Measurements and modelling of eutrophication processes in Lake Rotoiti, New Zealand

Deterioration of water quality is a common problem for aquatic systems globally, which is accelerated by factors such as urban settlement, farming, forestry and recreation. Spatial variability of water quality in these systems hinders a more advanced understanding of their dynamics, to better enable strategies to be developed to combat their deterioration. Understanding the drivers for spatial variability is fundamentally important for predicting how lake ecosystems will respond to management scenarios and which management actions are most likely to be successful to improve lake health. Through a field study in a morphologically diverse lake in New Zealand, and the application of a lake ecosystem model, this study examined the spatial and temporal variability of phytoplankton biomass and made a detailed consideration of the performance of three-dimensional lake ecosystem models. To gain insight in the spatial variability in phytoplankton productivity, surface phytoplankton productivity measurements were carried out at three stations in morphologically complex Lake Rotoiti, North Island, New Zealand, with the objective of defining variations between sites and seasons, and the dominant environmental drivers of these variations. There was no overarching statistical relationship between measured environmental variables and primary productivity or specific production. Inorganic nutrient concentrations at the surface of the shallow station were low throughout the whole year but at the other two stations they showed a typical pattern for monomictic lakes of higher levels during winter mixing and declining concentrations during thermal stratification. The high variability between the three sites indicates that it is important to account for local differences in productivity in morphologically diverse lakes, and that whole-lake productivity estimates may vary greatly depending on the location and depth at which measurements are made. To gain understanding of the spatial variability, a higher resolution three-dimensional (3D) model, ELCOM-CAEDYM, was used to simulate the time-varying horizontal and vertical variations in water quality over one year (May 2004-May 2005). The main inflow to Lake Rotoiti arises via the Ohau Channel from adjacent eutrophic Lake Rotorua. Highly spatially resolved field data were collected monthly to validate the model performance for simulations of temperature, dissolved oxygen and chlorophyll a. The model was configured to simulate a geothermal heat source in the deepest part of the main lake basin. Model simulations of temperature, dissolved oxygen and chlorophyll a were highly correlated with measurements of these variables but simulations of spatial variations in nutrient concentration showed relatively low correlation coefficient values, in particular at a station located in a shallow embayment. An examination of the behaviour of a conservative tracer introduced into the major inflow of Lake Rotoiti, the Ohau Channel, confirmed previous findings that this inflow could enter the lake as a surface inflow, interflow or underflow, depending on temperature gradients between the inflow and the lake water column. The results showed that ELCOM-CAEDYM is capable of reproducing highly spatially resolved field data in a complex, geothermally-influenced lake, and can provide important insights into the fate of heat and constituents in major inflows. A wall to divert Ohau Channel water directly towards the outflow of Lake Rotoiti was implemented in August 2008. Based on parameter values calibrated for the time period before the inflow diversion (May 2004-May 2005), ELCOM-CAEDYM was applied to the post-diversion wall period of August 2008-August 2009. The model showed good fit with observed data for temperature, dissolved oxygen and chlorophyll a, but the model showed a poor fit when simulations were compared with in situ nutrient concentrations at all stations. Simulations with ELCOM-CAEDYM suggest that model accuracy may be improved when a simplified dynamic sediment diagenesis model is available. This could provide for less sensitivity to sediment nutrient release rates of the model and better model fit to in situ data. Ecosystem modelling is likely to play an increasingly important role in lake management and scientific understanding of lake processes as computer speed increases and models undergo further refinements. The model development is in a phase of relative maturity in which water quality simulations provide an efficient and rational tool to compare water quality outcomes and cost effectiveness of lake improvement techniques, to provide for preservation and improvement in water quality in the future