Saturday, April 4, 2015

IBM-Weather Company Alliance: The Benefits and Limitations of Weather-Driven Analytics


Make no mistake about it, combining IBM and The Weather Company (TWC) is about as good as it gets for companies looking to improve their responsiveness to today’s extremes of weather.  This is not just because IBM’s Watson Analytics promises to add in-depth business insights and additional predictability to your basic average weather forecast.  In fact, The Weather Company is if anything the more important partner, because it has established a solid track record for anticipating some of the more unusual effects of such events as Hurricane Sandy – and Jeff Masters of Weather Underground is an effective analyst with a good appreciation of the effects of climate change on weather patterns.
And yet, the recent announcement of the alliance also serves to show how today’s business analytics is severely handicapped in its ability to incorporate likely future changes in climate and their effect on business for all but the next few days.  To put it another way, IBM/TWC can help a Boston-based firm anticipate and adjust on the fly to a 2-foot snow 5 days from now; but to make the necessary repairs to buildings to avoid roof collapses and leaks, and to make plans to truck snow to more remote locations because local “snow farms” can’t handle so much snow, when there has never been a need before – well, that IBM/TWC probably won’t help with.   However, increasingly, that’s where much of the additional weather-related business cost these days resides.  (And of course, what goes for IBM/TWC goes for other analytics solutions in spades)
The real problem here is the lack of a global climate model that operates on the basis of the most likely climate changes over the next 30 years, not just the ones that can be verified by conservative scientific analysis.  To take one example, we know that melting permafrost is likely to have a significant effect in warming the globe by releasing carbon and methane over the next 30 years, but since we can’t rule out the possibility that it will have no effect, today’s climate models do not include permafrost melting effects. 
As a result, even if Watson were to incorporate such a model in its projections of weather over the next couple of years at business locations, it would be very likely to fail to anticipate the extremes of weather that would actually occur.  Thus, Heidi Cullen, another meteorologist savvy about climate change, wrote a book a few years back in which she projected that NYC would see a major hurricane with a dangerous rise in the water level that threatened the city – but with a near-miss in 2017 and with a greater Hurricane Sandy equivalent happening in 2041.

A Quick Fix


So what’s a poor business to do?  Well, as of now, someone within the organization needs to develop enough expertise in today’s climatology to be able to create an informal “model” based on current research that can project possible regional climate changes for business and sales locations over the next five years – assuming non-zero global-warming effects of “known unknowns.”  Then, the business needs to run “what-if” weather scenarios that include these new possible extreme weather events – high winds, new hurricane locations, extreme rain and snow, extreme drought – and also their ancillary effects – water shortages, stressed utilities, difficulties getting to stores and work, impoverished consumers.  TWC’s expertise would be critical in determining the “penumbra” of related weather – geographical size and track of an extreme storm, frequency/length of extreme droughts – while IBM’s analytics would scrutinize the unfolding scenario for business impacts of ancillary effects.  This approach would do a far better job of adequately preparing the firm for the medium term (say, 2-5 years). 
Ah, but then there’s the longer term, 5-35 years out.  What prepares the firm for medium-term adaptation may very well prove to be more costly than adaptation that gives up certain locations as lost causes – and far, far more costly than the firm taking its fair part in effective global mitigation.

Getting Specific


What follows is necessarily speculative and US-centric, but should give an example of both medium-term and long-term business thinking.  Today’s “new climate” as it plays out over the next 2 ½ years is characterized by a modified “el Nino/la Nina” cycle and a new Arctic impact on weather, especially in winter.  High temperatures along various parts of the Pacific have attenuated el Nino and delayed its onset, but over the next 2 years, and possibly for longer, we should be in an el Nino phase characterized by higher temperatures than normal, especially in the western US.  At the same time, extensive warming of the Arctic combined with a mild el Nino should lead to significantly higher Arctic temperatures in the summer (and winter), and may redirect the cold and “extreme snowfall” of this winter to the European side of the Atlantic.
However, at the same time, Greenland ice melting has been putting a fresh-water “cap” on the area where the Gulf Stream dives down to the deep ocean, slowing and weakening the “force of flow” of the Gulf Stream.  At the same time, the greater increase in winter temperatures in the Arctic due to global warming compared to the continental US has made it easier for Arctic cold to “punch through” the jet stream and park over the eastern US for weeks at a time.  The increased warmth of that Arctic air means increased incidence of “extreme snowfalls” either in the eastern US or (in the other phase of the “Arctic oscillation”) in northern Europe.
Combining the two effects above with ongoing global warming, and we might project winters similar to the last 2 in the eastern US, but with a shorter snow season and slightly warmer temperatures.  The southern and western US may see significantly higher summer temperatures, and we cannot anticipate that the drought in the California area will be overcome by enough rain to replenish reservoirs, especially with little or no snowpack in the Sierras.  Finally, we might anticipate that there may be another out-of-season Hurricane Sandy or 75-80-mph nor’easter in the Northeast, with higher-wind-speed hurricanes in the South.  Overall, the globe should see a faster increase in temperatures year-round, leading to record high temperatures over much of the globe.  Businesses should review not only their dependence on water but also the reliability and cost-effectiveness of their air conditioning.
Beyond that, in the next 5-35 years, another main actor is sea level rise.  It is plausible that on average, global sea level will rise almost a foot over the next 35 years, and the slowing of the Gulf Stream means that sea level rise along the East Coast may even double that.  Sooner than expected, then, cities such as Miami and perhaps even Boston are under threat, and increased wave levels during more-violent storms mean faster coastal erosion and destruction.  Businesses should rethink their investments in locations less than 25 feet above sea level, in areas of high drought (western and possibly southern US), and in areas requiring high levels of air conditioning costs.  Analytics such as that available via IBM/TWC plus “what-if” scenarios should allow fine-tuning these shifts in investments within regions.

The Bottom Line


One line credited, I believe, to an IBM employee way back when, said of programs in general what could well be said of “business as usual” weather-related analytics:  “Garbage in, garbage out.”  In other words, if you assume that future weather will be like the past, with today’s climate change, that assumption is untrue “garbage”; and the result of your analytics will be either a 5-day horizon that gives you that long to make major changes in your current setup at various locations, or a very misleading guide to your multi-year strategy’s chance of avoiding major unnecessary weather-related costs – a medium-term forecast that is “garbage” as far as effective planning is concerned.
I have suggested one approach to overcome this problem; but, really, the first step is to recognize that dealing “just in time” with a weather problem is far less than half the battle, and to do something about it.  If you do create an approach along the lines I have suggested, however, IBM/TWC is an exceptional partner, because TWC “gets” climate change and because IBM via Watson and the like can dig deeper for related business effects, once the data it is fed is no longer “garbage.”
One final thought:  I and others like Joe Romm have said that mitigation (avoidance of further carbon emissions and hence far greater climate change) was and is far less expensive than adaptation such as the steps we have talked about here.  Unfortunately, in the 5-7 years since this point was made, most businesses have failed to have much positive impact on mitigation, either in their investments or politically.  As a result, the costs of mitigation have not increased significantly, but the costs of the adaptation that has now become necessary have risen quite a bit.  To put it another way, ignoring mitigation has guaranteed much more in the way of adaptation costs – and we may anticipate the same happening over the next 5-35 years, unless most businesses change both their political and investment practices in a much more drastic way.