4th International Conference on Integrating GIS and Environmental Modeling (GIS/EM4):
Problems, Prospects and Research Needs. Banff, Alberta, Canada, September 2 - 8, 2000.

Modeling Wetland Loss in the Richmond Catchment:

an example of historical vegetation modeling

GIS/EM4 No. 250

Owen Earley

Abstract

Wetlands occupy about six per cent of the Earth's land surface. They vary according to their origin, geographical location, water regime, chemistry, and dominant plant species. Wetlands include some of the most productive ecosystems in the world. On a national scale, wetlands are a limited resource in Australia with permanent wetlands being largely restricted to the coastal regions. As the majority of Australia' population reside in the coastal regions there is a great amount of pressure placed on wetland resources. It is estimated that 50 per cent of the freshwater inland and coastal wetlands in New South Wales (NSW) have been lost since European settlement.

A predictive GIS model was developed to assess the extent of wetland loss in the Richmond catchment, northern NSW. The model was based on multi-attribute data collected by the NSW Department of Land and Water Conservation (DLWC). The multi-attribute data was captured from 1:25,000 aerial photographs. The data set provides a comprehensive assessment of landscape attributes over the entire catchment.

Wetlands presently occupy 4.5 per cent of the Richmond catchment. Modelling indicates that prior to the clearing and draining, wetlands would have occupied between 13.6 to 20.7 per cent of the catchment. This represents the loss of between 67,199 to 116,867 hectares of floodplain wetlands since European occupation of the area.

Keywords

Wetland, clearing, drainage, ecological modeling, historical extent, GIS, predictive model.

Introduction

Wetlands occupy about six per cent of the Earth's land surface (Gosselink and Maltby, 1990). They vary according to their origin, geographical location, water regime, chemistry and dominant plants, and they are usually sustained by water sources other than direct rainfall. They include some of the most productive ecosystems in the world.

On a national scale, wetlands are a limited resource in Australia with permanent wetlands being largely restricted to the coastal regions (Adam et al., 1985). As the majority of Australia' population reside in these coastal regions there is a great amount of pressure placed on wetland resources. While the exact number of wetlands and their area in New South Wales (NSW) are not yet known, the NSW Wetland Management Policy (NSW Government, 1996) estimates that there are approximately 4.5 million ha of wetlands, about six per cent of the state's geographic area.

Problem statement

It is estimated that 50 per cent of the freshwater inland and coastal wetlands in NSW have been lost since European settlement (Sainty & Associates, 1996). Remaining wetlands are often subject to degradation problems affecting both water and land such as water pollution and vegetation clearance.

A key objective of The NSW Wetlands Management Policy is to halt or where possible reverse the loss of wetland vegetation (NSW Government, 1996). To accomplish this objective information is required concerning the state of existing wetlands, and their past extent (before clearing and draining occurred).

Background

The Richmond River catchment is located on the Far North Coast of New South Wales, Australia (Figure 1). The catchment has an area of just over 7,000 square kilometres and has a population of about 110,000. The Richmond catchment is made up of four subcatchments.

Figure 1: Location of the Richmond Catchment and its subcatchments.

The Richmond River catchment is a large floodplain surrounded by undulating ridges and creek valleys, with several mountain ranges forming the boundary of the catchment (Figure 1). The catchment is made up of several large subcatchments and many smaller coastal watersheds. Large subcatchments include those of the Richmond and Wilson's rivers and Bungawalbin Creek. Creeks originating in the coastal watersheds generally flow into the Richmond River on the lower floodplain.

Annual rainfall varies between 2,300 mm in the mountainous areas in the north, to less than 1,100 mm inland areas in the west of the catchment (Resource and Conservation Assessment Council, 1996). Rainfall shows a high annual variation due to major climatic influences such as tropical cyclones and the El Nińo - Southern Oscillation (ENSO). Average annual temperature within the catchment varies from 15°C to over 19°C with temperature being closely linked to elevation. Average maximum daily temperatures range from 26°C in the northern ranges to 32°C in the southern areas. Average minimum daily temperatures range from 4°C in the northern ranges to 12°C on the coast.

Wetlands in the Catchment

Wetlands comprise about 4.5 per cent of the total catchment area. Most of the wetlands occur on the floodplains of rivers and large creek systems, and on the coastal flats. Wetland types include swamp complex, swamp, open fresh waters, dunal waterbodies, dunal swamp, estuarine waters, mangrove forest, upland swamp and upland waterbodies.

The dominant wetland type in the Richmond River catchment is swamp complex, which makes up about 61 per cent of the total wetland area. Most of the swamp complex is located in the south of the catchment. Open areas of swamp make up 18 per cent of the total wetland area in the Richmond River catchment. The majority of these areas are grazing lands, which are subject to seasonal inundation. Wetland types such as estuarine waters and mangrove forests are restricted to the brackish areas of the Richmond River catchment. Estuarine waters account for eight per cent of the total wetland area in the Richmond River catchment. Mangrove forests fringing the estuarine rivers and creeks make up a further two per cent of the total wetland area.

Dunal swamps are restricted to coastal areas of the catchment where they occur behind the foredunes along much of the coastline. Dunal swamps comprise about six per cent of the total wetland area in the Richmond River catchment. The remaining wetlands in the catchment include open fresh water, upland waterbodies and upland swamps. Open fresh water accounts for just under two per cent of the total wetland area and occurs in the form of lakes and billabongs. The majority of these are located on the floodplains of the southern and central areas of the catchment. Upland swamps are extremely limited in the Richmond River catchment, covering a total area of only 35 hectares.

Overall, about 90 per cent of the wetlands in the Richmond catchment occur in the Bungawalbin subcatchment and the coastal subcatchments. Central areas comprise a further eight per cent of the total wetland area. The remaining two per cent are located in the northern and north-western areas of the catchment. In these areas the steep topography is highly restrictive to the formation of wetlands.

Methods

The methods employed in this project were developed to estimate the historical extent of wetlands. This was achieved through the use of an environmental model highlighting the physical conditions required by wetlands.

Wetland Extent Model

The wetland extent model is based on the principle that a vegetation community will form due to specific local physical conditions. For example, over time at any geographical location certain species will be selectively favoured due to the physical conditions at that location. Such correlations have been shown by Ingwerson (1983), who noted high influence of terrain on the distribution of eucalypt species. Prior to extensive human influence on vegetation distribution vegetation communities would have correlated closely with localised physical conditions.

To assess the possible extent of wetland communities prior to clearing and draining an analytical model was developed. This analytical model is based on the combination of Multi-Attribute Data modeling, drainage proximity modeling and flow accumulation modeling (Figure 2). The result of the model is the wetland suitability index.

Modeling was designed to highlight areas that were likely to have been floodplain wetlands. These include open wetlands such as reed swamps and forested wetlands such as swamp complex. The wetland suitability index serves two purposes:

1. to estimate the losses of floodplain wetlands in the catchment; and

2. to highlight the most suitable locations for any wetland rehabilitation or restoration works.

Figure 2: Component relationships in the wetland suitability index.

The following sections describe the methods used in each stage of modeling during the development of the index.

Multi-Attribute Data Model

The first stage in the development of the wetlands suitability index was a model based on the Multi- Attribute Data. This model was designed to select the physical landscape features that would be either favourable or unfavourable to the existance of a wetland. Areas of known historical wetlands were sampled to determine the physical parameters showing in the data set. The favourable (positive) parameters used were certain terrain, geology and soil types. Unfavourable (negative) parameters included slopes greater than two per cent and the existance of non-wetland species vegetation types. Parameters have not provided in this paper as they are specific to both the Richmond Catchment and the Multi-Attribute Data.

Flow Accumulation Model

To calculate flow accumulation a Digital Elevation Model (DEM) was used. This process involves determining the accumulated flow, or number of upslope cells. Relatively flat areas with a large number of upslope cells are highlighted through this process. Modeling was carried out using the ArcView Spatial Analyst Hydrology extension. This extension provides useful tools for delineating drainage basins and calculating basin parameters to be used in runoff modeling.

Pre-processing included the filling of ‘sinks’ or ‘pits’ in the DEM. These are a set of grid elements surrounded by higher terrain that, in terms of the DEM, do not drain. These features are extremely rare in natural topography and are generally assumed to be artifacts arising during the development of the DEM (Pack et al., undated). The Spatial Analyst Hydrologic Extension includes capabilities for identifying and filling ‘sinks’ prior to modeling.

The results of the flow accumulation modeling were converted to a vector polygon coverage. Areas with a value greater than one standard deviation from the mean value were included. Theses areas of high flow accumulation were given a value of one, areas with limited flow accumulation were given a value of zero. The resulting coverage was clipped with the results of the Multi-Attribute model and then by subcatchment.

Drainage Proximity Model

The second stage of the wetlands suitability model was an assessment of the proximity to drainage features. Drainage features in areas positively identified by the Multi-Attribute model were included in the second stage of modeling. All drainage features outside of this area were excluded. This was achieved by clipping the drainage coverages with the area positively identified in the Multi-Attribute model. Buffering was then used to highlight areas with close proximity to drainage features.

The extent of land affected by a drainage feature is dependent on many factors such as drain depth, local topography, soil type, sub-surface flow rates and variable factors such as water table height. This information is largely unavailable for the large number of drains in the catchment. Due to these restrictions, an arbitrary buffer interval was used.

The drainage theme to be used in the proximity model was developed by joining all of the 1:25,000 drainage coverages to create a single catchment drainage coverage. This was then clipped with the areas positively identified in the Multi-Attribute Data modeling. The drainage proximity theme was developed by buffering the drainage theme at distances of 150, 300 and 450 metres. Buffering of the drainage coverage was completed using the geoprocessing capabilities of ArcView. The final coverage was then re-clipped to remove areas that were not identified in the Multi-Attribute Data modeling.

All drainage coverages include both artificial and natural drainage features. No distinction is made between artificial and natural drainage features in the coverage. It was seen to be beyond the constraints of the project to distinguish natural and artificial drainage channels. This was due to the large number of drainage features within the study area and the amount of time needed to edit the dataset. However, this was not seen to reduce the quality of the model as any natural drainage features occurring within the areas identified in stage one of the model are likely to support the existance of wetland communities. The presence of artificial drainage features shows that at an area has been drained.

Wetland Suitability Index

The wetland suitability index combines the results of the Multi-Attribute modeling, flow accumulation modeling and drainage proximity modeling. The index was created by overlaying the results from each of the component models. Values generated by each component model were summed to create the final index score (Table 1).

Table 1: Components of the Wetland suitability index value.

Prior to overlaying, model components were divided into subcatchment areas. This was done to reduce computer processing requirements and to provide results based on subcatchment areas. The catchment was divided into subcatchments through on-screen digitising in ArcView. A digital elevation model with hill shading and a catchment drainage theme were used to identify subcatchment boundaries. The catchment was divided into four subcatchments for hydrologic and drainage proximity modeling.

Estimating Wetland Loss

The Multi-Attribute modeling was designed to highlight the maximum area suitable for floodplain wetlands. This can be seen as the maximum theoretical loss for floodplain wetlands (Equation 1). However, it is unlikely that wetland communities occupied the maximum potential area due to competition from other vegetation communities. In areas marginally suitable to wetland vegetation other communities such as wet sclerophyll forest may have dominated.

To counter this uncertainty, a range of estimates was developed for floodplain wetland loss. The high loss estimate (index values two, three, four and five) includes areas that are moderately suited to wetland vegetation (Equation 2). These areas are either within 450 metres of natural or constructed drainage channels and may be within an area of flow accumulation. The mid loss estimate (index values three, four and five) include areas that are quite suited to wetland vegetation (Equation 3). The majority of these areas are within 300 metres of a drainage channel. They may also include areas within 450 metres of a drainage channel, which are also in an area of flow accumulation. The low loss estimate (index values four and five) includes areas that are highly suited to wetland vegetation (Equation 4). There is a high level of certainty that these areas were formerly wetland communities. All areas included in the low estimate are within 300 metres of drainage channels. All estimate ranges are within the limits defined by the Multi-Attribute modeling.

The percentage of wetland lost in each subcatchment was also calculated. This calculation is based on the amount of wetland remaining (Table 2) and the predicted losses from Equations 12, 13 and 14 (above). Values were calculated for the high, mid and low loss estimates for each subcatchment.

Table 2: Subcatchment size and existing wetland vegetation area.

Model Limitations

The wetland suitability index should be seen as a guide to the potential location of wetlands prior to draining and clearing. Detailed local assessments of landscape, hydrology and historical records would be required to determine the exact amount of wetland loss throughout the catchment. The index does not attempt to determine the type of wetland that may have been present in positively identified areas. Some broad estimates can be made based on the contributing factors used in the modeling process.

The flow accumulation function of the Hydrological Extension relies upon having elevation information with measurable vertical relief. As vertical relief is limited on floodplains, it is difficult to determine flows across floodplain areas on a DEM. The flow accumulation model should be regarded as an assessment of contributing area on a catchment wide basis.

The drainage network is a highly complex system. Due to the limitations in the drainage data set, simplifications had to be made when determining the area of influence of a drain. At present, mapping of the constructed drainage network is being carried out in NSW by the DLWC. This project is nearing completion in the Richmond catchment (Hallinan, pers. comm.). Drainage mapping includes attributes such as drain size, which would increase the accuracy of drainage modeling in determining landscapes that may have supported wetlands.

Findings

Wetland suitability modeling indicates that between 67,199 and 116,867 hectares of floodplain wetland vegetation has been cleared or drained in the Richmond catchment (Table 3). The potential extent of wetlands Richmond catchment prior to clearing and draining is shown in Figure 4. The majority of wetland clearance has occurred in the Richmond subcatchment (25,060 to 50,569 ha) and the coastal subcatchments (22,584 to 29,322 ha). Considerable clearance has also occurred in the Bungawalbin subcatchment (10,102 to 20,203 ha) and Wilson subcatchment (9,454 to 16,773 ha).

Table 3: Wetland loss estimates for subcatchments within the Richmond catchment.

 

Background image: DEM with hillshading
a) Existing Wetlands	b) Wetland Loss

Figure 4: Existing wetlands and predicted extent prior to clearing and draining.

Modeling indicates that percentage loss is highest in the Wilson subcatchment with figures between 97.8 and 98.8 percent (Table 3). Percentage loss is also very high in the Richmond subcatchment (95.0 to 97.4 percent). These figures are very high due to the limited extent of remaining floodplain wetlands in the Richmond and Wilson subcatchments (Table 3). Figures are lower for the coastal subcatchments (67.6 to 73.0 percent loss) due to the presence of several large protected wetlands. The Bungawalbin subcatchment has the lowest percentage loss (39.5 to 56.6 percent) due to the presence of many large wetlands.

Discussion

Wetlands presently occupy 4.5 per cent of the Richmond catchment. Modeling indicates that prior to the clearing and draining of wetlands, this area would have been between 13.6 to 20.7 per cent. This represents the loss of between 67,199 to 116,867 hectares of wetlands since European occupation of the area. Due to the scale of continuing wetland clearance in the Richmond catchment, all remaining wetlands should be considered as valuable.

The large coastal wetlands in the Richmond catchment are protected to some degree. This includes protections such as State Environmental Planning Policy 14 (Coastal Wetlands) and as National Parks or Nature Reserves. However, the vast majority of non-coastal wetlands in the Richmond catchment have little protection. Of particular concern is the Bungawalbin subcatchment, which contains about half of the wetlands remaining in the Richmond catchment.

Evaluation of Methods

The methods used to produce the wetlands inventory was effective in determining the current extent of wetlands in the Richmond catchment. It is expected that the data quality is relatively high as it was captured from 1:25,000 scale aerial photography. The format of the DLWC’s Multi-Attribute Data facilitated the querying of landscape elements to identify wetland features.

Due to its comprehensive nature, Multi-Attribute Data is very time-consuming to develop. This has limited the area covered by the Multi-Attribute Data to the Far North Coast of NSW. In some areas, such as the Clarence River catchment, only some attributes have been captured. This limits the use of the exact project methodology to a relatively small geographical area. However, concise datasets based on wetland project requirements could be developed more rapidly. The application of other techniques, such as satellite image classification, may provide a more effective means of assessing and monitoring large catchments.

The wetlands inventory GIS dataset provides useful baseline data for the development of satellite imagery interpretation techniques. The dataset could easily be used to identify training areas for satellite image classification. This image classification method could then be applied to rapidly assess the extent of wetland types and monitor changes over time. Unfortunately, these methods were not developed during the project due to time constraints and existing workload.

Modeling was effective in providing estimates on the amount of wetlands that have been cleared. The reconstruction of the past state of floodplain wetlands within the Richmond catchment is a difficult task as there has been widespread clearing of the floodplains for agriculture. There has also been considerable alteration to floodplain hydrology through an extensive artificial drainage network, road levees, floodgates and canal developments.

Overall, the methods were highly effective in meeting project objectives. However, due to the limited spatial extent of Multi-Attribute Data, different methods may be more appropriate in other areas. To create a wetland inventory, a specialised GIS based on wetland boundaries captured from aerial photograph interpretation or satellite imagery is appropriate (Lee and Lunetta, 1995). However, a more comprehensive data set of landscape attributes is often required for modeling purposes.

Conclusion

Geographical Information Systems with appropriate data can provide valuable insights into natural systems.  In floodplain environments that have been extensively modified by humans, the modeling of physical characteristics may help determine the past extent of wetlands. This type of modeling also highlights suitable areas for wetland rehabilitation or restoration.

Recommendations for future research

Throughout this study, the need for further information was apparent. However, actions to conserve the remaining wetlands in the catchment should not be delayed due to lack of information. There is already considerable evidence regarding the importance of the wetlands remaining in the Richmond catchment to warrant their conservation. Further studies would provide useful information to validate the accuracy of modeling and to assist in wetlands management. There are several key opportunities for further studies:

1. Validating the model results with detailed historical information such as maps, diaries, letters and survey marker tree records;

2. Determining the state of existing wetlands at a local scale. The best means to achieve this objective is through the development of a standard appraisal method;

3. The need to establish standard monitoring procedures for wetlands has been identified (Jensen, 1997). Satellite image interpretation provides an opportunity to rapidly assess the extent of wetlands and monitor changes over time; and

4. Wetlands are inherently dynamic systems. The spatial extent of floodplain wetlands is constantly responding to hydrological conditions. An assessment of the spatial variability of floodplain wetlands would provide valuable information, particularly when linked to climatic patterns such as El Nino.

Acknowledgements

The author wishes to express his gratitude to staff at Southern Cross University, NSW Fisheries and the NSW Department of Land and Water Conservation for assistance throughout the project. Particularly Prof. Leon Zann (Southern Cross University) for supervising the project. The author would also like to thank his family for all of the time taken beyond work hours to complete this project.

This project was completed as part of a Master of Science degree at Southern Cross University, Lismore, New South Wales, Australia.

References used

Adam, P., Urwin, N., Weiner, P., and Sim, I. (1985) Coastal Wetlands of New South Wales, Coastal Council of New South Wales, Sydney.

Gosselink, J.G. and Maltby, E. (1990) 'Wetland Gains and Losses', in: Williams, M. (ed.) Wetlands: A Threatened Landscape, Basil Blackwell, Oxford.

Hallinan, M. (Project Officer – Acid Sulphate Soil Drainage Network Mapping) personal communication, 19/05/1999.

Ingwerson, F. (1983)Numerical analysis of the timbered vegetation in Tidbinbilla Nature Reserve, A.C.T., Australia, Vegetatio, Vol. 51 pp 157-179.

Jensen, A. (1997) ‘Australian Wetland Policies, Research and Management’, in: Williams, W.D. (ed.) Wetlands in a Dry Land: Understanding for Management, Workshop Proceedings, 29-30 September, Albury, NSW.

Lee, K.H. and Lunetta, R.S. (1995) 'Wetlands Detection Methods', in: Lyon, J.G. and McCarthy, J. (eds.), Wetland and Environmental Applications of GIS, Lewis Publishers, New York.

NSW Government (1996) The New South Wales Wetlands Management Policy, Prepared by the Department of Land and Water Conservation, Sydney.

Pack, T.R., Tarboton, D.G. and Goodwin, C.N. (undated) SINMAP: A Stability Index Approach to Terrain Stability Hazard Mapping - User's Manual, Utah State University, Logan.

Resource and Conservation Assessment Council (1996) Regional Report of Upper North East New South Wales: Physical Attributes, New South Wales Government, Sydney.

Sainty and Associates (1996) Review of the Efficiency of Compensatory Wetlands, Report to the Department of Urban Affairs and Planning, Sydney.

Authors

Owen Earley, Senior Land Information Officer
Central Land Council, PO Box 3321, Alice Springs, Northern Territory, Australia 0871.
Email: owen.earley@clc.org.au, Tel: +11-08-8951-6340, Fax: +11-08-8952-1590.