Development of a stakeholder-driven spatial modeling framework for strategic landscape planning using Bayesian networks across two urban-rural gradients in Maine, USA. Ecological Modelling, 291:42-57, 11, 2014.
Development of a stakeholder-driven spatial modeling framework for strategic landscape planning using Bayesian networks across two urban-rural gradients in Maine, USA [link]Website  abstract   bibtex   
Land use change results from frequent, independent actions by decision-makers working in isolation, often with a focus on a single land use. In order to develop integrated land use policies that encourage sustainable outcomes, scientists and practitioners must understand the specific drivers of land use change across mixed land use types and ownerships, and must consider the combined influences of biophysical, economic, and social factors that affect land use decisions. In this analysis of two large watersheds covering a total of 1.9 million hectares in Maine, USA, we co-developed with groups of stakeholders land use suitability models that integrated four land uses: economic development, ecosystem protection, forestry, and agriculture. We elicited stakeholder knowledge to: (1) identify generalized drivers of land use change; (2) construct Bayesian network models of suitability for each of the four land uses based on site-level factors that affect land use decisions; and (3) identify thresholds of suitability for each factor and give relative weights to each factor. We then applied 12 distinct Bayesian models using 99 spatially explicit, empirical socio-economic and biophysical datasets to predict spatially the suitability for each of our four land uses on a 30m×30m pixel basis across 1.9 million hectares. We evaluated both the stakeholder engagement process and the land use suitability maps. Results demonstrated the potential efficacy of these models for strategic land use planning, but also revealed that trade-offs occur when stakeholder knowledge is used to augment limited empirical data. First, stakeholder-derived Bayesian land use models can provide decision-makers with relevant insights about the factors affecting land use change. Unfortunately, these models are not easily validated for predictive purposes. Second, integrating stakeholders throughout different phases of the modeling process provides a flexible framework for developing localized or generalizable land use models depending on the scope of stakeholder knowledge and available empirical data. The potential downside is that this can lead to more complex models than anticipated. The trade-offs between model rigor and relevance suggest an adaptive management approach to modeling is needed to improve the integration of stakeholder knowledge into robust land use models.
@article{
 title = {Development of a stakeholder-driven spatial modeling framework for strategic landscape planning using Bayesian networks across two urban-rural gradients in Maine, USA},
 type = {article},
 year = {2014},
 identifiers = {[object Object]},
 keywords = {Bayesian networks,Conservation,Land use planning,Land use suitability,Natural resource management,Stakeholder engagement},
 pages = {42-57},
 volume = {291},
 websites = {http://www.sciencedirect.com/science/article/pii/S0304380014003056},
 month = {11},
 id = {8adb7754-2183-324d-8933-8e5102931288},
 created = {2015-04-11T19:51:57.000Z},
 accessed = {2015-01-27},
 file_attached = {false},
 profile_id = {95e10851-cdf3-31de-9f82-1ab629e601b0},
 group_id = {71a29c65-85d2-3809-a3a1-fe4a94dc78d2},
 last_modified = {2017-03-14T14:27:45.955Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {true},
 hidden = {false},
 private_publication = {false},
 abstract = {Land use change results from frequent, independent actions by decision-makers working in isolation, often with a focus on a single land use. In order to develop integrated land use policies that encourage sustainable outcomes, scientists and practitioners must understand the specific drivers of land use change across mixed land use types and ownerships, and must consider the combined influences of biophysical, economic, and social factors that affect land use decisions. In this analysis of two large watersheds covering a total of 1.9 million hectares in Maine, USA, we co-developed with groups of stakeholders land use suitability models that integrated four land uses: economic development, ecosystem protection, forestry, and agriculture. We elicited stakeholder knowledge to: (1) identify generalized drivers of land use change; (2) construct Bayesian network models of suitability for each of the four land uses based on site-level factors that affect land use decisions; and (3) identify thresholds of suitability for each factor and give relative weights to each factor. We then applied 12 distinct Bayesian models using 99 spatially explicit, empirical socio-economic and biophysical datasets to predict spatially the suitability for each of our four land uses on a 30m×30m pixel basis across 1.9 million hectares. We evaluated both the stakeholder engagement process and the land use suitability maps. Results demonstrated the potential efficacy of these models for strategic land use planning, but also revealed that trade-offs occur when stakeholder knowledge is used to augment limited empirical data. First, stakeholder-derived Bayesian land use models can provide decision-makers with relevant insights about the factors affecting land use change. Unfortunately, these models are not easily validated for predictive purposes. Second, integrating stakeholders throughout different phases of the modeling process provides a flexible framework for developing localized or generalizable land use models depending on the scope of stakeholder knowledge and available empirical data. The potential downside is that this can lead to more complex models than anticipated. The trade-offs between model rigor and relevance suggest an adaptive management approach to modeling is needed to improve the integration of stakeholder knowledge into robust land use models.},
 bibtype = {article},
 author = {},
 journal = {Ecological Modelling}
}
Downloads: 0