This data release contains the standard statistical suite (version 1.0) daily streamflow performance benchmark results for the National Water Model Retrospective (v2.1) at streamflow benchmark locations defined by Foks and others (2022). Modeled hourly timesteps were converted to mean daily timesteps. Model error was determined by evaluating predicted daily mean streamflow versus observed daily mean streamflow using various statistics; the Nash-Sutcliffe efficiency (NSE), the Kling-Gupta efficiency (KGE), the logNSE, the Pearson correlation coefficient, the Spearman correlation coefficient, the ratio of the standard deviation, the percent bias, the percent bias in flow duration curve midsegment slope, the percent bias in the flow duration curve high-segment volume, and the percent bias in flow duration curve low-segment volume. Two climatological KGE benchmarks are included that are calculated using daily mean streamflow observations and interannual daily mean or median flows. Additionally, KGE uncertainty estimates have been added as a separate csv file including the standard error of jackknife, standard error of bootstrap, the 5th, 50th and 95th percentiles of the estimates, the jackknife score, the bias of jackknife, the bias of bootstrap, and the standard error of jackknife after bootstrap.

This data release contains the standard statistical suite (version 1.0) daily streamflow performance benchmark results for the National Hydrologic Model Infrastructure application of the Precipitation-Runoff Modeling System (NHM-PRMS) version 1 "byObs" calibration with Muskingum routing computed at streamflow benchmark locations defined by Foks and others (2022). Model error was determined by evaluating predicted daily mean streamflow versus observed daily mean streamflow using various statistics; the Nash-Sutcliffe efficiency (NSE), the Kling-Gupta efficiency (KGE), the logNSE, the Pearson correlation coefficient, the Spearman correlation coefficient, the ratio of the standard deviation, the percent bias, the percent bias in flow duration curve midsegment slope, the percent bias in the flow duration curve high-segment volume, and the percent bias in flow duration curve low-segment volume. Two climatological KGE benchmarks are included that are calculated using daily mean streamflow observations and interannual daily mean or median flows. Additionally, KGE uncertainty estimates have been added as a separate csv file including the standard error of jackknife, standard error of bootstrap, the 5th, 50th and 95th percentiles of the estimates, the jackknife score, the bias of jackknife, the bias of bootstrap, and the standard error of jackknife after bootstrap.

This data release contains the standard statistical suite (version 1.0) daily streamflow performance benchmark results for the National Water Model Retrospective (v2.1) at streamflow benchmark locations defined by Foks and others (2022). Modeled hourly timesteps were converted to mean daily timesteps. Model error was determined by evaluating predicted daily mean streamflow versus observed daily mean streamflow using various statistics; the Nash-Sutcliffe efficiency (NSE), the Kling-Gupta efficiency (KGE), the logNSE, the Pearson correlation coefficient, the Spearman correlation coefficient, the ratio of the standard deviation, the percent bias, the percent bias in flow duration curve midsegment slope, the percent bias in the flow duration curve high-segment volume, and the percent bias in flow duration curve low-segment volume. Two climatological KGE benchmarks are included that are calculated using daily mean streamflow observations and interannual daily mean or median flows. Additionally, KGE uncertainty estimates have been added as a separate csv file including the standard error of jackknife, standard error of bootstrap, the 5th, 50th and 95th percentiles of the estimates, the jackknife score, the bias of jackknife, the bias of bootstrap, and the standard error of jackknife after bootstrap.

This data release contains the standard statistical suite (version 1.0) daily streamflow performance benchmark results for the National Water Model Retrospective (v2.1) at streamflow benchmark locations defined by Foks and others (2022). Modeled hourly timesteps were converted to mean daily timesteps. Model error was determined by evaluating predicted daily mean streamflow versus observed daily mean streamflow using various statistics; the Nash-Sutcliffe efficiency (NSE), the Kling-Gupta efficiency (KGE), the logNSE, the Pearson correlation coefficient, the Spearman correlation coefficient, the ratio of the standard deviation, the percent bias, the percent bias in flow duration curve midsegment slope, the percent bias in the flow duration curve high-segment volume, and the percent bias in flow duration curve low-segment volume. Two climatological KGE benchmarks are included that are calculated using daily mean streamflow observations and interannual daily mean or median flows. Additionally, KGE uncertainty estimates have been added as a separate csv file including the standard error of jackknife, standard error of bootstrap, the 5th, 50th and 95th percentiles of the estimates, the jackknife score, the bias of jackknife, the bias of bootstrap, and the standard error of jackknife after bootstrap.

This data release contains the D-score (version 0.1) daily streamflow performance benchmark results for the National Hydrologic Model Infrastructure application of the Precipitation-Runoff Modeling System (NHM-PRMS) version 1 "byObs" calibration with Muskingum routing (Hay and LaFontaine, 2020) computed at streamflow benchmark locations (version 1.0) as defined by Foks and others (2022). Model error was determined by evaluating predicted daily mean streamflow versus observed daily mean streamflow. Using those errors, the D-score performance benchmark computes the mean squared logarithmic error (MSLE), then decomposes the overall MSLE into orthogonal components such as bias, distribution, and sequence (Hodson and others, 2021). For easier interpretation, the MSLE components can be passed through a scoring function as described in Hodson and others (2021).

This data release contains the D-score (version 0.1) daily streamflow performance benchmark results for the National Hydrologic Model Infrastructure application of the Precipitation-Runoff Modeling System (NHM-PRMS) version 1 "byObs" calibration with Muskingum routing (Hay and LaFontaine, 2020) computed at streamflow benchmark locations (version 1.0) as defined by Foks and others (2022). Model error was determined by evaluating predicted daily mean streamflow versus observed daily mean streamflow. Using those errors, the D-score performance benchmark computes the mean squared logarithmic error (MSLE), then decomposes the overall MSLE into orthogonal components such as bias, distribution, and sequence (Hodson and others, 2021). For easier interpretation, the MSLE components can be passed through a scoring function as described in Hodson and others (2021).

This data release contains the D-score (version 0.1) daily streamflow performance benchmark results for the National Water Model (NWM) Retrospective version 2.1 computed at streamgage benchmark locations (version 1) as defined by Foks and others (2022). Model error was determined by evaluating predicted daily mean streamflow (aggregated from an hourly timestep) versus observed daily mean streamflow. Using those errors, the D-score performance benchmark computes the mean squared logarithmic error (MSLE), then decomposes the overall MSLE into orthogonal components such as bias, distribution, and sequence (Hodson and others, 2021). For easier interpretation, the MSLE components can be passed through a scoring function as described in Hodson and others (2021). References: Foks, S.S., Towler, E., Hodson, T.O., Bock, A.R., Dickinson, J.E., Dugger, A.L., Dunne, K.A., Essaid, H.I., Miles, K.A., Over, T.M., Penn, C.A., Russell, A.M., Saxe, S.W., and Simeone, C.E., 2022, Streamflow benchmark locations for conterminous United States (cobalt gages): U.S. Geological Survey data release, https://doi.org/10.5066/P972P42Z. Hodson, T.O., Over, T.M., and Foks, S.S., 2021. Mean squared error, deconstructed. Journal of Advances in Modeling Earth Systems, 13, e2021MS002681. https://doi.org/10.1029/2021MS002681.

This data release contains the D-score (version 0.1) daily streamflow performance benchmark results for the National Water Model (NWM) Retrospective version 2.1 computed at streamgage benchmark locations (version 1) as defined by Foks and others (2022). Model error was determined by evaluating predicted daily mean streamflow (aggregated from an hourly timestep) versus observed daily mean streamflow. Using those errors, the D-score performance benchmark computes the mean squared logarithmic error (MSLE), then decomposes the overall MSLE into orthogonal components such as bias, distribution, and sequence (Hodson and others, 2021). For easier interpretation, the MSLE components can be passed through a scoring function as described in Hodson and others (2021). References: Foks, S.S., Towler, E., Hodson, T.O., Bock, A.R., Dickinson, J.E., Dugger, A.L., Dunne, K.A., Essaid, H.I., Miles, K.A., Over, T.M., Penn, C.A., Russell, A.M., Saxe, S.W., and Simeone, C.E., 2022, Streamflow benchmark locations for conterminous United States (cobalt gages): U.S. Geological Survey data release, https://doi.org/10.5066/P972P42Z. Hodson, T.O., Over, T.M., and Foks, S.S., 2021. Mean squared error, deconstructed. Journal of Advances in Modeling Earth Systems, 13, e2021MS002681. https://doi.org/10.1029/2021MS002681.

This data release consists of 5390 streamflow gages within the conterminous United States that will serve as version 1.0 of streamflow benchmark locations for hydrologic model evaluation and benchmarking.

This data set serves to archive the data, analysis and models of the associated publication entitled “Calibration of the USGS National Hydrologic Model in Ungauged Basins Using Statistical At-Site Streamflow Simulations” as published in the Journal of Hydrologic Engineering. The input data files included here as comma-separated values contain measured streamflow, streamflow simulated by the Precipitation-Runoff Modeling System calibrated to measured streamflow, streamflow simulated by the Precipitation-Runoff Modeling System calibrated to streamflow simulated by pooled ordinary kriging, and streamflow simulated by pooled ordinary kriging at 1,410 streamgage locations across the United States. These data sets, built on previously published models, are assessed in the included analysis script (R programming language) to reproduce the findings of the associated manuscript. The manuscript argues that statistically generated daily streamflow can be used to support the ability of physical models to represent hydrologic processes at ungauged locations. The objective of this study was to determine the feasibility of using simulations in place of measured streamflow to calibrate physical models in ungauged basins. Calibrating with statistically simulated streamflow produced performances within 23% of applications with knowledge of at-site measurements. Furthermore, statistically generated streamflows produced accurate timing information, which, when combined with alternative data sets (e.g., evapotranspiration, recharge, etc.), can be used to improve representation of hydrologic processes at ungauged locations.