DEVELOPMENT OF A GIS-BASED ESTUARY SEDIMENTATION MODEL

Eleanor Bruce (1), Peter J. Cowell (1), David Stolper (2)

(1) School of Geosciences, University of Sydney (AU)
(2) United States Geological Survey, Woods Hole, MA (US)

1. Introduction
Anthropogenic activities have resulted in a direct loss of coastal wetland ecosystems in Australia and internationally. Due to the remnant nature of remaining wetland habitats they are highly vulnerable to human induced change such as sea-level rise. Increasing concern by researchers and environmental managers for the potential impacts of projected climate change to coastal environments highlights the importance of comprehensive model-based approaches for predicting and assessing longer-term change (Capobianco et al., 1999).
Sea-level rise (SLR) results in increased inundation and coastal erosion (Bryan et al., 2001), which has major implications for coastal aquatic ecosystems such as saltmarshes (Simas et al., 2001). Ecosystems which form the transition zone between land and sea, including mudflats, mangroves and saltmarshes, are highly vulnerable to rises in sea-level. Complex biophysical interactions stimulated by SLR will influence the type and rate of ecosystem response. The response of coastal saltwater wetlands to relative SLR is dependent on sedimentation rate and consequently the maintenance of relative elevation (Reed, 1990). If sedimentation rate exceeds SLR a positive balance exists however, in the reverse of this situation, erosion occurs resulting in inundation. Preliminary modeling paradoxically shows intertidal areas may diminish with excessive sedimentation rates (Stolper, 1996). This highlights the importance of investigating the morphological implications of different scenarios. Under gradual SLR increased submersion time results in higher sediment deposition and reduced soil compaction allowing saltmarsh ecosystems to gradually respond (Allen, 1994). However, the level of submersion time associated with rapid SLR can reduce saltmarsh production and associated organogenic sedimentation and result in overall loss (Simas et al., 2001). There is a need to examine the spatial relationship between differing levels and rates of SLR on the elevation profiles of coastal habitats.
Under differing SLR scenarios the extent and geographical location of intertidal areas and associated habitats will vary. In determining the adequacy of current reserve areas and planning mechanisms for protecting future coastal ecosystems from adverse development impacts it is critical to have an understanding of habitat distribution scenarios. Will current infrastructure developments restrict natural processes of ecological succession and limit the potential geographical range of important coastal habitats? The current research involves the development of a GIS based tool to provide insight into the rate of change and assist coastal planners assess future implications of current planning decisions.

2. Project Background
This paper presents the initial research phase of a longer-term collaborative project between the School of Geosciences and SOPA established in January 2003 to develop a spatial model for predicting the impact of SLR on remnant saltwater wetland ecosystems. The aim of the investigation presented in this paper was to integrate the Estuary Sedimentation Model (ES) (Stolper, 1996) into a GIS environment using Homebush Bay as a case study site.
The Millennium Parklands form part of the Homebush Bay environs and is one of the world's largest urban parks covering an area of 432 ha. The site is located approximately 14 km west of the Sydney CBD adjacent to the Parramatta River (Figure 1). Since the 1880s significant areas of mangrove forests and wetlands in the surrounding area have been lost to landfill and the dumping of domestic and industrial waste. In the 1990s an extensive remediation program involved the consolidation and isolation of 9 million tons of waste (Hudson et al., 2000). The Parklands support a range of distinct ecosystems including the saltwater wetlands comprising mudflats, saltmarsh and mangroves. Although the Millennium Parklands are a highly modified environment the saltwater wetland ecosystem represents some of the last remaining saltmarsh communities in Sydney Harbour (Burchett et al., 1998). The study site was selected due to the conservational significance of the remnant wetland ecosystems and growing management concern for the fragility of these saltwater habitats. Wetland sites within the Parklands are included on the Australian Heritage Commission's interim list of the Register of the National Estate.
The ESM was developed initially by Stolper (1996) to simulate aggradation of intertidal zones under conditions of SLR and varying rates of sediment supply. The model however, ws not integrated into a GIS. Incorporation of the ESM algorithms (Stolper, 1996 and Stolper, 2002) into ArcGIS through the VBA (Visual Basic for Applications) development functions provides a spatial visualisation tool for coastal managers to examine 'what if' ecosystem response scenarios and dynamic stochastic simulation to manage data uncertainty (Cowell and Zeng, 2003).

3. Estuarine Sedimentation Model
Simulation of estuary evolution in the ESM is governed by three factors: (i) changes in sea level (rise or fall); (ii) elevation-dependent accommodation space available for the deposition of sediment; and (iii) inundation-dependent vertical accretion of sediment (aggradation). In this phase of GIS model development the impacts of estuarine management practices such as the emplacement of artificial structures were not included.

3.1 Sedimentation Rate
In the ESM sedimentation is represented as an increase in estuary bed-elevations, including intertidal ground elevations. Thus the subtidal areas become shallower (for a given sea level) at a uniform rate throughout the estuary. The rate of deposition is determined by an elevation-dependent sedimentation function defined for specific estuaries to which the model is applied. This rate can be specified through input sedimentation curves.
The sedimentation rate is less for intertidal areas than for subtidal areas. Intertidal sedimentation rates decrease with elevation of the intertidal surface. In the intertidal zone the sedimentation rate decreases with increased elevation as higher elevations are inundated less frequently and have less access to suspended sediment (Stolper, 2002). Intertidal sedimentation thus depends, in addition to the rate of sediment supply to the estuary, on the tidal characteristics of a given site.
The rate of siltation (shoaling or infilling) of the estuary relative to mean sea level depends also on changes in sea level itself. So, rising sea levels tend to offset the shoaling induced through sedimentation in maintaining water depths throughout the estuary. The following simple implications ensue for siltation of the estuary:

The total area of land available for different intertidal habitats (ie, different ranges in intertidal elevations) is governed by the interplay of sedimentation rates, SLR, and the initial terrain (Stolper, 1996). The initial terrain includes the topographies of the initial intertidal zone and the surrounding land surface. The influence of the land surrounding the estuary is introduced through its control over the amount of area inundated during a given SLR. The size of this area in turn affects the extent of sedimentation possible in the upper parts of the intertidal zone. This area may be altered by the construction of artificial levees (including revetments and bund walls). However, in the current model-development phase the influence of such estuarine-management interventions on estuary evolution is not included.

3.2 Model Inputs
The three main inputs to the ESM include: (1) Digital Elevation Model (DEM); (2) sea level curve and; (3) sediment supply curve. The DEM is a continuous surface representing terrestrial, intertidal and subtidal elevation levels. Due to the transitional nature of this environment generation of the DEM requires integration of data sets that are often compiled from disparate sources. Model results are limited by the resolution of data obtained for the DEM. Although data uncertainty remains a predictive constraint, consideration should be given in the selection of surface interpolation techniques used to derive the DEM. In this study contour data (1 metre interval) derived from high resolution aerial photography and hydrographic survey of the Homebush Bay and Haslams Creek study site were used to interpolate the DEM surface.
The model application has been customised to incorporate alternative sea-level scenarios (represented as sea level curves) that can be specified by the user. The sea-level curve adopted in the Homebush study is the IPCC (2001) median projection starting from year 2001. The sea-level curve S(t) is specified as a set of coordinates {t,S}. The set of time coordinate values in the curve need not correspond to the set of simulation time-step values specified for the model by the user. The ESM automatically interpolates for each time step during initialisation of a simulation. This permits choice of different settings for different simulations using the same sea-level curve.
The maximum rate of sedimentation within the estuary is defined by the sediment supply curve. However, the deposition level occurring for each grid cell is determined by the elevation-dependent sedimentation function. Sedimentation is incorporated as an input to the model through a file containing the proportion of maximum sedimentation rate at corresponding elevations. These data points are linearly interpolated to create a function relating elevation and sedimentation. Data sources for the elevation-dependent sedimentation function include tidal gauge data.
The elevation-dependent sedimentation function assumes that the rate of deposition at each elevation is directly proportional to the duration of inundation. Further research is needed to refine this and also to incorporate the influence of river flooding. Radiometric dating, intertidal sedimentation erosion tables or other marker horizon methods are required to improve model estimates. However, in this study the adopted sediment curve and associated elevation-dependent sedimentation values were compiled from research on late Holocene averages estimated from radiometrically calibrated seismic and core data presented in Roy (1983). These data were collected for the Sydney Harbour but is assumed applicable to the Homebush Bay site.

3.3 Model Procedures: interpolation of sea level and sedimentation increments
The GIS model runs in a series of uniform time steps in which discrete sea level and sedimentation increments are applied. The number (N) and length ( years) of time increments are specified by the user, through an interactive menu option, prior to running the model.
Default values are N = 10 and = 10 years.
The duration of the simulation then is

and the time elapsed after n steps is

Estuarine sedimentation processes are simulated in the model through a series of computations. During each time increment the following model computations are completed:

  1. The vertical reference frame is transformed to account for sea-level change. A grid layer distinguishing subtidal, intertidal and supratidal zones under the imposed sea level conditions is generated.
  2. The area of accommodation space available for sediment deposition under the sea level derived in computation (1) is calculated and delineated in a grid layer.
  3. The elevation-dependent sediment function is applied to calculate the sediment aggradation level determined by cell elevation for cells within the accommodation space through the use of raster based map algebra.
  4. A new DEM is generated in which sediment aggradation levels have been added to the cell elevation values of the input DEM.
  5. The morphological change resulting from the derived sediment aggradation levels is displayed to provide the user with time incremented visualization.

3.4 Model Outcomes
In simulating intertidal sedimentation the GIS model produces outputs at each time increment or iteration. Resulting terrain surfaces are stored in DEM format, which are displayed as 2.5 dimensional surfaces and summarised in histogram plots. Rate of change is also depicted through sedimentation time series plots generated by the model. These results are presented during the simulation as screen displays.
Figure 2 shows the simulated evolution of Homebush Bay in five equal time steps over the next 100 years. These model results were calculated using a constant reference aggradation rate of 0.3 mm per year. The predictions are based on an SLR mid-range estimate of 0.86 m to year 2100 (IPCC, 2001). Figure 2 depicts the distributions of sub-, inter- and supra-tidal elevations at the SOPA Homebush Bay field site (with initial conditions based on a 1:4000 scale base data). The model probably applies better to Homebush Bay than the Parramatta River (Figure 1). As evident in Figure 3, the former water body is less horizontally constrained (ie. channelised) than the latter. The simulation is a first order approximation ignoring effects of wind wave fetch, tidal currents and river flows, all of which are responsible for remobilization of sediments (thus limiting deposition). These processes will be incorporated in subsequent phases of ESM development (Section 4).
Figure 3 shows the model results applied to the Homebush Bay study site after 100 years. This figure places the sedimentation in the management context of the cadastre, current land uses and landscape. Another limitation of the first approximation ESM is evident: sedimentation occurs in the model behind engineering structures that preclude deposition in reality.

4. Limitations and Future Considerations
This paper presents the preliminary application of the ESM within a GIS environment. The current GIS application provides an interactive simulation of sediment deposition in intertidal zones under varying sea level rise scenarios. It provides a management tool to quantify, spatially represent and visualise predicted changes in the elevation profile of intertidal coastal environments. However, this first approximation simplifies very complex geomorphic processes and does not include biophysical interactions. There are several assumptions and limitations inherent in the initial stage of model development that provide the focus for continuing GIS-based ESM research. Limitations associated with the GIS model presented here include the simplification of sedimentation processes, spatial variation in data uncertainty associated with DEM inconsistencies within intertidal areas and the absence of human infrastructure impacts. The fine scale resolution of estuary sedimentation processes also highlights the need to assess spatial data accuracy and model sensitivity. In addition to these model refinements the future requirement in the broader project is the incorporation of ecosystem response.
Currently the model assumes sediment deposition in all intertidal and subtidal areas and does not account for erosion and variations in biogenic sedimentation, both influenced by rate of SLR. Sedimentation is only incorporated in the model as a function of elevation which does not account for differing rates of deposition associated with geomorphic unit (for example flood tide delta, mud basin and fluvial delta). This highlights the need for further work to refine the elevation-dependent sedimentation function under different ecological and geomorphological conditions and determine rates of sediment loss.
Establishing a suitable datum for examining geographical phenomena that extend the intertidal domain is a fundamental problem when implementing GIS in coastal geomorphology (Raper, 1999). The sensitivity of coastal geomorphic processes to absolute elevation emphasises the importance of vertical resolution in GIS model development. Problematically the domain of focus for the ESM, the intertidal zone, represents the area of greatest data uncertainty. Temporal inconsistencies in the image capture of remotely sensed data used to derive height data, which need to be reconciled with tidal records, can result in multiple representations of the shoreline. The foundation data layer in the ESM is a DEM that integrates hydrographical and terrestrial elevation data. Commonly data capture for these areas are the responsibility of different agencies potentially resulting in discrepancies in datum, resolution, time of data capture and compilation standards. Although resolving these sources of potential error bias is difficult, it may be possible to examine the implications of error sources through spatial sensitivity analysis.
Assessment of model reliability is often neglected in many GIS based environmental models (Zeng and Cowell, 1999; Bartlett and Bruce, 2001). However, decision makers require an understanding of the certainty of information on which land-use planning and environmental management decisions are based. The effect of data uncertainty in the input spatial data sets and model parameters will be determined and incorporated through the application of fuzzy logic in the resulting GIS layers (Cowell and Zeng, 2003). Spatial sensitivity analysis will also be adopted to examine the model outcome when the input parameters are systematically modified. This will assist in identifying data sets, such as the sediment curve, requiring further refinement and prioritising fieldwork.
Predicting ecosystem response to modeled sea-level conditions will provide an understanding of future coastal habitat distributions. The next challenge in implementing the ESM within GIS is to develop methods for characterising environmental conditions of saltwater wetland habitats. Spatial analytical techniques within GIS allow species and ecosystem prediction mapping based on multiple environmental variables including substrate conditions, nutrient levels, exposure (eg, to waves and currents), and other geomorphic determinants. GIS-based ecological modeling approaches (such as the application of Bayesian theory) will be applied to predict future habitat ranges based on the presence and absence of contemporary habitat distributions. This research will incorporate data from other Australian estuarine areas ranging in level of human modification.

5. Conclusion
GIS implementation of the ESM is directed at the provision of a simulation tool readily accessible for educational and research use. The application provides an interactive simulation environment for users to examine how: (a) total intertidal area may alter under different SLR scenarios and sedimentation conditions; (b) extension of the intertidal area under predicted SLR conditions may be restricted by coastal management interventions and; (c) management practices affecting the biophysical relationships operating within coastal wetland environments can best ensure ecologically sustainability. The initial phase of GIS implementation provides insight into sedimentation rates under different inundation durations. The next phase of model development involves extending the model algorithms to incorporate dynamic tidal parameters and the influence of ecological and geomorphic units.

Acknowledgements
The Sydney Olympic Park Authority (SOPA) provided project funding in addition to GIS and field monitoring data. The research work was undertaken in collaboration with SOPA staff (John Hudson, Edwina Laginestra and Thomas Zeng).

Figure 1. Location of Homebush Bay study site and Millennium Parklands.

Figure 2. Depicts change in the distribution of sub-, inter- and supra-tidal zones over a 100 year period at 20 year increments for the SOPA Homebush Bay study site. (Data Source: SOPA 2002)

Figure 3. Modelled sediment accumulation over a 100 year time projection (A.). Aerial photograph for the Homebush Bay study site for comparison of current environment (B). (Data Source: SOPA 2002)

References