Automated runoff calculation engine
Rational Method, SCS/NRCS curve numbers, NOAA Atlas 14 IDF lookups, pre/post-development comparison and detention sizing — returned in under four seconds from a parcel polygon.
Phase 1Computational hydrology platform
Where every drop accounts.
StormwaterIQ is the only platform combining hydraulic computation, terrain analysis, computer vision, and automated permit verification in a single system — purpose-built for stormwater utilities, public works, consulting engineers, and watershed organizations.
01 — The gapOther platforms track compliance.
Other platforms track compliance.
We compute it.
The category has matured around documentation: SWPPP storage, inspection forms, permit lockers. None of the incumbents perform the calculations. StormwaterIQ begins where the spreadsheet ends.
| Capability | ComplianceGo | SW² | iWorQ | CloudCompli | SwiftComply | StormwaterIQ |
|---|---|---|---|---|---|---|
| Automated runoff calculations | — | — | — | — | — | Rational / SCS |
| Elevation & grading (DEM) | — | — | — | — | — | MapBox + 3DEP |
| Watershed delineation | — | — | — | — | — | Auto-generate |
| Computer vision / BMP AI | — | — | — | — | — | On-device ML |
| Photo deficiency detection | — | — | — | — | — | AI-flagged |
| MS4 permit DB verification | — | — | ~ | — | ~ | All 50 states |
| Native mobile (offline) | ~ | ~ | ~ | ~ | ~ | iOS + Android |
| Predictive maintenance score | — | — | — | — | — | ML model |
02 — Eight modules
Independently shippable. Inseparable in practice.
Each module is engineered to stand on its own and to compound when paired. Hydraulics inform CV. CV informs predictive scoring. Predictive scoring informs routing. The platform is the network of effects.
Elevation, grading & terrain analysis
USGS 3DEP DEM tiles, sub-meter elevation queries, watershed delineation up to 10 sq mi in ten seconds, slope rasters, flow-path tracing, and DXF grading-plan diff.
Phase 1Computer vision for BMP inspection
On-device classification across 20+ BMP types, deficiency detection (silt-fence tear, undercutting, turbid discharge), erosion severity scoring — inspector confirms or overrides every finding.
Phase 2MS4 permit verification
All 50 state NPDES registries, nightly. NOI coverage check, expiration alerts at 90/60/30/7 days, LLM-parsed permit PDFs, and one-click annual report generation.
Phase 2Native iOS & Android field apps
Offline-first via WatermelonDB. Live CV overlay on the camera, GPS-tagged photos, voice-to-text notes, biometric login, digital signature, and conflict-safe sync on reconnect.
Phase 1Real-time weather intelligence
NOAA NWS plus MRMS radar at 1 km². Configurable rain triggers auto-create inspection tasks within twenty minutes of qualifying precipitation. Named-storm escalation built in.
Phase 1Predictive maintenance scoring
XGBoost trained on inspection history, rainfall, BMP age, slope, and soil. Risk score per asset, portfolio dashboards across 500+ sites, and route-optimized inspection days.
Phase 3PostGIS-native spatial data model
Sites, watersheds, BMPs, inspections, permits, and weather events as first-class spatial entities. Row-level security by org. Queries run where the geometry lives.
Phase 103 — Performance targets
Numbers that hold in review.
The contract isn't the demo — it's the SLA. These are the p95 targets the platform ships against, in production, under load.
// p95 target
4 s
Hydrology API — full Rational Method including Atlas 14 lookup, p95 under load.
// p95 target
10 s
Watershed delineation for catchments up to ten square miles.
// p95 target
1.5 s
On-device BMP classification on iPhone 12 / Pixel 6 class hardware.
// p95 target
20 min
MRMS data availability to inspector notification, end to end.
// p95 target
≥88 %
Top-1 BMP classifier accuracy across twenty trained classes.
// p95 target
3 s
Mobile cold launch to inspection-ready state.
// p95 target
50 states
NPDES / MS4 permit registry coverage at full launch.
// p95 target
0.78 AUC
Predictive maintenance model floor before production deployment.
04 — Eighteen months
Foundation, then force multipliers.
Phase 1 puts the spatial data model and field apps in customers' hands. Each subsequent phase increases the leverage every active site already has.
Foundation
Spatial data model, MapBox base, USGS 3DEP tiles, React Native scaffold with offline sync, NOAA weather + inspection triggers.
Months 1–4Hydrology & vision
Rational Method engine, NRCS auto-lookups, Atlas 14, watershed delineation, YOLOv8 BMP classifier, on-device Core ML / TFLite export.
Months 5–9Permits & analytics
EPA ECHO integration, 15 priority state connectors, LLM permit parser, multi-state annual reports, predictive maintenance scoring.
Months 10–14Intelligence & launch
All-50-state coverage, MQTT IoT ingest, active-learning CV retraining, 3D terrain, public compliance portal, App Store launch.
Months 15–1805 — Frequently asked
Questions worth answering directly.
If something here doesn't answer your question, get in touch. We prefer conversations to forms.
What exactly does StormwaterIQ do?
It is a computational hydrology platform. You bring your service area — terrain, infrastructure, land cover, permits — and we build a continuously running model of how water behaves in it: where runoff goes, which BMPs are failing, which permits are about to expire, and which sites a storm just made urgent.
Will it stand up to regulatory and engineering review?
Yes. The methods underneath are physics-based and align with the conventions practicing engineers expect — Rational Method, SCS curve number, NRCS soil classification, NOAA Atlas 14 IDF, recognized time-of-concentration approaches, and well-documented hydraulic routing. Every output ships with the inputs and data sources visible.
Who is this built for?
Stormwater utilities, public works departments, consulting engineers, watershed organizations, and floodplain administrators. Anyone who has ever had to defend a runoff number, an inspection finding, or a permit deadline.
How does the on-device computer vision work?
A compressed Core ML / TFLite model runs entirely on the inspector's device — under 1.5 seconds per photo, no network required. On sync, the full resolution image is re-analyzed by a server model. Every CV finding is advisory: the inspector confirms or overrides each one before it becomes a compliance record.
What does the deployment look like for a municipality?
Phase 1 ships in roughly four months: spatial data model, MapBox-backed cartography, native mobile inspection apps with offline sync, and weather- triggered task creation. Most teams begin seeing leverage in week one, and the platform compounds from there.