Say Hello to RAP's New Forecasting Talent: The Auto-Nowcaster

The automated thunderstorm nowcaster, or Auto-nowcaster, is the product of several years of work by a team of RAP meteorologists and software engineers on how storms are generated and maintained. Supported by Federal Aviation Administration funding, the nowcaster was unveiled at the American Meteorological Society's annual meeting several weeks ago. Later this year, it will undergo its first real-world testing in Memphis, Tennessee, accompanied by a human nowcaster from RAP to monitor its performance.

Sandra Henry and RAP's new Auto-nowcaster. (Photo by Carlye Calvin.)

Trying to follow and predict thunderstorms is a taxing affair for even the best forecasters. A single cumulus cloud can blossom into a thunderstorm in 10 to 20 minutes. Standard operational computer models have proven effective in the 12- to 48-hour time frame, but they are of little use in the 0- to 60-minute time frame. According to a recent document on the Auto-nowcaster, "Significant benefit can be derived from automating the [thunderstorm prediction] process, because making forecasts for this short a time period is very labor-intensive and requires considerable expertise by local forecasters."

Jim Wilson is project leader for the Auto-nowcaster. He, along with Cindy Mueller and Wendy Schreiber-Abshire, pioneered the art of storm-initiation forecasts during the middle and late 1980s by studying the evolution of gust fronts and other convergence lines depicted on Doppler radar. Storms can develop quickly at points where the boundaries (seen as thin lines on Doppler displays) intersect or collide, pushing air upward.

Jim and Cindy devised rules by which forecasters can anticipate the development of storms by up to 60 minutes. Some of this work was disseminated to the forecasting community beginning in 1991 through an interactive module produced by the Cooperative Program for Operational Meteorology, Education and Training (COMET). Says Jim, "It is quite common now for National Weather Service forecasters who have been through the COMET training to issue very short period thunderstorm forecasts to the public based on tracking convergence lines on NEXRAD [Next Generation Weather Radar]."

"Jim is the brains behind a lot of this research," says Sandra. "He and Cindy did a lot of the initial work on devising the basic rules. I've stepped in to try to get these rules automated." Sandra is working with RAP programmers and engineers Laura Stoltz, Mike Dixon, Laurie Carson, Frank Hage, and Gary Blackburn in deciding how the various components should interact. These include:

• A machine-intelligent gust front algorithm developed at the Massachusetts Institute of Technology's Lincoln Laboratories. Using fuzzy logic, it establishes a point-to-point correspondence between consecutive radar scans to predict where convergence boundaries will move and intersect.

• A three-dimensional model that predicts the position, strength, and depth of convergence lines; moisture convergence; and stability. This model, developed by Andrew Crook (RAP/Mesoscale and Microscale Meteorology Division), uses surface and upper-level wind reports coupled with Doppler data. The wind field used to initialize the model uses a single Doppler retrieval technique called TREC, or tracking radar echoes by correlation, as modified by John Tuttle (MMM/RAP).

• A routine to find regions where the lifted index (the difference in temperature between the air at an altitude having 500 millibars of pressure and a parcel lifted from ground level to that height) is positive. In searching for likely storm birthplaces, the nowcaster rejects areas of positive lifted indexes because the air is not buoyant enough to support convection.

• A software package for thunderstorm identification, tracking, analysis, and nowcasting (TITAN). Developed by Mike Dixon, TITAN uses radar data to find and follow individual storms, defined as regions of a certain size and radar reflectivity (typically 50 cubic kilometers and 40 decibels reflectivity). Past storm trends are used to predict future location, intensity, and size.

In its first real-time testing last summer at RAP, the Auto-nowcaster fulfilled Sandra's expectations. "Our objective was to have it running completely on its own. We had everything in place. We still need to optimize the forecast itself, but at times it performed very well." Next comes the first real-world test, scheduled to start in May at the FAA's Integrated Terminal Weather Support site near Memphis. Meteorologists from Lincoln Labs, the National Severe Storms Laboratory, and NCAR will collaborate over the summer to examine the Auto-nowcaster's output.

"It has been a major team effort to get the Auto-nowcaster ready for field testing," says Jim, "but we are now very anxious to get exposure and feedback from this summer's test in Memphis. We hope to soon get it in front of FAA controllers and NWS operational forecasters. We expect the Auto-nowcaster will be issuing aviation weather advisories on its own in the near future." --BH

In this issue...
Other issues of Staff Notes Monthly