Estimating river accommodation capacity for organic pollutants in data-scarce areas

Globally, water quality degradation severely threatens the security of water resources. Understanding a river's capacity to accommodate pollutants (or water environmental capacity: WEC) can help efficiently protect rivers. However, the requirement for comprehensive ground-observed hydrological and water quality data in previous methods makes it difficult to estimate WEC in areas with limited ground observations. This paper proposes a new framework for WEC estimation in data-scarce areas based on remotely sensed skin water temperature and limited ground observations. Two new models were developed to calculate the two critical parameters for WEC estimation: water temperature, and integrated pollutant degradation coefficients (k). Images of ASTER Surface Kinetic Temperature (AST_08) 90 m grid product were used to retrieve water temperatures. The above results were subsequently used to calculate a river's capacity to accommodate pollutants, or WEC, in agriculturally dominated areas. The use of remote sensing techniques enables the methods to be applied over large spatial scales and to areas with limited ground observations. The application and testing of the framework in four rivers, including two Chinese rivers (the Huai and the Wei Rivers) and two Australian rivers (the Ovens and the Gwydir Rivers), suggest that the models performed well to calculate the real-time water temperature and the coefficient k based on limited ground-observations. Uncertainty analysis on water temperature calculated from remotely sensed land surface temperature and ground-observed meteorological air temperature suggests that remotely sensed water temperature had high concurrence with ground observations (RMSE = 3.08 °C with R2 = 0.88), while the sparse-spatially distributed meteorological stations reduced the accuracy in estimating water temperature (RMSE = 4.39 °C with R2 = 0.91). We found that the coefficient (k) increased with water temperature over different seasons in an exponential form but in a logarithmical form with streamflow velocity. Comparison with previous research and other models with abundant data revealed the practicability and effectiveness of our models, which can be easily applied to rivers with insufficient ground observations across the globe.