Derwent: Biogeochemistry

The Derwent Biogeochemical Model

The objective of this project was to implement a high resolution 3D biogeochemical model of the estuary, calibrate the model against observations taken throughout the region and better characterise the cycling of carbon, nitrogen, phosphorus and dissolved oxygen in the estuary to inform managers and stakeholders. The calibrated biogeochemical model would also be available for scenario simulation of alternative management strategies and to reconstruct former conditions in the estuary prior to urbanisation.

The CSIRO EMS (Environmental Monitoring Suite) includes a 3D coupled hydrodynamic, sediment and biogeochemical model. In 2005 the hydrodynamic and sediment models were implemented for the Derwent Estuary and calibrated against observations made in 2003 to simulate a seasonal cycle of hydrodynamics, sediment transport and absorption/desorption of zinc. In this project we augment the existing models with the biogeochemical model in EMS to simulate the cycling of carbon, nitrogen, phosphorus and associated dissolved oxygen, through dissolved and particulate organic and inorganic phases. The model includes 4 types of phytoplankton, 2 types of macrophytes, 2 types of zooplankton and 4 types of particulate detritus; dissolved organic and inorganic nutrients and carbon are also included.

Model parameters were derived from observations, literature values and previous model simulations. The model ran from January 2003 for 14 months with tracer concentrations initialised from observations of nutrients, phytoplankton and dissolved oxygen; other model variables were initialised with uniform low concentration. The hydrodynamical model was forced with Derwent River flow, local meteorology and incident irradiation. For the biogeochemical model, boundaries at New Norfolk and across the estuary at Iron Pot were implemented with an upstream condition for inflowing concentrations of model tracers specified from time series derived from observations. Point source nutrient loads into the estuary in 2003 from industry and STPs were estimated from data supplied by the Derwent Estuary Program (DEP), local industry and local councils. Stormwater loads were derived from catchment model results (DEP) and observations. After marine influx and river load, STPs supplied the largest quantity of nitrogen and phosphorus into the estuary whilst industry effluent provided most carbon.

The model was validated against observations made throughout the estuary in 2003 obtained from the DEP database. Observations of nitrate, ammonia, dissolved inorganic phosphorus (DIP), chlorophyll and dissolved oxygen, in surface and bottom waters were directly comparable with model output. There were no observations of macrophytes, phytoplankton group assemblages or zooplankton for 2003, although some information on broad patterns was gathered. Validation criteria were set for the conservation of mass and reproduction of the observed timing and amplitude of the seasonal cycle in dissolved nutrients, chlorophyll and dissolved oxygen. Poorly constrained parameters were varied within known ranges during calibration to optimise the simulation of observed biogeochemical substances. The model achieved all validation criteria and simulated the observed biogeochemical dynamics of nitrate, ammonia, DIP, chlorophyll, DOC and dissolved oxygen in most parts of the estuary very well. In the upper estuary, complex channel bathymetry was not well resolved by the relatively coarse model grid and model results should be treated with more caution. In some side bays with very high nutrient loads (e.g. Prince of Wales Bay) the model was not able to reproduce the full range of observed values possibly due to sub-grid scale gradients in observed concentrations and/or under-estimation of actual nutrient input. The modelled succession of plankton species, zooplankton abundance and distribution of macrophytes broadly agreed with ancillary data except that favourable conditions for seagrass growth were simulated in Ralphs Bay, where none is currently found.

Model results show a persistent salt wedge structure in the upper estuary which intersects the sea bed upstream of Elwick Bay (near DEP station U7). Modelled nutrient concentrations were greatest in the bottom waters of the mid-estuary adjacent to the salt wedge front. Nutrients appear to accumulate in this area from point source loads and remineralisation of organic material which re-circulates in the estuarine currents. Simulated nutrient concentrations were elevated in winter and reduced in surface waters in other seasons due to phytoplankton assimilation. DIP concentrations exceed Redfield ratio in summer indicating that modelled primary production in the estuary is controlled by access to nitrogen and irradiance for photosynthesis. During 2003 the model simulated a number of high rainfall events. In March model results show the formation and dispersal of long plumes of nitrate originating from STP and stormwater discharge into Elwick Bay over a 10 day period.

Modelled chlorophyll concentrations were highest in the mid-estuary and along the shoreline in regions of elevated nutrient supply. Sustained periods of high chlorophyll occur in all seasons in sub-regions of the estuary depending on the modelled availability of light and nutrients. In the upper estuary coloured dissolved organic matter (CDOM) and opaque industry effluent limit the propagation of light and photosynthesis through the water column and modelled chlorophyll concentrations are generally low. Simulated phytoplankton biomass showed seasonal succession with dinoflagellates dominating in summer and autumn, large phytoplankton in winter and mixed populations in spring, throughout much of the estuary. In the model grazing by small zooplankton was tightly coupled with production by small phytoplankton whilst large zooplankton grazing responded more slowly to increases in large phytoplankton and dinoflagellates. In some areas modelled spring and autumn peak chlorophyll concentrations persisted longer than observed possibly due to under representation of large zooplankton growth rate. Modelled dissolved oxygen levels were reduced in bottom waters in the upper estuary and the mid and lower reaches of the estuary, particularly in autumn. Regions of low dissolved oxygen saturation were simulated adjacent to the salt wedge front, similar to the distribution of elevated nutrient concentration and likely associated with local remineralisation of organic material.

Modelled photosynthetically active radiation reaching the epi-benthos was greatest in the shallow waters of the lower estuary and Ralphs Bay, Elwick Bay and in shallow waters of the upper estuary. The model favoured macrophyte growth in these areas, however it does not resolve gradients in substrate type, disturbance or recruitment and results should be interpreted as potential rather than actual areas of macrophyte growth. The model simulated potentially favourable conditions for seagrass growth in Ralphs Bay whilst there was the potential for epiphytic macroalgae to dominate in the mid and upper estuary due to elevated water column nutrients. With access to more detailed observations of species present, typical biomass levels, substrate type, growth, disturbance and recruitment rates, the model could be improved to resolve macrophyte dynamics better.

Modelled surface sediment dissolved oxygen concentrations were lowest in the mid and lower reaches with 10 percentile monthly concentrations falling below 40% saturation in autumn and spring. In March 2003 simulated surface sediment dissolved oxygen concentrations fell to 20% saturation for 3 days in a small area close to the Tasman Bridge. Modelled denitrification flux was highest in the upper estuary and mid-estuary corresponding to regions with high sediment ammonia and low dissolved oxygen saturation. In the vicinity of Bridgewater Bridge and Ralphs Bay the simulated denitrification flux was low due to higher dissolved oxygen saturation resulting in part from the shallow bathymetry and in part from simulated photosynthesis of local macrophytes. There were no observations of sediment properties in 2003-4 to validate the simulated sediment biogeochemistry and these results should be treated only as a hypothesis of possible conditions. Recent observations in 2008 have shown high spatial and temporal variability in local sediment conditions due in part to bio-turbation and bio-irrigation of sediment by in-fauna. The impact of sediment in-fauna on pore water biogeochemistry is poorly constrained in the model, due to lack of local observations and parameterisation of these processes, and is a priority area for future model improvement.

The modelled nitrogen budget for the estuary showed that in 2003 the depth-integrated daily flux of nitrogen across the marine boundary was the largest flux into the region (44%), followed by the Derwent River (29%), STP inputs (18%), stormwater (6%) and industrial loads (3%). The largest loss term from the estuary was denitrification (59%) with depth-integrated daily flux of nitrogen across the marine boundary accounting for 41% of export. During 2003 the net accumulation of nitrogen in the estuary was a minor ~44 tN/y which suggests the estuary was in near steady state.

Modelled annual mean chlorophyll concentrations in the top 0-11 m were used to classify the estuary by area as 18.3% mesotrophic and 81.7% eutrophic. The modelled mesotrophic areas (with annual mean chlorophyll 1-3 mg m^-3) include the upper estuary where light limits phytoplankton growth, and the lower estuary and southern Ralphs Bay, where near-surface nutrient concentrations were depleted for much of the year. The modelled eutrophic region (with annual mean chlorophyll >3 mg m^-3) included the mid- and lower estuary and the remainder of Ralphs Bay.

Recommendations for future work include utilising modern instrumentation in the estuary to collect biogeochemical observations over a greater diversity of time and space scales. In addition, observations of phytoplankton, zooplankton and macrophyte properties would allow these aspects of the model to be better constrained. This study suggests denitrification plays a key role in maintaining the ‘health’ of the ecosystem and it would be good to validate the algorithms and parameterisations included in the model with detailed observations of these (as yet unvalidated) processes. The current modelling study is limited to a specific year and set of environmental conditions, it would be wise to extend the simulated period to place it in the context of natural inter-annual variability. This could be efficiently achieved through the implementation of a near real time operational biogeochemical model which is routinely updated with the most recent advances in science understanding.