MIT’s un­sci­en­tific, cat­a­strophic cli­mate fore­cast

National Post (Latest Edition) - - FINANCIAL POST - KESTEN C. GREEN J. SCOTT ARM­STRONG Fi­nan­cial Post

When we drive on a long bridge over a river or fly in a passenger air­craft, we ex­pect the bridge and the plane to have been de­signed and built in ways that are con­sis­tent with proven sci­en­tific prin­ci­ples. Should we ex­pect sim­i­lar stan­dards to ap­ply to fore­casts that are in­tended to help pol­i­cy­mak­ers make im­por­tant de­ci­sions that will af­fect peo­ple’s jobs and even their lives? Of course we should. Such stan­dards ex­ist. But are they be­ing fol­lowed?

The Fi­nan­cial Post asked us to look at a re­port last month from the Mas­sachusetts In­sti­tute of Tech­nol­ogy (MIT) Joint Pro­gram on the Sci­ence and Pol­icy of Global Change, ti­tled “Prob­a­bilis­tic Fore­cast for 21st Cen­tury Cli­mate based on un­cer­tain­ties in emis­sions (without pol­icy) and cli­mate pa­ram­e­ters.”

The MIT re­port au­thors pre­dicted that, without mas­sive gov­ern­ment action, global warm­ing could be twice as se­vere as pre­vi­ously fore­cast, and more se­vere than the of­fi­cial pro­jec­tions of the United Na­tions’ In­ter­gov­ern­men­tal Panel on Cli­mate Change (IPCC). The MIT au­thors said their re­port is based in part on 400 runs of a com­puter model of the global cli­mate and eco­nomic ac­tiv­ity.

While the MIT group es­pouses lofty-sound­ing ob­jec­tives to pro­vide lead­er­ship with “in­de­pen­dent pol­icy anal­y­sis and pub­lic ed­u­ca­tion in global en­vi­ron­men­tal change,” we found their pro­ce­dures in­con­sis­tent with im­por­tant fore­cast­ing prin­ci­ples. No more than 30% of fore­cast­ing prin­ci­ples were prop­erly ap­plied by the MIT mod­ellers and 49 prin­ci­ples were vi­o­lated. For an im­por­tant prob­lem such as this, we do not think it is de­fen­si­ble to vi­o­late a sin­gle prin­ci­ple.

For ex­am­ple, MIT fore­cast­ers should have shrunk fore­casts of change in the face of un­cer­tainty about pre­dic­tions of the ex­plana­tory vari­ables; in this case the vari­ables pos­tu­lated to in­flu­ence tem­per­a­tures. More gen­er­ally, they should also have been con­ser­va­tive in this sit­u­a­tion of high un­cer­tainty and in­sta­bil­ity. They were not.

We rec­og­nize that judge­ment is re­quired in rat­ing fore­cast­ing pro­ce­dures. Ev­i­dence for our prin­ci­ples, how­ever, is in the form of find­ings from sci­en­tific ex­per­i­ments com­par­ing rea­son­able al­ter­na­tive meth­ods, and ac­cepted prac­tice (see link be­low).

So what’s re­ally wrong with the MIT re­port? The phrase “global en­vi­ron­men­tal change” pro­vides a clue. The group’s ob­jec­tive im­plic­itly re­jects the pos­si­bil­ity of no or unim­por­tant change or, de­spite men­tion of un­cer­tain­ties, the pos­si­bil­ity of un­pre­dictable change. Peo­ple who do re­search on fore­cast­ing know that a fore­cast of “no change” can be hard, if not im­pos­si­ble, to beat in many cir­cum­stances. A fore­cast of no change does not mean that one should nec­es­sar­ily ex­pect things not to vary. Such a fore­cast can be ap­pro­pri­ate even when a great deal of change is pos­si­ble but the di­rec­tion, ex­tent or du­ra­tion is un­cer­tain.

When one looks at long se­ries of Earth’s tem­per­a­tures, one finds that they have gone up and down ir­reg­u­larly, over long and short pe­ri­ods, on all time scales from years to mil­len­nia. More­over, sci­ence has not been able to tell us why. There is much un­cer­tainty about past cli­mate changes and about the strength and even di­rec­tion of causal re­la­tion­ships. To wit, do warm- ing tem­per­a­tures re­sult in more car­bon diox­ide in the at­mos­phere or is it the other way round — or maybe a bit of both? Does warm­ing of the at­mos­phere re­sult in neg­a­tive or pos­i­tive feed­back from clouds? There are many more such ques­tions without an­swers. All this strongly sug­gests that a no-change fore­cast is the ap­pro­pri­ate bench­mark long-term fore­cast.

With Dr. Wil­lie Soon of the Har­vard-Smith­so­nian Cen­ter for As­tro­physics, we found that sim­ply pre­dict­ing that global mean tem­per­a­tures will not change re­sults in quite small fore­cast er­rors. In our val­i­da­tion study that cov­ered the pe­riod 1851 to 2007, we com­pared the no-change fore­cast with the IPCC global warm­ing fore­cast that tem­per­a­tures will climb at a rate of 0.03C per year. We com­pared the IPCC pro­jec­tion of 0.03C per year with what ac­tu­ally hap­pened af­ter 1850. The er­rors from the IPCC pro­jec­tion were 12 times larger than no-change bench­mark. Con­sider the ac­cu­racy of the no-change model: On av­er­age the 50-years ahead fore­casts dif­fered by only 0.24C from the global mean tem­per­a­ture as mea­sured by the Hadley Cen­tre in the U.K.

Based on our anal­y­sis, we ex­pect the an­nual global mean tem­per­a­ture for ev­ery year for the rest of the 21st Cen­tury to be within plus-or-mi­nus 0.5C of the 2008 mean.

The MIT ap­proach to fore­cast­ing is in sub­stance the same as the ap­proach adopted by the IPCC. Our fore­cast­ing au­dit of the IPCC ap­proach and its con­clu­sion there­fore ap­plies as well to the MIT fore­cast­ing ef­fort: The fore­cast­ing pro­ce­dures were not valid and there is no rea­son for pol­i­cy­mak­ers to take their fore­casts se­ri­ously. It also leads to the con­clu­sion that the MIT fore­cast er­rors will be much larger even than the IPCC’s fore­cast er­rors.

Pol­i­cy­mak­ers and the pub­lic should be made aware that the fore­casts from the MIT mod­ellers, as well as those used by the IPCC, are merely the opin­ions of some sci­en­tists and com­puter mod­ellers. It is not proper to claim that th­ese are truly sci­en­tific fore­casts.

Dr. Kesten C. Green is a se­nior re­search fel­low of the Busi­ness and Eco­nomic Fore­cast­ing Unit at Monash Uni­ver­sity in Aus­tralia. Dr. J. Scott Arm­strong is Pro­fes­sor of Mar­ket­ing at The Whar­ton School, Uni­ver­sity of Penn­syl­va­nia. Arm­strong and Green are co-direc­tors of the pub­lic ser­vice Web site fore­cast­ing­prin­ci­ples.com spon­sored by the In­ter­na­tional In­sti­tute of Fore­cast­ers.

Comments

Newspapers in English

Newspapers from Canada

© PressReader. All rights reserved.