Atmospheric Research - NCAR & UCAR
photo Home Our Organization Community Tools News Center Our Research Education Libraries Community Tools

Summertime, and the Forecasts Aren't Easy

NCAR scientists are probing the ebb and flow of warm-season rain for clues to better forecasts

  The great Midwest flood of 1993, which killed 50 people and wreaked $15 billion in damage, was produced by repetitive summer storms, a weather regime now under scrutiny by NCAR scientists.

When your local forecast calls for a 30% chance of rain, it disguises a good deal of uncertainty about just where and when those raindrops will fall. A major challenge for scientists is determining what factors in the sky or on land may cause a storm to form over one town instead of another just a few miles away. Equally important is the question of why some storm clouds release rain more quickly or more intensely than others.

Because there is a certain degree of randomness in our atmosphere, we may never know with absolute certainty where and when storms will strike. But research at NCAR and other organizations may eventually lead to forecasts so detailed that, for example, commuters might know to expect a storm along a certain stretch of highway during evening rush hour.

Here are some of the ways NCAR scientists are probing the ebb and flow of warm-season rain.

What makes summer rain overstay its welcome?

When a summer storm hits, it usually passes through within an hour or two. But sometimes a region will be hit day after day by torrential, flood-triggering rain. There's more than simple persistence behind this serious weather concern. A group of storm researchers at NCAR is looking into what drives multiday episodes of heavy rain across the central and eastern United States.

"Knowing that it's raining in one place, now, drastically changes the likelihood that it's going to rain somewhere else later on," says team member Christopher Davis. He and his colleagues are poring over radar and satellite images, tracing the fingerprints of showers and thunderstorms over days and weeks. If each storm is like a tree, their goal is to understand the meteorological forest that includes them all.

This summary of radar data for July 1998 shows a corridor of repeated showers and thunderstorms arcing from Colorado across Kansas to Tennessee (solid arrow). In other years, July storms tended to track across the Northern Plains (1997) and from New Mexico to Michigan (1999). (Illustration courtesy John Tuttle and David Ahijevych, NCAR.)

On a typical summer day, thunderstorms build in the early afternoon over the Rocky Mountains, driven by intense high-altitude solar heating. They spread eastward, often gaining strength through the afternoon and overnight as they move into more humid parts of the Great Plains and Midwest. As these storms die out in cooler morning air, the next episode may be starting just hours later.

Building on groundbreaking work by NCAR's Richard "Rit" Carbone, the team has found that an east-west "corridor" sets up during some summers, with storms favoring this atmospheric highway and bypassing areas to the north or south. During the summer of 1993, more than 70 rounds of storms passed through a corridor straddling the Missouri and mid-Mississippi river basins, causing historic floods.

"We don't necessarily understand the linkages between individual events in these episodes," says Stanley Trier. Davis adds, "The episodes move more quickly than you would expect a single thunderstorm to move." The extreme regularity of these episodes, once they set up, seems to be unrelated to external factors such as ocean processes, El Niño–La Niña, or hurricane activity.

  Corridors of Concern
  Learn more about patterns of repetitive rainfall in this multimedia feature (~12 MB, requires high-bandwidth connection and Flash 6 or later).

While the NCAR team has yet to find all the answers, they know that the still-mysterious phenomenon they've identified must be understood in order for computer forecast models to do a better job handling summer storms. The team is now working with colleagues to analyze similar summertime patterns in the Himalayas, Europe, Australia, and sub-Saharan Africa. "We want to do more expansive climatology studies all over the globe," says Davis.

Read the Research
Groundbreaking paper on warm-season precipitation episodes

Putting the forecast spotlight on storms, one by one

Although national-scale computer models are a linchpin of modern weather forecasting, most of them don't depict showers and thunderstorms directly. Instead, they parameterize storms. This means they estimate what percentage of each square in a checkerboard grid drawn from coast to coast will be covered by storms. Until now, scientists haven't had the computer power or the model sophistication to produce a nationwide forecast of storms with the same level of detail as a radar image you might see on TV.

Now, such a forecast is being produced each night by one of the world's largest supercomputers. For the second year in a row, NCAR is serving as a testbed for a new model, the first one ever to predict individual showers and storms across the bulk of a continent (see map).

The model being put through its paces is a research version of the Weather Research and Forecasting Model (WRF), a collaboration among six partners. The first generation of WRF will be adopted by the National Weather Service this fall. It tracks weather every 10 kilometers (6 miles). While that's a far better resolution than previous national-scale models, it's still not enough to portray individual thunderstorms, some of which may be only 2 or 3 miles wide.

  NCAR's high-precision forecasts this summer will extend from the Great Basin to the Appalachians.

The second generation of WRF—the one under testing—has a resolution of 4 km (2.5 miles). With that kind of precision, the new version can tell forecasters whether an evening's storms are likely to form along a squall line, in a large cluster, or in a set of small, discrete cells. Even if storm locations aren't completely accurate, just knowing what storm type to expect is an invaluable aid to forecasters. Though regional models have produced such output for several years, they can't match the continent-wide picture provided by WRF.

Each night from April through July 2004, an IBM supercomputer at NCAR nicknamed Bluesky will spend about six hours generating a forecast for the next day's activity. Those forecasts will appear on the Web by early morning, allowing National Weather Service forecasters (and the public) to see how well the new model performs. A similar round of 4-km runs will be tested at the National Severe Storms Laboratory, using a different approach to setting the initial conditions for WRF.

In its first test last summer, the 4-km WRF got rave reviews. "We were pleasantly surprised at how well it was doing," says Morris Weisman.

This year, the model's territory is being extended further west, to capture more of the storm genesis areas in the Rocky Mountains. "We'll be getting feedback from forecasters on a daily basis," Weisman says. Next winter, the model will get a similar round of testing for winter weather. After several more years of evaluation and refinement, a 4-km version of WRF may be adopted by the National Weather Service as a standard forecast tool.

WRF 36-hour Forecast   National Weather Service Radar

When flood follows fire

How does the ground affect the sky? That, in essence, is the focus of an NCAR research project that aims to eventually incorporate soil conditions into weather predictions.

“Soil moisture and temperature are important for summer weather,” explains Fei Chen. “It can be a deciding factor in storms.”


Researchers have long known that dry soil can exacerbate heat and drought, whereas wet soil sometimes provides moisture that rises into the lower atmosphere and helps to fuel precipitation. But soil may have differing impacts depending on atmospheric conditions, while other small-scale ground features, such as vegetation, hillsides, and housing or industrial areas can also have subtle impacts that can determine the exact boundary and intensity of storms.

In a case study, Chen teamed up with Thomas Warner and Kevin Manning to look at the impact of the land surface and terrain on the 1996 flash flood that struck Colorado's Buffalo Creek, in the foothills southwest of Denver. Among their findings: the site of an earlier forest fire received particularly heavy rainfall during the flood, possibly because the denuded site transferred more heat into the atmosphere than an area filled with living trees would. That extra heat would enhance a local storm by adding to the buoyancy of updrafts that pull in nearby moist air.

Chen is working with the National Centers for Environmental Prediction on a community land surface model that will incorporate ground conditions into weather forecast models. The land surface model includes information on the type of ground cover (such as grasslands, croplands, or forest) and the degree to which the ground is saturated or dry. In time, the model may have a resolution as fine as 500 meters, potentially helping forecasters to better predict the exact locations of summer storms.

Learn More
RAP land-surface modeling

“One of the greatest opportunities to make progress in summer convection forecasting is to better represent surface processes,” Chen says.


This document can be found at
rss icon Subscribe to our News Feeds at