A modelling approach to assist with managing water quality in a catchment subject to rapid urbanisation: Lake Rotokauri, Hamilton, New Zealand

The objective of this study was to apply a coupled hydrodynamic-ecological model to a peat-stained and shallow (~4 m maximum depth) eutrophic lake whose catchment is likely to be subject to urban expansion associated with the development of Hamilton city, Waikato, New Zealand. The in-lake modelling was designed to increase understanding of the lake ecosystem and potentially to influence planning and management decisions associated with the prospective urban development project being undertaken by the Hamilton City Council (HCC). The overarching goal of the development is to accommodate urban expansion whilst retaining and enhancing the existing natural resources of Lake Rotokauri and Waiwhakareke Lake, and to restore the ecological value of the Rotokauri catchment. The main objective of this study was to understand the relationship between lake water quality and the effects of change of land-use from pastoral to urban within the Rotokauri catchment. This study incorporated results from a twelve-month programme of field work undertaken independently to the present study, into empirical calculations and computer modelling related to the catchment water budget and nutrient load, as well as the lake water quality. The fieldwork included the collection of water samples at set depths from Lake Rotokauri for the analysis of total and dissolved nutrients, chlorophyll a and dissolved oxygen concentrations, and water temperature. On each sampling occasion a Secchi depth was measured. The surface flow measurements and nutrient loadings via the inflows were obtained as part of a water budget calculation for the lake as well as from previous studies that used both field measurements and models to derive nutrient concentrations and loads. An empirical water budget for Lake Rotokauri was developed to estimate the groundwater and outflow discharge as there were no gaugings that could be applied to input these variables into the lake model. Meteorological data for Lake Rotokauri was obtained from the National Institute of Water and Atmosphere Limited database, based on measurement at the Ruakura meteorological station. Meteorological data, inflows (including empirically estimated groundwater and measured surface water discharges to the lake) and the calculated outflow were entered as daily inputs to the DYRESM-CAEDYM lake model for the period of 2009. The available data relating to 2009 were looped for 2010 to check the stability of the model and its ability to capture repeated inter-annual dynamics that would be expected with identical annual forcing data input. DYRESM is a one-dimensional hydrodynamic model that predicts the vertical distribution of temperature, density and salinity. CAEDYM is an aquatic ecological model which was coupled with DYRESM as its hydrodynamic driver to simulate transport and mixing, and output temperature and biogeochemical parameters associated with lake water quality. The model satisfactorily simulated both the surface (0 m) and bottom (3 m) water temperature and the seasonal trends including the occasional stratification periods observed through spring to autumn. The model simulations showed greater departures from field data in simulating the dynamics of biogeochemical variables, particularly the seasonal dynamics of phytoplankton. The conceptual seasonal succession in phytoplankton communities depicts dominance of cyanobacteria in summer and diatoms in winter. In the observed data for Lake Rotokauri diatoms were found to be the dominant group throughout the year. The calibrated model was able to show diatoms to be the dominant group over cyanobacterial blooms. The agreement between concentrations of nitrate and dissolved reactive phosphorus in the water column was better than for chlorophyll a, and the observed magnitude and seasonal fluctuations at both depths (0 and 3 m) were captured reasonably well by the model simulations. The total nitrogen (TN) and total phosphorus (TP) concentrations were under and over-estimated, respectively. Dissolved reactive phosphorus (PO4) was overestimated perhaps as a result of insufficient uptake of phosphorus by the two phytoplankton groups. As the present model does not contain a dynamic description of sediment dynamics, the sediment phosphorus release rates were influenced by user-defined maximum phosphorus release rate, temperature and the oxygen concentration in the overlying waters. Concentrations of ammonium were underestimated but it represented a relatively small proportion of TN. Due to wind-induced mixing and sediment resuspension, as well as convective sediment-water heat exchanges, phosphorus may be released from the bottom sediments where it has previously sedimented out. The model simulations may not have captured these internal loads of phosphorus adequately as sediment resuspension, for example, was not explicitly included in the model configuration. To depict the future water quality of Lake Rotokauri when subjected to urbanisation, three scenarios were developed which involved simulations with altered nutrient loads to DYRESM-CAEDYM and comparisons with the calibrated model which represented a ‘base’ or present case of water quality. The scenarios considered the water quality that could evolve during and after urban development, and with a range of mitigation measures, from relatively modest treatment to best management practices to reduce nutrient loads and attenuate water flows to the lake. The predicted nutrient load contributed from future urban run-off was less than the nutrient load from the pastoral run-off in all scenarios. The model indicated that the nutrient loading from a future catchment with little or modest treatment of the urban area (Scenario I) would be only slightly poorer water quality than Scenario II which examined the water quality during the construction phase. Scenario III (treated water) was most effective in reducing nutrient loads to Lake Rotokauri. At 3 m depth dissolved oxygen (DO) concentrations showed large fluctuations throughout the year for the both the base and untreated discharge scenarios. Chlorophyll a (chl a) concentrations for the untreated scenario were greater than in the base scenario. The timing of peak chl a concentrations between base and untreated discharges differed by a few days. The TP, TN and nitrate (NO3-N) concentrations of the base scenario were greater than the untreated scenario. Scenario II represented the intermediate stage towards Scenario 3 which was the optimal treatment case for the catchment. The greatest difference in DO at 0 m between the base case and scenario II was in March (i.e., base-intermediate = 2.76 mg L-1). At 3 m depth, Lake Rotokauri was predicted to be anoxic on 4 July 2011 (0.18 mg L-1) for scenario II. Chlorophyll a concentrations for scenario 2 were lower than the base case and PO4-P concentrations were higher. Concentrations of NO3-N and NH4-N at 3 m depth for scenario II were lower than the base case. Scenario III involved simulating water quality from with best management practices implemented. These practices included detention basins (grass-lined), constructed wetlands, biofiltration swales and floodways. At 3 m depth, fluctuations in DO concentration for both the base and scenario III were similar at the beginning of the simulated period, but for the months of May to November DO was lower in Scenario III than the base case. The maximum chl a for scenario 3 peaked at 30.8 μgL-1 compared to 38.9 μg L-1 observed in the base model. The TP and TN concentrations were substantially lower in Scenario III than the base model. Concentrations of PO4-P at 3 m depth were low for most of the year except in March. At 0 m depth the NH4-N concentrations were greater than the base model from mid-June to July. Concentrations of NO3-N for the treated scenario at 0 m depth were approximately 25% less than the base model. Future studies should consider an ongoing comprehensive and consistent monitoring plan that would emphasise any change in the water quality of Lake Rotokauri during and/or after high-density urban developments within the catchment. Future works should involve regular monitoring that would not only limit the uncertainties in the data but also account for any effects that may be attributable to the management plan. Restoration plans should also be considered to explore the effects of biomanipulation and re-establishment of submerged vegetation. The DYRESM-CAEDYM model may also be used to examine the effects of climate change on in-lake processes and external loads to the lake