Using Width-Based Rating Curves from Spatially Discontinuous Satellite Imagery to Monitor River Discharge

One of the significant technical hurdles to remote estimation of river discharge using river widths is the acquisition of suitable imagery. Most high-resolution, radar-based sensors acquire imagery infrequently (e.g. once or twice per month), while optical sensors are subject to the effects of clouds. Many more images that are partially cloud-free or cover only part of a study area are often available, but their use is problematic because the spatial extent of observable subreaches differs from image to image. Here, a method is presented that allows spatially discontinuous remote observations of river width to be successfully used to estimate discharge despite partial cloudiness and/or inconsistent spatial coverage. The method is demonstrated over 62 km of the Tanana River downstream of Fairbanks, AK. This reach is divided into 1 km subreaches and a separate discharge-width rating curve is produced for each subreach. For any given image, final discharge is estimated as the median discharge calculated from all visible and cloud-free reaches. Application of the method to 28 RapidEye images at 5 m spatial resolution and useable coverage ranging from <20% to 100% suggests that it produces highly accurate estimates of discharge (Root Mean Squared Error (RMSE=6.7%, r²=0.92) when compared with in situ observations. Similar but slightly less accurate results are evident from eleven Landsat TM and ETM+ images (RMSE=12.6%, r²=0.86). These results suggest that the method used here produces estimates of discharge comparable in error characteristics to ground-based measurements, at least when high resolution imagery is used over a river reach with substantial temporal width variability.