Author(s)

ANDRÉ ST-HILAIRE, CLAUDINE BOYER, NORMAND BERGERON, ANIK DAIGLE

Abstract

Water temperature is a key variable affecting important water quality parameters such as dissolved oxygen. In Eastern Canada, iconic fish species such as Atlantic salmon (Salmo salar) can be affected by increase in temperature associated with climate change. A major endeavour is underway to establish and optimize a water temperature monitoring network in this region. This network, called RivTemp, includes temperature data from over 600 stations in 277 streams or rivers. These data are being used to develop/adapt methods network optimization, temperature interpolation and modelling/forecasting. Different approaches to interpolate water temperature at ungauged sites using data from monitoring stations are being compared. More recently, two regression approaches that are often used when collinearity is present among predictors, the ridge regression and the LASSO regression were compared. Results show that the LASSO regression is more parsimonious than the ridge regression and provides adequate estimates of daily average water temperature.

Keywords

water temperature, network, monitoring model regression