Streamflow forecasts for a scenario of persistent dry weather can be used by water and other natural resource managers to assess the earliest date when streamflow could reach levels that will trigger actions including reservoir releases or the curtailment of withdrawals. Low-flow forecasts can also be used for drought mitigation activities that require advanced planning such as leasing water for instream flows or switching from surface water to groundwater for water supplies. This website provides access to baseflow forecasts for river and streams across the United States using the Baseflow Separation Model, BFS (Konrad 2022). The model is described in the report BFS—A non-linear, state-space model for baseflow separation and prediction. The model code and site-specific parameter values are available from the BFS Model data release (Konrad 2020). The baseflow forecasts represent the lowest streamflow that is likely to be observed over the next 90 days assuming no rain and decreasing or no snowmelt.
Baseflow is the reliable, slowly varying component of streamflow that sustains water availability for human uses, aquatic habitats, and riparian vegetation during dry periods. In many rivers, groundwater discharge sustains baseflow but meltwater from snowfields and glaciers or outflow from lakes are also important sources of baseflow in some rivers and streams. Baseflow represents the lower limit on streamflow because it does not include contributions of runoff from rainfall or rapid snowmelt. Baseflow is predictable because it varies slowly over time, so it can be forecast using relatively simple models and current streamflow conditions.
The BFS model calculates baseflow from an observed streamflow record and will forecast baseflow for periods of dry weather with decreasing or no snowmelt. Baseflow forecasts represent the lowest streamflow likely to be observed. Every week, the BFS model forecast baseflow at sites with real-time streamflow measurement where the model has been calibrated. Refer to the map to see where these sites are located. Baseflow is forecast for the next 90 days assuming no additional rainfall, increased snowmelt, or reservoir releases that would increase streamflow. The forecasts are based on minimum observed streamflow over the past 30-60 days and the typical rates that streamflow decreases at a site during dry periods. Baseflow forecasts are NOT probabilistic forecasts of streamflow. The probability that streamflow will have a given value in the future depends on current conditions but also the likelihood of rainfall, increasing snowmelt, or reservoir releases, which are not incorporated in the baseflow forecasts produced by the BFS model.
The BFS model was calibrated at 13,208 sites using an automated routine to maximize baseflow without exceeding daily streamflow using records available for water year 1981 to 2015. The full list of sites, including those where forecasts are not available, with their parameter values is available from this data release. Model calibration does not assure reliable forecasts and model error for a site should be evaluated to understand model limitations for any use. Hydrographs for a portion of the calibration period can be viewed by a selecting a site from the map. Model parameters for a site recalibrated to forecast baseflow for particular conditions of interest including extreme dry periods rarely observed during the calibration period or if there have been recent changes in the watershed of a site that could affect runoff or recharge including but not limited to:
Baseflow forecasts require a recent streamflow observation (within the last 60 days) that is primarily baseflow. The BFS model generally does not provide reliable forecasts when streamflow is:
You should understand the uncertainty of forecasts from the BFS model at a site and determine whether the forecast uncertainty at a site is acceptable for your purposes by assessing the five factors affecting forecast reliability (above), examining the hydrograph from model calibration as an example of model performance at a site (accessible from the map), and considering recent conditions at a site. The forecast hydrograph will include a dashed line labeled "Minimum streamflow for model calibration" indicating the lowest observed streamflow used for model calibration. Forecasts below this line have high uncertainty.
Baseflow forecasts can be accessed for individual sites and major river basins from the map. The controls for viewing layers are in the box in the upper right corner of the map. The box with +/- controls the zoom level. The open hand icon allows you to pan (move) the map by holding down your primary mouse button. The hand icon with one finger raised allows you to select a site or major river basin by clicking your primary mouse button. A list of major river basin with links for downloading forecasts are also available here.
Check "Forecast Sites" in the layer controls, zoom/plan to an area of interest, and select a site from the map. A pop-up will appear with links to information available for the site:
Baseflow forecasts will not be available if the site does not have real-time streamflow or recently had no flow.
All available forecasts in a major river basin can be downloaded by checking the "Major River Basin" in the layer control and clicking anywhere in the basin that is not a site. A pop-up will appear with the name of the major river basin, the 4-digit hydrologic unit code, and links to download data as a comma delimited text file in either "long" format (all forecasted baseflows in a single column with the site identification number in a column) or "wide" format (each site in column with the site identification number as the column label).
Results for all sites in the US are available as a comma delimited text file from this link: bfprj_all.csv.
The BFS model produces a daily time series of baseflow with confidence intervals indicating the likely range of streamflow given baseflow and no rain. The output can be viewed as a hydrograph (see example to the right) or downloaded as a comma delimited, ASCII text file. The download file for a site will have the name "bfprj_[SiteID].csv" where SiteID is a character string used by the operating agency to identify the site (e.g., USGS station number).
The baseflow forecast file for a site will have five columns:
Date - Formatted as a character string, YYYY-MM-DD, where YYYY is the four digit year, MM is the two digit month, and DD is the two digit day of the month.
Qobserved.cfs - Real-time observations of mean daily streamflow in cubic feet per second for the past 60 days.
Baseflow.cfs - Daily forecasts of baseflow in cubic feet per second for the next 90 days assuming no rain.
StreamflowCB05.cfs - 5th percentile of streamflow in cubic feet per second for recessional periods during the calibration period when baseflow was similar to the forecast. The 5th percentile of observed streamflow given a baseflow value is used as the lower bound on the 90% confidence interval for future streamflow assuming persistent dry conditions.
StreamflowCB95.cfs - 95th percentile of streamflow in cubic feet per second for recessional periods during the calibration period when baseflow was similar to the forecast. The 95th percentile of observed streamflow given a baseflow value is used as the upper bound on the 90% confidence interval for future streamflow assuming persistent dry conditions.
The 5th to 95th confidence interval indicate the range of observed streamflow given predicted baseflow on days when streamflow was decreasing during the calibration period.
The file with baseflow forecasts for a major river basin in wide format has dates in the first column and daily values of baseflow for each site in a column labeled as "Baseflow.cfs.[SITE ID]". The long-format file for a major river basin or for the US has daily values for each site in a row and five columns with the headers:
Note that the output files for major river basins and for the US do not include observed streamflow.
The baseflow separation model BFS has been approved for release by the U.S. Geological Survey (USGS). Although the model has been subjected to rigorous review, the USGS reserves the right to update the model as needed pursuant to further analysis and review. No warranty, expressed or implied is made by the USGS or the U.S. Government as to the functionality of the model, the reliability of its output, and related material nor shall the fact of release constitute any such warranty. Furthermore, the model output is released on condition that neither the USGS nor the U.S. Government shall be held liable for any damages resulting from its authorized or unauthorized use. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.