This USGS data release contains daily-timestep and monthly-timestep estimates of baseflow at 49 reference stream gages located within 25 miles of the Delaware River basin watershed boundary. Estimates are provided for the available period of record of streamflow data at each site between 1950 and 2015. A two-parameter recursive digital filter was used to estimate baseflow at the selected stream gaging stations using U.S. Geological Survey Groundwater Toolbox (Barlow and others, 2017; Eckhardt, 2005). References cited: Barlow, P.M., Cunningham, W.L., Zhai, T., and Gray, M., 2017, U.S. Geological Survey Groundwater Toolbox version 1.3.1, a graphical and mapping interface for analysis of hydrologic data: U.S. Geological Survey Software Release, 26 May 2017, https://doi.org/10.5066/F7R78C9G. Eckhardt, K., 2005, How to construct recursive digital filters for baseflow separation: Hydrological Processes - An International Journal, v. 19, p. 50-515, https://doi.org/10.1002/hyp.5675.

This metadata record describes monthly estimates of natural baseflow for 15,866 stream reaches, defined by the National Hydrography Dataset Plus Version 2.0 (NHDPlusV2), in the Delaware River Basin for the period 1950-2015. A statistical machine learning technique - random forest modeling (Liaw and Wiener, 2018; R Core Team, 2020) - was applied to estimate natural flows using about 150 potential predictor variables (Miller and others, 2018). Calibration data used for the random forest model are available from (Foks and others, 2020). Each model was run twice, first using all potential predictor variables, which represents a "full" model run, and a second time using the top 20 predictors from the original run, which represents the "partial" model run. Model performance of the full and partial models were compared and identified to be similar. Therefore, predictions for all NHDPlusV2 stream reaches were made using the partial model. Methods used to calibrate the random forest models, and references to predictor data sources are detailed in (Miller and others, 2018). The R scripts used and directions to run the scripts are included in this data release. References cited: Liaw, A., and Wiener, M., 2018, Package 'randomForest': The Comprehensive R Archive Network, https://cran.r-project.org/web/packages/randomForest/randomForest.pdf. Miller, M.P., Carlisle, D.M., Wolock, D.M., and Wieczorek, M., 2018, A database of natural monthly streamflow estimates from 1950 to 2015 for the conterminous United States: Journal of the American Water Resources Association, v. 54, no. 6, p. 1258-1269, https://doi.org/10.1111/1752-1688.12685. Foks, S.S., Miller, M.P., and Hopple, J.A., 2020, Daily-timestep and monthly-timestep estimates of baseflow at 49 reference stream gages located within 25 miles of the Delaware River basin watershed boundary for the years 1950 through 2015: U.S. Geological Survey data release, https://doi.org/10.5066/P9XY70L4. R Core Team, 2020, R-A language and environment for statistical computing: R Foundation for Statistical Computing, https://www.eea.europa.eu/data-and-maps/indicators/oxygen-consuming-substances-in-rivers/r-development-core-team-2006.

This metadata record describes monthly estimates of natural baseflow for 15,866 stream reaches, defined by the National Hydrography Dataset Plus Version 2.0 (NHDPlusV2), in the Delaware River Basin for the period 1950-2015. A statistical machine learning technique - random forest modeling (Liaw and Wiener, 2018; R Core Team, 2020) - was applied to estimate natural flows using about 150 potential predictor variables (Miller and others, 2018). Calibration data used for the random forest model are available from (Foks and others, 2020). Each model was run twice, first using all potential predictor variables, which represents a "full" model run, and a second time using the top 20 predictors from the original run, which represents the "partial" model run. Model performance of the full and partial models were compared and identified to be similar. Therefore, predictions for all NHDPlusV2 stream reaches were made using the partial model. Methods used to calibrate the random forest models, and references to predictor data sources are detailed in (Miller and others, 2018). The R scripts used and directions to run the scripts are included in this data release. References cited: Liaw, A., and Wiener, M., 2018, Package 'randomForest': The Comprehensive R Archive Network, https://cran.r-project.org/web/packages/randomForest/randomForest.pdf. Miller, M.P., Carlisle, D.M., Wolock, D.M., and Wieczorek, M., 2018, A database of natural monthly streamflow estimates from 1950 to 2015 for the conterminous United States: Journal of the American Water Resources Association, v. 54, no. 6, p. 1258-1269, https://doi.org/10.1111/1752-1688.12685. Foks, S.S., Miller, M.P., and Hopple, J.A., 2020, Daily-timestep and monthly-timestep estimates of baseflow at 49 reference stream gages located within 25 miles of the Delaware River basin watershed boundary for the years 1950 through 2015: U.S. Geological Survey data release, https://doi.org/10.5066/P9XY70L4. R Core Team, 2020, R-A language and environment for statistical computing: R Foundation for Statistical Computing, https://www.eea.europa.eu/data-and-maps/indicators/oxygen-consuming-substances-in-rivers/r-development-core-team-2006.

The U.S. Geological Survey (USGS) operates a dense network of streamgages within the Delaware River Basin (DRB) that provides near real-time and daily mean streamflow values, many of which have a decades long period of record. This long-term, historical dataset of daily mean streamflows is crucial for prediction of flows at ungaged stream reaches by means of parameter-based regression equations and can assist in informing water management and use decisions within the DRB. The USGS computed updated flow-duration exceedance probabilities for 98 current (2022) reference streamgages within the DRB with at least 10 years of record and minimally altered (nominal anthropogenic activity such as mining, diversion, or impoundment) streamflow regimes. The Make Plotting Position (MkPP; Granato, 2009) program with Weibull plotting position was used to calculate flow-duration exceedance probabilities for 21 streamflow percentiles. Percent difference comparison between these resulting flow-duration exceedances were made with observed and regression equation predicted streamflow values published by Stuckey (2016) for the 1960 through 2010 time period. A time-series trend analysis of 1979–2022 daily mean streamflows at 19 percentiles was computed using a non-parametric Mann-Kendall trend test (Hirsch and others, 2015). Associated basin characteristic raster layers that are currently used in StreamStats for the DRB, including percentage sand in soil, percentage poorly drained soils, percentage urban land use, mean annual precipitation, mean winter precipitation (December–February), and soil hydraulic conductivity, are included in this data release. Citations: Granato, G.E., 2009, Computer programs for obtaining and analyzing daily mean streamflow data from the U.S. Geological Survey National Water Information System Web Site: U.S. Geological Survey Open-File Report 2008–1362, 123 p. on CD-ROM, 5 appendixes, https://pubs.usgs.gov/publication/ofr20081362 Hirsch R.M., DeCicco L.A., 2015, User Guide to Exploration and Graphics for RivEr Trends (EGRET) and dataRetrieval: R Packages for Hydrologic Data: U.S. Geological Survey Techniques and Methods 4-A10, https://pubs.usgs.gov/tm/04/a10/ Stuckey, M.H., 2016, Estimation of daily mean streamflow for ungaged stream locations in the Delaware River Basin, water years 1960–2010: U.S. Geological Survey Scientific Investigations Report 2015–5157, 42 p., http://dx.doi.org/10.3133/sir20155157

The U.S. Geological Survey (USGS) operates a dense network of streamgages within the Delaware River Basin (DRB) that provides near real-time and daily mean streamflow values, many of which have a decades long period of record. This long-term, historical dataset of daily mean streamflows is crucial for prediction of flows at ungaged stream reaches by means of parameter-based regression equations and can assist in informing water management and use decisions within the DRB. The USGS computed updated flow-duration exceedance probabilities for 98 current (2022) reference streamgages within the DRB with at least 10 years of record and minimally altered (nominal anthropogenic activity such as mining, diversion, or impoundment) streamflow regimes. The Make Plotting Position (MkPP; Granato, 2009) program with Weibull plotting position was used to calculate flow-duration exceedance probabilities for 21 streamflow percentiles. Percent difference comparison between these resulting flow-duration exceedances were made with observed and regression equation predicted streamflow values published by Stuckey (2016) for the 1960 through 2010 time period. A time-series trend analysis of 1979–2022 daily mean streamflows at 19 percentiles was computed using a non-parametric Mann-Kendall trend test (Hirsch and others, 2015). Associated basin characteristic raster layers that are currently used in StreamStats for the DRB, including percentage sand in soil, percentage poorly drained soils, percentage urban land use, mean annual precipitation, mean winter precipitation (December–February), and soil hydraulic conductivity, are included in this data release. Citations: Granato, G.E., 2009, Computer programs for obtaining and analyzing daily mean streamflow data from the U.S. Geological Survey National Water Information System Web Site: U.S. Geological Survey Open-File Report 2008–1362, 123 p. on CD-ROM, 5 appendixes, https://pubs.usgs.gov/publication/ofr20081362 Hirsch R.M., DeCicco L.A., 2015, User Guide to Exploration and Graphics for RivEr Trends (EGRET) and dataRetrieval: R Packages for Hydrologic Data: U.S. Geological Survey Techniques and Methods 4-A10, https://pubs.usgs.gov/tm/04/a10/ Stuckey, M.H., 2016, Estimation of daily mean streamflow for ungaged stream locations in the Delaware River Basin, water years 1960–2010: U.S. Geological Survey Scientific Investigations Report 2015–5157, 42 p., http://dx.doi.org/10.3133/sir20155157

This data release contains daily time series estimates of natural streamflow at 5,439 GAGES-II non-reference streamgages in 19 study regions across the conterminous United States from October 1, 1980 through September 30, 2017, using five statistical techniques: nearest-neighbor drainage area ratio (NNDAR), map-correlation drainage area ratio (MCDAR), nearest-neighbor nonlinear spatial interpolation using flow duration curves (NNQPPQ), map-correlation nonlinear spatial interpolation using flow duration curves (MCQPPQ), and ordinary kriging of the logarithms of discharge per unit area (OKDAR). NNDAR, MCDAR, NNQPPQ, and MCQPPQ estimates were computed following methods described in Farmer and others (2014), with updates to the flow-duration curve modeling which is described in Over and others (2018). OKDAR estimates were computed using pooled variograms for each study region following methods described in Farmer (2016). Daily streamflow estimation was conducted by study region (hydrologic unit code level-2 regions as defined in Falcone, 2011) by building statistical models using 1,385 GAGES-II reference streamgages from mostly undisturbed watersheds as index gages (Russell and others, 2020). Estimates were then made at GAGES-II non-reference streamgages. Location information and basin characteristics for study gages were obtained from the GAGES-II dataset (Falcone, 2011). Observed daily streamflow data were retrieved from the National Water Information System (USGS, 2019). This data release contains 19 separate zip files; one for each study region. Each zip file contains an individual tab-delimited text file for each non-reference streamgage in the study region. A text file summarizing period of record information for each non-reference streamgage is provided (non-reference_gages_summary.csv). This data release also contains a text file (Model_info.csv) of regional regression equations for 27 flow quantiles that were developed in each study region in order to implement the QPPQ methods and a text file (BC_transformations.csv) describing transformations made to the GAGES-II derived basin characteristics prior to use in the regression equations. The five sets of streamflow estimates represent expected natural streamflow conditions with minimal disturbance by human activities, in other words, without the effects of regulation, diversion, land development, or other anthropogenic activities. The observed streamflow records at the non-reference streamgages were compared to the five simulated streamflow records. These performance metrics are provided at each gage for all five statistical methods (NonRef_PMs_byStation.csv) and as summaries by region (NonRef_PM_summaries_byRegion.csv). References cited: Falcone, J.A., 2011, GAGES-II: Geospatial Attributes of Gages for Evaluating Streamflow [digital spatial dataset]: U.S. Geological Survey Water Resources NSDI Node web page, https://water.usgs.gov/lookup/getspatial?gagesII_Sept2011. Farmer, W.H., Archfield, S.A., Over, T.M., Hay, L.E., LaFontaine, J.H., and Kiang, J.E., 2014, A comparison of methods to predict historical daily streamflow time series in the southeastern United States: U.S. Geological Survey Scientific Investigations Report 2014–5231, 34 p., http://dx.doi.org/10.3133/sir20145231. Farmer, W. H., 2016, Ordinary kriging as a tool to estimate historical daily streamflow records, Hydrology and Earth System Sciences, 20, 2721-2735, https://doi.org/10.5194/hess-20-2721-2016. Over, T.M., Farmer, W.H., and Russell, A.M., 2018, Refinement of a regression-based method for prediction of flow-duration curves of daily streamflow in the conterminous United States: U.S. Geological Survey Scientific Investigations Report 2018–5072, 34 p., https://doi.org/10.3133/sir20185072. Russell, A.M., Over, T.M., and Farmer, W.H., 2020, Cross-validation results for five statistical methods of daily streamflow estimation at 1,385 reference streamgages in the conterminous United States, Water Years

These data include base flow separation estimates for 64 USGS streamflow gages in the Lake Superior watershed from 1945 to 2020, shapefiles of the gaging stations and watersheds for each gaging station, and a zipped folder of graphics of the base flow separation results. The base flow separation estimates were calculated using the U.S. Geological Survey Groundwater Toolbox (Barlow and others, 2014) for any complete water years of record for these gages from 1945 to 2020. The shapefile of the gaging stations includes the starting and ending years of data for each station, the number of years of record. The watersheds shapefile includes the source for the watershed delineation, the watershed area, and the number of upstream and(or) downstream gaging stations on the same river system. If there are upstream gaging stations in the river system, the watershed delineated is only the incremental part of the watershed between gaging stations. The baseflow separation estimates for each gaging station include daily, monthly, and annual output from the Groundwater Toolbox for six estimation methods included in the software (full references are available in Barlow and others, 2014): the baseflow Index-Standard method, HySep Fixed Interval, HySep Local Minimum, HySep Sliding Interval, baseflow Index-Modified, PART, and BFLOW. A summary of the annual baseflow estimates for all the gaging stations using all the methods is provided also is included in this data release. This data release is one of three child items under the overall data release at https://doi.org/10.5066/P9084UKQ.

As part of the U.S. Geological Survey’s Groundwater Resources Program study of the Appalachian Plateaus aquifers, estimates of annual water-budget components were determined at 849 continuous-record streamflow gaging stations from Mississippi to New York. Base flow, which can serve as a proxy for annual recharge, streamflow, and runoff were estimated from computer programs—PART (Rutledge, 1993), HYSEP (Sloto and Crouse, 1996), and BFI (Wahl and Wahl, 1988)—that are included in the hydrograph analysis component provided with version 1.0 of the U.S. Geological Survey Groundwater Toolbox. Only complete years (January to December) of record at each gage were used to determine annual estimates. Estimates of base-flow index, which is the percentage of streamflow from base flow, are included in the annual and average tables. Precipitation was estimated by calculating the average of cell values in the PRISM dataset intercepted by basin boundaries where previously defined in the GAGES-II dataset (Falcone, 2011). Estimates of evapotranspiration were then calculated from the difference between precipitation and streamflow.

As part of the U.S. Geological Survey’s Groundwater Resources Program study of the Appalachian Plateaus aquifers, estimates of annual water-budget components were determined at 849 continuous-record streamflow gaging stations from Mississippi to New York. Base flow, which can serve as a proxy for annual recharge, streamflow, and runoff were estimated from computer programs—PART (Rutledge, 1993), HYSEP (Sloto and Crouse, 1996), and BFI (Wahl and Wahl, 1988)—that are included in the hydrograph analysis component provided with version 1.0 of the U.S. Geological Survey Groundwater Toolbox. Only complete years (January to December) of record at each gage were used to determine annual estimates. Estimates of base-flow index, which is the percentage of streamflow from base flow, are included in the annual and average tables. Precipitation was estimated by calculating the average of cell values in the PRISM dataset intercepted by basin boundaries where previously defined in the GAGES-II dataset (Falcone, 2011). Estimates of evapotranspiration were then calculated from the difference between precipitation and streamflow.

This data release contains daily time series estimates of natural streamflow for 1,385 streamgages in 19 study regions in the conterminous U.S. from October 1, 1980, through September 30, 2017. These estimates are provided for gages from mostly undisturbed watersheds as defined by Falcone (2011), using five statistical techniques: nearest-neighbor drainage area ratio (NNDAR), map-correlation drainage area ratio (MCDAR), nearest-neighbor nonlinear spatial interpolation using flow duration curves (NNQPPQ), map-correlation nonlinear spatial interpolation using flow duration curves (MCQPPQ), and ordinary kriging of the logarithms of discharge per unit area (OKDAR). Location information and basin characteristics for study gages were obtained from the "Reference" gages of the GAGES-II dataset (Falcone, 2011, https://water.usgs.gov/lookup/getspatial?gagesII_Sept2011). Observed daily streamflow data were retrieved from the National Water Information System (NWIS) on September 7, 2018. NNDAR, MCDAR, NNQPPQ, and MCQPPQ estimates were computed following methods described by Farmer and others (2014), with updates to the flow-duration curve modeling which is described by Over and others (2018). OKDAR estimates were computed using pooled variograms for each study region following methods described by Farmer (2016). Daily streamflow estimation was conducted in a leave-one-out-cross-validation approach where each streamgage was treated as if ungaged and all the remaining streamgages in a study region were used to calibrate each method and perform estimations at the "ungaged" site. The observed streamflow records were compared to the five simulated streamflow records to help assess performance of each method. These performance metrics are provided at each gage for all five statistical methods. References cited: Falcone, J.A., 2011, GAGES-II: Geospatial Attributes of Gages for Evaluating Streamflow [digital spatial dataset] : U.S. Geological Survey Water Resources NSDI Node web page, https://water.usgs.gov/lookup/getspatial?gagesII_Sept2011. Farmer, W.H., Archfield, S.A., Over, T.M., Hay, L.E., LaFontaine, J.H., and Kiang, J.E., 2014, A comparison of methods to predict historical daily streamflow time series in the southeastern United States: U.S. Geological Survey Scientific Investigations Report 2014–5231, 34 p., http://dx.doi.org/10.3133/sir20145231. Farmer, W. H., 2016, Ordinary kriging as a tool to estimate historical daily streamflow records, Hydrology and Earth System Sciences, 20, 2721-2735, https://doi.org/10.5194/hess-20-2721-2016. Over, T.M., Farmer, W.H., Russell, A.M., 2018, Refinement of a regression-based method for prediction of flow-duration curves of daily streamflow in the conterminous United States; U.S. Geological Survey Scientific Investigations Report 2018–5072, https://doi.org/10.3133/sir20185072.