Recent history reveals trends in mine failures

“I’ve heard they’ve got 15 million ounces of gold,” the broker said. “The news is all over the Street.”

In the heady days of 1995 and 1996, it was something you heard all the time: ounces, ounces, ounces. Every gold deposit had to have them — preferably a lot of them — and potential for about twice as much again.

Resources and reserves were the measure of a gold company’s worth.

Market cycles in the exploration business are all the same, yet all different. They show many of the same broad patterns — the bargains that can be picked up early in the bull phase, the big discovery that proves the recovery is real, the hype-driven plays near the top. But they differ in detail — the fashionable commodity, the location of the land plays, and most of all in the expectations of the investors.

John Felderhof of Bre-X Minerals was the icon of the 90s boom, and his promise of “30 million ounces, plus, plus, plus,” rings in the ears of anyone who followed the market over those strange three years. And many trace the bust that followed to the revelation that Bre-X’s Busang deposit was a fraud.

It is intriguing, but perhaps not surprising, to note that Strathcona Mineral Services was the firm responsible for shutting down Bre-X Minerals.

Turn back the clock a decade, and you find Strathcona probing the wreckage of almost a score of gold mining operations that had failed during the flow-through years.

The boom, and subsequent bust, of the late 1990’s gold-mining stock market had a theme: your value was in the ground, in long mine lives, and in massive resource estimates. In contrast, 10 years before, gold miners were expected to be in production, and fast.

Graham Clow, now a senior vice-president of operations at Breakwater Resources, worked for Strathcona at the time and analyzed the failures in a 1991 article published in The Northern Miner Magazine. Clow’s observation — which put him at odds with many observers at the time — was that the deadline-driven developments of the flow-through years only aggravated the fundamental problem, which was poor mine planning. “It has become fashionable,” Clow wrote, “to blame the flow-through tax rules for the excesses that led to the mine development failures of the 1980s. . . . All of these failures came about because the unit cost of production was too high.”

High costs, and two closely related problems — shortfalls in grade and in recovery — certainly did grip the small gold producers of the 1980s like a cold hand. There were the dead: D’Or Val’s Beacon, where production lasted less than a year before being suspended due to low head grades; Seabright’s Beaver Dam, which went from initial production to remnant mining in the space of a year; Noramco’s Golden Rose, which closed after 13 months of making what was then termed a “significant loss”; Mirado, whose low head grades drove Golden Shield into bankruptcy; and Wrightbar, written off by Belmoral Mines after minimal production.

Then there were the wounded: Skyline’s Johnny Mountain, which produced for less than two years at low recoveries; Coxheath’s Tangier mine, where grades and tonnages shrank drastically as soon as the project was in commercial production; Canamax’s Ketza River and Corona’s Nickel Plate, both of which suffered from astronomical costs; Granges’ Tartan Lake, which never met its production targets; and Pioneer Metals’ Puffy Lake, which never even went into commercial production.

Clow’s study of flow-through pathology found that one common thread was over-optimistic ore reserve estimates, and his conclusion would seem to be borne out by a widespread phenomenon in these projects, the grades and tonnages that headed downward much faster than the mine could ever hoist ore.

Beaver Dam, for instance, reported an average grade of 24.7 grams gold per tonne before it went into production in 1988; a year later it was mining remnants at 4.9 grams. During 1989, Tangier saw its reserves shrink to 122,000 tonnes grading 7.2 grams from more than 500,000 tonnes grading 10.3 grams the year before. Tartan Lake’s average grade fell by a third in two years, taking much of the tonnage with it.

The 1990s, on the other hand, are less likely to be remembered for their production excesses as for the ounces of gold that never came out of the ground. Busang, of course, was the prime example, a confidence game that found willing marks in the investment houses of Toronto and New York; but it was the big wheel at the end of a long counter of pseudo-projects that — in retrospect — showed how irrational the exuberance had become. There was Timbuktu Gold, run by a man who was already persona non grata on the U.S.

securities markets; Naxos Resources (NAXOF-O), with its “unassayable” gold; and the egregious Delgratia Mining.

But beside the frauds and farces stand a number of honest failures that, like those of the 1980s, could leave an investor wondering, like Robert Browning, “what’s become of all the gold?”

There were a number of factors that helped to sink Pegasus Gold (PGU-M), and one of those had to be the disaster at its Mt. Todd gold project, south of Darwin, Australia. An Australian subsidiary of Pegasus, Zapopan, had developed Mt. Todd in 1993, and Pegasus, predicting great things for the deposit, had funded an expansion that was to include a large new mill. In 1994, Mt. Todd produced 61,000 oz. of gold, but at such a high cost that the project was never placed in commercial production. The following year, production was up to 70,000 oz. — still without an operating profit — and Pegasus gave the new mill the go-ahead, stating confidently that commercial production would run at an annual rate of 260,000 oz.

Production in 1996 fell to 62,613 oz. gold, as cash costs swelled to US$370 per oz. And still the mine had not met its cost forecasts, and was not in commercial production. Ultimately that never came: to the tune of a US$353-million writedown (later increased to US$398 million), Mt. Todd was placed on care-and-maintenance in November 1997. Two months later, so was Pegasus, which applied for protection under Chapter 11 of the U.S.

Bankruptcy Code.

The ordeal was not over even then. Pegasus Gold Australia, the unit that operated Mt. Todd, sued BKK, the contractor that had designed and constructed the project’s Phase II Mill. Pegasus Australia charged that BKK, a joint venture of Bateman Kinhill and Kilborn Engineering Pacific, had not met the specifications of its own feasibility study on Mt. Todd.

At least Mt. Todd’s putative reserves grew during that painful period. Not every operation could say the same: Hecla Mining (HL-N) and Great Lakes Minerals got an unpleasant surprise from a large hole they dug at Grouse Creek in Idaho. Hecla had brought the Grouse Creek mine into production with a reserve of 13.7 million tonnes grading 1.9 grams gold and 37.7 grams silver per tonne. The millhead grades from early stages of the open pit met expectations, but once the operators went after three higher-grade benches in the pit things started to go very wrong.

The three new benches produced only about half the gold that had been predicted by geostatistical modelling of the mineral deposit. Drill spacing (on 20-metre centres) proved to be too wide to catch discontinuities in the vein system. By the end of 1995, Hecla was quoting a reserve estimate of just 6.25 million tonnes grading 1.5 grams gold and 42.8 grams silver per tonne — less than half the tonnage with which it had started the year.

Rea Gold, now in bankruptcy, had a similar experience with its Mt. Hamilton gold project near Ely, Nev. Mt. Hamilton was actually two gold deposits, Centennial and Seligman, both of which had elements of structurally and stratigraphically controlled gold mineralization.

After a positive feasibility study that confirmed reserves of 8.2 million tonnes grading 1.8 grams gold and 13 grams silver per tonne, Rea started trucking ore from an open pit on Seligman in September 1994 and reached commercial production five months later.

Production never met the feasibility study’s forecast; first, there were proble
ms with trucks and scheduling, but it ultimately came to light that the biggest problem was a shortage of ore. The structure of the deposit had been misinterpreted and geological models of the Seligman deposit had overestimated both the grade and tonnage in the pit. Costs never got under control, and before the Centennial pit could be put into production, Rea declared bankruptcy.

How much of the blame for this decade’s failures lies at the door of poor reserve estimation? The point of estimating a reserve is to predict the grade of ore that will be delivered to the mill, and low head grades were once again at the centre of most projects’ problems.

Resource estimates at both Mt. Todd and Grouse Creek were developed using geostatistics, a methodology that uses specific mathematical formulas to predict grades from drill hole data. Geostatistics, developed in the 1950s and 1960s in France and South Africa, departed from the conventional discipline of mathematical statistics but found success in the mining industry because it offered the geologist a way to calculate grades, in theory at least, anywhere he didn’t have a drill hole.

Its popularity grew in the 1980s, as computing power became widely available, and it was taken up by a large number of engineering consulting firms. That tide coincided with the development of a large number of gold mines in Canada — many of which were the notable failures of the decade.

At the time, some observers fingered overzealous computer modelling of the orebodies as the culprit, or at least as the accomplice of starry-eyed resource estimators. Clow, while recognizing that computer modelling played a part in the debacles, saw that, in many cases, production decisions were based on “reserve” figures that did not take the minability of the mineralized zones into account. Nobody blamed geostatistics itself, but it was the technique of choice at many of the mines, as it was once again at Mt. Todd and Grouse Creek.

Little public information exists on how well production at failed mines was reconciled with geostatistical ore reserve estimates at the stope or bench level; but it is clear that the shrinking tonnages at Grouse Creek and the chronically low grades at a number of other gold deposits were a technical failure at the mine scale.

There were also deposits that simply never got to a point where they could be called failures yet still cost their operators money. The Mt. Milligan deposit northeast of Ft. St. James, B.C., was acquired for a cool $336.8 million by Placer Dome (PDG-T) in 1990, through a buyout of Vancouver-based Continental Gold and the acquisition of a 30% interest held by BP Minerals.

A prefeasibility study in 1991 estimated the minable reserve at Mt. Milligan at 299 million tonnes with grades of 0.22% copper and 0.45 gram gold per tonne.

The grades were a little thin, but the cost estimates were low, so planning went ahead; but by early 1992, Placer realized it was walking into a trap.

The investment was written down and plans were put on hold and ultimately abandoned.

Busang was a fraud, of course, but it also ranks as a technical failure because (whatever the Statements of Defence filed in recent lawsuits may say) it clearly could have been caught by a proper technical review.

Ultimately it was. Colin Jones, formerly of Freeport-McMoRan Copper and Gold (FCX-N), made that clear in an April 1998 paper on Freeport’s due diligence review of the Busang project.

“Database verification, assay variance work, and the site technical audit work highlighted serious discrepancies and concerns from the start,” wrote Jones. Later, the results of Freeport’s own sampling filtered in from the laboratories, which “showed almost no gold occurred in the core.” Similarly, one of the other rivals in the battle for control of the “world’s largest gold deposit” retained statistician Jan Merks to examine the data set it had been given to review. Merks concluded, from an examination of the data alone, that there was no convincing statistical evidence that a deposit actually existed at Busang. Yet geostatistical resource estimates made by Bre-X’s consultants, Kilborn Pakar Rekayasa, missed the problems; and the results of the mathematical models were also confirmed by Freeport.

In his 1991 article, Clow cautioned that “even the most ardent proponents of geostatistics will concede that their science encounters the most difficulty when applied to vein-type gold deposits.” But why is that?

One of the effects of geostatistical ore reserve estimation, pointed out in a number of technical papers by Merks, is that the distance-weighted, or “kriged,” estimate of grade at a given point converges to the mean grade of the data set at great distances. The consequence is that if results from widely spaced drill holes are used to predict grade at points between them, the point will inevitably be identified as mineralized whether or not the mineralization is actually continuous between the two drill holes.

In turn, this makes geostatistics, and other techniques of resource estimation that rely on extrapolation and interpolation rather than direct sampling, a powerful generator of tonnage and of “ounces in the ground.” This played into the theme of the 1990s gold boom, which rewarded the generation of reserve ounces rather than the ability to mine them productively and cheaply.