BEGIN:VCALENDAR VERSION:2.0 PRODID:-//pretalx//talks.osgeo.org//T9LLQV BEGIN:VTIMEZONE TZID:CET BEGIN:STANDARD DTSTART:20001029T040000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10 TZNAME:CET TZOFFSETFROM:+0200 TZOFFSETTO:+0100 END:STANDARD BEGIN:DAYLIGHT DTSTART:20000326T030000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3 TZNAME:CEST TZOFFSETFROM:+0100 TZOFFSETTO:+0200 END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT UID:pretalx-foss4g-2022-academic-track-T9LLQV@talks.osgeo.org DTSTART;TZID=CET:20220825T123500 DTEND;TZID=CET:20220825T124000 DESCRIPTION:1. INTRODUCTION\nIn a new initiative to deliver higher-quality data and support improved geospatial analysis\, the U.S. Geological Survey (USGS) is upgrading the elevation and hydrography datasets into the 3D Na tional Topography Model (3DNTM)\, which will include fully integrated hydr ography and elevation. The USGS 3D Elevation Program (3DEP) recently compl eted acquisition of interferometric synthetic aperture radar (IfSAR) eleva tion data at 5-meter spatial resolution for Alaska (USGS\, 2022). Other pa rts of the United States are being mapped at higher resolution with lidar- derived elevation data. \n\nUnder the 3DNTM\, new hydrography data are acq uired through methods that derive or extract the features directly from be st available 3DEP elevation data to ensure proper integration of the hydro graphy and elevation layers. By applying specifications for deriving 1:24\ ,000 or larger scale hydrography from high resolution elevation data (Arch uleta and Terziott\, 2020\; Terziotti and Archuleta\, 2020)\, a tenfold in crease in the number of features in the National Hydrography Dataset (NHD) is expected. Consequently\, highly automated machine learning methods to extract and validate the hydrography data collection are being investigate d.\n\nXu et al. (2021) demonstrated that the U-net fully convolutional neu ral network (Ronneberger\, Fischer\, and Brox\, 2015) is capable of extrac ting hydrography from lidar elevation data with 80 to 90 percent accuracy. Stanislawski et al. (2021) applied a similar U-net model using several If SAR and IfSAR-derived input layers to predict hydrography for a 50-watersh ed study area in northcentral Alaska\, where 68 percent average F1-score a ccuracies were achieved on test watersheds. Further work to refine U-net p redictions of hydrography using IfSAR for the same 50-watershed area in Al aska achieved average F1-scores for test watershed of better than 80 perce nt (Stanislawski et al.\, 2022). Research presented in this paper builds u pon this earlier work by testing transfer learning methods and scaling-up U-net predictions of hydrography from IfSAR for other areas of Alaska usin g workflows in high-performance computing environment. \n\n2. METHODS\nA w orkflow was developed to automate downloads and processing of IfSAR-derive d tiles of digital elevation model (DEM)\, digital terrain model (DTM)\, a nd orthorectified intensity (ORI) data for user-selected watersheds from t he 3DEP database. The workflow mosaics common tiles and derives several ra ster data layers from the DEM that are related to surface hydrology\, such as topographic position index and shallow water channel depth. Overall\, seventeen data layers are generated and coordinated with identical raster projection systems. The layers were used in U-net modelling for predicting hydrography for the 50-watershed Kobuk River study area (Stanislawski et al.\, 2022). In this study a transfer learning process begins with the Kob uk River U-net model and subsequently includes additional training data fr om outside the Kobuk area. Hydrography predictions are then generated from the transfer learning model and assessed. Several levels of refinements t o training data are tested and the accuracy of predictions are assessed. R eference data consist of vector hydrography features derived by USGS contr actors.\n\nThe data processing workflows are implemented with Python\, lin ux shell scripts\, and opensource software libraries such as the Geospatia l Data Abstraction Library (GDAL). Neural network modelling is implemented through TensorFlow\, and data processing is completed on a 12-node linux cluster and through the GPU nodes of the USGS Tallgrass computing faciliti es (https://hpcportal.cr.usgs.gov/hpc-user-docs/Tallgrass/Overview.html).\ n\n3. DISCUSSION\nMapping hydrography for the state of Alaska is a dauntin g task\, given its vast area and terrain that is difficult to navigate. Bi g challenges with large high-quality datasets are well suited to take adva ntage of recent advancements in neural networks (Usery et al.\, 2021). Thi s research demonstrates the tremendous potential to improve and speed up m apping of surface water features in Alaska\, and elsewhere in the world ha ving challenging terrain and limited resources. \n\nReported accuracy scor es measure how well a machine can reproduce hydrography generated with met iculous editing by numerous subject matter experts. It is not a score of h ow well the surface water features are mapped by the model. The human fact or in contemporary broad scale mapping efforts cannot be ignored and warra nts consideration as a source of uncertainty in the related accuracy metri cs. How well the maps fit what is on the ground can only be definitively c onfirmed by being on the ground at any given point in time\, as hydrologic conditions are constantly in flux. Thus\, the work here could be used as an aid to human cartographers in their efforts to interpret what is import ant to the map user. \n\nThis work could also benefit change detection eff orts. As new and better elevation data are collected\, automated strategie s such as the model presented here could be used to identify regions with significant changes in surface water distribution. This type of automation would be valuable to maintain an accurate national map over time and help address the numerous challenges that society faces related to hydrology. DTSTAMP:20260413T205250Z LOCATION:Room Hall 3A SUMMARY:Scaling-up deep learning predictions of hydrography from IFSAR data in Alaska - Larry Stanislawski URL:https://talks.osgeo.org/foss4g-2022-academic-track/talk/T9LLQV/ END:VEVENT END:VCALENDAR