By Matthew Owens March 13, 2013
March 6 was a snow bust for the D.C. area and is one of those events that leaves the less-informed grumbling about how stupid weather forecasters are. These grouchy skeptical people would say: If these "stupid" forecasters and their "stupid" forecasts can't even get something right that's a day away, how can we expect them to get something right like global climate - years from now!? After all, doesn't weather build on itself, leading to increased uncertainty the longer the projection is?
In short, such skeptics are blind to the difference between precision and accuracy. Let me give an all-American example to demonstrate the difference: You're on the shooting range, with your handy 22 rifle. You're using your fancy new scope, aiming a long distance down the field at your paper target. You pull the trigger and *pow!* you've put a nice little hole in the bright orange center - it's a bullseye! But the guy shooting next to you marches over and taps you on the shoulder "Hey! What do you think you're doing shooting my target?!" Your shot was extremely precise, hitting the bullseye as you intended. But it was terribly inaccurate, because you in fact were shooting the wrong thing. Precision is the spread of your results - so if you shoot 10 shots and they all hit the bullseye, your precision is excellent. And even if you shoot 10 times and you miss the bullyseye each time, but the grouping of your shots is still in a tight cluster, your shots are still precise, but not at all accurate (e.g. you've shot the other guy's target). Accuracy is how aligned your estimate or result is with the correct, true one.
So the weather forecast (not the snow forecast) was actually very accurate and it was on target: it got the fact that there would be a large precipitation event over the D.C. area, and that temperatures would be very close to freezing. It didn't predict sunny skies, or for the storm to hit somewhere else, at a different time. So in fact, from a certain perspective it was actually also very precise. But like all weather forecasts, it's subject to high levels of hyper-local imprecision: the variability that happens on hourly time-scales and within the geographic resolution of a few tens of miles. This is especially amplified with snow, and even more so when the temperature is expected to be within a degree or two of freezing.
So lo and behold, the forecasters who had for the most part carefully warned us that snow was likely - but not guaranteed - got ridiculed when temperatures were just barely warmer than modeled for just a few critical hours in D.C. and most of its suburbs. The snow melted quickly as it landed. There were however, still power outages, strong winds, and a few inches of accumulation in several western suburbs.
As for long term climate models, in many ways these are better than short-term weather models because their resolution is at a climate-level scale, so the highly variable local weather gets wrapped up in area averages. These climate models are built on the laws of physics: conservation of energy and mass and energy transfer dynamics. The same physics equations and principles have made our modern world possible - with electricity, phones, television, etc. What's more, climate models have been given a pretty solid test drive. The first ones came into use in the 1970's have been thoroughly tested and compared with real observations since, leading to refinement after refinement. There is high confidence that their results are valid and useful.
Although there are still uncertainties in the models, these are almost all exclusively weighted against our favor - leading to worse climate change than the models project. And the revisions to the models over the years has tended to increase the severity of their outcomes. What's more, the models leave out substantial feedback processes, which are accelerating warming, and will only continue to do so even more into the future. And finally, the model assumptions are probably niave as regards human greenhouse gas pollution going forward, meaning yet more warming.
[I should note that the models are increasingly capable of reproducing tipping points (which again, almost all push us towards more warming), but still not quite there yet. However, these tipping points can be manually entered into the model parameters to simulate what such events might look like. And many researchers have done just that with various aspects, such as thermohaline slow down, catastrophic methane clathrate release, and polar ice sheet disintegration.]
So what do the models show?
The graph at the start of this article is output from my work with the GISS II general circulation model, which has a resolution sufficient to let me easily run it on a powerful workstation computer. It depicts a "business as usual" scenario, by my definition (see What the models don't show, Part 1 and 2), which means that while solar and wind power generation increase substantially, so does fossil fuel use. It projects that sea ice (on a global basis) declines at a steady clip from before mid 20th century, until about now, and then takes a sudden turn, declining much more rapidly until it's virtually gone by the end of the century.
This model, like other climate models has underestimated the rate of Arctic sea ice decline, but what it does indicate is that there is a strong relationship between the increasing CO2 levels in the atmosphere and the ice coverage. CO2 has been rising exponentially, but until about now, that increase could be taken as linear. We've hit the "takeoff" point however, where the exponential nature of the trend is really becoming manifest.
Equally important, this model and others also project that this "takeoff" point in CO2 coincides with a sharp increase in rate of surface air temperature rise. This however, has not manifested itself yet. Temperatures are still rising at the same rate they have been for several decades. Considering that climate models follow the laws of physics and thus conserve energy, could it be possible that the sooner-than-modeled decline in sea ice is where the "missing" increase in surface air temperature energy is going? If so, it could be that when the sea ice is gone, air temperatures will spike higher, at a rate faster than the model projects. Considering current trajectories, most researchers involved with the Arctic sea ice are now projecting it's elimination (during summers) in just a few years.
The fundamental issue is that greenhouse gases trap heat; more greenhouse gas means more heat kept in. That heat has to go somewhere. The only way for an equilibrium to reestablish is for average air temperature to increase, and thus increase the outgoing radiation. It's just that simple. So don't let anyone tell you that clouds or declining water vapor or any other hogwash contortion of reality or wishful thinking is going to keep climate change from happening. It's already here, in the Arctic, and everywhere else - and it is taking off now - as we can clearly see in the melting of the Arctic sea ice. Expect more big changes, soon.
If we're going to tackle this rapidly growing threat, we must have an open mind to making large changes to the structure of our social and economic system.
Above image compiled from Cryosphere Today archives; showing sea ice near minimum summer extent. The 2012 image includes snow and ice cover on land (as white areas), while the 1982 image only includes ice cover.
Below: historical sea ice extent (a combination of observed and modeled data for the pre-satellite era) compared to atmospheric CO2. The relationship indeed appears to be a strong one (data set unfortunately ends at 2007, I plan to update when new data becomes available).