Have you ever wondered how polar bears survived the ice ages? Yes, ice ages! The question arises because scientists have found that when spring conditions are more than usually icy, fewer ringed seal pups—the bears' favorite food—are born. With less food available for the mother bears, fewer bear cubs are born and survive.
You might also ask: How did the ice-loving polar bears survive periods much warmer than we are currently experiencing—times when there was little or no ice around the Arctic basin and Hudson Bay area? The most recent such period occurred between 6,000 and 9,000 years ago and it was even warmer between 110,000 and 130,000 years ago.
The bears not only survived these periods of dramatically different climate and environment, but provided an invaluable source of food, clothing, and raw materials for tools and trade goods for peoples living in the Arctic regions. In more recent times, during the 1950s and 1960s in particular, hunting with the help of modern technology and in excess of subsistence requirements reduced the population to perhaps as few as 5,000 bears. As their survival as a distinct species for as long as 250,000 years suggests, however, polar bears are robust. Once hunting restrictions were enforced the population grew quickly and there are now estimated to be as many as 27,000 bears; enough of them to pose a danger to Alaskan townsfolk.
Given the historical facts about polar bears, then, we were surprised when we learned that a team of experts commissioned by the U.S. Geological Survey had predicted that the population of bears would fall by two-thirds by the year 2050. These predictions were made "...to Support U.S. Fish and Wildlife Service Polar Bear Listing Decision," under the federal Endangered Species Act. We wondered whether the bear experts' forecasting methods could be trusted.
Fortunately, the trustworthiness of the bear experts' forecasting methods is not just a question of opinion. Scientific research on forecasting has been conducted since the 1930s and has led to a set of evidence-based principles (rules or guidelines) that dictate which procedures are appropriate for the conditions. The forecasting principles have been published, and are easily available at forecastingprinciple.com.
Using an Internet search, we found roughly a thousand published papers that addressed the problem of forecasting polar bear numbers. None of them made reference to the scientific literature on forecasting. Most importantly, neither did the nine government reports prepared in support of the listing decision.
We judged two of the reports (Steven Amstrup was the lead author of one and Christine Hunter was the lead author of the other) to be the most relevant forecasting documents. Both Amstrup's and Hunter's forecasting procedures started with the assumption that the sea ice predictions from the General Circulation Models (GCM) that are favored by some climate researchers are valid. They are not. The Models do not constitute scientific forecasting methods and do not deal correctly with what is known about the physics of ice. Since the underlying assumption that the GCM sea ice forecasts are valid is false, the polar bear forecasts are of no value.
We nevertheless used forecasting principles to audit the forecasting procedures used by Amstrup and by Hunter in order to determine whether their procedures would be useful for making conditional forecasts of bear numbers. That is, what would be the bear population in 2050 if low ice conditions prevailed over the intervening decades?
Amstrup's forecasts were the product of a complex set of assumptions. We could not rate his procedures against 26 relevant forecasting principles because his report did not contain sufficient information. Of the 90 relevant principles against which we were able to rate Amstrup's procedures, 73 were violated. Some of the violations were sufficient by themselves to render the forecasts invalid.
One polar bear expert specified variables, relationships, and inputs. The same expert then made adjustments until the forecasts conformed to his expectations. In effect, then, Amstrup's forecasts were the opinions of a single expert (himself) unaided by forecasting principles. Much research has shown that unaided expert opinions are not valid for forecasting in situations with high complexity and much uncertainty, as is the case with the polar bear population.
Hunter's forecasts were also the product of a complex set of assumptions. Complexity makes errors hard to detect, and when knowledge of the situation is weak, erroneous assumptions multiply and lead to large errors. Hunter's procedures violated 80 out of the 105 principles against which we were able to rate them. Amazingly, Hunter and her colleagues extrapolated the polar bear population nearly 100 years into the future on the basis of five years of data. Even the five years of data were of doubtful validity.
As far as predicting the future of the polar bear population is concerned, the opinions of polar bear experts have no value without the aid of scientific forecasting procedures.
A decision to list polar bears as endangered would be expensive for the government authorities to police. It would also lead to new burdens for businesses and for people living in Alaska. Recently, villagers from the town of Noorvik, Alaska killed a polar bear that was threatening residents. Would they have been allowed to do this if polar bears were an Endangered Species? Would it be fair to ask people to abandon their homes if a polar bear decided to include them in its range?
Polar bears are magnificent creatures and the cubs look cute. But decisions about listing a species as threatened or endangered should not be based on emotional responses. Without scientific forecasts of a substantial decline in the polar bear population and of net benefits from feasible policies arising from listing them, a decision to list would be irresponsible.
The authors (JSA, KCG, WS) are the team of inter-disciplinary scientists that wrote the paper "Polar bear population forecasts: A public-policy forecasting audit" soon to be published. Draft copies are available at publicpolicyforecasting.com.