This page contains 15 estimated quantiles for 9,203 level-12 Hydrologic Unit Code in the Southeastern United States for the decades 1950-1959, 1960-1969, 1970-1979, 1980-1989, 1990-1999, and 2000-2009. A multi-output neural network was used to generate the estimated quantiles (Worland and others, 2019). The R scripts that generated the predictions are also included along with a README file. The 15 quantiles are associated with the following 15 non-exceedance probabilities (NEPs): 0.0003, 0.0050, 0.0500, 0.1000, 0.2000, 0.3000, 0.4000, 0.5000, 0.6000, 0.7000, 0.8000, 0.9000, 0.9500, 0.9950, and 0.9997. The quantiles were calculated using the Weibull plotting position (more details can be found in the accompanying manuscript). In addition to the median estimate of the quantiles, 68th, 95th, and 99.7th percentile intervals are also included in .csv file. The percentile intervals were estimated using Monte-Carlo dropout for 500 forward passes of the neural network. The intervals are represented in the .csv file as p0.0015, p0.0250, p0.1600, p0.5000, p0.8400, p0.975, and p0.9985 which indicates the 68th, 95th, and 99.7th percentile intervals. The median (p0.5000) and the mean estimate should be used if only a single realization of the estimated quantiles is needed. The neural network was trained using streamflow data at sites with records that contained only non-zero streamflow values. However, the model was used to make predictions for every HUC12 pour point. Some of these predictions are likely for sites that have streamflow values equal to zero. Worland, S. C., Steinschneider, S., Asquith, W., Knight, R. and Wieczorek, M., 2019, Prediction and inference of flow-duration curves using multi-output neural networks, Water Resources Research , submitted.

Using previously published (Robinson and others, 2019) no-flow fractions and L-moments of nonzero streamflow from decadal streamflow flow-duration analysis (daily mean streamflow), probability distributions were fit to provide 27 estimated quantiles of decadal flow-duration curves, and hence the probability distributions are a form of parametric modeling that ensures monotonicity of the quantiles by non-exceedance probability (NEP). For both U.S. Geological Survey streamflow-gaging stations (streamgages) and level-12 hydrologic unit code (HUC12) catchments, as defined by Crowley-Ornelas and others (2019), the 27 quantiles were estimated and tabulated in this data release. Three probability distributions were used and are summarized by Asquith and others (2017): the asymmetric exponential power (AEP4) (4-parameter), generalized normal (GNO) (3-parameter log-normal), and kappa (KAP) (4-parameter). A summary of the mathematics for these distributions is provided in the README files within this data release and close consultation of the mathematical discussion in Asquith and others (2017) also is suggested. The lmomco R package (Asquith, 2020) was used for distribution fitting and the technically-demanding implementation for a single location is archived in the RESTORE/fdclmrpplo software release within file fdclmrpplo/scripts/pred_fdc_ref/pred_fdc_ref.R (Asquith and others, 2020). The implementation for the streamgages is archived in the RESTORE/fdclmrpplo software release within file fdclmrpplo/scripts/pred_fdc_gage/pred_fdc_gage.R, and the implementation for the HUC12s is archived file fdclmrpplo/scripts/pred_fdc_huc12/pred_fdc_huc12.R and README files therein. For a given data set of no-flow fraction and L-moments, the three distributions will have similar results in the central parts of NEP and differences will be largest in the far left (low flow) and far right (flood flow) tails. No opinion that a particular distribution is more suitable than another is provided with exception that the GNO is fit to the first three L-moments and the AEP4 and KAP are fit to the first four L-moments. As a result, it is logical to state that more information on the distribution of streamflow is retained by the AEP4 and KAP distributions than the GNO. The availability of three distributions with the data release is considered a feature because a semi-quantitative assessment of model error (uncertainty attributed to choice of model) can be made. Asquith, W.H., 2020, lmomco—L-moments, censored L-moments, trimmed L-moments, L-comoments, and many distributions: R package version 2.3.6, https://CRAN.R-project.org/package=lmomco. Asquith, W.H., Kiang, J.E., and Cohn, T.A., 2017, Application of at-site peak-streamflow frequency analyses for very low annual exceedance probabilities: U.S. Geological Survey Scientific Investigation Report 2017–5038, 93 p., https://doi.org/10.3133/sir20175038. Asquith, W.H., Knight, R.R., and Crowley-Ornelas, E.R., 2020, RESTORE/fdclmrpplo—Source code for estimation of L-moments and percent no-flow conditions for decadal flow-duration curves and estimation at level-12 hydrologic unit codes along with other statistical computations: U.S. Geological Survey software release, Reston, Va., https://doi.org/10.5066/P93CKH92. Crowley-Ornelas, E.R., Worland, S.C., Wieczorek, M.E., Asquith, W.H., Knight, R.R., 2019, Summary of basin characteristics for National Hydrography Dataset, version 2 catchments in the Southeastern United States, 1950–2010: U.S. Geological Survey data release, https://doi.org/10.5066/P9KXTDU4. Robinson, A.L., Asquith, W.H., and Knight, R.R., 2019, Summary of decadal no-flow fractions and decadal L-moments of nonzero streamflow flow-duration curves for National Hydrography Dataset, version 2 catchments in the southeastern United States, 1950–2010: U.S. Geological Survey data release, https://doi.org/10.5066/P9Z4PM55.

Two methods of calculating hydrologic alteration were applied to modeled daily streamflow data for 9,201 12-digit hydrologic unit code (HUC12) pour points draining to the Gulf of Mexico (Robinson and others, 2020). The first method is a new modified method of calculating ecosurplus and ecodeficit called hydro change. For this project, ecosurplus and ecodeficit have been combined to assess overall hydrologic regime change. The second method is the confidence interval hypothesis test (Kroll and others, 2015). The first method is a means of quantifying hydrologic alteration while the second is a hypothesis test to simply determine if statistically significant alteration has occurred. Both methods are employed to determine which is best at analyzing alteration of the hydrologic regime in the Gulf Coast Ecosystem Restoration Council (RESTORE) study area. Statistical analysis was done in RStudio (2020). The data release includes four attached files: (1) metadata .xml file, (2) csv with the p-values for each HUC12, (3) csv with results from the hydrologic change analysis, and (4) the shapefile of the pour point locations for the HUC12s used in the analyses.

Two methods of calculating hydrologic alteration were applied to modeled daily streamflow data for 9,201 12-digit hydrologic unit code (HUC12) pour points draining to the Gulf of Mexico (Robinson and others, 2020). The first method is a new modified method of calculating ecosurplus and ecodeficit called hydro change. For this project, ecosurplus and ecodeficit have been combined to assess overall hydrologic regime change. The second method is the confidence interval hypothesis test (Kroll and others, 2015). The first method is a means of quantifying hydrologic alteration while the second is a hypothesis test to simply determine if statistically significant alteration has occurred. Both methods are employed to determine which is best at analyzing alteration of the hydrologic regime in the Gulf Coast Ecosystem Restoration Council (RESTORE) study area. Statistical analysis was done in RStudio (2020). The data release includes four attached files: (1) metadata .xml file, (2) csv with the p-values for each HUC12, (3) csv with results from the hydrologic change analysis, and (4) the shapefile of the pour point locations for the HUC12s used in the analyses.

This data release provides the data and R scripts used for the 2018 publication titled "Improving predictions of hydrological low-flow indices in ungaged basins using machine learning", Environmental Modeling and Software, https://doi.org/10.1016/j.envsoft.2017.12.021. There are two .csv files and 14 R-scripts included below. The lowflow_sc_ga_al_gagesII_2015.csv datafile contains the annual minimum seven-day mean streamflow with an annual exceedance probability of 90% (7Q10) for 224 basins in South Carolina, Georgia, and Alabama. The datafile also contains 231 basin characteristics from the Gages II dataset (https://water.usgs.gov/lookup/getspatial?gagesII_Sept2011). The "all_preds.csv" file contains the leave-one-out cross validated predictions for all the models. The paper associated with the data release compares the ability of eight machine-learning models (elastic net, gradient boosting, kernel-k-nearest neighbors, two variants of support vector machines, M5-cubist, random forest, and a meta-learning ensemble M5-cubist model) and four baseline models (ordinary kriging, a unit-area discharge model, and two variants of censored regression) to generate estimates of the 7Q10 at 224 unregulated sites in South Carolina, Georgia, and Alabama.

This dataset provides numerical and categorical descriptions of 48 basin characteristics for 9,314 ungaged basins coinciding with 12-digit hydrologic unit code (HUC12) pour points that drain to the Gulf of Mexico. Characteristics are indexed by National Hydrography Dataset (NHD) version 2 COMID (integer that uniquely identifies each feature in the NHD) and HUC12 identifying number. The variables represent mutable and immutable basin characteristics and are organized by characteristic type: physical (5), hydrologic (6), categorical (12), climate (6), landscape alteration (7), and land cover (12). Mutable characteristics such as climate, land cover, and landscape alteration variables are reported in decadal increments (for example, average percent forest for the decade 1950-1959, 1960-1969, etc). The majority of basin characteristics in this dataset were calculated using divergence-routing methods and are often referred to as “network-accumulated”. This method uses a modified routing database to navigate the NHDPlus reach network to aggregate (accumulate) the values derived from the reach catchment scale (Schwarz, G.E., and Wieczorek, M.E., 2018, Database of modified routing for NHDPlus version 2.1 flowlines: ENHDPlusV2_us: U.S. Geological Survey data release, https://doi.org/10.5066/P9PA63SM ). In four instances, values are also provided for the entire catchment above a site and area designated using the “CAT_” prefix.

This data release contains fractional intersectional weights used to crosswalk data from the National Water Use Program to the National Integrated Water Availability Assessment (IWAAs) projects. The Watershed Boundary Dataset (WBD; https://www.usgs.gov/national-hydrography/watershed-boundary-dataset) is a companion dataset to the National Hydrography Dataset and contains polygons that define the spatial boundaries of hydrologic units (i.e., the area of land the landscape that drains into a portion of the stream network). These are periodically updated as these boundary definitions are refined by incorporating better, more localized data. When aggregating data from multiple sources that rely on data from the WBD, a situation can occur where different datasets rely on different versions (or “snapshots”) of the WBD. This was the case for the IWAAs National Report which relied upon data using a version of the WBD found in the Mainstem Rivers data release (https://doi.org/10.5066/P92U7ZUT) as well as data that relied upon a version used by the National Water Use Program (https://doi.org/10.5066/P9FUL880). This dataset is the output of a pipeline of R code published as a software release (https://doi.org/10.5066/P1UANON8) and contains the fraction of spatial overlap (i.e., weights) between the subwatershed (HUC12) boundaries from these two versions of the WBD. These weights can be used as a crosswalk between the two snapshots of the WBD.

The location of a point (or pixel) within the conterminous U.S. can be assigned based on its position relative to the Nation’s stream network. Two metrics are recognized: lateral position (LP) and distance from stream to divide (DSD). And given that a point can have different positions in different hydrologic orders the term multi-order hydrologic position (MOHP) is used to describe the ensemble of hydrologic positions. LP and DSD were developed for nine hydrologic orders across the conterminous U.S. (Belitz and others, 2019; Moore and others, 2019). Four case studies are presented here that were used for evaluating the utility of MOHP in the context of random forest machine learning (Belitz and others, 2019). Two of the case studies evaluate categorical response variables: geomorphic province in the Central Valley of California (Faunt, 2009) and physiographic province in the conterminous U.S. (Fenneman and Johnson, 1946). The other two case studies evaluated depth to the water table (DTW), which is a continuous variable. DTW for these two cases were determined from: 1) a numerical simulation model of the groundwater flow system in the Fox-Wolf-Peshtigo area located to the west of Lake Michigan (Juckem and others, 2017); and 2) observed values in Wisconsin (Fan and others, 2013). The point data for each of the four case studies include: land surface elevation, the 18 MOHP metrics (LP and DSD for nine hydrologic orders), and the appropriate response variable. Latitude and longitude are also included for the purposes of plotting. The case studies show that some MOHP metrics serve as indicators of hydrologic process and others as indicators of location. MOHP is shown to have utility as a predictor variable across a large range of scales (50,000 to 8,000,000 square kilometers). Four comma-separated values (.csv) data tables are included in this data release: 1) CVAL_sampsites_mohp.csv -- Central Valley, California 2) FPR_sampsites_mohp.csv -- Fenneman Physiographic Regions 3) FWP_sampsites_mohp.csv -- Simulated depth-to-water in the Fox-Wolf-Peshtigo model area 4) WIOBS_sampsites_mohp.csv -- Observed depth-to-water throughout Wisconsin

This data set has been superseded by huc2m. This file contains hydrologic unit boundaries and codes for the conterminous United States along with Alaska, Hawaii, Puerto Rico and the U.S. Virgin Islands. It was revised for inclusion in the National Atlas of the United States of America, and updated to match the streams file created by the USGS National Mapping Division (NMD) for the National Atlas of the United States of America. For the most current data and information relating to hydrologic unit codes (HUCs) please see http://water.usgs.gov/GIS/huc.html. The Watershed Boundary Dataset (WBD) is the most current data available for watershed delineation. See http://www.nrcs.usda.gov/wps/portal/nrcs/main/national/water/watersheds/dataset

This file contains hydrologic unit boundaries and codes for the conterminous United States along with Alaska, Hawaii, Puerto Rico and the U.S. Virgin Islands. It was revised for inclusion in the National Atlas of the United States of America, and updated to match the streams file created by the USGS National Mapping Division (NMD) for the National Atlas of the United States of America. For the most current data and information relating to hydrologic unit codes (HUCs) please see http://water.usgs.gov/GIS/huc.html. The Watershed Boundary Dataset (WBD) is the most current data available for watershed delineation. See http://www.nrcs.usda.gov/wps/portal/nrcs/main/national/water/watersheds/dataset