Melanie K. Vanderhoof William Keenan Wayana Dolan 20250321 csv Denver, CO U.S. Geological Survey https://doi.org/10.5066/P14WYWSY Melanie K. Vanderhoof William Keenan Wayana Dolan Heather E. Golden Charles R. Lane Jay R. Christensen Kylen Solvik Adnan Rajib TBD Publication (journal) TBD TBD TBD TBD Globally, many waterbodies and floodplains have been lost, degraded, or are at risk for further loss, which may have unintended consequences for rivers, including exacerbating flood and drought conditions. We explored how including surface water storage time series in deep learning models influences our ability to predict river discharge. We utilized Sentinel-1 and Sentinel-2 algorithms to generate time series of surface water extent. Surface water extent (m2) was converted to storage (m3) using topographic estimates of depression probability and depth. These surface water storage estimates were then tested with meteorological data and catchment characteristics in four Long Short-Term Memory (LSTM) models, each containing a different combination of variable groups, to simulate daily river discharge (2016-2023) for 72 watersheds across the conterminous United States. The purpose of this study was to explore how surface water storage dynamics may influence daily river discharge. Files in Data Release: The data consists of a csv (“surface_water_storage_by_gage_2016_2023.csv”) which provides the daily surface water storage variable values included in the LSTM models, for each of the 72 watersheds (2016-2023). Spatial bounding coordinates represent the geographic bounds of the 72 watersheds. Estimates of surface water storage are provided in meters cubed. fn=focal network or stream network; fp=floodplain Attributes: 1: date; 2: United States Geological Survey gage number; 3: total surface water storage (“all”); 4: stream-connected surface water storage (“fn_yes”); 5: stream-disconnected surface water storage (“fn_no”); 6: surface water storage within the floodplain (“fp_yes”); and 7: surface water storage outside of the floodplain (“fp_no”) 20160101 20231217 Ground Condition Complete None planned -123.358451446 -75.5840683428 48.9995054876 28.9579118358 None time series surface water storage wetlands lakes floodplain non-floodplain stream-connected stream disconnected USGS Metadata Identifier USGS:67a39325d34e63325c2b6f92 None United States Oregon California Idaho Montana Arizona New Mexico Texas Mississippi Arkansas Missouri Kansas Nebraska Iowa Illinois Wisconsin Minnesota South Dakota North Dakota Georgia South Carolina North Carolina Virginia Maryland Pennsylvania Delaware None Unless otherwise stated, all data, metadata and related materials are considered to satisfy the quality standards relative to the purpose for which the data were collected. Although these data and associated metadata have been reviewed for accuracy and completeness and approved for release by the U.S. Geological Survey (USGS), no warranty expressed or implied is made regarding the display or utility of the data on any other system or for general or scientific purposes, nor shall the act of distribution constitute any such warranty. Sentinel-1 and Sentinel-2 imagery was accessed through Google Earth Engine from the image collections COPERNICUS/S1_GRID and COPERNICUS/S2, respectively. Sentinel-1 and Sentinel-2 based algorithms are published in: Vanderhoof, M.K., Alexander, L., Christensen, J., Solvik, K., Nieuwlandt, P., Sagehorn, M. (2023) High-frequency time series comparison of Sentinel-1 and Sentinel-2 for open and vegetated water across the United States (2017-2021). Remote Sensing of Environment. 288, 113498, https://doi.org/10.1016/j.rse.2023.113498. Version 6.2 (Build 9200) ; Esri ArcGIS 10.3.1.4959 The Sentinel-1 algorithm has a documented omission and commission error of 3.1% and 0.9% for open water, and a 28.4% and 16.0% commission error for vegetated water, respectively, while the Sentinel-2 algorithm has an omission and commission error of 3.1% and 0.5% for open water, and a 10.7% and 7.9% commission error for vegetated water, respectively, when validated against 36 high-resolution images (i.e., WorldView-2, WorldView-3, PlanetScope) (Vanderhoof et al., 2023, https://doi.org/10.1016/j.rse.2023.113498). When consolidated at a monthly time-step to a Sentinel-1 and Sentinel-2 water, non-water classification, and validated against 64 PlanetScope images, errors of omission and commission for monthly surface water extent averaged 1.9% and 6.5%, respectively (Vanderhoof et al., 2024; https://doi.org/10.1029/2023EF004106). The conversion from surface water extent to an estimate of surface water storage have not been independently validated. We note that the data provided should be used for informational purposes only. Inaccuracies in the data are to be expected. Reference: Vanderhoof, M.K., Christensen, J., Alexander, L.C., Lane, C.R., Golden, H.E. (2024) Climate change will impact surface water extents across the central United States. Earth’s Future, 12(2), e2023EF004106. No formal logical accuracy tests were conducted. Data are complete. The horizontal accuracy was not formally evaluated. No vertical positions were reported. Surface water extent over time at each of the 72 sites was derived using previously published algorithms based on Sentinel-1 (C-band synthetic aperture radar in dual polarization mode, 20 x 22 m resolution) and Sentinel-2 (optical sensor with l3 spectral bands, 10-20 m resolution) (Vanderhoof et al., 2023, https://doi.org/10.1016/j.rse.2023.113498). These algorithms used gradient boosted classifiers to classify images at 20 m resolution into non-water, open water (e.g., lakes, ponds, rivers) or vegetated water (e.g., emergent wetlands, forested/shrub wetlands, riparian corridors). The two algorithms were applied to all imagery acquired over the 72 watersheds between 2016 and 2023. The classified Sentinel-1 and classified Sentinel-2 time series were consolidated at a two-week timestep. Within each timestep, pixel values were assigned as the majority classification (non-water or water (open water plus vegetated water). If observations of water and non-water were equal, then, depending on the classified observations present, open water was selected over vegetated water and non-water, and non-water was selected over vegetated water. For pixels where no data was available, each raster was gap-filled using the classified timestep before and after. To convert surface water extent to storage we calculated the depth of sink which is the depression fill depth, from the U.S. Geological Survey, 3D Elevation Program 10-meter resolution digital elevation model (DEM). Since many waterbodies, including most wetlands, are shallower than the reported vertical accuracy of the DEM, we also used a stochastic depression tool to generate the probability a pixel was within a depression. Where the depth of sink was <1 m and the pixels were classified as open water, we used a 1:1 ratio to convert the depression probability to a depth in cm (range = 1 to 100 cm). Where the depth of sink was <1 m and pixels were classified as vegetated water, we used a 1:0.5 ratio (range = 1 to 50 cm) to convert depression probability to depth. Lastly, where the probability of a depression was <10%, open water and vegetated water pixels were assigned a nominal depth of 10 cm and 5 cm, respectively. Surface water storage for each watershed at each time step was summarized as (1) total surface water storage (i.e., across the entire watershed; “all”), (2) surface water storage within the floodplain (“fp_yes”), (3) surface water storage outside of the floodplain (“fp_no”), (4) surface water storage that was continuous with the stream network (“fn_yes”), and (5) surface water storage that was non-continuous with the stream network (“fn_no”). The floodplain was defined as the 100-year floodplain (Woznicki et al., 2019). Surface water storage at each time step was classified as stream connected if it was adjacent to or connected with the rasterized (20 m) National Hydrography Dataset (NHD) high resolution flowlines classified as intermittent or perennial streams, named artificial paths, or NHD waterbodies that occurred along the stream network (USGS, 2022). References: U.S. Geological Survey (2022). USGS TNM Hydrography (NHD). U.S. Geological Survey, Reston, VA, https://apps.nationalmap.gov/services/ (Last accessed August 4, 2022). Woznicki, S. A., Baynes, J., Panlasigui, S., Mehaffey, M., and Neale, A.: Development of a spatially complete floodplain map of the conterminous United States using random forest, Sci. Total Environ., 647, 942-953, 2019. 2024 D_WGS_1984 WGS_1984 6378137.0 298.257222101 FILE NAME: “surface_water_storage_by_gage_2016_2023.csv” FILE DESCRIPTIONS: Estimates of surface water storage in meters cubed at a daily timestep for 72 watersheds (2016-2023), including: 1: date; 2: United States Geological Survey gage number; 3: total surface water storage (“all”); 4: stream-connected surface water storage (“fn_yes”); 5: stream-disconnected surface water storage (“fn_no”); 6: surface water storage within the floodplain (“fp_yes”); and 7: surface water storage outside of the floodplain (“fp_no”) Sentinel-1 and Sentinel-2 based algorithms are published in: Vanderhoof, M.K., Alexander, L., Christensen, J., Solvik, K., Nieuwlandt, P., Sagehorn, M. (2023) High-frequency time series comparison of Sentinel-1 and Sentinel-2 for open and vegetated water across the United States (2017-2021). Remote Sensing of Environment. 288, 113498, https://doi.org/10.1016/j.rse.2023.113498. US Geological Survey gage ID: U.S. Geological Survey. (2019). The StreamStats program, online at https://streamstats.usgs.gov/ss/, accessed on (August 4, 2022). U.S. Geological Survey ScienceBase U.S. Geological Survey ScienceBase mailing and physical

Denver Federal Center, Building 810, Mail Stop 302

Denver CO 80225 USA 1-888-275-8747 sciencebase@usgs.gov Any use of trade, product or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Geological Survey. Although this information product, for the most part, is in the public domain, it also contains copyrighted materials as noted in the text. Permission to reproduce copyrighted items for other than personal use must be secured from the copyright owner. This database has been approved for release and publication by the Director of the USGS. Although this database has been subjected to rigorous review and is substantially complete, the USGS reserves the right to revise the data pursuant to further analysis and review. Furthermore, it is released on condition that neither the USGS nor the United States Government may be held liable for any damages resulting from its authorized or unauthorized use. Although these data have been processed successfully on a computer system at the U.S. Geological Survey, no warranty expressed or implied is made regarding the display or utility of the data on any other system, or for general or scientific purposes, nor shall the act of distribution constitute any such warranty. The U.S. Geological Survey shall not be held liable for improper or incorrect use of the data described and/or contained herein. comma delimited (.csv) https://doi.org/10.5066/P14WYWSY None. No fees are applicable for obtaining the data set. 20250321 U.S. Geological Survey, Geoscience and Environmental Change Science Center Melanie Vanderhoof Research Geographer Mailing

PO Box 25046, DFC, MS 980

Denver CO 80225 US 303-236-1411 303-236-5690 mvanderhoof@usgs.gov FGDC Content Standard for Digital Geospatial Metadata FGDC-STD-001-1998 local time