top border

Please note, not all links may be active. This site is a snapshot of an earlier time.

Warmth to Continue; Gradually Increasing Humidity
Low confidence in late-weekend forecast

Dan Stillman @ 8:40 AM

Our warm and rain-free weather pattern is expected to continue into the early part of the weekend (Aug. 26 is the last time National Airport recorded measurable precipitation). But forecast confidence plummets come Sunday, when an area of low pressure currently developing off the Southeast coast may or may not make its way toward the region, possibly as a tropical system.


Forecast Confidence: Very HighMostly sunny, warm. Today should be similar to yesterday, with mostly sunny skies and highs in the upper 80s to near 90. Humidity should still be on the low side, but not as dry as yesterday when dewpoints dropped into the 40s. Winds will be light from the southeast at 5-10 mph. Tonight, partly cloudy with increasing humidity and lows from the mid 60s in the suburbs to near 70 in the city.

Tomorrow and Friday

Forecast Confidence: HighWarm, more humid. Another pair of pleasant days, but the increase in humidity is likely to be noticeable. Expect partly to mostly sunny skies and moderately humid air with highs 85-90. Tomorrow night and Friday night, mostly clear with lows from the mid 60s in the suburbs to near 70 in the city.

The Weekend

Forecast Confidence: Low-MediumA Sunday Stumper. Saturday looks like more of the same -- partly to mostly sunny with highs near 90 and moderate humidity. Saturday night may bring increasing clouds with lows in the upper 60s to mid 70s. Sunday is the big question mark. A low pressure system currently spinning off the Southeast coast is poised to strengthen, possibly into a tropical system, and could potentially move up the East Coast and affect the MidAtlantic. But at the moment, models are all over the place with the track and timing of the system. Thus, the Sunday forecast ranges anywhere from mostly sunny and highs near 90, to cloudy and rainy with highs in the 70s.

Pictured: A developing area of low pressure off the Southeast coast may become tropical in nature. Uncertainty in its track and timing means a low-confidence forecast for the latter part of the weekend. Courtesy AccuWeather.

History-Making Hurricanes

Hurricane Felix made landfall in Nicaragua yesterday, becoming the second Category 5 storm to slam into the coast of Central America in two weeks. This is the first time on record that two Category 5 Atlantic hurricanes -- Felix and Dean -- have made landfall in the same year. Felix is now quickly weakening after causing extensive damage in Nicaragua and Honduras. Also of note, Hurricane Henriette made landfall yesterday across the tip of Baja California, marking the first time on record that Atlantic and Pacific hurricanes made landfall on the same date.

Straight from the Source: The Real Deal on Climate Models

Two weeks ago I posted graphics comparing model forecasts and the actual track of Hurricane Dean, the Category 5 storm that recently devastated parts of the Yucatan Peninsula before making a second landfall in Central Mexico. One model -- the GFDL -- did a particularly poor job in predicting Dean's track. In his Aug. 20 post, blogger Joe Sobel implies that the inaccurate hurricane forecast by the GFDL -- a model developed at the Geophysical Fluid Dynamics Laboratory in Princeton, N.J. -- should breed skepticism about long-range predictions from climate models that are developed at the same laboratory and used in this year's International Panel on Climate Change report.

Sobel writes, "There was one computer model that insisted that the storm was coming farther north. That outlier was the GFDL which is run in Princeton, N.J., and is considered by many forecasters, including those at the National Hurricane Center, to be one of the most, if not the most, reliable hurricane model that we have. It is interesting to note that at this point, it looks like it is going to turn out to be one of the worst computer models for this particular hurricane. It is also interesting to note that this model was developed at the same place that many of the 'super climate' models were developed, which of course are the basis for most of the concern about carbon-dioxide induced global warming and climate change."

A commenter who called our attention to Sobel's post was more explicit in expressing Sobel's implication, saying that the GFDL's performance "does cause me great concern about the reliability of the modeling coming out of the Princeton Group."

Is there any validity to linking the accuracy of the GFDL hurricane model with that of the laboratory's climate models? I put the question via e-mail straight to the folks at the laboratory. The reply, which came from "The GFDL Climate Model Info folks," reads as follows:

"The GFDL Hurricane Model is separate and distinct from the GFDL Climate Model. The models were developed by different people, with different model code, using different algorithms to address the different scientific questions being posed. So, any attempt to link the performance of one (good or bad) to the other is nonsensical."

The GFDL group's response went on to point out an important difference between short-range models that simulate day-to-day weather or hurricanes, and long-range climate models.

"In many important ways, modeling weather differs from modeling decadal to century time scale climate. ... the quality of a single numerical weather forecast is dependent both upon the ability of the computer model itself to represent real world physical processes and upon the details of the data used to initialize the model," they said. "In contrast to the short-range Dean forecast, a long-range climate model projection of the mean climatic state approximately 100 years from now is relatively unaffected by the data uncertainties" in observations of the environment surrounding hurricanes and other phenomena that occur on short time scales.

As evidence of how sensitive hurricane models are to the initial conditions fed into them, a version of the GFDL run by the U.S. Navy, called the GFDN, "nailed the track and took Dean on a west-northwest track across the Yucatan Peninsula. These two versions of the model code are very similar, but they draw their initial conditions from two different global models," according to the GFDL team.

The take-home point from the GFDL modelers is that weather models (including hurricane models) and climate models are two different beasts, and that the accuracy of one does not reflect the accuracy of the other. So while there are many factors that make predicting long-range climate a difficult task, the misconception that climate predictions for decades from now can't be trusted if weather models can't even get the forecast right for this weekend is just that -- a misconception that has no basis in science.

Comments are closed for this archived entry | Link | email post Email this post