4.   MINERAL RESOURCE AND RESERVES

4.4.1  

Hillendale Mine

The Hillendale Mine deposit consists entirely of Berea Formation red sands; this is consistent with the relative elevation of this dune relative to the present sea-level. The sediments are generally well-sorted and fine to medium grained and also include a significant volume of clays that indicate extensive decalcification and kaolinisation within a tropical weathering environment. Within the Hillendale deposit, silt (– 45µm) proportions range between 15% and 45% and the sand thickness ranges from 3m to 45m; the average thickness is between 18m and 21m.

The footwall of the orebody is defined by falling THM grades. Within the Hillendale dune complex, some variations in sand qualities have been noted within the drillhole data. In the central and northern sector of the Hillendale dune, the western half of the dune contains sporadic occurrences of coarse-grained orange sands characterised by low silt grades. The origin of these sand units remains equivocal and the relationship between these units and the red Berea sands remain unclear. In the central parts of the dune a grey semi-indurated calcic sandstone unit may be developed between 10m and 25m beneath the surface. Beneath the Berea sand, the Hillendale dune complex is underlain by a kaolinitic clay unit that is green to yellow-brown. This unit is interpreted to represent lagoonal muds developed within a low-energy environment landward of the dune complex.

Hillendale deposit was initially sampled during the late 1980s by NMS, who used a power auger to explore the deposit following visual identification of heavy minerals within the Hillendale dune complex. A limited amount of large diameter auger drilling has also been completed in the Hillendale deposit. In addition, continuous-auger drilling has also been undertaken. Reverse-circulation drilling has also been undertaken at Hillendale and comprises the majority of the sample database.

The central portion of the Hillendale deposit has been mined. The southern section of the deposit is predominantly sampled using RC drilling, at a nominal drillhole spacing of 50m by 50m. In July 2004 most of the northern sector, north of the mined block, was re-drilled by Wallis drilling using aircore drilling technology (WAC); this drilling was completed at a nominal drillhole spacing of 50m by 50m. The most northerly sector of the deposit has not been covered by Wallis drilling and remains sampled by RC drillholes. In total, inclusive of grade control drillholes, there are 1,279 drillholes within the Hillendale deposit; 440 of these holes represent NMS and IHM sampling operations consisting of Power Auger (124 holes by NMS), RC (105 holes by NMS and 493 by IHM and Ticor),WAC (409 by Ticor), six core auger holes (IHM) and a set of Large Diameter Auger (LDA) holes (52 by NMS and 90 by IHM).

The geology and mineral resources of the Hillendale deposit is described by three separate block models:

• Area 1 is the northernmost block and Area 3 is the southernmost. Area 1 is the unchanged 2001 vintage model;
  
• Area 2, the central sector of the deposit that contains most of the Wallis aircore drilling, is covered by a 2004 model; and
  
• Area 3, which includes the mined-out block, is described by a model of 2003 vintage.
  

Exploration samples for Hillendale and Fairbreeze Project have been analysed at two laboratories; these are the Old Blen/IHM Laboratory, which was situated at Hillendale; and currently the CPC Laboratory, which has ISO 17025 accreditation for its final product analysis. The analytical process applied to exploration drillhole samples includes a number of consecutive and dependent stages:

Samples are screened at a 1 mm size to remove the oversize component. This process is not part of the initial work recorded at Hillendale and appears to be most specifically undertaken at Fairbreeze Project.

Samples are subject to attrition and then desliming, in which the – 45µm material is screened off and the remaining dried sample mass is determined: this permits the silt mass percentage to be estimated

The heavy mineral content is determined by undertaking a Tetra-bromo-ethane (“TBE”) heavy liquid separation: the material that sinks is recovered and weighed to yield the THM content.

The THM fraction is subjected to magnetic separation within a Carpco Magnetic Separator; initially the high susceptibility magnetics (essentially magnetite) is removed, followed at higher gauss settings by crude ilmenite, the magnetic “others” and the non-magnetic heavy mineral fraction.

Zircon and rutile report to the non-magnetic heavy mineral fraction. Leucoxene may report here, as well as in the magnetic “others” fraction.

For Hillendale, a set of samples was subjected to grain counts on the various magnetic separation concentrates and a set of constant ratios were determined for the non-magnetic fraction mineralogy. The determined proportions of the minerals of interest that were subsequently used were: zircon 56%, rutile 26% and leucoxene 6%.

Rig duplicate samples have been collected for several of the sampling campaigns, typically at a frequency of between 5% and 2% of the total samples, and these data have been used to examine the precision of these data pairs utilising two techniques, namely a modified Thompson-Howarth process and the Absolute Relative Deviation (also known as the Mean Percent Difference). These two methods have been applied to field duplicate data from CPC, Old Blen and IHM Laboratories and these laboratories yield similar results. Additionally the two precision measures also yield convergent results; at the 95% confidence limit, relative precision for THM is of the order of 27%, for Silt, 23%. The precisions for the magnetic separates are significantly higher because errors in the process are cumulative and at each stage in the process smaller masses are considered, so relative errors are expected to increase. Magnetite experiences a precision of ±39%, ilmenite ±40%, magnetic – “others” ±80% and non-magnetic fraction, ±68%.

In addition to monitoring reproducibility, control samples (internal standard samples whose values are not certified) are submitted within exploration samples to attempt to monitor accuracy. Kumba have submitted control samples at a frequency of 4% with routine exploration sample submissions to the laboratory. From laboratory returns for exploration campaigns conducted at Hillendale (2001 – 2002 – CPC Laboratory) and Fairbreeze Project (June – September 2002 – CPC Laboratory) and Fairbreeze Project C (2003 – IMP Laboratory) it is evident that the about 82% of the data fall within 1 standard deviation of the mean value of the control sample. The co-efficient of variation of the standard values and the relative scale of this variation does, however, vary between the different sample programmes. As an example, the 1o¯ limits in the Fairbreeze Project C exploration programme (CPC Laboratory) for the Ilmenite determinations are ±2.1%, whereas the same limits within the Fairbreeze Project C Extension exploration programme (IMP Laboratory) are ±1.2%. Analytical results for the Hillendale Ilmenite values at the CPC Laboratory have 1o¯ limits of ±0.44%. These results show that there are significant differences between the Relative Standard Deviations of the determinations undertaken at the two laboratories within the three programmes; the IMP Laboratory programme has the lowest RSD (11%), followed by the CPC Lab results at Hillendale (11.5%) and the CPC Laboratory at Fairbreeze Project C (19%).

Details of quality control data for pre-1995 drill programmes are not available and it is necessary to assume that the analytical quality of previous sampling campaigns probably approximates the present sampling data. Some of the drillhole logs record evidence of duplicate sampling having taken place; however the duplicate results have not been captured electronically and have not been subjected to any systematic analysis that would provide a measure of analytical repeatability for earlier drilling and sampling campaigns. One issue of note, however, concerns a change in the analytical procedures that appears to have taken place in April 2001.

Prior to this date, magnetic separations were undertaken using a current setting of 0.1A for magnetite, 0.5A for ilmenite and 3A for magnetic others/non-magnetics. After 2001 the settings for the magnetic separators were 0.05A for magnetite, 0.8A for ilmenite and 2.4A for magnetic-others and non magnetics. The implications of these changes are unclear, but some variance between older exploration data and newer information may become apparent.

With regard to other data aspects, there have been some issues with respect to drill collar information. Within the Fairbreeze Project data set drill collar locations have either been surveyed or have been scaled-off from orthophotographs. Some of these drillholes had collar elevations that differed from the elevation determined from an airborne laser survey conducted in December 2002. The collar elevations have been ‘corrected’ according to the DTM. Data requiring a correction of greater than 4m are excluded from all geological modeling processes; data where the collar correction exceeds 2m and the hole was not surveyed, but was scaled from orthophotographs are used in modeling, but not for grade interpolation.

The Hillendale deposit is described by three separate block models. The block size used for modeling is 10m x 10m x 5m. A wireframe surface has generally been created to depict the floor of the mineralisation and this wireframe surface attempts to delineate the sample cut-off grade of 1.5% THM. Crude ilmenite corresponds to approximately 55% of the THM grade. The THM grade cut-off is determined considering the average grade of the last two samples. It is unusual to see a large number of grade ‘spikes’ beneath the modeled footwall surface to the mineralisation. In some cases, drillholes do terminate in ‘ore’, particularly some of the older drillholes that might have intercepted more indurated sand zones that retarded the drillhole progress. Modeling of these older holes has given rise to ‘high toes’ in the modeled footwall within older models of the Hillendale deposit and mining has exposed ore grade materials beneath many of these features. The topography of the Hillendale deposit has been surveyed at a 2m resolution using differential GPS instrumentation to a ±1 cm accuracy.

The Area 2 Model is informed by Wallis Aircore drillholes at a 50m x 50m drillhole spacing; on the southwestern periphery of the deposit some RC drilling data are retained within the model, which was completed in April 2005. Two separate geological domains have been considered within the Area 2 Model; the western ‘core’ of the dune complex consists of low-silt coarser-grained sand, which is also characterised by a high proportion of ‘magnetic-others’ and the remaining volume of red Berea Formation, high silt sands.

Variographic analysis has been undertaken on the two domains and the variogram ranges are used to develop search strategies for the estimation. Estimation has been undertaken using inverse-distance cubed weighting. Grade interpolation takes place in four stages. The first estimate is based on the search ranges derived from the variogram parameters (search ranges of 2/3 variogram ranges) and an octant based search strategy is applied in which a minimum of between 10 and 15 composite samples is required to be located within the search neighbourhood for the block to be estimated.

Blocks that are not estimated with the first search are subject to a second pass estimate in which the search ranges are increased (search ranges are doubled to be equal to 1.5 times the variogram ranges). Blocks that are not estimated using this second search are estimated using a nearest neighbour estimate based on either drillhole samples or the nearest estimated blocks. Blocks are flagged to record which search was used to estimate the block grade, also with the number of samples located within the search procedure.

At present only the Hillendale is in production. Reconciliation work between the block model and the PWP recoveries has been undertaken on a monthly basis over a period of 29 months (January 2003 to May 2005). In general the tonnage values correspond quite well, although there are some large variations between the actual results and the estimates within the model (as large as 25%) on a monthly basis. Since late 2004 the tonnage treated by the PWP divided by the tonnage depleted factor has been close to 1.

For ilmenite the results are quite stable, although over the 29 months, the model has understated the production results by 17% with respect to ilmenite, by 28% for zircon and 22% for rutile. Leucoxene recovered values are on average 220% of the predicted values.

SRK considers that these observed biases within the estimates are highly likely to be a result of the sampling methodology, most specifically the use of reverse circulation drilling. Mining of the area sampled by Wallis aircore drilling includes only the months of March 2005 until May 2005. Indications are that the reconciliation between the model and the production plant may be improving, although the sample is too small to demonstrate this behaviour conclusively.

Rig split duplicates reveal that the total precision associated with the sampling and analytical procedures is actually quite poor for the non-magnetic fraction, but are acceptable for THM and Silt determinations.

Accordingly, the zircon, rutile and leucoxene estimates are expected to be quite poor and this is clearly indicated by the reconciliation data for the Hillendale deposit.

SRK considers that the Mineral Resources are based on an acceptable number of data values, with acceptable quality to permit the classification of Mineral Resources that has been applied by Kumba to this deposit.

Hillendale has drilled off significant sections of the remaining Mineral Resource using Wallis Aircore drilling, which shows better sample recovery than RC drilling that dominates the Mineral Resource database in the current mined out areas. Given the fact that the Wallis Aircore drilling is considered to provide better sample information than the older RC drilling and that the majority of the remaining Hillendale Mineral Resource is drilled off predominantly by Wallis Aircore drilling, SRK has not used any upgrade factors in the preparation of the FM and has used the in-situ ore grades as reported from the Hillendale Mine Surpac block models, against the mine plans provided by Ticor SA to SRK.

4.4.2

Fairbreeze Project

The Fairbreeze Project deposit is divided into five separate blocks: Fairbreeze Project A, B, C, C Extension and D Blocks. Fairbreeze Project deposits consist of Berea formation sands, although at Fairbreeze Project these sands occur on a variable bed-rock surface. In the extreme north, Vryheid Formation rocks (sandstones and shales) crop out southeast of the Fairbreeze Project C deposit. Between the A and B blocks Natal Group lithologies (sandstones and grits) are exposed. In this area, Berea sands are generally absent and the overburden developed above the bedrock consists of fine-grained, silt poor wind blown sands.

The Fairbreeze Project A and B deposits are considered to represent two sections of a set of strandline deposits. Strandlines represent tabular zones of concentrated heavy mineral accumulations that are preserved by gradual marine regression that leaves the strandline above the level of marine erosion.

Fairbreeze Project C deposit has been interpreted to be a beach deposit formed on the low energy side of an ancient headland that projected into the sea. The heavy minerals accumulated against the headland, whilst the lighter minerals were continually remobilised by wind transportation, resulting in concentration of the heavy minerals. Fairbreeze Project D deposit has been interpreted to represent a set of strandlines deposited to the east of Fairbreeze Project A, B and C deposits during progressive sea-level regression. The Fairbreeze Project deposits have been sampled at variable spacings and using a variety of sampling techniques. The initial stages of exploration involved the use of shallow hand auger holes by NMS to undertake a reconnaissance sampling programme over the dune complexes to identify presence of heavy minerals. Mittal Steel continued sampling of the Fairbreeze Project deposits after acquisition of NMS and have made use of auger techniques including power auger, large diameter auger and core auger. Reverse circulation drilling was undertaken at Fairbreeze Project A (at a nominal drillhole spacing of 200m x 200m) and Fairbreeze Project D deposits (15 drillholes).

Substantial areas of the Fairbreeze Project C deposit have also been explored using RC drilling techniques, at a nominal drillhole spacing of 50m x 50m. Wallis aircore drilling has been undertaken at Fairbreeze Project C Extension at a nominal drillhole spacing of 100m x 100m. A limited amount of Wallis aircore drilling has been completed in the Fairbreeze Project A deposit and Fairbreeze Project C deposit, specifically for metallurgical purposes and to gain additional data relating to the mineralogical zoning of the deposit.

This work has been concentrated within the areas planned for initial mining. The Severin Development Corporation did complete some RC drilling within the Fairbreeze Project C Extension deposit; because of uncertainties relating to the position and analytical procedures employed for these holes, Kumba have not retained these earlier data in the Fairbreeze Project C Extension estimation. Fairbreeze Project Mineral Resource estimation has been undertaken within two modeling exercises. SRK were provided with two block models for the purposes of review; one model created in 2002 contained data relating to Fairbreeze Project A, B, C and C Extension deposits. A second model created in 2004 contained estimates for the Fairbreeze Project C and C Extension orebodies.

Staff of Kumba reported that the initial estimation at Fairbreeze Project was undertaken using a block size of 10m x 10m x 5m. The block model has been constrained between a topographic surface and a footwall surface that that was derived from prevailing costs and assumptions relating to an economic grade at the time of the generation of this model. In essence an incremental grade cut-off was applied, that approximates 1.5% ilmenite. In situ grades for THM, silt, ilmenite, magnetic-others and non-magnetics were interpolated using inverse-distance cubed weighting. Given that the drill hole data are located on a grid, SRK are of the opinion that inverse distance-cubed weighting, whilst not preferable to kriging, should give a globally unbiased result.

The deposit was subdivided into two separate domains; estimations were undertaken within an upper zone and a lower zone of the deposit, where a high THM sand unit and a low THM sand unit, respectively, have been identified. Zircon, rutile and leucoxene were determined from abundances of these minerals observed within grain counts performed on composite samples of the non-magnetic heavy mineral fractions.

In 2004 Kumba undertook a re-estimation of the Fairbreeze Project C and Fairbreeze Project C Extension model after completion of additional drilling; in 2002 103 RC drillholes were developed at a nominal drillhole spacing of 50m x 50m within the northern half of the Fairbreeze Project C deposit. Conventional analyses were completed on 3m composite samples. In addition, XRF determinations and mineralogical examinations were completed on composite samples of the crude ilmenite and non-magnetic fractions. The southern half of the deposit is covered with 100m x 300m spaced RC drillholes. In 2003, 157 Wallis aircore drillholes were developed at a nominal spacing of 100m x 100m over the Fairbreeze Project C Extension deposit. Analyses were completed on 3m composites and XRF and mineralogical analyses were undertaken on composite samples of the crude ilmenite concentrates as well as on the non-magnetic fractions of the THM. A 14-hole, large-diameter auger programme was also undertaken during May 2003 on the Fairbreeze Project A, Fairbreeze Project B, Fairbreeze Project C and Fairbreeze Project C Extension to generate samples for pilot plant test work.

Fairbreeze Project C and Fairbreeze Project C Extension deposits were sub-divided into domains on the basis of the THM values and the magnetic-others component within the THM. Variography was undertaken to establish acceptable search criteria. The majority of components examined displayed nested spherical variograms. Values were interpolated using inverse distance cubed estimation with the search ranges equivalent to the ranges of the first structures of the variograms. Blocks that were not estimated in the first search were estimated using a second search, where the search ranges were equivalent with the longer range structures of the variograms. Blocks that were not estimated using these two search procedures were estimated using nearest-neighbour methods.

Silt was examined on a global basis, whilst within the domains the THM and other components were examined. Kumba elected not to use the in-situ values for Ilmenite, and the other components of the THM, but rather to work with the percentage of Ilmenite within the THM value. This approach was then applied to all other components as well. In-situ ilmenite values were recovered by multiplying the THM estimate with the Ilmenite percentage-of-THM estimate. The estimated magnetite, ilmenite, magnetic-others and non-magnetic values were summed and then normalised against the THM estimate, the normalisation factors derived from this process were then applied to each of the estimated components in turn to ensure that the sum of the estimated components is equal to the estimate of the sum. Mineralogical grain counting work was undertaken on selected composite samples of the magnetic separates of the THM concentrates to determine the ratio between leucoxene and rutile. The orebodies were zoned according to these results and the rutile and leucoxene values were determined from factors determined from mineralogical and XRF analyses of the composite samples of the non-magnetic fraction. This total ZrO₂within the non-magnetic fraction has been assumed to be hosted within zircon and the zircon grades have been determined on this basis.

Surface models were created from the drillhole data, representing cut-off grade isopachs that would effectively define the base of the mineralised orebodies. An isopach surface representing the 1.5% THM boundary defines the bottom of the orebody. Material between this surface and a 1.5% ilmenite surface has been classed as ‘low grade material’, whilst ‘Ore’ is classed as material above the 1.5% ilmenite surface model.

The recent sampling (post-2003 in Fairbreeze Project C and Fairbreeze Project C Extension) within portions of the Fairbreeze Project deposit is subject to the same analytical imprecision (particularly with respect to the non-magnetic fraction). Because of the drilling methodology employed in Wallis aircore drilling there is an assumption that data developed from sampling of Wallis aircore holes presents samples that are ‘of higher quality’ than conventional RC samples.

SRK’s experience in Mineral Sands exploration leads to a concurrence with this opinion. SRK consider that the Fairbreeze Project C and C Extension Mineral Resource estimates are based on suitable numbers of acceptable samples.

4.4.3

Limpopo Province – Gravelotte

The Limpopo Province assets were explored and delineated by IHM. The primary orebody consists of alluvial mineralisation that has accumulated above the recent unconformity developed above the Rooiwater Igneous Complex lithologies. Some hard rock mineralisation has been included within the Mineral Resource as well.

Cut lines were developed at a 200m spacing over the project area and detailed mapping was carried out along the cutlines. Some diamond drilling was undertaken; two holes were drilled on each section line spaced 400m apart, except on the Farm Quaqqa 759LT, where the drill sections are located 800m apart on strike. The sand unit was sampled within pits that were dug at 50m spacings along the 200m spaced cut-lines. The samples were analysed at Mittal Steel Laboratories (SANAS accredited, registration number TO165) and the ilmenite content of the samples includes all ilmenite, even that material with a grain size of less than 45µm; approximately 14% of the total ilmenite is located within the fines component of the orebody. No quality control data are available to demonstrate the appropriateness of the sub-sampling techniques or the resultant sample grade values that are included within the Mineral Resource estimate. An in situ density value of 1.75t.m^-3 has been derived from testwork conducted in the sand mineralisation, using a sand replacement method.

The pebble and hard rock mineralisation has been explored using diamond drilling techniques and 1m downhole sampling intervals were employed. Half core samples were analysed by whole rock XRF methods. The 50m x 200m data grid confirms the relatively uniform grade and thicknesses of the Gravelotte Sand resources, and testwork reported by Shepherd (1996) suggests that similar estimation results are accessible from subsets of the drillhole and trench sampling data.

The Gravelotte Sand resource has been estimated within the Surpac modeling software, using inverse distance cubed estimation. Depth and percentage TiO₂ have been interpolated into a two-dimensional grid model with blocks of 200m x 50m. Michael Shepherd (1996) reports the comparison between the Mittal Steel Mineral Resource estimate for the Begin property within the Gravelotte Resource and his own independent estimate conducted on the Mittal Steel data; there is a favourable comparison.

The analytical technique used and the subsequent estimation method applied in the Gravelotte Sands deposits accounts for all ilmenite, including the fines material that would not normally be recovered using a sand-type circuit. In many instances early removal of the fines is preferable to enhance the efficiency of gravity concentration equipment. Mittal Steel have modeled the exclusion of this component of the Resource within their feasibility study work completed on the Gravelotte deposits.

The Gravelotte Resource and Reserve Statement is SAMREC compliant.

4.4.4

Eastern Cape Deposits

The Eastern Cape deposits consist of Pleistocene age sands. There are four principal heavy mineral deposits; Sandy Point Old, Sandy Point Recent, Wavecrest and Kobonqaba. The Sandy Point Old, Wavecrest and Kobonqaba deposits consist of a belt of undulating fixed dunes aligned parallel to the coast and are typically 4km to 5km wide and up to 80m thick. They are considered to be aeolian in origin. The Sandy Point Recent deposit consists of younger dunes and beach sands, which occur as a belt of light, coloured sand separating the older fixed dunes from the sea. This second setting has a lower silt content, 4% compared to 20%, as weathering has been less prolonged. THM content varies between 5.8% and 9.5% with a valuable fraction of 62% to 82%.

4.4.5

Ticor SA Resource and Reserve Statements

Table 4.21 lists the audited Mineral Resources estimates for the Ticor SA deposits as per 1 January 2006. Fairbreeze A and B, classified as Reserves, are valued as exploration properties and therefore not included in Table 4.21. SRK is of the opinion that sufficient work has been done on these resources to demonstrate that the resources can be profitably mined and recovered in the future. Ticor SA intends to undertake a Definitive Feasibility Study (“DFS”) on the mining of the Fairbreeze Project A and Fairbreeze Project B deposits in the medium term future. The Gravelotte and Letsitele deposits are also included in the Table 4.21.

Table 4.21  

4.4.6

Ticor

In general Resources and Reserves are classified using a combination of drillhole spacing and knowledge of the geological continuity, assisted by variography. A particular drill hole spacing alone does not define a classification. A typical drill spacing for Measured Resources across strike is 20m with 10m edge definition. Along strike spacing for Measured resources can vary considerably and can be up to 200m depending in the continuity of the HM grades.

SRK has concerns regarding the lack of quality assurance procedures for classification of Resources and Reserves, however reconciliation studies show that overall the HM tonnages predicted by the models are within 3% of the production tonnages.

Head grade: Grade factors of approximately +10% for HM and +16% for Valuable Heavy Minerals (“VHM”) are used for production forecasting and long term planning. The reason for the HM grade discrepancy is not well understood by Tiwest JV but is thought to be in the assaying process rather than the block estimation procedures. SRK supports this view. Tonnage factors are also in use for production forecasting. Reconciliation data from nine months of 2005 shows production figures for HM tonnes exceed the block model Reserves by 2.9% and VHM tonnes exceed the block model Reserves by 8% inclusive of dilution and losses.

Ore Loss and Dilution: Floor and top of ore are surveyed and analysed on a monthly basis against block model and production data. Current monthly reconciliation is attempting to quantify some of the actual dilution and mining loss parameters. SRK calculations from nine months of reconciliation data show that the combined effect of dilution and losses results in a 2.6% loss of HM tonnes and a 2.6% loss of VHM tonnes across the North and South mines combined when compared to the original reserves.

Resource and Reserve Inventory: The planning ‘reserves’ are essentially Resource material with an allowance for dilution within ‘conceptual pits’ allocated to a long-term schedule.

Table 4.22