Identifying river ice on the Tibetan Plateau based on the relative difference in spectral bands

Remote sensing can potentially be used to monitor river ice, which is of significance for water resource utilization and hydrological research, especially on the Tibetan Plateau, over which field observations are severely limited. However, methods to monitor the river ice distribution over a large scale while overcoming snow interference have not been reported yet. To monitor river ice on the Tibetan Plateau on a large scale, we present a relative difference river ice (RDRI) identification method based on the differential spectral characteristics of river ice. The difference between the red and near-infrared (NIR) band reflectance of river ice is divided by the sum of the NIR reflectance and infrared band reflectance. The resulting values are notably different for river ice and other neighboring similar landscapes, such as snow cover. This method can overcome the interference caused by snow cover and other surrounding landscapes and enable the monitoring of river ice in different forms at different elevations on the Tibetan Plateau. The RDRI method is applied to Landsat and Sentinel-2 images. The validation results show that the mean overall accuracy and Kappa coefficient were 99.76% and 0.94, respectively. The comparison of the RDRI and normalized difference snow index (NDSI) methods indicates that the accuracies of both methods for river ice monitoring are similar in the absence of any snow disturbance. The RDRI can enable more accurate remote sensing of river ice on a large scale, which can support river ice distribution research over the Tibetan Plateau.