How the distributed model works

Forecasting hourly streamflow in large basins is difficult because model inputs such as elevation and other basin characteristics are heterogeneous and become averaged over a large modeling unit. This poses a challenge in routing the timing and magnitude of water through the basin. Additionally, a lumped model has limitations on the amount of information it can incorporate into predictions; it cannot take advantage of real-time data from streamflow gauges throughout the basin. For this reason, our new distributed model divides catchments up into smaller sub-basins based on spatial elevation data and locations of publicly available flow gauges within the larger catchment.

Basin delineated with the distributed model

How the distributed model predicts flow

To predict flow, the distributed model: 

• Predicts runoff at each hydrologic response unit (HRU) or sub-basin
• Predicts the delay time, spread, and gain/loss of water as it flows through the river network
• The above information is then used to route flows through the river network, starting with the upstream most sub-basins and adding those to the next layer of downstream sub-basins
• As water is routed through the catchment, total flow is computed at every point

As the model is routing flow through the river network, it is also assimilating to any flow gauges within the larger catchment in real-time. This is particularly impactful in regulated basins, where the model can learn certain altered behavior by utilizing gauge information during the training period. As a bonus, the distributed model can also utilize an organization’s outflow observations and forecasted releases to produce accurate regulated flow forecasts.

Interact with sub-basin specific information

In the dashboard, click on each sub-basin to view weather inputs and surface observations averaged across the sub-basin. To view flow forecast points and weather inputs at specified forecast points, click on the blue dot.