Department of Mining Engineering

Mackay School of Mines

Reno, NV

During the spring of 2001 a survey was sent out to a number of gold mines to gather information about sampling and resource/reserve estimation practices in the industry. The survey targeted surface gold mines mining deposits having micron sized gold particles. It contained questions on borehole spacing, sampling and resource/reserve estimation practices in both the exploration stage and the production stage. A total of 50 questionnaires were mailed to surface gold mines in the US, Canada and Australia. Of these only 10 responded for a 20% return rate. Of the 10, 3 responses were from Australia and 7 from the US. One of the responses was from an underground mine, since so few responses were received the relevant portions of the survey returned by the underground mine were included in the results. In general it was found that inverse distance and geostatistical methods were used the most in calculating resources and reserves in the exploration stage. This was usually done by a team consisting of a geologist, a mining engineer and a metallurgist/geotechnical engineer. In some cases an outside consultant reviewed the calculated resources/reserves. During production the polygonal method was most widely used followed by geostatistical methods. Most mines indicated that duplicate samples were assayed and that statistical analysis was done to check the validity of the sampling and assaying practices during both the exploration and the production stage. The majority of the mines surveyed indicated that a gold content versus particle size distribution study was carried out on the blast hole cuttings.

The survey was divided in three parts:

1. A general section containing questions on the type of operation, the material mined and the deposit type.
2. A second section containing questions on sampling and resource/reserve estimation practices in the exploration stage.
3. A third section containing questions on sampling and resource/reserve estimation practices in the production stage.

Two of the 10 responses indicated that both silver and gold were being mined. The correlation coefficient between silver and gold was about 0.16 and 0.25. The gold particles were micron sized except for one mine that also had coarse gold and used gravity methods to separate its gold. This mine indicated that lateritic type ore was mined. The general deposit type was a hydrothermal-hydrothermal disseminated type.

The general method of sampling was a combination of diamond coring and reverse circulation drilling. Other sampling methods used were channel sampling and trenching. The most widely used borehole spacing in the US was about 100-150ft (30.5-45.7 m). In Australia the spacing was about 65.6 ft (20 m). The spacing of drill holes depends on the deposit characteristics. Half of the mines said that they were using geostatistics to determine the "optimum" drill hole spacing. Some used the kriging variance and others the range of the semivariogram. Both regular and irregular drill hole grids were used (see Tables 1, 2 and 3).

Table 1: Sampling method in the exploration stage

Number 1-7=US, Number 8-10=Australia

Table 2: Use of geostatistics to determine optimum drill hole spacing in the exploration stage

Table 3: Type of sampling grid in the exploration stage

The size of the samples taken for assaying was usually determined by experience, only one mine used Gy’s formula. Most of the mines had set procedures in place for reducing field size samples to samples used in assaying. This usually involved pulverizing to about -125 to -200 mesh. The most common assaying method was fire assaying (FA) with either a gravimetric finish or an AA finish. One mine used FA with a gravimetric finish only for samples with a higher grade, while an AA finish was used for lower grade samples.

Two thirds of the mines indicated that the number of duplicate samples varied from "all mineralized samples" to "20 samples/month" to "10% of all samples", however, running 10% of the samples in duplicate seemed to be the most common. The duplicate samples were taken either in the field at the drill or in the laboratory at the end of sample preparation. No matter what sampling method is used, it is important to have set procedures in place for systematically monitoring the quality of the process. The simplest way is by collecting duplicate samples. Collecting duplicate field samples gives an indication of reproducibility of the original sampling; it gives an indication of precision of the entire sampling and assaying process. The precision measured includes any variability due to the method of sampling, the sample preparation process, the assaying method and the sample’s inherent variability. Taking duplicate samples during the sample preparation process (split rejects) only measures the variability due to the sample preparation procedure. Also taking duplicate samples from the pulp at the end of sample preparation measures only the variability due to the assaying process. The mine should also send standards to measure assay accuracy and blanks to measure contamination during the sample preparation process. Duplicate samples were almost always sent blind to another laboratory.

All the mines did check assays on sample rejects, typically 1 in 10 samples. Assaying in the US was performed by both company personnel and an independent laboratory while mines in Australia indicated that assaying was usually done by an independent laboratory.

Analysis of variance, plotting scatter grams of measurements from one laboratory versus measurements from another laboratory and correlation analysis were usually done to check validity of sampling and assaying practice. Samples were not usually tested to determine liberation characteristics (see Table 4).

Analysis to determine the statistical distribution of gold grade was always carried out. The most common distributions encountered were the normal and the lognormal.

Table 5 shows that one third of the mines indicated that a team consisting of a geologist, a mining engineer and a metallurgist/geotechnical engineer was responsible for the estimation of resources/reserves. The other two thirds used either a company engineer and a company geologist or a company geologist and an outside consultant. Most companies required a "competent" person to be responsible for the resource/reserve estimates. The majority of the mines indicated that they adhere to the Australasian Joint Ore Reserves Committee (JORC) Code to determine competency.

The computer programs used to perform resource/reserve calculations were a mixture of commercial packages and in-house software as indicated in Table 5.

Table 5 indicates that mines with a normal grade distribution tended to use ID methods, while kriging was used where the distribution was lognormal (Table 5)

The inverse distance methods were used the most in determining resources/reserves followed by geostatistical methods as seen in Table 5. ID squared, cubed, fourth and fifth were used. The reason given for using the ID method was mostly because of better reconciliation with production. What was unusual was that even mines using ID methods used variography to determine the search radius. The total number of samples used for estimation when using the ID method varied from 2-8, 6-30 and 4-10. Some mines used both geostatistical and ID methods. This gives the advantage of being able to compare methods, to see which one works best with reconciliation. It can also be used to point out inconsistencies in either method.

Multiple indicator kriging followed by ordinary kriging was most widely used. Statistical analysis to determine geostatistical domains was always done where geostatistical estimation methods were used. Indicator semivariograms were modeled the most followed by general and pairwise relative semivariograms. There was only one mine that modeled log semivariograms. It was interesting to find that some mines preferred using the general relative semivariogram while others used the pairwise relative semivariogram. The general relative and pairwise relative semivariograms are used when a proportional effect was present and the samples were preferentially clustered in areas with high values. The nugget to sill ratio for the variograms varied from 10-50%. The models were usually validated for change of support.

Confidence levels were usually not determined in the US (Table 6). There were only two mines in the US that used confidence levels, and even then only on the grade. Confidence limits in Australia were usually done on both the tons and the grade, 90% and 95% confidence levels were typical values. The confidence levels were determined:

• By looking at average distance of samples used to make the estimate
• By looking at the kriging variance
• As a percentage of the range.

The criteria used to categorize ore into the measured, indicated and inferred category were:

• Experience over time
• Drill hole spacing and search criteria
• The kriging variance
• Blocks categorized by % of range.

The JORC guidelines for resource/reserve reporting were most widely used.

Through the deck sampling is the most common method, the average blasthole spacing ranged from 14-20 ft (4.3-6 m) in the US. The sampling density is dependent on the blasting pattern used (Table 7)

The size of the samples taken is usually determined by "experience" (Table 8). Set procedures were used to reduce field size samples to a size suitable for assaying. This usually involved pulverizing to about –150 to -175 mesh. Fire assaying with either a gravimetric finish or AA finish was typically used. Duplicate samples were usually taken for assaying with about 10% of the samples run as duplicates. These samples were generally taken in the field at the drill or in the lab during sample preparation. 50% of the mines indicated that duplicate samples were generally not sent blind to another laboratory. More than half of the mines indicated that check assays were done on sample rejects.

Analysis of variance, correlation analysis and paired t tests were usually done to check validity of sampling and assaying practice. About 50% of the mines said that they tested samples to determine liberation characteristics. And about 80% said that they had done a gold content versus particle size distribution study on the blasthole cuttings. Assaying in the US was done by either both company personnel and an independent laboratory or only company personnel while mines in Australia indicated that assaying was usually done by an independent laboratory.

The ore-waste boundaries were usually determined using the polygonal method. Only three mines used ordinary kriging. One mine used conditional simulation. Local semivariograms were generally calculated if kriging or conditional simulation was used (Table 9).

Production results were always compared to estimates. The results were can be seen in

Table 10.

Table 10: Reconciliation results.

Number 1-7=US, Number 8-10=Australia

It seemed generally accepted that a difference of 10% between production (grade and tons) and estimates (grade and tons) is considered to be significant. According to the responses a difference between production and estimates for both tons and grade greater than 10% warrants investigation of:

• The data
• The geological model.
• Assaying techniques
• Sampling procedures.

Grade control was usually done by the company engineer in the US while in Australia it was usually done by the company geologist.

Most mines indicate that duplicate samples were taken during or at the end of the sample preparation process. This only measures the variability due to the assaying and sample preparation process. It does not say anything about the sampling process in the field. Taking duplicate samples during or at the end of the sample preparation procedure is not enough. A good quality control program uses:

• Field duplicates: measures total variability and thus the precision of the entire sampling, sample prep and assaying process.
• Lab duplicates (split rejects): these measure the variability due to the sample preparation process
• Check standards: measure accuracy.
• Blanks: measure contamination.

Although geostatistical methods are used in determining resources/reserves in the exploration stage, it seems that this has not contributed to better reconciliation with production (see Table 5 and Table 10). On the contrary it seems that those mines using geostatistical methods have a harder time getting good reconciliation.

ID to the fourth and fifth power seems to be used the most at the exploration stage. Why? Is it just fudging with the factor to get a better reconciliation?

1 The major change to be seen is that in 1988 most mines collected duplicate samples but no statistical analysis was done. Although the current survey does show that statistical analysis is carried out now, the locations in the sampling flowsheet where duplicate samples are taken needs improvement. Duplicate samples taken during or at the end of the sample preparation process only measure the variability due to the sample preparation and assaying process but say nothing about the field sampling procedure.

2. Another change is that most mines currently use geostatistical methods for resource/reserve estimation during the exploration stage where in 1988 the polygonal method was commonly used.
3. The polygonal method is still used the most in determining reserves in the production stage.

Kwa B.L and Mousset-Jones.P. 1988, Mineral Estimation of Gold Deposits—A Survey of Practices. Proceedings of Bulk Mineable Precious Metal Deposits of the Western United States Symposium. GSN, Reno, NV. Pp 691-718