Lake upwelling-induced stream temperature variations

The Fraser River Basin supports around 60% of British Columbia’s salmon populations, where river temperature influences migration success and survival. During the summer, sustained, strong winds over stratified lakes can induce upwelling, releasing cold hypolimnetic water into downstream rivers. This study investigates upwelling-induced cooling events in the Quesnel River downstream of Quesnel Lake to quantify their characteristics and identify their physical drivers. Lake and river temperature data (2016–2025) were analyzed to isolate cooling events from seasonal and daily variability. More than 90 cooling events were identified, with an average temperature drop of 4°C over 1.5 days. These events occurred every year, approximately once per week during the summer stratified season, indicating that rapid temperature fluctuations recur in the Quesnel River. Cross-correlation analysis revealed strong correlations between river and lake temperatures at a depth of 8 m in the West Basin of Quesnel Lake, with coefficients ranging from 0.91 to 0.95. A broad range of highly correlated lags indicates complex upwelling dynamics influenced by stratification, wind variability, and basin geometry. Analysis of wind-field simulations from a 3D regional atmospheric model (WRF) identified sustained, lake-wide easterly winds (> 3 m/sˉ¹), aligned with the lake’s thalweg, as the primary forcing mechanism. Wedderburn number-based models for internal seiche amplitude were used to estimate upwind thermocline displacement. Near full upwelling, nonlinear internal seiche models predicted upwind-end thermocline displacements 1.5–2.3 times those of a linear model, highlighting the importance of nonlinear effects in internal seiche modeling. Morphology-aware models showed statistically insignificant improvement over box-shaped models, likely because the lake’s elongated geometry makes 2D and 3D morphological effects smaller than the 1D influence of long wind fetch. When the internal seiche models were linked with a withdrawal layer model, the river temperature was modeled for a single stratified season with a mean absolute error of 1.4°C. Although the model captured the river temperature variability, its performance was limited by the assumptions of the Wedderburn number framework and input data quality. These findings highlight the importance of incorporating lake dynamics into watershed-scale thermal models to improve predictions of salmon habitat conditions under changing environmental conditions.