APPENDIX 10. LAND USE AND COVER MAPPING IN SCOTLAND
The first direct mapping of land use in Scotland, as opposed to land cover or sectoral inventories, was that of Dudley Stamp (1948), in his 1" to the mile land utilisation survey, comprehensively described at the 'Land of Britain' WWW site ( http://riga.iso.port.ac.uk/django_projects/home/). Since then, the mapping, modelling and interpretation of changes in land use in Scotland have focused on maps relating to specific sectors of land use. Examples include:
The mapping of land cover for Scotland, allows some interpretation with respect to specific land uses. Mapping programmes for land cover have been undertaken over a number of years, but by different means and for different purposes. For example, the Land Cover of Scotland 1988 ( www.macaulay.ac.uk/mscl/gis2_dataset_4a.php) was a census of land cover which has underpinned several other studies, including changes in land cover and modelling of land uses ( e.g. Aspinall, 1993). Scotland has also been covered in programmes of mapping of land cover and land classifications for Great Britain, such as the Countryside Survey (1976, 1984, 1990, 2000 and 2007; www.countrysidesurvey.org.uk/), and the Land Cover Map of Great Britain ( LCMGB) for 1990 and 2000, and the National Countryside Monitoring Scheme. These have provided inputs to estimates of changes in land cover (for example, see www.snh.org.uk/strategy/Landcover/home.asp), but not explicitly land use. Recently, the results for the Countryside Survey 2007 were published for Scotland and can be found at www.countrysidesurvey.org.uk/scots_reports2007.html).
Other sources of data seeking to describe elements of land use include the Scottish Natural Heritage ( SNH) Trends series, Forestry Commission statistics ( www.forestry.gov.uk/forestry/infd-7aqknx), and land cover and data related to land use in the Scottish Environment Statistics ( www.scotland.gov.uk/Topics/Statistics/Browse/Environment/seso/Q/TID/13).
Recent modelling of potential land-use change in Scotland has considered some of these sources of data as inputs. For example, the agricultural census provided one basis for modelling potential changes in agricultural land use, with its provision of data such as existing land use, ownership and tenure. Aalders and Aitkenhead (2006) used these data in three approaches to modelling the probability of selected land uses ( e.g. improved grassland, woodland, rough grazing) in parishes, using neural networks, Bayesian belief networks and decision trees.
As land use-change is commonly driven by both biophysical and human processes, the challenge of modelling land-use change using empirical data is that it is a complex process which requires the integration of quantitative and qualitative data representing the two driving factors. It represents the result of interactions, over time and space, between humans and their environment, with changes consequent on complex interactions between social, economic, political and environmental factors ( IGBP, 1995; Medley et al., 1995; Pijanowski et al., 2002; Vesterby and Heimlich, 1991).
In reviews of modelling, and their work on categorising the approach to modelling land-use change, Lambin et al. (2000) summarise the categories of land-use change models (Table A10.1). Briassoulis (2000) also identifies categories of land use modelling approaches: statistical and economic, spatial interaction models, optimisation models, and integrated models.
Table A10.1. Summary of categories of model used in land-use change (modified from Lambin et al., 2000).
Model Category | Preliminary knowledge on Land Use and Cover Change | Form of knowledge of Land Use and Cover Change | Modelling Approach |
|---|
Stochastic | Past location and rate of change | When in the future (short-term) | Transition probability |
Empirical, statistical | Why in the past (proximate causes) | Multivariate statistical |
Where in the future (short-term) | Spatial statistical/ GIS-based |
Process-based, mechanistic | Past location, rate of change and drivers | When in the future (long-term) | Behavioural and dynamic simulation models |
When and where in the future (long-term) | Dynamic spatial simulation models |
Analytical, agent-based, economic | Why in the future (underlying causes) Why in the future (underlying causes, scenarios) | Generalised Von Thunen models Deterministic and stochastic optimisation models |
The commonly-used methods for empirical modelling landscape processes are quantitative in nature and are unable to facilitate the inclusion of qualitative data. In addition, due to the complexity of the interactions that lead to land-use change, there is often an understanding of the relationship within the process but modelling is generally constrained by the absence of consistent and comprehensive information relating to the process and its drivers. The determinants of land-use change ( i.e. drivers) can be from either the human activities or land use.
The modelling of spatial interactions can use information of human interactions in space, including travel patterns, migration, information and commodity flows (Haynes and Fotheringham, 1990). Although the process of change is often not the same for a given place, the direction of change may be guided by overarching drivers or constraints, such as policy, planning or economic rewards.
Table A8.1 summarises the modelling approaches best suited to take account of past location, rate of change and drivers. The results from the qualitative and quantitative surveys provide the types of information that influence decisions about land use at the level of the land manager, and the prospect of populating models designed to inform the user about possible changes in land use in Scotland. Two relevant approaches have been used recently as applied to Scotland: agent-based modelling and Bayesian modeling.
Modelling tools which can exploit spatial patterns and distributions of land uses can also be used to simulate changes in land use under scenarios of different drivers, types of decision-makers and the information available. Such models can include consideration of drivers of change, such as technology ( e.g. uptake of precision-farming, new crop cultivars, changes in husbandry, and new agricultural inputs; Aspinall, 2009).
The use of agent-based models in the study of coupled human-natural systems provides one means of representing behaviours of individual and interacting land managers while taking account of cognitive, social, economic and environmental constraints (Hare and Deadman, 2004). Dynamic feedback due to changes in the spatial pattern or distribution of land uses can be incorporated to simulate changes in land use under scenarios of different drivers, types of decision-makers and the information available (Parker et al., 2008). The modelling of choices of different land uses using hypothetical environmental contexts can utilise information on factors with different spatial and temporal extents, from constant over a region ( e.g. climate and economic factors), to progressive fragmentation of land ownership and biophysical characteristics ( e.g. Gotts et al., 2003). The incorporation of social influences on decisions, such as imitation (Gotts and Polhill, 2009), are now being tested in Scottish conditions using qualitative field research ( e.g. Burton et al., 2008) with agent-based models as part of the RERAD Programme on Environmental Stewardship and Land Management.
Bayesian belief networks (also known as belief networks, causal nets, causal probabilistic networks, probabilistic cause effect models, and graphical probability networks) provide a second means to utilise a combination of limited empirical data and expert knowledge (Bacon, 2002). The modelling process involves the development of a graphical conceptual model consisting of nodes and links that represent system variables and their cause-and-effect relationships (Jensen, 1996, 2001) (Figure A10.1 and Figure A10.2; Aalders and Williams, 2008). The relationships between the nodes is quantified based on expert knowledge or preferably empirical evidence in conditional probability tables ( CPT). While the use of expert knowledge rather than empirical data to build a model is often criticised, developing the BBN in an evolving process of knowledge development can be very useful in the early phases of exploring complex processes. In order to ensure the credibility of the model, Marcot et al. (2006) propose a three-staged process of model development: initial BBN, i.e. the creation of the graphical conceptual model; peer-reviewed BBN; evidence-based tested and updated BBN.
With the BBN populated with the best available information and data, it is possible (bearing in mind the state of the available evidence incorporated in the BBN) to explore the causal relationships between the modelled variables through diagnostic and predictive reasoning (Korb and Nicholson, 2004). By identifying the value of each known variable represented by a node, probabilistic inference updates the beliefs for the other unknown variables. This allows us to explore the behaviour of the BBN for scenarios represented by a combination of known values of selected variables.
The new data provided in this study offer a new opportunity to explore the spatial implications of land use decisions by land managers in Scotland, using a number of modelling tools. The two approaches summarised above are currently being used in the RERAD programme and could exploit the new data and information as part of their implementation.
References
Aspinall R.J. (1993) Use of Geographic Information Systems for interpreting land-use policy and modeling effects of land use change. In: Landscape Ecology and Geographic Information Systems (Ed. R. Haines-Young, D.R. Green and S.H. Cousins): 223-236.
Bacon P.J., Cain J.D. and Howard D.C. (2002) Belief network models of land manager decisions and land use change. Journal of Environmental Management 65:1-23.
Burton R.J.F., Kuczera C. and Schwarz G. (2008) exploring farmers cultural resistance to voluntary agri-environmental schemes. Sociologia Ruralis 45(1): 16-37.
Gotts N.M., Polhill J.G. and Law A.N.R. (2003) Aspiration levels in a land use simulation. Cybernetics and Systems 34: 663-683.
Gotts N.M. and Polhill J.G. (2009) When and how to imitate your neighbours: lessons from and for FEARLUS. Journal of Artificial Societies and Social Simulation 12(3): 2. http://jasss.soc.surrey.ac.uk/12/3/2.html
Hare M. and Deadman P. (2004) Further towards a taxonomy of agent-based simulation models in environmental management. Mathematics and Computers in Simulation 64: 25-40.
Jensen F.V. (1996) An Introduction to Bayesian Networks. University College London Press, London, UK.
Jensen F.V. (2001) Bayesian Networks and Decision Graphs. Springer, New York, USA.
Korb K. B. and Nicholson A. E. (2004) Bayesian artificial intelligence. Eds J Lafferty, D Madigan, F Murtagh, and P Smyth. Series in Computer Science and Data Analysis. London: Chapman & Hall/ CRC.
Lambin E.F., Rounsevell M.D.A and Geist H.J. (2000) Are agricultural land use models able to predict changes in land use intensity? Agriculture, Ecosystems and Environment 82: 321-331.
Marcot B., Steventon J., Sutherland G. et al. (2006) Guidelines for developing and updating Bayesian belief networks applied to ecological modeling and conservation. Canadian Journal of Forest Research 36: 3063-74.
Parker D.C., Hessl A. and Davis S.C. (2008) Complexity, land-use modeling, and the human dimension: Fundamental challenges for mapping unknown outcome spaces. Geoforum 39(2): 789-804.
Figures A10.1. Bayesian Belief Network on land transactions (Source: Aalders and Williams, 2008).
Figure A10.2. Factors influencing land values in Scotland (Source: Aalders and Williams, 2008).
