A changing climate requires new methods of forecasting

Communities around the country are reeling from the effects of climate change on water. In just the last few months, the news in the West has ranged from the 1,000 year megadrought that we’re in to historic flooding in California and back to the negotiations for securing the Colorado River’s diminishing water supply. Throughout this whiplash, the professionals who ensure that cities, farms, and ecosystems stay safe and can access the water they need have had to deal with more volatility than ever before. These experts coordinate electricity production at hydropower plants, manage flood control at reservoirs, and release drinking and irrigation water supplies to cities and farms.

Current Forecasting is Falling Short

The work of a water manager is not getting any easier. Climate change is water change, and the bottom line impact is: supply is less reliable. The reasons and compounding effects are complex, but the writing is on the wall: it is rare to finish a day’s news cycle without a story on how extreme(r) events, shifts in snowpack and rainfall seasonality, or a lack of flood early warning systems are impacting society and our environment. Take the example of California, which for the last three years until the start of 2023 was the driest 126 years on record¹, and after a month of multiple atmospheric river events and catastrophic flooding², is still in drought according to some environmental indicators³. This kind of whiplash in hydrology is becoming more frequent, and current public forecasts are missing the mark.

Given these increasingly difficult challenges, the ability to look further ahead into what water flows might look like in the coming days, weeks, and months is more important than ever. State agencies, like California’s Department of Water Resources, are overhauling their forecasting systems⁴, while NOAA’s National Weather Service is in the early stages of developing its next generation forecasts. However, advancements in streamflow forecasting methods remain largely untapped by those managing most of our global water resources, despite the enormous benefits of supporting adaptation across sectors, including:

• Transforming our energy grid: a sustainable renewable energy grid is dependent on hydropower for providing a critical flexible resource.
• Meeting growing municipal water supply in the coming decades: as historical patterns no longer indicate how the future unfolds, new tools and approaches are needed.
• Supporting habitat and fish livelihoods: current gauging networks do not support the density required to understand habitat and fish health, and with fewer dollars put toward physical gauges, we need innovative tools for understanding what changes are occurring in critical habitat watersheds.
• Allocating water under limited supply: the Colorado River basin’s resources are being stressed to unseen levels; for those in the lower basin, annual allocation is critical for planning and operations, yet difficult to predict with current methodologies and sources.

Enter HydroForecast

Inspired by these grand challenges and with a perspective to improve the access and quality of water information globally, Upstream Tech launched a streamflow forecasting system, HydroForecast. Running operationally on four continents, HydroForecast uniquely blends hydrologic theory with machine learning, and leverages advancements in cloud computing, remote sensing data, and the latest academic research to build a more accurate set of tools for water resources managers.

Forecast accuracy unlocks more flexibility for safe dam operations, hydropower generation, supply planning, and a slurry of other benefits. In a live forecasting competition co-hosted by the US Bureau of Reclamation and the CEATI Hydropower Interest Group over water year 2021, across 19 sites in North America and 25 categories of performance metrics, HydroForecast took 23 of the 25 first place wins and was the strongest overall forecast in each region. For hydropower users, this accuracy can increase annual revenue by up to 6%, according to a study by the Idaho National Laboratory. HydroForecast’s seasonal (90-day) water supply forecast outperformed the California Department of Water Resources over the past two spring periods at the five largest reservoirs that control much of California’s water supply and allocation.

Many ask what makes HydroForecast more accurate over alternative forecasting systems. The answer is long, but can be summarized into a few main points:

• The model takes advantage of recent developments in machine learning that are particularly well-suited for making sense of the complicated relationships between the variables that affect flows, like precipitation and temperature. Our peer-reviewed research shows improved accuracy over traditional forecasting approaches. 
• HydroForecast receives information from multiple weather forecast sources, including NOAA and its European equivalent’s leading predictions. This gives it a more thorough picture of what might happen in the coming days and allows it to understand when there is uncertainty ahead, like whether or not a big storm will hit a specific place.
• Satellite imagery and incorporation of any available streamflow measurements show the model near real-time information about the current conditions on the ground in a watershed. This means that its predictions are regularly ‘ground-truthed’ to live conditions and allow it to adapt automatically to changes in the basin that affect how water flows, including conditions like forest fires and urbanization.

Unless we invent a reliable way to create water from other abundant materials and can control when and where it is stored, water managers need better information and tools to squeeze out efficiency where they can. The sector at large is tackling this problem on many fronts, with advancements in monitoring groundwater, satellites that can detect surface water movement and storage, and changes to include forecasts in operational protocols of large dams. Better forecasts are one critical piece of the story, shrinking the uncertainty and gap between what is expected and what arrives, and at longer lead times so that planners can focus on the decisions rather than mitigating error. We can’t paint the complete picture of hydrology twenty years from now, but we can utilize advancements in hydrologic research and operational systems to be robust and prepared to support in however the future unfolds.

Sources

¹ https://patch.com/california/across-ca/california-parched-historic-drought-2022-review

² https://www.scientificamerican.com/article/why-california-is-being-deluged-by-atmospheric-rivers/

³ https://californiawaterblog.com/2023/01/22/is-the-drought-over-reflections-on-californias-recent-flood-drought-combo/

⁴ https://calmatters.org/environment/2022/02/california-water-climate-change-snowpack/