The potential for native woodland in Scotland:
the native woodland model Management

Natural Heritage Management

• Executive Summary
• Introduction
• Assumptions and general principles underlying the Native Woodland Model
• The value of modelling potential woodland cover
• Methodology
• How accurate is the NWM?
• How potential woodland types correspond to HAP types
• Use of the Native Woodland Model – case studies
• References
• Appendix 1 The base datasets
• Appendix 2 The National Vegetation Classification
• Appendix 3 Allocation of woodland types to different site conditions
• Appendix 4 Correspondence between NWM and HAP types

Methodology

The approach adopted in the NWM, although developed within a GIS, differs little from the approach that a practical forester or ecologist would adopt when planning a new woodland on the ground. Indeed the model seeks to replicate the approach, but over large areas without the need for ground survey. On the ground, information on site factors which influence woodland growth, such as geology, soil conditions (drainage, fertility, acidity), climate (temperature and exposure), topography (slope and rockiness) and current land cover and use would be collected and assessed. These would then be compared with the conditions suited to different woodland types, enabling recommendations to be made on the most appropriate woodland type(s) for the site.

The NWM follows this logical approach, but with the key difference that the information on ‘site’ conditions has already been collected and is stored within the soils and land cover geospatial datasets. When combined, these data provide information on geology, soil moisture and nutrient regime of the component soils, topography and terrain type and the precursor vegetation species. All these are important for the assessment of site suitability for native woodlands. We consider that a cost-effective and practicable approach is to use landscape scale modelling to gain a strategic understanding of the potential distribution of different woodland types under current environmental conditions, and to refine this by detailed surveying of targeted sites. The advantages of computer modelling at the landscape scale are that it:

• provides an opportunity for rapid appraisal of large areas
• reduces the need for costly site survey
• allows site survey effort to be targeted
• cuts across landownership boundaries and provides opportunities for investigating wider benefits or costs of different woodland schemes.

Base data

Two digital data sources are used in the model: the 1: 250 000 scale national soils map (MISR 1982) and the 1: 25 000 scale Land Cover of Scotland 1988 (LCS88) dataset (MLURI 1993). Because of the scales of these datasets, we advise that the model should not be used at scales more detailed than 1:50 000. For individual woodlands, a more functional ecological approach, incorporating landscape character, landscape ecology, planning and design, is required. Details of the datasets are given in Appendix 1.

Dataset integration

The datasets were overlaid within a Geographic Information System to produce a number of combined soil/land cover types describing the present site conditions. Each of these types was allocated to an NVC community (Appendix 2), or a mosaic of communities, based on the relationships between biophysical properties and woodland requirements. It must be stressed that, although there is considerable knowledge about the relationships between most NVC woodland types and site conditions, this requires careful translation when applied to the integrated dataset derived from the soils and land cover data. By the very nature of what these data describe, they are imprecise. Consequently, they must be interpreted with an underpinning understanding of the opportunities and constraints of the data.

Land cover data add valuable extra information to the soils data in a number of ways:

• The LCS data was captured in a consistent fashion using air photo interpretation, whereas the methodology for collecting soils data evolved as the surveyors gained more experience of Scottish soils.
• The much larger scale of the LCS data adds detail to the soil units and allows some soil complexes to be disaggregated into their component parts in the landscape. Being on a larger scale, the digitised LCS data are also likely to be much more accurate.
• The LCS can be used to locate cultivated land, thus highlighting soils where nutrient or moisture status may have been altered by ploughing and fertilisation (which affects the suitability of the site for woodland development).
• It also allows the delineation of montane vegetation types and the separation of the montane scrub zone from land considered unsuitable for trees/scrub, which is not possible using the soil data alone.

Figures 1 and 2 illustrate how integration of the soils and land cover datasets provides extra detail on the relative proportions of soils within complex terrain, such as rocky ground or moundy moraine, both typical landforms in the Highlands of Scotland. We call these ‘soil landscapes’ in which the pattern, location and proportions of different soils are strongly influenced by the physical landscape.

Figure 1, rocky landscapes, shows how data integration can be used to identify whether peat is present and, if so, the proportion of the landscape which it occupies. The soil map indicates a mosaic of closely related soils occurring in a loosely defined pattern within a complex rocky landscape. The finer-scale land cover data has been used to reveal more of the detail of this mosaic. To the left, the land cover data indicates that blanket bog, and therefore the peat on which it grows, is absent. In contrast, on the right, the presence of blanket bog indicates a more diverse landscape with numerous basins and channels of peat.

Figure 1

In moundy moraine landscapes (Figure 2), the land cover indicates where the peat is dominant (to the right), sub-dominant (to the left) or in some instances absent. Figure 3 illustrates how this differentiation has been made on the moundy moraine landscape of Rannoch Moor.

Figure 2

Figure 3: Use of integrated data to enhance woodland predictions

Some key points regarding data integration:

• The NWM predicts ‘climax’ NVC woodland types. Wet woodland types, which depend on water level more than soil type (W1-W3 and W5), and ‘seral’ lowland scrub (W21-W25) are not predicted by the NWM
• Most of the soil/land cover combinations were logical and made ‘ecological sense’, but slivers of mismatching soils and land cover are inevitable considering the difference in scale between the two datasets. The slivers have been allocated to the site type indicated by the land cover data
• There is no digitised information about the original vegetation cover of the 12% of Scotland covered by coniferous plantations. In these areas the semi-natural land cover type most often associated with each soil type was used to determine the potential native woodland type
• The model is restricted to those parts of Scotland with semi-natural soil types. The central belt of Scotland and other lowland areas are excluded because the soils have been so modified by cultivation and industry that they could now support many types of woodland.

The predicted woodland types

The model only predicts the composition of the vegetation that would be expected if woodland were to develop naturally on a site. It tells us nothing about the structure or condition of any woodland already present. For example, SSNWI shows all woodland over 0.1ha in size. If the semi-natural polygons are extracted and overlaid with NWM predictions of where pine is the most suitable woodland type, we should get a reasonable idea of the distribution of woods dominated by pine. However, this would tell us nothing about the condition of these woods, which may be in good condition with much regeneration, or senescent and heavily grazed with no seedlings and little structural diversity, for example.

The woodland types used in the NWM are listed in Appendix 2. Single woodland types are predicted where possible. Where the soil pattern renders this approach inappropriate, mosaics or interchangeable types are predicted. Categories defined as open woodland, or scattered trees/scrub are mapped where trees may potentially cover only 10-30% of the area. These may occur as single categories or as mosaics with closed woodland NVC types or with non-NVC types. These open woodland and scattered trees categories, and mosaics containing them, account for over half of the total NWM output, indicating the great variability in suitability for tree cover across Scotland.

Mosaics

Over much of upland Scotland, the soil and land cover patterns vary on a scale below that of the model. In such cases mosaics of NVC types have been defined.

Different woodland types have been matched to the different components of such landscapes. For example on moundy moraine, W18 would be suited to the peaty podzols on the mounds whereas peatland with scattered trees/scrub would be suited to the peat in the intervening hollows and channels. This is indicated on the map legend as ‘W18 + peatland with scattered trees/ scrub’ with W18 being the dominant component and so listed first.

Interchangeable types

Classifications have to ‘pigeon hole’ elements of the physical environment, such as vegetation and soils. In reality such elements are part of a continuum, and woodland types grade into each other. Where the bedrock changes abruptly, boundaries between types can be distinct, but most are more gradual and subtle due to small changes in the base status of the bedrock or drift. This means that there are large areas where two (or more) woodland types are considered equally well suited to the site conditions present. In order to make this clear, such areas are classed as ‘interchangeable’, to distinguish them from ‘mosaics’, as described above. For example, ‘W17/W18’ indicates that site conditions are considered suitable for either of these NVC types.

Allocation of woodland types to different site conditions

This is an interpretative process which involves matching the site conditions, as described by the integrated dataset, with the known site requirements of the range of NVC and other woodland and scrub types that the model is seeking to predict. The basis of the decisions taken, and the criteria used to allocate different woodland types to different sites, are explained in detail in Appendix 3. Woodland modelling and the role of soil landscapes in heterogeneous environments are also detailed.

Potential woodland cover in different NWM categories

During the development and early application of the NWM it became clear that, in many areas of Scotland, the potential for native woodland cover would be restricted to particular topography and soils within very complex landscapes, rather than densely covering whole areas. We therefore estimated the percentage tree cover for different woodland categories to aid in subsequent model interpretation. Most of the single and interchangeable categories were assessed as having the potential for 80-100% cover (although this might not necessarily be desirable), whereas the W4 with open ground and peatland with scattered trees/scrub categories are considered to have only 30% and 10% potential tree cover, respectively.

This has important implications for assessing the actual woodland potential of these categories and for the woodland mosaics which have these categories as components. For example, within the ‘W4 Birch (with open ground) + Peatland with scattered trees’ mosaic, as little as 23% of the land area is predicted to have woodland potential, the remainder staying as open ground. This is based on a 65:35 split between the components of the mosaic, and a potential tree cover of 30% and 10% for the two components respectively. It is important to stress that this gives only strategic guidelines as to potential tree cover; on-site survey would be required to accurately determine this proportion on the ground.