Читать книгу The Best-Laid Plans - Randal O'Toole - Страница 8

2. Garbage In, Gospel Out

Оглавление

Writing a 150-year plan (or even a 10-year plan) for a 1.5- to 2.0-million-acre forest that produces a wide variety of goods (and bads) that often compete and conflict with one another requires more than the back of an envelope. In the 1970s and 1980s, managers of both public and private forests increasingly turned to computers to help plan their lands, schedule timber harvests, and coordinate resource management.

A number of competing computer models were available for forest planning, and some might have been more appropriate for certain forests while others would work best on other forests. But as part of its growing centralization, the Forest Service directed all forests to use the same computer program, which was called FORPLAN, short for forest planning. FORPLAN allowed planners to enter information about the forest into the computer and then ask questions, such as “What is the maximum amount of timber that can be cut?” or “How much timber would be cut each decade if the forest were managed to earn maximum profits?” Forest officials gushed that FORPLAN would simultaneously allocate land and schedule timber cutting for the next 150 years.1

To build their FORPLAN models, planners would break their forests up into hundreds of different zones based on such factors as vegetation, the age of the timber, wildlife habitat, steepness of slope, whether the zone had roads, or any other criteria that seemed important. For each zone, planners had to identify management costs, timber values, timber yields, and the values and yields of other resources such as recreation, water, forage, and specific species of wildlife. Planners would set a goal such as maximizing timber or net economic value and could also set constraints, such as floors or limits on the amount of timber that could be cut. They could then give FORPLAN a goal, such as maximizing timber or profits, and it would allocate zones to timber, recreation, and other prescriptions and tell how much timber could be harvested from the forest for each of the next 15 decades.

As parodies of soviet planning, the forest plans were quite humorous. The data used in the models were often erroneous or fabricated. Many plans assumed, for example, that timber was worth 50 to 100 percent more than timber companies were actually paying for it. Others assumed that trees could grow far faster than is realistic. One actually projected that trees could grow 650 feet tall, nearly twice the height of the tallest trees in the world.

Many people described FORPLAN as a black box, that is, a machine whose inner workings were too complicated for most people to understand.2 Technically, FORPLAN used linear programming methods to find the optimal solution to any problem the planners gave it. But FORPLAN’s inner workings were much less important than the quality of the data planners entered into the computer. As an outsider, I suspected that the Forest Service would bias FORPLAN models toward timber, and I set a goal of reviewing at least a third of the plans to find out if this was true. Ultimately, I collected and read every draft and final plan and reviewed the actual FORPLAN computer runs and background data for well over half the 120 forest plans.

To write the forest plans required by the National Forest Management Act, the Forest service hired hundreds of recent graduates in economics, planning, operations research, and other technical fields. These people enthusiastically and often idealistically embraced the opportunity to prepare objective plans that would determine the future of nearly 10 percent of the nation’s land. Almost immediately, however, they ran into serious obstacles.

Data collection is one of the most important early steps in any planning process. Forest service rules required planners to use the “best available data”—but the emphasis was on available. Historically, the Forest service usually did a complete forest inventory before each 10-year timber management plan. An inventory would not measure every tree in a forest but would measure randomly or systematically selected plots scattered across the forest. In one common inventory procedure, one plot was measured for every 1,850 acres, so each plot was assumed to represent that many acres. If 10 plots were found to have 100-year-old Douglas fir trees, planners assumed the forest had 18,500 acres of 100-year-old Douglas firs.

Inventory specialists planned to measure the same plots every 10 years, providing information on how fast trees were growing and other changes in the forest. Reinventories also made it possible to identify and correct any errors in the previous inventory. In the 1970s, managers of one Oregon forest realized that one of its inventory crews had made a serious mistake in the 1960s: Contrary to directions, if a plot fell in a meadow or a lake, they moved the plot to the nearest forest. This led managers to underestimate the number of acres of meadows and lakes and overestimate the number of acres of productive forest.

Given planning deadlines and the fact that they were spending so much money on computer runs and newly hired experts, forest planners in the 1980s were rarely able to do new inventories. So they relied on data that were anywhere from 10 to 30 years old. These data were updated by subtracting the volume of timber cut in that time and adding the amount that planners thought trees would grow in that time. Obviously this meant they had no opportunity to correct errors in earlier inventories or their growth projections.

The few forests that did new inventories often took shortcuts to save time and money. Previous inventories collected a huge amount of data, including the height, age, diameter, and species of every tree in each plot, plus more general information such as the steepness of the slope, the direction the slope faced, and the species of shrubs growing under the trees. The computer age is supposed to enable people to consider and analyze ever-greater quantities of data. But FORPLAN could deal with only a limited number of variables, so planners decided not to collect any data FORPLAN couldn’t handle. This saved money in the short run, but reduced the reliability of the inventory and made it impossible to compare the inventory results with any future inventories that did collect more data.

Other forests completed their reinventories only after forest planning was well under way. A reinventory of Oregon’s Malheur National Forest found that trees measured in the previous inventory subsequently “shrank” in both diameter and height. This prompted speculation that the person in charge of the previous inventory had inflated the numbers to get answers more in keeping with the Forest Service’s timber goals.3 Since the reinventory was completed in the midst of forest planning, planners continued to use the older discredited data in FORPLAN.

Given information, however unreliable, about how much timber was standing in the forest, the next question planners had to answer was how fast trees could grow. Under the nondeclining flow policy, forests that had lots of old-growth timber couldn’t cut that timber any faster than the next generation of trees could grow. So second-growth yield tables that projected rapid growth allowed for more cutting of old-growth trees today.

The first plan I reviewed was for the Okanogan National Forest. Though located in arid eastern Washington, it based most of its growth projections on yield tables written for moist western Washington, which receives as much as four times the rainfall. Timber inventory data collected by the forest revealed that Okanogan growth rates were only about 60 percent of the rates projected by the western Washington yield tables.4

The Santa Fe National Forest itself discovered that actual timber volumes were only 80 percent of the numbers it had entered into FORPLAN. Rather than reenter all the yield tables, it decided to simply reduce the timber harvests proposed by FORPLAN by 20 percent. This seemed simple enough—except that planners asked FORPLAN to maximize the forest’s net economic value. Given the overestimated volumes, FORPLAN calculated that timber cutting was more lucrative than it really was. The higher volumes made it appear that only 11 percent of the forest would lose money on timber sales. I found that correcting the volumes increased this to 48 percent.5

Some forests had already cut much of their old growth, so—if you believed their second-growth yield tables—the main limiting factor to timber-cutting levels was the growth rate of the old growth. California national forests used yield tables that stunningly predicted old-growth forests would double in volume in as little as 20 years.6 Since old growth is normally considered to grow very slowly, these predictions were not credible and greatly distorted the plans.

Other national forest yield tables were even more absurd. University of Montana forestry professor Alan McQuillan found computer-generated yield tables used by Idaho’s Clearwater National Forest that predicted trees could grow 650 feet tall in 150 years.7 That’s nearly twice as tall as the tallest tree in the world and close to three times as tall as the tallest trees in Idaho.8 No one on the forest noticed the error, and planners didn’t correct it after McQuillan pointed it out.

When they weren’t overestimating timber growth rates, many planners overestimated timber prices. Timber prices had risen rapidly during the 1970s, partly due to speculation fueled by contracts that allowed purchasers to pay for timber up to five inflation-filled years after they bid.9 Many forests presumed that prices would continue to rise at similar rates for the next 50 years. Since FORPLAN did its calculations in decades, planners applied the prices predicted for the midpoint of each decade to that entire decade. Yet planning was taking place during the deepest recession in the second half of the 20th century, and actual prices had crashed well below the levels of the late 1970s. So planners found themselves in the odd position of using prices for the first decade that were higher than the forests had ever received at a time when actual prices were their lowest in years. Average FORPLAN prices were often two or three times actual prices, and in some forests they were more than five times as high as the highest amounts ever paid for timber on those forests.10

Even in the long run, planners’ predictions of future timber prices were unlikely. The price surge of the 1970s was due to timber purchasers’ discovery that, in an inflationary era, they could take advantage of Forest Service timber contracts that gave them years to cut and pay for the trees without indexing the price they paid to lumber or other wood values. At 5 percent inflation on a five-year contract, purchasers could bid 25 percent more than the trees were worth at the time of sale and still make a profit. This led to furious speculation as some companies that didn’t even have processing facilities started bidding in anticipation of reselling the logs to some other buyer several years down the road.11

All of this came to a screeching halt when the Federal Reserve Board raised interest rates enough to curb inflation, effectively shutting down the homebuilding industry in the process. Purchasers found themselves holding contracts for trees they couldn’t afford to cut at the time they bid on them, much less after interest rates rose. But in the world of FORPLAN, many forest planners pretended that the party would continue another 50 years.

While timber price trends—an assumption of increasing timber prices over time—were built into many FORPLAN models, costs and competing resource values were usually assumed to remain constant. In reality, costs were rising and recreation values had historically risen even faster than timber values. Forest planners in Oregon and Washington were so disturbed by this “strong and unjustifiable bias” toward timber that they wrote a memo to the Washington office arguing that the trends were “unstable” and of questionable accuracy.12 The Washington office replied that “it would be an error to ignore the substantial documentation of long-term real prices” for timber and ordered them to use the trends.13 Of course, the reason the Forest Service had documentation of future timber prices and none for the future prices of other resources was that the agency had directed its researchers to study the one and not the others.

Even if the researchers’ work was valid, many forests misapplied it. For example, the research predicted that lodgepole pine prices in western Montana would reach $116 per thousand board feet by the year 2030. The Bitterroot National Forest told FORPLAN that this value would exceed $350 by the same year.

Rather than justify timber sales using overestimated timber yields or prices, some forests relied on the relationships between timber and other resources. Typically, they would tell FORPLAN that recreation was very valuable, but that no recreation could be produced without roads. The only way FORPLAN could get roads was by cutting trees. As one forest planner derisively wrote, “The mechanic has just modeled a nation of people who like to camp in clearcuts!”14 Timber values on some of these forests were so negative that, without the recreation values, FORPLAN would cut no timber. But with recreation, FORPLAN would cut over the entire forest.

Research showed that, on virtually all forests, the use of roaded recreation at the current price—which was zero—was far less than the supply offered by the 300,000 miles of roads that laced through the forests. The recreation that was in short supply was wilderness and roadless recreation. Yet some planning teams designed FORPLAN to calculate that even wilderness recreation would increase after more roads were built.

One of the first forests to include these recreation-timber relationships in FORPLAN was Indiana’s Hoosier Forest. The formal documentation included a table claiming that the demand for roaded recreation greatly exceeded both the supply and current usage. Since demand (technically, the quantity demanded) cannot exceed use when the price is zero, I asked to see all the background documents relating to recreation demand. The planners brought me a stack of papers about three feet high.

Reading through the papers, I found several earlier demand projections that were much lower. Near the bottom of the stack, I found a memo from the regional office in Milwaukee questioning the later, higher demand numbers. A note was written on the memo in hand-writing that I recognized as belonging to the forest’s recreation specialist. “I would agree that my 7/7/82 calculations are high!” said the note. “I was told by the forest planning team to make sure that demand was higher than our capability. I did as I was told.”15 By that time, he had transferred to another forest and nervously refused to talk with me on the telephone about the demand figures.

Two years later, when I was reviewing the Mark Twain Forest plan in Missouri, I found a memo from the Eastern Regional Office of the Forest Service to all forests in the Midwest and Northeast. The memo said that the Hoosier Forest had found a way to make FORPLAN cut trees even where timber sales lost money: simply tell FORPLAN that there was a huge, unmet demand for roaded recreation. Samuel Clemens no doubt would find it amusing that the Mark Twain Forest joined most of the other forests in the region in following the Hoosier’s example.

Planners’ assumptions about the relationships between timber and wildlife were also often questionable. The grizzly bear, a threatened species in Montana, is vulnerable mainly to humans. To minimize conflicts, biologists recommended isolating the bear by building few new roads and closing existing ones. So the Gallatin National Forest, which borders Yellowstone Park, put over 120,000 acres in “grizzly-timber emphasis.” This prescription called for building new roads, selling timber, then using revenues from the timber sales to close both the new and the existing roads.

In practice, however, Gallatin timber values were so low that few, if any, timber sales sold by the forest generated enough revenues to close any roads.16 The result was that each new timber sale built more roads that increased the vulnerability of the grizzly to humans. Planners didn’t consider the possibility of simply closing existing roads, which would have cost less than the subsidized timber sales, because Congress freely appropriated funds for timber sales but disliked funding road closures.

The Flathead National Forest, which borders Glacier National Park, went a step further and presumed that a mere decision not to build roads in a roadless area would lead to a 20 percent increase in grizzlies, thus allowing more timber cutting somewhere else without threatening overall grizzly populations.17 Seemingly, forest planners believed that grizzlies would read the plan and increase their numbers as soon as the decision was made to not build roads.

Another important question was how much land was suited for timber management. Congress had directed the Forest Service to exclude from timber management land that was not physically or economically suited for timber production. The Forest Service neatly sidestepped the question of economic suitability through its definition of the term. Each forest was given a timber target based on what the forest had previously produced or what forest managers thought—before doing any economic analyses—it could produce. Economically suitable land, said forest planning rules, included all the land needed to meet that target. That this might require taxpayers to lose money on most or all of a forest’s sales was irrelevant to the Forest Service.

Questions of physical suitability were often sidestepped as well. Planners on the Mt. Hood National Forest carefully studied 141 randomly selected plots and calculated that 16.4 percent of the land that had previously been included in the forest’s timber base was basically just rockslides and talus slopes, incapable of producing any wood at all. Since these areas were too small to map in FORPLAN, they asked for permission to reduce yield tables by 16.4 percent. However, regional timber staff approved only a 10 percent reduction, suggesting that wishful thinking prevailed over the analysts.18

The problems with the data used in FORPLAN were pervasive, yet Forest Service officials based important decisions on how to manage land and how much timber to cut on this computer program. One Forest Service economist told me that his FORPLAN motto was “Garbage in, gospel out.”

Most of these errors could be explained away by bias on the part of Forest Service officials. They do not necessarily prove that there is anything wrong with the concept of planning itself. Nor do they explain the fake forests uncovered by Cherry DuLaney. But it turned out that both the fake forests and the garbage-like data in the computer models were closely related.

The Best-Laid Plans

Подняться наверх