How Tomorrow.io is building a new generation of weather intelligence through rapid-refresh microwave observations, operational forecasting applications, and global agency partnerships
Weather forecasting has always depended on one critical ingredient: observations. But today, the global observing system faces mounting pressure. Many operational microwave sounders — the instruments responsible for feeding temperature and moisture data into numerical weather prediction models — are aging out of service. At the same time, billions of people remain outside reliable weather radar coverage, leaving major gaps in forecasting and early warning infrastructure.
In a recent webinar, leaders from Tomorrow.io’s science, engineering, and government partnerships teams shared how the company’s Microwave Sounder Constellation is helping address those gaps. The session explored everything from the satellite technology itself to forecast applications, data assimilation studies, and operational integration pathways for agencies around the world.
Below is a recap of the conversation, broken down by speaker and topic, with recommended moments for embedded video clips throughout the article.
Rei Goffer: Why the Global Weather Observing System Needs More Data
Rei Goffer, Co-Founder and Chief Strategy Officer at Tomorrow.io, opened the webinar by framing the broader challenge facing weather forecasting today: the world increasingly needs more atmospheric data at the exact moment many legacy observing systems are aging out.
Microwave sounders have served as a foundational component of global forecasting systems for decades. These instruments provide critical temperature and moisture observations that feed numerical weather prediction models around the world. According to Rei, much of today’s forecasting skill ultimately traces back to microwave sounding observations.
The issue is that most operational microwave sounders currently in orbit belong to government agencies, and many are approaching end of life. Several have already begun aging out of operational service, while replacement systems are not being launched fast enough to maintain — let alone expand — global coverage.
At the same time, Rei highlighted another major challenge: approximately 5 billion people live outside reliable ground-based weather radar coverage. Large portions of Africa, South America, oceanic regions, and developing nations lack the radar infrastructure that supports real-time precipitation analysis and severe weather monitoring in more developed regions.
Traditionally, microwave sounders were not considered useful for nowcasting or real-time precipitation monitoring because revisit rates were too slow. Satellites might only pass over a location a few times per day, making the data less actionable for rapidly evolving weather systems.
That changes with a constellation.
By deploying multiple microwave sounders across several orbital planes, Tomorrow.io is now achieving sub-hourly revisit rates globally. Instead of isolated snapshots, forecasters can monitor atmospheric evolution continuously throughout the day.
Rei explained that this frequent coverage transforms microwave sounding from a primarily model-assimilation tool into something much more operationally useful for precipitation monitoring, storm tracking, and rapid-update forecasting.
He also emphasized the growing role of AI-based forecasting systems. Both traditional numerical weather prediction models and emerging AI weather models require enormous amounts of observational input to improve forecast skill. More observations — especially frequent global observations — directly support better model performance.
One statistic stood out during Rei’s presentation: at any given point in time, roughly 70% of the most recent microwave sounding observations available globally now come from Tomorrow.io’s constellation.
The webinar also included a striking animation showing two days of global microwave coverage from the constellation, revealing tropical cyclones, frontal systems, and atmospheric moisture structures evolving in near real time.
Joe Munchak: Inside Tomorrow.io’s Microwave Sounder Technology
Joe Munchak, Tomorrow.io’s VP of Mission Science, then shifted the webinar into the technical foundation behind the constellation itself.
Joe introduced the Tomorrow Microwave Sounder (TMS), a 6U CubeSat microwave sounder derived from NASA-funded TROPICS mission heritage. While the original TROPICS instruments were 3U CubeSats roughly the size of a shoebox, Tomorrow.io expanded the design into a larger platform that allowed for additional calibration targets and improved receiver stability.
To date, Tomorrow.io has launched 11 microwave sounders into orbit.
Joe spent significant time explaining an important distinction between microwave sounders and traditional weather satellites.
Visible and infrared satellites primarily observe cloud tops, surface temperatures, and atmospheric water vapor patterns near the tops of clouds. Microwave instruments operate at much longer wavelengths, allowing them to penetrate cloud layers and observe atmospheric structure beneath them.
That means microwave sounders can:
- Detect temperature profiles throughout the atmosphere
- Measure atmospheric humidity
- Observe precipitation cores within storms
- Infer cloud structure
- Sense surface conditions beneath many cloud layers
Joe demonstrated this using imagery comparisons between visible, infrared, and microwave channels. While visible imagery showed cloud patterns and infrared imagery highlighted cloud-top temperatures, microwave imagery revealed precipitation signatures and moisture structures hidden underneath those clouds.
TMS includes 12 channels specifically designed to retrieve atmospheric temperature and humidity at different vertical levels.
Joe also walked through the constellation architecture itself. The satellites operate across six orbital planes, combining sun-synchronous and inclined orbits. This hybrid approach allows the constellation to avoid the temporal clustering common among traditional government microwave sounders.
Historically, many operational sounders have occupied only a few orbital planes, creating large temporal gaps between observations. TMS fills in those gaps by distributing observations across different local times.
The result is approximately hourly average revisit globally, with even more frequent coverage in some regions.
Joe also discussed validation and calibration efforts in detail.
Tomorrow.io evaluates TMS observations against:
- ERA5 atmospheric reanalysis
- NOAA’s Advanced Technology Microwave Sounder (ATMS)
- NASA’s GPM Microwave Imager (GMI)
- Radio occultation profiles
These comparisons help assess instrument noise, calibration consistency, geolocation accuracy, and atmospheric retrieval performance.
One particularly important finding was that TMS observations often agree more closely with other satellites than with atmospheric reanalysis systems — suggesting the constellation is capturing atmospheric information not fully represented in existing models.
Joe also highlighted several data products available from the constellation:
- Level 1 brightness temperatures
- Intercalibrated products
- Resampled products for machine learning applications
- Level 2 atmospheric retrievals
- Level 3 gridded cloud-native Zarr datasets
The data pipeline itself is highly automated and cloud-native, with products often generated within roughly 10 minutes of satellite downlink.
Ryan Honeyager: Forecast Applications, Tropical Cyclones, and Data Assimilation
Ryan Honeyager, who leads Numerical Weather Prediction at Tomorrow.io, focused on how TMS data can be applied operationally.
One major theme throughout Ryan’s presentation was the value of frequent microwave observations for tropical cyclone monitoring.
He showcased examples from Typhoon Sinlaku and Cyclone Norell, demonstrating how microwave imagery reveals storm structures invisible to traditional visible and infrared imagery. TMS can infer warm-core structures, precipitation intensity, upper-level moisture patterns, and storm-top proxies.
For Cyclone Norell, the constellation captured 84 direct overpasses over 10 days — an unusually dense dataset for microwave observations.
Ryan explained that because TMS satellites operate across multiple orbital inclinations, observations occur at many different times throughout the day. This creates a more evenly distributed observing system compared to traditional government constellations clustered around a few sun-synchronous orbital planes.
That temporal diversity becomes especially valuable for precipitation nowcasting.
Historically, precipitation retrieval algorithms relied heavily on geostationary infrared imagery. But infrared observations mainly infer precipitation indirectly from cloud-top temperatures.
Microwave observations provide much more direct insight into precipitation processes inside clouds.
Tomorrow.io’s nowcasting system fuses microwave observations with geostationary imagery to improve precipitation retrieval skill. According to Ryan, microwave observations continue adding predictive value for roughly 70 minutes after an observation is captured.
Because TMS revisit rates are now sub-hourly globally, there are long periods where forecasts never need to fall back to geostationary-only retrieval skill.
Ryan also spent considerable time discussing data assimilation.
Tomorrow.io has developed reference workflows for assimilating TMS observations into numerical weather prediction systems using JEDI (Joint Effort for Data Assimilation Integration), an emerging operational framework used by agencies worldwide.
Their studies examined:
- Observation error characterization
- Clear-sky and all-sky assimilation methods
- Quality control approaches
- Bias correction
- Data thinning and superobbing techniques
The company conducted data denial experiments comparing forecast impact when removing TMS observations versus removing NOAA ATMS observations.
Results showed that while ATMS still performs better in some temperature channels due to its larger instrument design, TMS demonstrated strong and physically consistent forecast impact — especially for water vapor fields.
Ryan also highlighted independent evaluations conducted by:
- NOAA’s Commercial Microwave Data Pilot
- The Joint Center for Satellite Data Assimilation (JCSDA)
- NOAA, NASA, and Air Force research groups
These evaluations continue assessing the operational value of TMS observations in real forecasting systems.
Cesar Benetti: How Agencies and Researchers Access TMS Data
Cesar Benetti, Solutions Engineer at Tomorrow.io, focused on the practical side of accessing and integrating TMS data.
For individual users and researchers, Tomorrow.io provides sample Level 1 and Level 2 datasets directly through its online data catalog. Researchers can also submit requests for archived data covering specific time periods and geographic regions.
Government agency access follows a more structured operational workflow.
The process typically begins with:
- Discovery meetings
- Scientific briefings
- Product and domain definition
- NDA or MOU execution
- Trial data delivery
- Evaluation and validation
- Operational integration
Data can be delivered through APIs or cloud-based storage systems such as S3 buckets.
Tomorrow.io also supports multiple data formats, including NetCDF and other WMO-compatible standards.
Cesar emphasized that technical onboarding support is a major part of the engagement process. The company helps agencies configure data assimilation systems, review calibration reports, and integrate TMS observations into operational forecasting pipelines.
Ari Davidov: Building Long-Term Operational Partnerships
Ari Davidov, Director of International Government Development, closed the webinar by discussing how agencies engage with Tomorrow.io operationally.
He emphasized that Tomorrow.io’s offering is not simply a raw data feed. Instead, the company aims to provide a broader operational capability that combines:
- Data delivery
- Technical integration support
- Mission-specific services
- System optimization
- Managed infrastructure layers
Ari explained that different agencies have different operational requirements. Some may need only archived research data, while others require low-latency operational streams integrated directly into forecasting systems.
As a result, licensing and pricing structures are flexible and tailored to each agency’s mission requirements.
He also highlighted the company’s move toward managed service architectures, which introduce:
- Secure policy-driven access
- Operational compliance layers
- Defined latency and uptime guarantees
- Distributed deployment capabilities
- Centralized oversight
The goal, Ari explained, is to evolve from simply providing satellite observations to delivering mission-ready weather intelligence infrastructure.
Closing the Global Observation Gap
Throughout the webinar, one theme remained constant: the future of weather forecasting depends on more frequent, more global, and more operationally accessible observations.
Rei Goffer framed the global challenge. Joe Munchak explained the science and engineering behind the constellation. Ryan Honeyager demonstrated the forecasting applications. Cesar Benetti outlined operational access pathways, and Ari Davidov described how agencies can scale these capabilities into mission-critical infrastructure.
Together, the presentations illustrated how Tomorrow.io’s Microwave Sounder Constellation is helping build a new layer of global weather intelligence — one designed for continuous sensing, rapid refresh forecasting, and greater resilience worldwide.
