Estimating discharge from Landsat based on water occurrence derived rating curves.

Abstract:

Satellite remote sensing can provide spatially-distributed estimates of river flow to supplement the gauge record. The past two decades has seen considerable development of remote sensing of discharge algorithms with improvements in accuracy and spatial coverage. This study presents a simple approach to efficiently estimate river discharge from Landsat imagery. The approach relies on the development of reach-level width-discharge rating curves derived from Landsat water occurrence data and modeled historical river flow. The modeled river flow estimates are from the Global Reach‐level A priori Discharge Estimates for Surface Water and Ocean Topography (GRADES) dataset and the water occurrence data are from the Joint Research Centre Global Surface Water (GSW) dataset. We developed reach-level width-based rating curves by spatially joining flow frequency distributions from GRADES and width occurrence distributions from GSW along river cross sections generated from the Global River Widths from Landsat (GRWL) database. We then estimated discharge of North American rivers wider than 100 m from Landsat scenes using the width-based rating curves from 1984-present. For a preliminary validation effort, we used daily discharge measurements from 28 USGS gauges along the lower Mississippi and Missouri rivers, which produced a median RRMSE of 54%, a median NRMSE of 65%, a median Relative Bias of -27%, and a median RMSE of 1154 m3/s. An alternative remote sensing of discharge approach, which uses only same-day width-discharge matchups to generate the reach-level rating curves, yields a marginal performance improvement but at the cost of computational efficiency. These results suggest that our approach produces discharge estimates that are within the accuracy range of published remote sensing of discharge algorithms. The method presented in this study provides a simple approach to efficiently estimate river discharge with reasonable accuracy, and can be readily applied at the global level.