Stream water temperature modelling in forest catchments.

摘要

Water temperature influences most physical, chemical and biological processes of the river environment. It plays an important role in the distribution of fishes and on the growth rates of many aquatic organisms. Therefore, a good understanding of the thermal regime of rivers is an essential tool for the management of fish habitat. The modelling of water temperatures is key to the understanding of river thermal regimes as well as being invaluable for environmental impact assessments. This study deals with the modelling of river water temperatures using four different models: a deterministic model, a stochastic model, a simplified deterministic model, and an energy reference model.; The objective of the study consists of the development of a new and simplified deterministic model based on the equilibrium temperature concept in addition to the development of an energy reference model. These newly developed models were compared to the more classic deterministic and stochastic models. The equilibrium temperature model was based on a simplified function of meteorological parameters explaining the equilibrium temperature, which was thereafter used to calculate total energy flux at the water surface. This energy component was subsequently used to relate variations in water temperatures using a heat exchange coefficient. The energy reference model was based on the long-term meteorological parameters, and thus represents the long-term energy. This long-term energy component was then used with the corresponding annual component to predict river water temperatures.; Following the development of the models, they were applied to two thermally different river systems in a similar meteorological area, namely Catamaran Brook and the Little Southwest Miramichi River (NB). Catamaran Brook is the smaller of the two systems (10 m wide), with a mostly closed riparian canopy. By contrast, the Little Southwest Miramichi River is a larger and wider river (80--100m), which is more exposed to environmental conditions. Results from the present study showed that all models performed relatively well with root-mean-square error of between 1.26°C and 1.61°C (1992--99). Nash coefficients were observed in the range of 0.92 to 0.95 for all models (1992--99). It was concluded that differences in the modelling performances were related to model concept, data requirement, hydrometeorological conditions as well as timing within the year (e.g., early spring and late summer).