MODELING OF HOURLY STREAM TEMPERATURES WITHIN TWO FORESTED
Water temperature is a key physical habitat determinant in lotic ecosystems as it influences many physical, chemical and biological properties of rivers. Hence, a good understanding of the thermal regime of rivers is essential for effective management of water and fisheries resources. This study deals with the modeling of hourly stream watertemperature using a deterministic model, an equilibrium temperature model and an artificial neural network model. The water temperature models were applied on two thermally different streams, namely, the Little Southwest Miramichi River (LSWM) and Catamaran Brook (Cat Bk) in New Brunswick, Canada. The deterministic model calculated the different heat fluxes at the water surface and from the streambed, using different hydrometeorological conditions. Results showed that microclimate data are essential in making accurate estimates of the surface heat fluxes. Results also showed that for larger river systems, the surface heat fluxes were generally the dominant component of the heat budget with a correspondingly smaller contribution from the streambed (90%). As watercourses became smaller and as groundwater contribution became more significant, the streambed contribution became important (20%). The equilibrium temperature model is a simplified version of the deterministic model where the total heat flux at the surface is assumed to be proportional to the difference between the water temperature and the equilibrium temperature. The poor model performance compared to the other models developed in this study suggested that the air and equilibrium temperature did not reflect entirely the total heat flux at an hourly scale. The model’s best performance was in autumn, where the low water level permitted a more efficient thermal exchange, whereas the presence of snowmelt conditions in spring resulted in poorer performance. An artificial neural network (ANN) was also developed to predict hourly river water temperatures using minimal and accessible input data. The results showed that ANN models are effective modeling tools, with similar or better results to comparable modeling studies. The ANN model performed best in summer and autumn and had poorer, but still good, performance in spring, explained by the high water levels.