The systematic underestimation of nutrient load variability in coupled streamflow-water quality models: effects on lake cyanobacteria bloom simulations

Many studies have suggested that the severity of cyanobacteria blooms in freshwater bodies may worsen under projected future climate changes. To mitigate these blooms, watershed and lake managers need accurate time series of nutrient loads entering freshwater bodies. Frequently, calibration residuals associated with models of streamflow and nutrient concentration time series are omitted once calibration is deemed adequate. This practice can cause the variance of simulated time series to be lower than observations, an outcome that may compromise simulations of cyanobacteria blooms during and following extreme hydrometeorological events.

Through an integrated assessment model (IAM) of Lake Champlain’s Missisquoi Bay on the US-Canada border, we examine the extent to which the treatment of calibration residuals in streamflow and water quality models affects estimates of daily phosphorus loads and cyanobacteria blooms. First, we examine the extent to which re-introducing calibration residuals into simulated time series improves the performance of process-based models. Next, we compare these results with statistical models that consider fluctuations in river phosphorus concentrations dependent on discharge, seasonality, short- and long-term hysteresis, as well as the temporal structure of their residuals. We also discuss the importance of reproducing the variability of loads at different timescales when blooms are driven by external and internal loading. Finally, we consider the effects that modeling choices have on compliance with water quality standards, health and economic impacts of blooms, and examining impacts of potential changes in extreme hydrometeorological events.