From the NVIDIA edition of October 26, 2026
Colorado State University (Oct 26): researchers test NVIDIA-based Hail-StormScope nowcasts alongside CSU-MLP outlooks, warning the approaches are not yet tested together
Colorado State University · Higher education and public weather research · Colorado, United States
- Publisher
- Colorado State University News and Media Relations
- Original publication
- 2026-10-26 (CSU news release)
- Source retrieved
- 2026-10-27
- Event date
- 2026-10-26
What happened
CSU says its researchers are developing AI hail forecasting on two horizons: Hail-StormScope, a specialised version of NVIDIA Earth-2's StormScope architecture trained and deployed in a project a CSU researcher led with NVIDIA, for 2 to 3 hours ahead, and the CSU-MLP machine-learning system for 1 to 8 days ahead. The work sits within NSF's ASCEND programme (up to $160 million over 10 years), notes the team's models are accurate for areas of about 1.3 km and plans Vaisala RM60 hail sensors from spring 2027 for verification. Researchers have not decided whether to merge or keep the two systems separate, and a researcher cautions the approaches have not been tested extensively or together.
Why it matters
Relevant to state emergency management, transportation and agriculture agencies and to public universities doing applied weather research; it shows a public university building on an NVIDIA model with explicit verification plans.
Evidence and measured results
University news release quoting researchers; describes plans and early development, not verified forecast skill.
Limitations and uncertainty
Release from the institution doing the work. No skill scores for hail are reported. The release says the hail model came from a project led with NVIDIA but gives no agreement terms or NVIDIA funding details.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-27; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- State agencies need earlier, more precise hail warnings to protect people, crops and infrastructure.
- Stakeholders
- State emergency management
- State transportation department
- Agriculture agencies
- University atmospheric science departments
- Discovery questions
- How do you use hail forecasts today, and at what lead time?
- Would you participate in verification of experimental nowcasts?
- Value hypothesis
- Interpretation: AI nowcasts could extend useful hail lead time; skill is not yet verified.
- Potential engagement
- Watching brief and possible verification partnership, not a product sale.
- Unsupported claims
- Any measured hail forecast skill
- Operational readiness
Pre-sales engineering
Role takeaway
- Fit
- Research and verification settings with radar and satellite data access.
- Architecture and integration
- StormScope-based nowcaster for 2 to 3 hours, CSU-MLP for 1 to 8 days, verified with ground hail sensors.
- Prerequisites
- GPU compute for training and inference
- Radar and satellite data feeds
- Hail sensor network
- Constraints
- Not tested extensively or together
- Sensors arrive spring 2027
- Base model not evaluated on hail
- Security
- Public data; standard research-computing controls apply.
- Proof of value
- Season-long comparison of nowcasts with observed hail
- Comparison with existing forecaster guidance
Delivery
Role takeaway
- Work
- Track verification results and sensor deployment; scope a pilot only after skill is published.
- Dependencies
- CSU publications
- NSF ASCEND timeline
- Sensor deployment
- Ownership
- Agency forecasting or emergency management lead would own any pilot.
- Skills and adoption
- Forecasters need guidance on interpreting AI nowcasts.
- Governance checkpoints
- Review published verification before pilot
- Decide on merged or separate systems
- Acceptance criteria
- Proposed: hail nowcasts compared with observed reports over a full season
- Proposed: comparison with existing guidance documented
- Risks
- Adopting before verification
- Resolution and hail adaptation untested
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Combines a vendor-originated nowcasting model with a university-developed medium-range system; verification depends on new ground sensors arriving in 2027.
Governance
Who approves, reviews and stays accountable for outcomes?
Operational use by forecasters or emergency managers would need skill verification against observed hail before adoption.
Security and privacy
What data, permissions and controls need testing?
No specific security issues raised; public weather data.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Forecasters will need training to interpret AI hail nowcasts alongside existing guidance.
Procurement
What should contracts, pricing and exit terms secure?
Agencies interested in hail nowcasts should wait for published verification results rather than procure on the basis of this announcement.
Operating model
Which teams own the service once it runs?
Open question whether short-range and medium-range systems will be merged or run separately.
Publication history
- 2026-10-26NVIDIA · Issue 518 resources
Stable resource ID: colorado-state-university-ai-hail-forecasting-stormscope-20261026