Dig it: Mining models increase value
The inherent risks in the business prompted eight companies, representing half the mining capitalization in the world, to each invest $50,000 a year in Dimitrakopoulos' Stochastic Mine Planning Laboratory at McGill. One of them, BHP Billiton of Queensland, Australia, says its use of the Montreal lab's analysis system has led to a 5% to 20% improvement in the value of its mining operations. BHP Billiton is the second-largest producer of copper in the world and one of the top three nickel producers; it also mines aluminum, manganese, and titanium.
"Stochastic" means uncertain, and Dimitrakopoulos uses uncertainty as an integral part of his analysis models. In terms of mining, it works like this: At a mine, samples are taken at regular intervals and their yields used to build a model of the ore body. To derive an average yield from each sample would provide a more precise map of a mine, but the projected yield would never be exactly right. Dimitrakopoulos uses histograms, a form of charting that allows multiple data points to represent a sample rather than one, to show a range of possible values without specifying which is right.
The lab translates its research findings into Fortran or C++ functions that illustrate how to map uncertainty factors into existing mining systems. The mining companies either rewrite them into their own applications or use the lab's contributions as plug-ins to their existing software. Dimitrakopoulos is available as an adviser or consultant to train IT employees how to use the code.
He and the lab's staff use their knowledge of the occurrence of ores in geological strata, and the data from the samples provided by a mining company, to map the possible yield of a mine's ore body. For example, a few ounces of gold can be extracted from tons of copper ore. From a small-scale model, "we can infer the grades and percentage of recovery" of each metal in the process, he says. Then they try to match yields to markets, whose uncertainties they also include in their models.
"Booms in economies increase the demand for base metals," Dimitrakopoulos says. "China and India have absorbed a lot of raw materials, such as copper and iron, in recent years." Since 2003, steel prices have stayed high longer than economists had predicted, as construction in Shanghai and other Chinese cities consumes much of the world's output. Recently, prices have started to fall.
When a mine's management single-mindedly extracts the highest yields as fast as possible, the result may prove self-defeating. Mines can be left with surplus high-grade output that's gone down in value, because of a flood of such metal onto the market. The lab's models can recommend getting part of the output from more uncertain or low-yielding areas of the mine, matching low output to low market demand--or vice versa. Doing so increases the overall value of the mine, Dimitrakopoulos says.
A hidden cost factor, which also figures into the lab's equations, is the loss in value of metals that are extracted and then held off the market. With money inflating at a slow but steady rate, the loss in value of currency by the time the metal is sold may chew up the profit that existed at the time it was mined.
The mine-planning system helps companies meet their reporting requirements. "Every year, a mining firm reports to the stock exchange what the reserves are in the mine, what its potential is, and what production has been. These factors lead to an assessment of lifetime value," reflected in the company's stock price, Dimitrakopoulos says. To underestimate the full value of the mine is to rob the stock of part of its value.
"We understand the uncertainties of mining mineral deposits," Dimitrakopoulos says. His stochastic analysis systems help turn that uncertainty into gold for the mines' owners.