Grêt-Regamey Adrienne
Participatory land-use decision modeling with bayesian networks
Project Number: CH-5838
| Project Type: |
Dissertation |
| Project Duration: |
07/01/2010 - 07/01/2014 project completed |
| Funding Source: |
other , |
| Leading Institution: |
ETH Zürich |
| Project Leader: |
Prof. Adrienne Grêt-Regamey
ETH HIL H 51.3
Phone: +41 (0) 44 633 29 57 ; +41 (0) 44 633 29 81
FAX: +41 (0) 44 633 10 84
e-Mail: gret(at)ethz.ch
http://www.plus.ethz.ch/ |
Research Areas:
Disciplines:
Keywords:
bayesian networks, land use, participatory
Abstract:
Land-use change occurs worldwide at different rates and thus influences the provision and quality of ecosystem goods and services, such as biomass production, pollination, flood regulation capacity, and the identification with a certain landscape. While land-use change is one driver of the supply of these goods and services, in turn, land- use change is driven by direct and indirect causes that mutually interact, forming complex socio-ecological systems. The understanding of such systems is key for a sustainable development of ecosystem goods and services in spatial planning processes. Identifying drivers, their importance, and their influence characteristics enables strategic planning. In these systems, local actors as owners or managers of land are one important driver. Therefore, the goal of this thesis is to elaborate on the integration of actors (local/regional/cantonal decision-makers) into land use modeling (LUM) to enhance the understanding and credibility of the model and its output. To cope with diverse sources of influence on land use and its inherent uncertainty, the modeling approach was based on the implementation of Bayesian networks (BNs) in a spatially explicit manner. The objectives were approached in five papers completed during this PhD project:
Paper I presented the set-up process and the validation and evaluation of modeling outputs. Furthermore, the spatially explicit updating of the BN was introduced. The core of the approach consists of three BNs for the land- use sectors of agriculture, forestry, and settlement, which calculate for each time step the probability of land-use occurrence. The results showed the importance of different drivers of land-use change. The generated modeling outputs, envisioned as the probability of land-use occurrence, can enable decision-makers to obtain insights into land-use change processes.
Validation is a crucial part of a model building process. Whereas in Paper I the quantitative approach mainly focused on using allocation and quantity disagreement as indicators, in Paper II, the focus was on subjective (qualitative) validation. This paper showed the result of a validation workshop with experts elaborating on three exercises: (1) BN validation, (2) validation of the probabilities of land-use occurrence in different time steps, and (3) the validation of simulated land-use maps framed by scenario story lines. Subjective validation showed that experts understood the mechanisms of the modeling framework and could assess the BNs and the model output maps.
A first application of the modeling framework (Paper III) focused on the role of local actors. The possibility of incorporating local actors’ characteristics in the modeling approach was the foundation for the quantification of farmers’ influence on land-use change processes. We analyzed different scenarios based on the probabilities of land-use occurrence to find the influence of local actors for the model outputs. Two farmer types were defined (production-oriented, ecology-oriented), and comparisons showed that local actors were more important in terms of future land-use change than changing from the present to a new agricultural policy scheme.
Whereas Paper III has shown the importance of local actors for land-use change, in Paper IV, we further investigated their decision-making and the relation to the policy scheme. In a questionnaire, farmers were asked to judge decision-making factors for their importance to decision-making regarding their farms. To differentiate between a farmer’s individual characteristic and the policy scheme, two case study areas in different countries were chosen, and identical questionnaires were used to investigate a farmer’s perceptions of his decision-making. Results revealed that respective policy schemes may explain certain characteristics of decision-making perceptions; the differentiation into full-time and part-time farmers was particularly valuable for gaining insights into the relation of policy scheme and decision-making.
The LUM approach applied in this thesis shows land-use change as raster maps. Maps are valuable to gain an overview of a study area, but the emotional components inherent in landscapes are suppressed. By contrast, 3D visualizations are more easily understood because they are closer to the real perception of a landscape. Therefore, raster maps were visualized in 3D and used in an interactive survey tool to find preferences for future landscapes and today’s feasible policy measures (Paper V). Besides participants’ preferences for landscape visualizations and policy measures, the specific set up of the tool revealed the trade-off process between different policy mixes. Results showed how respondents change their choice for specific policy measures if they are forced to trade one policy mix for another.
The main results of this thesis provide insights for further development of the modeling approach as well as for a more productive integration of quantitative modeling into spatial planning and landscape development processes.
Publications:
Celio, E. (2014): Participatory land-use decision modeling with bayesian networks. Dissertation. ETH Zürich.
pdf Dissertation
Last update: 7/18/17
Source of data: ProClim- Research InfoSystem (1993-2024)
Update the data of project: CH-5838
Go Back