Principal Investigator: Frederick "Fritz" Policelli (NASA's Goddard Space Flight Center)

This project is a partnership between NASA's Goddard Space Flight Center, the Alaska Satellite Facility, the University of Arizona, and the University of Maryland. The goal of this project is to develop a dataset of river width measurements for use in training machine learning models to derive river flow rate estimates and for use in related models. The measurements will be made using ESA  (European Space Agency) Sentinel-1 C-Band Synthetic Aperture Radar (SAR) data, which can provide data day and night and in all-weather conditions, nominally at a six-day revisit time (with two operating satellites).

Project Objectives

  • Provide a dataset of river width measurements to the research community through NASA's Physical Oceanography Distributed Active Archive Center (PO.DAAC)
  • Provide a capability for the research community to map surface water at 10 m resolution using Sentinel-1 data distributed by the Alaska Satellite Facility
  • Demonstrate the utility of the river width measurement data for deriving river flow rate estimates

Update

The goal of this project is to develop a dataset of effective river width measurements (surface water area/river reach length) using ESA Sentinel-1 SAR data for use in training machine learning models to derive river flow rate estimates and for use in related models. The workflow for generating the effective river width measurements is being done on the Alaska Satellite Facility (ASF) OpenScienceLab System and will consist of three primary components: preprocessing using the standard ASF Sentinel-1 methods and programs, a surface water extent mapping program, and a river width measurement program, the latter of which uses the surface water maps as input. The system will also have a module to filter out Sentinel-1 scenes that do not include river reaches of interest (those in the NASA Surface Water and Ocean Topography [SWOT] Mission SWOT River Database [SWORD]).

The project is evaluating three surface water mapping algorithms using ESA Sentinel-1 SAR data: HydroSAR (developed by the Alaska Satellite Facility), the Equal Percent Solution (developed by the PI), and a machine learning algorithm developed by the University of Arizona, and is considering two additional algorithms: the NASA Observational Products for End-Users from Remote Sensing Analysis (OPERA) Project algorithm, and a hybrid algorithm containing elements of the Equal Percent Solution and the OPERA algorithm. The project is evaluating the water maps produced by the selected algorithms using hand-labeled water maps produced from commercial high resolution data from the Planetscope constellation. The algorithm evaluated as the best will be incorporated into the ASF processing system and will be made available for use by the research community through an interface provided by ASF. The effective river width measurement program will be a modification of the RivWidth Cloud Program, developed by the University of North Carolina.

In addition to enabling the use of the Sentinel-1 water maps, the modifications will include interfacing with the SWORD database to automate measurement of effective river width for the nodes and reaches in that database. We will again use hand measurements based on Planetscope data to evaluate the accuracy of the effective river width data. Finally, we will evaluate the utility of the effective river width measurements for use in deriving river flow rates in a machine learning model and the SWOT Mission GeoBAM model.

Major Accomplishments

  • Development of a new algorithm for mapping surface water using ESA Sentinel-1 data. The algorithm is called "The Equal Percent Solution", so-called because it forces a radar backscatter threshold value (between water and not-water pixels) such that the percentage of false positives are approximately equal to the percentage of false negatives
  • Development of a new high resolution dataset of hand labelled water maps based on Planetscope data and being made available to the research community for training machine learning models and evaluating water maps
  • Implementation of an end-to-end system for measuring river width using Sentinel-1 SAR data. Regrettably, there was an issue releasing the river width measurement code to the public, and the code was removed from the system, however replacement code is in development

Publications and Presentations

Mukherjee, R., et al. (2023). A globally sampled high-resolution hand-labeled validation dataset for evaluating surface water extent maps. Submitted to Earth System Science Data.

Gangodagamage C., et al. (2023). Harmonized Sentinel-1 SAR Global River Geometry and Inundation Database. American Meteorological Society (AMS) 103rd Annual Meeting, January 8-12, 2023, Denver, CO, USA.

Gangodagamage, C., et al. (2022). Harmonized Sentinel-1 SAR Global River geometry and Inundation database. AGU Fall Meeting, December 12-16, 2022, Chicago IL, USA, Abstract #OS22A-26.

Lamont, S., et al. (2022). Evaluation of Regional Physically-based Flood Model Outputs using Flood Extent Maps from Remote Sensing for Flood Events in Japan. AGU Fall Meeting, December 12-16, 2022, Chicago IL, USA, Abstract #H46D-05.

Wang, et al. (2022). Validating surface water detection methods using a hand-labeled PlanetScope dataset. AGU Fall Meeting, December 12-16, 2022, Chicago IL, USA.

 

Last Updated