Hurricane Ida was an extremely powerful Atlantic hurricane that moved through the Southern United States at the end of August, 2021. When Hurricane Ida hit, the additional heavy rainfall increased the operational challenges for organizations in a region that is already heavily rain-driven, where flashy storms can occur quickly particularly in the summer months. Examples of the operational challenges augmented by heavy rain include avoiding lost revenue due to spill, preventing overtopping, and ensuring the safety of downstream communities in flood situations.
To improve outcomes during such events, HydroForecast uses a unique combination of statistical modeling constrained by physical processes in a neural network framework. This combination allows the model to quickly interpret a wide variety of data sources and deliver highly accurate forecasts even in scenarios that lie outside of the historical record.
Below is an example of how HydroForecast predicted the impacts of Hurricane Ida in the Emory River at Oakdale, Tennessee. The event occurred between August 31 - September 3, 2021.
Emory River at Oakdale, Tennessee. Basin size =~2,000 km2, characterized by rugged and hilly slopes on the Cumberland Plateau. Rainfall comes from hurricanes, summer convective currents, and fronts that move across the Eastern U.S.
The black line in the plot below shows the flow that occurred when Hurricane Ida hit. The Emory River reached a peak flow rate of 8,553 cfs on September 1, 2021.
4 days ahead of the event, HydroForecast interprets discrepancies among weather forecasts to capture the eventual flow close to the median forecast, and within its 50% confidence interval.
Two days ahead of the peak flows, the HydroForecast median predicted the ultimate magnitude of peak flows with less than 2% error and continued to predict within this error percentage through the event.
The HydroForecast median accurately predicted the flow recession after the peak.
HydroForecast’s unique combination of physical and statistical modeling creates a powerful tool that helps organizations to de-risk decisions leading up to difficult-to-forecast events such as heavy rainfall in a rainy basin. Moreover, its neural network core enables it to assess a wider breadth of factors such as MODIS satellite data and multiple weather forecasts, and weight each factor to present as accurate a forecast as possible.
If you’re interested in learning more about HydroForecast and wish to view other examples, get in touch with our team at firstname.lastname@example.org or visit our website at https://www.upstream.tech/hydroforecast.