HomeNewsArticle Display

Machine-learning radar closing the gap to operations

Graphic representation for artificial intelligence

For the last several months the 557th Weather Wing has been assisting Massachusetts Institute of Technology / Lincoln Labs by evaluating the operational feasibility for a potentially groundbreaking Machine Learning / Artificial Intelligence capability called Global Synthetic Weather Radar. (Original graphic created by Stefanie Pidgeon)

OFFUTT AIR FORCE BASE, Neb. --

This is the third in a three-part series covering the innovative work by 557th Weather Wing Airmen for the ongoing development efforts into machine-learning for a weather radar depiction across the globe, designated the Global Synthetic Weather Radar (GSWR).

Airmen of the 557th Weather Wing have been participating in and providing qualitative feedback for Massachusetts Institute of Technology Lincoln Laboratory’s cutting-edge machine-learning initiative called Global Synthetic Weather Radar.

GSWR was developed by MIT/LL to provide weather-related predictions where radar sensors are not available.

The concept was originally envisioned for the Federal Aviation Administration as a means to create synthetic radar-like mosaics in areas just offshore of the continental United States with no radar coverage. Air Force leaders recognized the potential and sponsored MIT/LL to produce global radar-like mosaics in regions currently without radar data.

“A huge limitation to our current tactical ground-based weather radar programs are their myopic scope of coverage, which is often restricted further by nearby land features,” said Maj. Daniel Muggelberg, 2d Combat Weather Systems Squadron operations officer. “Ground-based radars require significant logistical and maintenance support to work properly.”

The vision for GSWR being expanded to a global capability required the improved ability to remotely sense data via a global sensing network comprised of terrestrial and space-borne sensing systems. Satellite imagery, lightning, rainfall rates and radar reflectivity are just a few of the components that GSWR will use in its mosaic creations. 

It is important to note that the GSWR platform doesn’t replicate the full capabilities of a radar system and therefore isn’t meant to replace radar systems but instead is intended to supplement them. For a variety of reasons, radar systems may not be available in certain regions and this is where the GSWR would be most valuable.

In order to achieve the high standards set by Air Force leaders, the 557th Weather Wing was called upon to provide the substantial amount of data required by the machine-learning GSWR platform and also to confirm the results were suitable for real-world operations.  Although promising, GSWR is not yet ready for operations.

“We anticipate a need for greater involvement in shaping continuous improvement efforts leading to an eventual fielding decision,” said Jeffrey Fries, 1st Weather Group chief of operations, standards and tactics. “We tend to lean on terms like Initial Operational Capability and Full Operational Capability, but they are somewhat dated concepts in an era of machine learning [where systems] have the potential to learn constantly and are paired with continuous integration and delivery of software.”

On some confusion regarding how and when the GSWR is fielded, the final phase to this journey is to determine if GSWR is a viable tool for 557th WW Airmen to utilize in order to meet the needs of combatant commanders.  While capabilities such as GSWR may not follow traditional acquisition constructs and thus forego official IOC and FOC declarations, the capability must still demonstrate its ability to address gaps in operational information environment.  As such, the capability must successfully meet defined performance measures as demonstrated in an Operational Test and Evaluation process.

This is where the Airmen from 2 CWSS leverage their expertise in OT&E.  These Airmen will engage with weather forecasters performing a variety of DoD missions to understand how GSWR will meet their mission needs. They will ensure those who will operate the software are represented. The operational testing may involve numerous users and take several iterations to get right. Once satisfied, 2 CWSS will provide a fielding recommendation to an Integrated Test Team.

The ITT consists of members from the test community, program office and the lead command. The ITT will jointly consider factors like funding, mission urgency, program deficiencies, accuracy, and benefit to operations related to fielding the GSWR capability. The ITT system of checks and balances will produce the best solution for our warfighter community. If approved, the GSWR capability will transition to operations.

“Once operationalized, the 16th Weather Squadron is in a position to build upon and update the training set and re-train the AI/ML model used by GSWR,” said Richard Butler, 2d Weather Group Enterprise Management. “For example, as the 557WW upgrades its Global Air-Land Weather Exploitation Model (GALWEM), the model output would be fed to GSWR’s machine-learning model which should in-turn translate into an improvement in GSWR’s 3 and 12-hour forecasted / synthesized global coverage radar mosaics.”

Location, cost and infrastructure in remote or contested regions will continue to be cost-prohibitive to building a physical RADAR sensing capability. In the end, the GSWR capability could increase trans-oceanic mission effectiveness, reduce the logistical and maintenance burden for the ‘first in’ deployment teams, and afford a myriad of other operations benefits.

The attributes of this program and how it has been approached by the Air Force is unique in that it is not a physical weapons system. It is rather a “smart” system that consumes available data from multiple sources, tunes the data and then provides a forecast. And just like forecast models today, the output requires Airmen to analyze and assess utility and resultant operational impacts tailored for the thousands of missions 557 WW supports each day.  This system may very well be the foundation for machine-learning advancements and the 1 WXG evaluations show promise, however, the 2 CWSS operational test team assesses just how ready it is for operations and sustainment.

“If GSWR can operate in the manner envisioned by the Air Force, I can see where future Airmen might look back at GSWR as being one of the defining points in history,” said Col. Patrick Williams, 557th WW commander. “If nothing else, the GSWR program has taught us to keep an open mind to new and exciting innovations.”