Regional estimation of river water temperature at ungauged locations

River water temperature measurement networks suffer from an inadequate spatial coverage and a lack of data. No methods exist for the regional estimation of river water temperature at ungauged sites based on data series from gauged sites. The development of such methods is hence of significant importance. It is proposed in this study to develop a Temperature-Duration-Curve (TDC) based method to estimate river water temperature at ungauged sites on a real-time basis. A Generalised Additive Model (GAM) based method is used to estimate TDCs at ungauged sites. The estimated TDCs are then used in combination with a spatial interpolation method to obtain daily temperature estimates at ungauged sites. Results are compared with a simple method based on the geographical distance weighted average of neighboring stations. The approaches are applied to 126 river thermal stations located on Atlantic salmon rivers in eastern Canada. Leave-one-out cross validation results indicate that the TDC based methods are robust and outperform the geographical distance weighted method.