How Good Are the Climate Models?

This blog has been highly critical of the computer models that are used to forecast future climate. A review article, just out, says:

There should be little confidence in climate models. The model simulations fail in their attempts to provide credible quantitative estimates of future climate change, on regional, or continental, or global scales. The models have shown little to no ability to reproduce observed features of current climate and past climate changes. Confidence in model estimates is greatly overstated by the IPCC for the most common of climate variables (e.g., surface temperature) used to present the supposition of manmade global warming. 

Details here.

Undoubtedly, climate models will begin to show skill at some point in the future. But, at this point, there is no reason to give them any credibility.

More on the South Central Rain Chances

Here are two computer runs (U.S. and European) and the amount of rain they are forecasting late next week in the drought-striken areas of the central U.S. These models are from AccuWeather.com’s Professional site: 

This is fairly encouraging given that the models typically under forecast rainfall this far out.  While these amounts are not large, it would be good just to get the dry pattern to break.  Keep your fingers crossed!

More Power, Please

The IBM Power 6 computer used by the National Weather Service to run its computer simulations of the atmosphere performs 17,100,000,000,000 calculations per second and it is not fast enough to run all of the models as quickly as we would like.

For example, the latest high resolution model used in extreme weather prediction (available here) is often an hour or two behind because we do not have enough computer power!

Promising, and Not So Promising, Tornado Research

Yesterday’s Wall Street Journal had an article about tornado research in the wake of the recent storms.

In general, I’m in favor of meteorological research because better weather forecasts and warnings have a huge benefit to cost ratio.  This research, in some cases, performs a dual role in helping better understand climate. That said, research dollars are not infinite. So, I’m going to critique the projects mentioned in the article.  Keep in mind that the goal of the taxpayer-funded public relations and lobbying arms of these government laboratories is to pry more tax money from Congress. Most of them never met a research program they didn’t like, regardless of its real-world merits.

First of all, the premise of the article: “Better” warnings on April 27 in the South would have saved lives. Based on the evidence I have examined (more on that in a blog posting tomorrow), the forecasts and warnings of those storms were excellent. The loss of life was exaggerated in this case because of the extensive power failures 12 to 16 hours before the tornadoes arrived. The lack of power prevented people from receiving the warnings via TV, internet, etc. Improving warnings makes no difference if people cannot get them because the power is out.

Now, the areas where the interviewees want to perform more research, spend more money, or create new programs, in the order in which they appear in the article:

NOAA’s Budget Cut This Year
I believe this was unfortunate, but it had nothing to do with the loss of life on April 27. The warnings and forecasts were excellent.

WSR-74C radar used by NWS in Topeka.
I used one of these for more than 20 years and it could survey the atmosphere for
tornado detection every 20 to 30 seconds. 

Radars That Don’t Survey the Lower Atmosphere Often Enough to Detect Tornadoes
This part of the article is misleading. There is nothing inherent in the National Weather Service’s radars that prevent them from surveying the lower atmosphere much more frequently. That is software and it could be changed by modifying a few lines of code if the decision was made to do it.

In fact, up until the Doppler radars were installed from 1991 to 1996, radars surveyed the lower atmosphere every 20 to 30 seconds, which is plenty often to find tornadoes. The reason they don’t today is because the radars were part of a three agency program: NOAA, the Department of Defense, and the Federal Aviation Administration.  The latter two agencies wanted better coverage for aviation and, when the radars were installed, we thought they were going to be able to detect turbulence aloft. To do so, the radars spend a lot of time probing the upper atmosphere — not very useful for detecting tornadoes. So far, turbulence detection has not worked as anticipated. I believe there are compelling reasons, in tornado situations, to ditch the lengthy “volume scans” for aviation and run the radars so they survey the lower atmosphere more often.

That said, the current generation of radars is approaching twenty years old and will need to be replaced. I’m not sold on the article’s proposed solution of “phased array radars” because, at least so far, they present a picture that is more “blurry” than the existing radars. Detail is critical when trying to find tornadoes.

Something the article does not mention is “gap filling” radars. There are inexpensive current generation radars that could be put in areas where the current radar coverage is poor (northeast Missouri, for example). That would do more to save lives than phased array would at the current state of the art.


Better Computer Models and Vortex II
Needed and welcome!

Lightning Detection
There are good reasons to get better lightning data from both ground and satellite-based sensors, but it is not clear whether better tornado warnings are among them. Since at least 1955, meteorologists have been trying to figure out the link, if any, between lighting and tornadoes. While I have my ideas on the subject, they have not been tested in rigorous studies. This is worth pursuing.

Infrasonic (sound) Detection of Tornadoes
The theory is that, if a tornado made it to the ground without warning, we would “hear” the roar made by the tornado in contact with the ground so we could get a warning out for people downwind. This idea has been around since at least the 1970′s. It is not needed anymore due to the fact that in 1999 meteorologists discovered the “debris ball” radar signature. This signature will be more useful with the NWS’s new dual-polarization capability being installed at present in its existing radars.

A weather balloon. From top, the balloon, a parachute, and the
instrument package that measures the atmosphere.

Aircraft Weather Measurements
As noted, this is already being done and it is great if we can expand it. That said, it should never replace weather balloons. #1) An airplane might not be near the location when and where a measurement is needed. #2) The vertical measurements that can only be done by balloons are needed for climate studies.

Climate Model Uncertainty

For those who have seen my global warming presentation, I discuss the high-level of uncertainty associated with climate models. If we can’t forecast weather in a consistently accurate manner ten days in the future, there is no basis to believe we can do so ten years in the future. For those who would like to learn a little more about model uncertainty, go to this lightly technical article. The quote below sums things up nicely.

But for all that was controverted, the basic physical uncertainties were not disputed. It turns out that uncertainties in the energetic responses of Earth climate systems are more than 10 times larger than the entire energetic effect of increased CO2.15 If the uncertainty is larger than the effect, the effect itself becomes moot. If the effect itself is debatable, then what is the IPCC talking about? And from where comes the certainty of a large CO2 impact on climate?

Why the SEC’s Decision is Misleading, and Meaningless

Many non-specialists are unaware of a recently-created industry of companies that use sophisticated models to try to predict hurricane losses and losses from other types of storms.

In the previous post, I talk about the SEC’s nonsensical decision to “encourage” companies to disclose their exposure to ‘climate change.’ One method of doing that is to use these computer modeling services.

While the models can and do provide useful information when a storm is imminent (i.e., Hurricane X will move into location Y with 120 mph winds and a 15 ft. storm surge), they have shown no detectable skill at forecasting losses years (the timescale for “climate”) into the future. Another new paper shows this to be the case. Roger Pielke, Jr. has more here.  


So, if there is no way to forecast the losses caused by storms, how are companies supposed to — in advance — disclose their losses?