Water quality & streamflow forecasting
Weather forecasting, built for water.
OpenCurrent turns legacy monitoring records and modern sensor networks into predictive, real-time intelligence, empowering environmental managers to anticipate change instead of reacting to it.
What we do
Predictive intelligence for changing waters
Harmonize the past
Unify fragmented, multi-decade sampling archives into clean, analysis-ready datasets that give predictive models the ground truth they need.
Forecast in real time
Blend satellite remote sensing, in-situ sensor networks, and physics-informed machine learning to forecast turbidity, nutrients, and flow days before a problem reaches the intake.
Build for ownership
Deliver open, maintainable code and a gradual handoff, with documentation, walkthroughs, and time with your staff, so your team can run its own forecasting pipelines.
Our approach
Merging the historical record with modern technology
Water management is shifting from periodic monitoring toward continuous data and models. Both still matter. Decades of field sampling and gauge records are the ground truth any credible model has to answer to, and modern sensors and satellites see what a monthly site visit never could.
OpenCurrent brings the two together. We bring weather-style forecasting to water quality and quantity by anchoring modern machine learning in historical ground truth, so the record you already have makes new data more useful, and new data makes the record more valuable.
We are candid about the methods we use, where they fall short, and how much confidence to place in each estimate. And we build with our partners, so the teams that rely on a forecast can understand, run, and own the code behind it.
Where we come from
Capabilities
What this looks like in practice
Satellite-to-stream water quality
Calibrating remote sensing algorithms against historical grab samples to map lake and reservoir dynamics across space and time.
Predictive modeling workflows
Training catchment-scale models that forecast acute water-quality events, such as post-fire sediment runoff or bloom conditions, from upstream precipitation and weather forecasts.
Sensor networks, made useful
We integrate commercial in-stream sensors with lab records and weather data, turning raw readings into estimates a manager can act on.
Partnerships that hold together
Structuring projects so that landowners, conservation interests, and governments each come out ahead, which is usually the difference between a pilot and a program that lasts.
Current work
What we are building now
Fire-to-faucet forecasting for Denver Water
As water-system lead in the NSF ASCEND Engine's ARID program, we are assembling a public benchmark dataset that links wildfire, remote sensing, hydrology, and water-quality records across the western US, training models to forecast post-fire turbidity, organic carbon, and algal conditions, and building a prototype decision-support tool for Denver Water's supply network.
Hourly bacteria nowcasts from in-stream sensors
For a sensor manufacturer, we run an hourly nowcast feed that combines optical sensor readings with lab-confirmed samples at dozens of recreational-water sites in North America and Europe. Every prediction is scored against the lab result once it arrives, and each site is served by whichever method has the best track record there.
Environmental forecasting you actually own.
Tell us what you need to anticipate and when. We will tell you what is feasible and what it would take.