantly NQ (54.8mm) and BQ (42.1mm) sized. Core is first verified in terms of depth and recovery, and then logged on a palmtop computer. The log is transferred to a Sable database and a paper copy of the log is made and verified. Recoveries are recorded on the log sheets and are typically greater than 95%. Samples are marked out in 1m lengths ensuring samples do not cross lithological boundaries and a 1m sample of waste is taken adjacent to the ore lithologies. The core is photographed before being split with a diamond saw and one-half bagged for analysis.

On receipt of the samples at Sishen laboratory sample preparation area, all samples, whether originating from percussion or diamond drilling, are then reduced to –2mm by a gyroll crusher and then split to a 0.6 – 1.4kg sample for analytical procedures using a rotary splitter. If the sample list indicates a field duplicate then a second sample is also taken, for submission to the laboratory. The sample numbers are allocated to Sishen South by the Laboratory, but assigned to samples by the Field geologist and the control samples are therefore blind to the laboratory. Field duplicates are submitted for every ten samples and a control sample made up from a 150kg sample of homogenised Sishen ore is included with every batch of 50 samples sent from the exploration site. Pellets are then made from the pulverised samples, which are then automatically analysed for FeO total, K₂O, P₂O₅, Al₂O₃, MnO, MgO, TiO₂, Ba, S, Sr, Na₂O and CaO by XRF methods.

Results are automatically recorded by the analytical instruments, captured electronically and sent to the Geologists as digital data files.

The density of every sample pulp is measured with a MINDENS density meter, which is calibrated on a daily basis. A bulk density determination programme has also been undertaken, which determined an average density for each orebody.

Geological and assay drillhole data are extracted from the drillhole database and transferred to geological modelling software. Vertical east west sections are created every 50m and detailed interpretations of the geology made on these. These are checked on north south sections before being signed-off. The interpretation includes waste lithologies where they are of significant thickness, and consistent across drillholes. These interpretations are transferred to a second geological modelling package to conform to Sishen standards, where certain of the lithologies are combined to create a simplified 3D wireframe model for each lithology in each orebody. Wireframe solid shells are created by linking up the interpretation on each section.

The drillhole assays were composited to 1m for the estimation process, and omni-directional horizontal semi-variograms calculated from this data for the following elements Fe: SiO₂, Al₂O₃, K₂O, P, S and for RD, as well as shorter range vertical semi-variograms. A block model was created for each orebody, with block dimensions of 10m x 10m x 5m in the X, Y and Z direction, respectively. The wireframe solids were used to assign a percentage of the block within ore, where the whole block is not contained within the wireframe. Grades and relative densities are calculated as a length weighted average of all the intersections within a lithology type, and assigned to the waste blocks within the model. The 1m composites are used to populate the ore lithologies using ordinary kriging. Where there was sufficient data, only the diamond drill core composites were used in the estimation, however, where the data was more widely spaced, the percussion composites were used as well. Blocks that remained uninformed by the Kriging estimation were populated by using the estimated blocks and assigning a value based on the nearest neighbour. The relative density values were Kriged for the Kapstevel – Welgevonden model, but were calculated using an algorithm based on the Fe value for the remaining models. The Resource model estimates have been reviewed by various external consultants, including SRK and Snowden, for the purpose of validation and have been accepted by them.

Ore lithologies are considered for inclusion in the Mineral Resource and only material within those lithologies that has a Fe content of greater than 60% are included. The classification of the Mineral Resources is based on a combination of the density of the drilling and the Geologist’s interpretation of the complexity of the geological structures in the area. The category may be downgraded if the Geologist considers the area more geologically complex. Structurally complicated areas are judged by changes in dip, faulting, folding and erratic waste intersections within the orebody. Areas with a drill grid spacing of greater than 200m will be categorised as Inferred Resources. Areas with a drill grid spacing of closer than 200m but more than 70m would be classified as Indicated Resources. Areas with a closer Drill grid spacing are considered as Measured Mineral Resources.

For purposes of valuation, the Sishen South Project has been divided into two phases:

• Phase I consists of Reserves within the LoM Plan and which have been valued on a DCF basis; and
  
• Phase II consists of Resources that lie outside the LoM Plan and which have been valued as exploration assets.
  

The Mineral Resource and Mineral Reserve Statement for Sishen South that has been reviewed by SRK is contained in Table 4.2.

4.2.3

Thabazimbi Mine

Mineral Resource estimates at Thabazimbi Mine are based predominantly on percussion drill hole sample information. A few diamond drill holes have been developed for the purposes of acquiring geological, geotechnical and geometallurgical data. Reverse circulation drilling is used only where the geological understanding of the orebody is low and where low-grade or waste zones within orebodies are being examined. Geologists at Thabazimbi Mine recognise the potential for down-hole contamination particularly in deeper drill holes. Accordingly, when lithological interpretations are made, using deep drill holes, there is a tendency to make these interpretations in a conservative fashion. All drill hole data are stored in a single database.

The percussion drill holes are sampled using a one-metre downhole interval. These samples, consisting of drill cuttings are logged by geologists, with particular reference to lithology, extent of weathering, rock colour and if possible any structural information that might be accessible from the sampling. Cuttings collected from the drill head are bagged and transported to the geological core shed facility for splitting; a small sub-sample of the cuttings is taken at the drill head for immediate lithological identification.

At the core shed, the samples are split using a riffle splitter into three samples: one sample is retained as an analytical sample, one is retained as a reference and the remaining sample is used to facilitate the logging. This sample is washed to remove dust and allow the lithologies of the retained fragments to be examined in detail and logged. Samples are bagged and labelled for assay at the mine laboratory, situated in the main process plant.

Blast holes (250mm diameter) are also sampled to assist in short-range grade control procedures. Technical staff of the mine are responsible for the sampling of all blast holes. The sample lithology is considered more important in this particular case than the sample grade. Blast hole samples are composited by lithology and single samples of each lithological unit encountered within each blast hole are prepared. One aspect of this sampling technique that must receive comment is the relative sample size and the sub-sample dimensions. Blast hole samples yield large volumes of cuttings per metre drilled and comparatively small sub-samples are retained. The granulometry of the blast holes is also fairly coarse. There has not been any application of modern sampling theory to the determination of the sampling characteristics of these ores and SRK consider that the blast hole samples are likely to retain an unacceptably large sampling variance, largely as a result of the reduction of sample volumes. However, this situation is not problematic because grade control samples do not contribute to any Mineral Resource estimation process. Drill hole collars (both blast hole and percussion drill holes) are surveyed by qualified surveyors; percussion holes deeper than 60m depth are surveyed using down hole instrumentation to determine the drill trace.

Percussion drill hole samples are analysed at the mine laboratory. Labelled sample bags are delivered to the mine laboratory entrance. The bags are examined to ensure that the appropriate sample numbers are present. At present, the geological department does not submit any external quality control samples to the laboratory. All laboratory quality control is the responsibility of the laboratory manager.

On receipt of the bagged samples, the bags are inspected and if the samples are dry, approximately 50g of sample are scooped from the bag and milled. If the sample granulometry is considered to be too coarse, the sample may be crushed within an Osborne Crusher before being milled. The large samples received by the assay laboratory are not reduced to sub-samples through any mechanical process of sample reduction. The sub-sample that is manually extracted from the bag, via scooping, is mixed with a borax binding agent and milled within a closed circuit ring and puck mill and approximately 8g of the resultant pulp is extracted and compressed into a pressed pellet using a hydraulic press; the pellet is approximately 25mm in diameter. These pellets are then analysed using XRF to determine Fe total, SiO₂, Al₂O₃, MnO, P₂O₅, CaO, MgO, K₂O and TiO₂. In the past there have been some restricted programmes involving duplicated samples riffled in the geological core shed, but these have been of very short duration. At the time of writing, there is no active duplicate sample submission process and there has been no conclusive demonstration that the sampling protocols that are employed at Thabazimbi Mine do not result in high sampling variances. All analytical quality control is internal to the laboratory. Instrumental calibration is tested using reference materials that include carbon-steel samples and internal quality control materials. Mine production samples are used as internal quality controls and borate-fusion discs are prepared from sample material and circulated between Kumba Laboratories at Sishen Iron Ore Mine, Newcastle Steel Works and Saldanha Steel in an internal round-robin process. In addition, there have been tests run comparing fusion discs and pressed pellets.

Density of routine geological sample materials is not measured at the Thabazimbi Mine Laboratories. Within the sample database, the stored value for the density is estimated using a relationship that links density to the iron content. This relationship has been determined from extensive sampling conducted at Sishen Mine, using diamond cores. This relationship has not, however been validated with respect to the Thabazimbi Mine ores and the tonnages applied within the Mineral Resource estimation purposes are based exclusively on average density values applied to individual rock types. In the case of Thabazimbi Mine, ore material is assigned a density of 4.7t.m^-3. Low grade ore (ore with iron content between 55% and 60%) has an assigned density of 4.1t.m^-3.

Analytical results from the XRF determinations are compared against the visually determined lithology of the sample. If there is an obvious mis-match between the sample geochemistry and the lithology, the sample is resubmitted for analysis. In addition, the sum of the oxides determined by XRF is estimated. In general the sum of oxides is less than 100%, implying that other elements not determined by the XRF are present within the materials. Analytical data derived from the laboratory are provided in electronic format to the geological department.

The geological and geochemical data relevant to sampled drillhole intervals are stored within a specialised geological database that has several in-built data validation procedures that trap specific errors within the drill hole data. Data are stored within this database and exported to the generalised mining software package that is used for the geological modelling. Within this software package, a data validation script is available that tests each sample interval to check that the geochemistry and the lithology are consistent; instances where these are not, are flagged and are manually checked by the relevant pit geologist responsible for the specific area of the mine.

SRK note that there are no independent analytical quality control checks that permit the geologists responsible for the Mineral Resource estimation to certify that the analytical data are of sufficient quality on which to base the Mineral Resource estimate. Furthermore the geologists are unable to declare that the analytical data are unbiased and that this characteristic is then shared by the Mineral Resource estimates. Thabazimbi Mine has had an extensive production history and the analysis of the production samples against which payments are made are undertaken at the same laboratory. In addition, the round-robin analyses conducted between the Kumba Laboratories on borate fusion discs of production samples are considered to provide a minimum level of quality control that implies that the standard geological samples are probably acceptable for use within the Mineral Resource estimation process. The use of a single density value cannot be considered to be an example of best practice, particularly when the laboratory does not routinely determine sample densities from either the few diamond drill holes that may be present or from routinely collected pit samples representative of intact specimens of ores.

Geological interpretation is undertaken on individual section planes drawn through drill holes. These section planes are digitised and captured within the generalised mining software employed for the geological modelling. In addition to the sectional interpretations, pit sidewall mapping is undertaken on suitable exposures using an MDL laser distomat instrument that records the three-dimensional co-ordinates of sample locations for geologically significant features. Location data for geological contacts mapped on any particular mine face can thus be included within the modelling process.

Wireframe modelling is undertaken on sections, delineating geological entities identified within the drill holes. The individual three-dimensional rings, developed within serial sections are linked together to form a single closed-form wireframe describing the interpreted shapes of the geological features that are being modelled. The features that are modelled include diabase dykes, the high-grade ore zones (+60% Fe) and the low-grade ore zones (ore with iron content between 55% and 60%). Prior to the modelling of the ore body lenses, data coding is undertaken. All one-metre long samples that include a contact between adjacent rock types are coded specifically. These ‘mixed’ samples are specifically excluded from all subsequent estimation since they are contaminated entities that do not belong to any specific lithological population. Block models with block sizes of 10m x 10m x 10m are developed for the purposes of grade interpolation. Originally, this block size was used in order to achieve a reasonable volumetric approximation to the wireframe forms used in the modelling of the orebodies.

Several years ago, Thabazimbi Mine implemented a new generalised mining software package for the purpose of orebody modelling; this programme had several significant advantages over the programme that it replaced, including the ability to sub-block against geological boundaries in order to better reproduce the ore volumes within the sub-blocked models. However, at the time of the change, Thabazimbi Mine did not re-examine the block size issue and retained the use of this block geometry. In reality, the 10m x 10m x 10m block size is significantly smaller than the typical drilling grid, which in well informed areas may approach 50m x 50m. The impact of estimating to small blocks has been the subject of several geostatistical studies and there is no doubt that small block estimates will be accompanied by biased grade-tonnage curves. Thabazimbi Mine do not report the ore estimates at cut-offs above those used to define the orebody envelopes and any attempt to do so would be erroneous; Thabazimbi Mine additionally do not base mining decisions on the grade estimates of individual blocks, but do consider qualities of aggregates of blocks.

Thabazimbi Mine make use of ordinary kriging and inverse distance squared interpolation methods for the development of the grade models used for pit planning, optimisation and the reporting of Mineral Resources. The grades of Fe, Al₂O₃, MnO, P₂O₅ and K₂O are interpolated using ordinary kriging for the ore (Fe>60%)and the low-grade ore (55%<Fe<60%); Mineral Resources are only reported from these two material types. Qualities for all other lithologies are interpolated using inverse distance squared weighting. Within the ore and low-grade ore SiO2 is estimated using an orthogonal regression residual approach, with the regression  between SiO2 and Fe utilised for this purpose.

This estimation approach considers the interpolation of the regression residual between SiO₂ and Fe and the development of the SiO₂ estimate from the Fe estimate, the regression equation between SiO₂ and Fe (SiO₂ = m.Fe + c) and the estimated residual. This approach is advantageous as it approximates a co-kriged estimate between SiO₂ and Fe, which enjoy a strong negative correlation within the ore and low-grade ore units; use of co-kriging ensures that the relationships between the estimated Fe and SiO₂ values are coherent.

Variograms for the kriging estimates are developed from a combined data set of low-grade ore and ore samples. These variograms are updated on an annual basis. The kriged estimates are developed using a nested three-part search strategy. The first search is characterised by the shortest search distances and the requirements for the greatest number of samples, ensuring that this is the highest quality estimate that is made. Sequentially, the search ranges increase and the minimum number of data required to complete an estimate reduce, permitting more blocks that are remote from the data, yet within the interpreted wireframes to be estimated. Blocks are flagged with the index of the search volume that was used to undertake the block estimate.

Thabazimbi Mine classify the Mineral Resources using definitions of the SAMREC Code. In determining the appropriate classification Geologists at Thabazimbi Mine consider the following factors relevant to the estimates:

• Distance separating the block centroid from the nearest sample used in the estimate;
  
• Total number of sample data used to develop the estimate;
  
• The Kriging Relative Standard Deviation; and
  
• Search volume used to derive the estimate.
  

These parameters have all been used to assist in determining the classification that is to be applied to a block estimate. Indicator variograms have been developed for IFe>55%, to assist in determining the “continuity” of low-grade and ore at each of the orebodies; these variographic parameters have been used to define ‘classes’ for each orebody that express the typical continuity of high grade lenses, in this manner appropriate values for the distance to the nearest sample can be derived for each block estimate. The total number of samples used in an estimate is reflective of the total confidence in the quality of the estimate. However if the ore occurs as thin lenses, then less local samples of ore are likely to be available to service an estimate. Accordingly the confidence attached to thin lenses must be lower than that attached to thicker, more robust ore bodies. Blackwell and Sinclair (1996) proposed the use of the kriging relative standard deviation as an aid to the classification of Mineral Resources, its use here is an attempt to include variographic characteristics into the classification and use them in a manner that does not purely consider variogram range versus sample-block separations. The nested search volumes used to estimate the block grades provide an additional perspective on the data density with respect to the search requirements. At Thabazimbi Mine, the classification has been automated, by defining a set of criteria values for each of the four factors considered in the classification. Values for the factors are determined and compared against a set of critical values that define the classification system.

SRK appreciate the desirability of having a classification system that can be automated. The advantages of such a system lie primarily in the reproducibility of the classification independent of personal sentiment that the Competent Person may harbour at the time of classification. In SRK’s opinion the classification system applied at Thabazimbi Mine covers the most salient features that would be required and when the results are examined, these certainly appear reasonable when the sample density distribution is examined as well. One minor observation, for example, is that in the case of the Buffelshoek West estimate, areas of Inferred Mineral Resources occur within largely Measured Mineral Resources. This scenario is unusual in the sense that it implies that within a large area that has been classified as being very well understood, are areas in which the mine staff are unsure that ore may even be present. This type of scenario implies that there may be some requirements to either modify the criteria applied in the classification of the Mineral Resources or to manually modify the classified Mineral Resources, in order to render the classified results more coherent with adjacent blocks. These features aside, the classification system considers a wide spectrum of the most relevant features that deserve consideration in a balanced classification system.

The estimates are based on ordinary kriged estimates of small blocks. In one sense it would be preferable to make use of larger blocks that are more appropriately sized with respect to the drill hole grid spacings. If this were the case, the individual qualities of the block estimates would improve and issues such as conditional bias would be reduced. In mitigation of these effects, these estimates are developed within envelopes that are relatively restrictive with respect to the grade ranges that can physically exist, at least as far as the estimates of iron are concerned. The real issue of concern with respect to the quality of the estimates centres on some of the deleterious elements, in particular Phosphorous that frequently does not respond significantly to beneficiation, unlike Al₂O₃ and K₂O, which frequently reflect shale components within the iron ore and which often respond very favourably under beneficiation. It is highly desirable to maximise the quality of the estimates of some of these components; SRK note that the variograms of these elements have also been developed from the raw data. In SRK’s experience, analysis of components like phosphorous frequently benefit from transformation of the data to a more tractable distribution compared to the usual highly-skewed distribution displayed by these elements.

Mineral Resources defined by Kumba are based on an iron ore quality of at least 55%Fe, appropriate mining and processing methods and a market for the product. SRK understands that Kumba are investigating the application of appropriate economic factors as outlined in the SAMREC Code for the definition of Mineral Resources. The application of economic factors may impact on the quantity of resources. In consideration of the work that is being undertaken and the likely increase in Kumba’s iron ore resources from the Phoenix Project SRK has not modified the Mineral Resources stated above.

The Mineral Reserve Statement includes the in-situ material from the Measured and Indicated Mineral Resource categories that has been converted to produce Mineral Reserves. The Mineral Reserves have been developed following an appropriate open pit optimisation and mine design exercise and reflect the material contained with the envelope of the final open pit outline. SRK has differentiated the Mineral Resources in terms of the categories of op1 and op2 to illustrate the likely Mineral Resource converted to Mineral Reserve and remaining Mineral Resource. The converted Mineral Resource is subjected to certain mining loss and dilution modifying factors. The remaining Mineral Resource (op2) principally reflects material that is available beyond the currently accepted final open pit envelope. The TEPs presented as part of this CPR reflect the Mineral Reserves of some 13Mt. There is no material derived from other sources such as modified Inferred Mineral Resource included in the LoM Plan.

Quality Assurance and Quality Control procedures are in SRK’s opinion inadequate. Notwithstanding these comments, SRK believes that the Resource and Reserve statements are SAMREC compliant. This is supported by:

• The fact that the laboratory makes use of Quality Assurance and Quality Control procedures in the determination of the final product. It must be noted that the Quality Assurance and Quality Control procedures do have an influence on the Reserve Estimations and the conversion of Resources to Reserves. In addition, sample material is circulated between Kumba Laboratories at Sishen Iron Ore Mine, Newcastle Steel Works and Saldanha Steel in an internal round-robin process.
  
• Historical mining and reconciliation since 1932, representing 73 years of operation.
  

The Mineral Resource and Mineral Reserve Statement for Thabazimbi Mine that has been reviewed by SRK is tabulated in Table 4.3.

4.2.4

Mineral Resources Estimate

Details of mineral resources estimates are contained in Table 4.4.