Identification of integrated models is still hindered by submodels’ uncertainty propagation. In this article, a novel identifiability and identification framework is applied to screen and establish reasonable hypotheses of an integrated instream (WASP) and catchment water quality (VENSIM) model. Using the framework, the models were linked, and critical parameters and processes identified. First, an ensemble of catchment nutrient loads was simulated with randomized parameter settings of the catchment processes (e.g. nutrient decay rates). A second Monte Carlo analysis was then staged with randomized loadings and parameter values mimicking insteam processes (e.g. algae growth). The most significant parameters and their processes were identified. This coupling of models for a two-step global sensitivity analysis is a novel approach to integrated catchment-scale water quality model identification. Catchment processes were, overall, more significant to the river’s water quality than the instream processes of this Prairie river system investigated (Qu’Appelle River).

A quasi-two-dimensional (quasi-2D) modelling approach is introduced to mimic transverse mixing of an inflow into a river from one of its banks, either an industrial outfall or a tributary. The concentrations of determinands in the inflow vary greatly from those in the river, leading to very long mixing lengths in the river downstream of the inflow location. Ideally, a two-dimensional (2D) model would be used on a small scale to capture the mixing of the two flow streams. However, for large-scale applications of several hundreds of kilometres of river length, such an approach demands too many computational resources and too much computational time, especially if the application will at some point require ensemble input from climate-change scenario data. However, a one-dimensional (1D) model with variables varying in the longitudinal flow direction but averaged across the cross-sections is too simple of an approach to capture the lateral mixing between different flow streams within the river. Hence, a quasi-2D method is proposed in which a simplified 1D solver is still applied but the discretisation of the model setup can be carried out in such a way as to enable a 2D representation of the model domain. The quasi-2D model setup also allows secondary channels and side lakes in floodplains to be incorporated into the discretisation. To show proof-of-concept, the approach has been tested on a stretch of the lower Athabasca River in Canada flowing through the oil sands region between Fort McMurray and Fort MacKay. A dye tracer and suspended sediments are the constituents modelled in this test case.

A quasi-two-dimensional (quasi-2D) modelling approach is introduced to mimic transverse mixing of an inflow into a river from one of its banks, either an industrial outfall or a tributary. The concentrations of determinands in the inflow vary greatly from those in the river, leading to very long mixing lengths in the river downstream of the inflow location. Ideally, a two-dimensional (2D) model would be used on a small scale to capture the mixing of the two flow streams. However, for large-scale applications of several hundreds of kilometres of river length, such an approach demands too many computational resources and too much computational time, especially if the application will at some point require ensemble input from climate-change scenario data. However, a one-dimensional (1D) model with variables varying in the longitudinal flow direction but averaged across the cross-sections is too simple of an approach to capture the lateral mixing between different flow streams within the river. Hence, a quasi-2D method is proposed in which a simplified 1D solver is still applied but the discretisation of the model setup can be carried out in such a way as to enable a 2D representation of the model domain. The quasi-2D model setup also allows secondary channels and side lakes in floodplains to be incorporated into the discretisation. To show proof-of-concept, the approach has been tested on a stretch of the lower Athabasca River in Canada flowing through the oil sands region between Fort McMurray and Fort MacKay. A dye tracer and suspended sediments are the constituents modelled in this test case.

Climate mediated warming water temperature, drought and extreme flooding are projected to shift the phenology of nutrients in receiving lakes and reservoirs further intensifying eutrophication and algal blooms, especially in temperate reservoirs. An emerging issue in reservoir management is the prediction of climate change impacts, a necessity for sound decision making and sustainable management. Lake Diefenbaker is a large multipurpose reservoir in the Canadian Prairies. In this study, the impact of climate change on nutrient speciation in Lake Diefenbaker is examined using loosely linked SpAtially Referenced Regression On Watershed attributes (SPARROW) and CE-QUAL-W2 models. Two climate mediated scenarios, RCP 8.5 representing the most extreme climate change, and climate induced streamflow were modelled. Nutrient levels are anticipated to double under the climate change and streamflow scenarios. Winter and spring were identified as hot moments for nitrogen pollution with a plausible saturation of nitrous oxides in the future. Of concern is a plausible recycling of nitrate through dissimilatory nitrate reduction to ammonium. Summer and fall on the other hand represent the period for phosphorus enrichment and internal loading with a probable succession of cyanobacteria in the summer. • Nutrient cycling in a large reservoir is investigated under two climate mediated scenarios. • Two loosely coupled models are forced with projected climate and streamflow changes. • Nitrogen pollution is projected to worsen during winter and spring during the 2040 decade. • Reservoir internal loading is anticipated to accelerate during the intermediate decade.