3 Restoration Plan GIS Analysis

• Remnant vegetation areas and (assistance in) delineation of priority remnant areas, categorised as High, Medium and Low priority for protection.
• Recommended priority stream restoration reaches with a table of appropriate minimum buffer widths for all stream orders.
• Recommended areas where perennial pasture is preferred as a hydrological control -showing property boundaries and main roads.
• High nutrient source areas within the catchment - with stream network overlaid.
• High nutrient export risk areas and potential extent of influence on associated waterways.
• Pathogen risk sources and potential waterways 'hot spots' - in conjunction with the Marbellup Protection Plan being prepared by Department of Environment.

The maps produced as a result of this analysis are listed in Appendix 2.

3.1 Nutrient Risk Modeling

• A map of the high nutrient source areas; and
• A map of high nutrient export risk areas.

In addition, nutrient risk modeling is an important part of identifying perennial pasture location and priority stream restoration areas, as each of these is at least in part a response to nutrient export risks. These four maps, in turn, will inform a large part of the entire Torbay Restoration Plan. We therefore undertook a significant modeling exercise in order to best identify the high nutrient source areas and, by combining these with other factors, the high nutrient risk areas. The process undertaken is discussed below.

Nutrient Risk Model approach

The approach taken follows on from the PEWOC approach described by Weaver et al. (2003), which was in turn adapted to examine nutrient management scenarios for the Peel-Harvey catchment (Neville, 2005), to develop a tool called the Support System for Phosphorus Reduction Decisions (SSPRED).

The nutrient export model used in SSPRED is a modification the P indicators tool of Heathwaite et al. (2003) which uses three model layers: loss potential, represented by nutrient inputs and soil mineralisation; transfer indicators (effective rainfall and erosion risk); and delivery indicators (land drainage or hydrological connectivity). These layers synthesise a number of existing modelling approaches in the UK. As we do not have other modeling approaches in the south west of WA to build on this approach was adapted using local data and research.

While SSPRED was developed for Phosphorus, the underlying model can be used for Nitrogen as well, although a variation on the nutrient loss tables - suitable for N pathways - is used for nutrient delivery.

The combined nutrient export model as developed - 'SSPNRED' - combines three layers:

• Nutrient Source: landuse related nutrient availability;
• Nutrient Delivery factors - based on susceptibility to nutrient loss from the most limiting soil/geomorphological factor (aka 'Nutrient Loss Risk'); and
• Connectivity/transport potential - based on waterway proximity, and sub-catchment related to in-stream processing of nutrients or assimilation.

Base map units

The modelling is carried out for a set of unique map units for the entire Torbay catchment, derived from an intersection of each of the model layers, shown in Figure 1.

Figure 1 - Mapping Units used in N & P Loss Modeling

Landuse & Nutrient Source

2 The farm gate balances carried out in WA do not yet include N fixation through leguminous pastures, so a significant shortfall in N surplus is actually to be expected. 3 As beef cattle is the largest landuse in the catchment, it had to be the major source of N for the two catchments - Torbay Main Drain and Marbellup Brook - where N was significantly under-estimated by the model. The calibration process allowed us to increase these levels to a point where N was being well-estimated in all catchments.

Available P and N was calculated initially using farm-gate nutrient audits, to derive an annual nutrient 'surplus' in kg/ha/year. Land use-based P & N surplus figures are sourced from land use farm gate nutrient balances for agricultural land uses (Neville, 2004A), Lance, 2005) or derived from published work for urban land uses (Gerritse, et al. (1990), Kelsey and Zammit (2003), Whelan et al. (1981)). Export rates are in kg/ha available for loss, and represent the gross annually-available nutrient. Note that as nitrogen is fixed in large amounts from the atmosphere, the N surplus figures had to be considerably modified during the calibration process for the model. This was particularly true ion the case of beef cattle, which had to have its N surplus set far higher than the farm gate balances indicated in order to provide the levels of N being measured at the gauging stations.^{2 3}

Figures were not available for some potentially important landuses, such as Dairy Sheds and Rubbish Disposal sites: the former can be estimated but is subject to management of effluent. The latter has not been measured for the landfill sites in the catchment.

The surplus figure for the wastewater-irrigated plantations in the Seven-mile creek catchment may be an overestimate, and would produce a higher site discharge than Water Corporation figures, although one still well below the Ministerial conditions for the site. However we arrived at this figure in order to try to improve the results in the Seven-mile creek: even so, this is the only sub-catchment where the modeled N and P are still well below gauged loads (71 and 67% respectively), indicating that either we have under-estimated the catchment nutrient surplus, or the in-stream processing of nutrients is less effective than our model suggests. Note that the only other significant nutrient producers here are beef cattle and mixed grazing, both of which have already had modifications to their export rates.

Nutrient generation rates thus estimated are an annual average figure, and have been modified to produce the best result from the model. Other factors (soils, geomorphology, stream assimilation) will influence the proportion of the production loss that reaches both local drainage and the lakes or estuaries.

Landuse nutrient surplus estimates are as follows:

Soil & Landform and Nutrient Loss Risk

Nutrient Delivery factors

An existing framework for P loss risk for WA soils (Van Gool et al. 2001) using soil and landform qualities (susceptibility to water erosion, flood risk and specific landforms/landscape location, soil PRI and water-table depth) has been modified to represent loss risk as weighted percentages for land units. This loss risk simulates the ability of the soil to store P or to lose P through leaching, surface runoff, loss of soil particles (Neville, 2004B).

The Van Gool et al. framework as modified is presented in Table 5:

Nitrogen

A new framework for N loss risk for WA soils was created in conjunction with staff in the Department of Agriculture, WA during the CCI project (see Neville, 2004B) and was used here. This uses soil and landform qualities (susceptibility to water erosion, flood risk and specific landforms/landscape location, soil PRI and water-table depth) to represent loss risk as weighted percentages for land units. This loss risk simulates the ability of the soil to store N or to lose N through denitrification, volatilisation, leaching, surface runoff and loss of soil particles. Additional work will be required on this model but this was not possible in the timeframe of the current project.

Each nutrient risk framework produces a nutrient export susceptibility risk (Low to Extreme) for each soil group. In the digital datasets developed by the Department of Agriculture, soil occurrence is represented as a percentage for each landscape mapping units, rather than being mapped directly. We therefore had to derive an area-weighted risk factor for each mapping unit.

This area-weighted risk factor is based on an estimate of probable P loss (a value ranging from 0% - no losses to 100% loss) for each phosphorus export susceptibility risk class. This estimation was done in the Peel-Harvey project) through an iterative process, which allowed us to vary estimated risk values for each class of P Export Risk class for the entire study area, and see the result on the correlation of the model results against monitored catchment loads (see Neville 2004B). The relative risk of each class (L<M<H<VH<E) was maintained throughout the process, and adjusted to provide overall loads equal to the current mean loads for the catchment while maintaining a good overall correlation.

Using these values as a starting point, we again carried out an iterative process to obtain the best fit for the model with the gauged data for the Torbay sub-catchments. The best fit was provided by the risk values shown in Table 7. Note the significant difference between the P and N indexes. By multiplying the proportion of each map unit in each class with the derived risk value we made an estimate of the % of P and N surplus for each map unit that could be available to the stream network.

Subcatchments and Assimilation

A digital elevation model (DEM) was created as part of the PEWOC project, and this was combined with recent surveying work of the lower Torbay catchment drainage system and existing 1:25000 topographic contour maps to develop a hydrologically-based catchment model for the Torbay Catchment.

A total of 19 modelling (or `fine') subcatchments were delineated for the area, within the seven major subcatchments that drain to Torbay Inlet (subcatch_fine_new_z50.shp).

Nutrient Assimilation

Flow times as used in this formula (based on catchment length, slope and area) were modified according to the proportion of catchment drainage which is artificial drains rather than natural waterways. The manning coefficient of these two types of channel was used to estimate an increased speed of flow in drains.

We calculated an assimilation value (ie a % reduction) for flows from each sub-catchment, and another value for flows passing through subsequent sub-catchments, as the two pathways are frequently different in length, slope and order. Where the subcatchment in question was also the receiving subcatchment (Lake Powell, Manarup or Torbay Inlet) we did not include the in-catchment assimilation in modeling, as this assimilation in the receiving waterbody is in effect the nutrient problem. In addition, where nutrients from a subcatchment were received in different quantities at different receiving points (eg in Lake Powell AND Torbay Inlet), we recorded the higher modeled nutrient risk.

Subcatchment assimilation

4 Where the longest stream flow path was not characteristic of the rest of the subcatchment this would not be measured. Instead, the longest stream path representative of the subcatchment would be measured.

This used a measurement of the longest typical⁴ stream flow path within the subcatchment, and the height at the beginning and the exit of that stream path from the subcatchment. Where the subcatchment represented the beginning of the routing chain (ie no streams flowed into the subcatchment) this value was the only assimilation value used. An example of the internal flow path used for the sub-catchment assimilation is shown in the figure below:

Figure 2 - Sub-catchment assimilation: identification of an internal flowpath

In some cases the longest sub-catchment path will also be the through path, and the same value will be used for both assimilations.

Through assimilation

Where streams flowed through a subcatchment, it had a second assimilation value established: the 'through' assimilation. This value would be used for nutrients coming from the next upstream sub-catchment and passing (routed) though the subcatchment.

The through assimilation used a `through length' measurement for the main stream channel passing though the subcatchment. The length of that major channel, and its height at entrance to and exit from the subcatchment, were used to calculated assimilation.

An example of the through flow path used for the through assimilation is shown in yellow in the figure below:

Figure 3 - Through assimilation: identification of a through flowpath

Final Model Unit Map

The nutrient model requires a classification of the entire Torbay Catchment according to the various components of the model: landuse type, soil/geomorphology and subcatchment. It provides for a division of the area into the mapping units shown in Figure 1. Each polygon contains a description according to subcatchment, landuse type and soil/geomorphology, as well as its calculated area. The base data is used in SSPNRED to calculate nutrient export.

Model Calibration

The model was constructed for calibration and testing in Excel, with a total of 10,224 separate mapping units, each represented as a single line in the table. Polygon values were calculated using the various model components:

• Land use-based P & N Surplus (in all cases, the basic P production layer);
• Soil & Landscape P & N Loss Risk; and
• Subcatchment Assimilation.

The various layers are constructed so that the formula is very simple:

P Loss Risk = P Surplus * (Soil & Landscape P Loss Risk / 100) * (Assimilation)
N Loss Risk = N Surplus * (Soil & Landscape N Loss Risk / 100) * (Assimilation)

In the full form, the model returns estimated P & N loss to the inlets as shown in Table 9 - Results of model for Gauged Catchments. These values (in kg/ha/year) represent an averaged annual loss, and are summed by subcatchment or groups of subcatchments as required.

A series of catchments within the Torbay Catchment area have been gauged and monitored over a number of years, and we have used median loads for these as a calibration set (See Table 8).

As Table 8 shows, the monitored catchments comprise a range of sizes, and include all of the main river catchments in the Torbay catchment. The Torbay Main Drain catchment is mainly cleared and has extensive artificial drains, while Marbellup catchment drainage is largely un-modified. Seven-Mile Creek has a large amount of natural vegetation.

When the model results are compared to the gauging station median loads, we can see a very good correspondence, as evidenced by R² values: 0.9388 for Phosphorus and 0.9874 for Nitrogen.

Figure 4 - Model Loads plotted against Gauging Station Loads

Calibration conclusion

The purpose of this calibration work was to assess the underlying model for the purpose of providing Nutrient Loss Risk mapping. Given that the purpose of this project is to provide relative results, we believe the results being achieved to date indicate that the model as developed is appropriate.

Final Model

For mapping purposes, the model was run in ArcGIS 9.1, using the base dataset derived from polygon overlays of all model datasets. Of the various maps have been produced from the model, we have used two outputs: nutrients delivered and nutrient export risk.

3.2 High Nutrient Source Areas

The areas indicated as having high nutrient delivery are likely to be more important nutrient contributors when nutrient-related problems are observed within the stream system.

3.3 High Nutrient Risk Areas

3.4 Remnant vegetation conservation priority areas

A number of factors can be considered in determining such areas, including:

• Rarity
• Representativeness,
• Condition status,
• Threats/current protection,
• Size and context of the area.

Only some of these factors could be considered in this study. No data was available on remnant vegetation condition or fencing status, nor has any ecological work been done on representativeness of the various remnant areas.

We were able to assess rarity, but only in terms of the extent of clearing of vegetation types according to the data available from the estimated extent at time of white settlement. This was calculated the % clearing of each vegetation type. In addition, we were able to assess the area of each contiguous patch of vegetation, in effect rejecting any area of vegetation less than 1 ha, and weighting our selection of areas for patches in the range of 5-30ha. We also measured the perimeter to area ratio of each as well, weighting our selection towards areas tending more towards square or circular - on the basis that areas of vegetation which are highly elongated return less for the effort of protecting them (eg through fencing).

A simple model was created to reflect this thinking (see Figure 5 and Figure 6):

Figure 5 - Vegetation Conservation Priority Model Diagram

Figure 6 - Vegetation Priority Model - Weighted Overlay Values

This model identified areas which have the highest priority for protection, which were mapped using a low, medium and high scale for conservation priority [VegConsPriority].

3.5 Recommended priority stream restoration

In addition, we should note that research on the south coast has investigated the relationship between stream order, riparian condition and water quality: "Amongst many factors, one factor influencing the change in water quality is that low order streams exhibit the poorest riparian zone condition and therefore have little capacity to moderate paddock nutrient runoff. ...low order streams ... hence represent a greater relative opportunity to moderate surface derived nutrients and sediment. ... low order streams ... are priority candidates for the purpose of minimising the downstream impacts of nutrients when limited funds are available." (Weaver et al. 2001:57)

The stream survey dataset clearly identifies recommended sites for fencing, and streams where riparian vegetation is degraded. Combining this with the modeled P and N exports to the inlet and previous research findings indicates priority sites: fencing and rehabilitation should initially be targeted for degraded and unfenced low-order streams in areas likely to be contributing high levels of P and N to the inlets.

It is not feasible to provide a map showing appropriate minimum buffer widths for all stream orders, as these are generally very narrow - 10-15m for erosion/sediment control in areas with low-moderate soil loss rates (Prosser & Karssies, 2001). Thus they would not really show at catchment map scales. In most instances fencing off streams will be done with buffers of 5m or more, which will fulfill the basic requirements for erosion/sediment control. Ecological requirements are not known for Torbay.

Figure 7 - Recommended grass filter strip widths (m) for typical values of annual soil loss and filter gradient under conditions of dispersed overland flow. (Table 1 from Prosser & Karssies, 2001:8)

3.6 Recommended areas for perennial pasture assessment

The research on the south coast (Weaver et al. 2001) concluded that low order streams exhibit the poorest riparian zone condition and therefore have little capacity to moderate paddock nutrient runoff." In this case we can use perennial pastures to reduce paddock nutrient losses in areas where riparian condition is poor. And again, such work should be concentrated in 1^st and 2^nd order catchments because they are primary sources of N and represent a greater relative opportunity to moderate surface derived nutrients and sediment.

We have therefore prepared a map showing the combined modeled P and N exports, overlaid by stream orders (1^st and 2^nd order streams highlighted) and existing/proposed fencing (from the stream surveys). This allows an initial identification of areas where perennial pasture would be indicated as a hydrological control.

3.7 Pathogen risk sources and potential waterways 'hot spots'

Our approach was to recognise that in general, a number of situations are likely to contribute to a higher risk:

• Grazing locations;
• Unfenced waterways;
• Low quality riparian vegetation;
• The presence of other potential sources - septics, rubbish tip; and
• Proximity to the water take-off point.

In order to allow the above factors to be combined by DoE officers, we have produced a map - 'Marbellup Brook Pathogen Risk Factors' - which combines the above factors for Marbellup Brook. Interpretation of this map in light of the above factors provides a tool for an initial assessment of pathogen risk factors. It clearly identifies the likely area of highest risk in the south-west of the catchment, where a large area of grazing is drained by un-fenced streams with little or no stream vegetation buffers. The area is reasonably close to the proposed area for water take-off.