SNAMP Publication #36: Exploring SNAMP discussion with Self Organizing Maps
Article Title: Evaluating collaborative adaptive management in Sierra Nevada forests by exploring public meeting dialogues using self-organizing maps
Authors: Shufei Lei and Maggi Kelly
- We wanted to know what important characteristics of a participatory process were revealed through time in discussions at public meetings.
- We wanted to observe the influence of collaborative adaptive management on the discussions among contentious stakeholders in natural resource management.
- We used self-organizing maps (SOM), a new unsupervised machine-learning method to process, organize, and visualize the public meeting notes from the Sierra Nevada Adaptive Management Project (SNAMP) from 2005-2012.
- We sought to reveal: 1) the general landscape of discussion topics at public meetings, 2) popular topics, 3) how these topics change over time, and 4) the relationships among topics.
- We found that public discussion remained focused on the project content, yet the more contentious and critical issues dominated the discussions through time. Integration across topics could be improved.
We used self-organizing map (SOM), which is a new method for analyzing and organizing text. It is an alternative to traditional content analysis. It is an unsupervised artificial neural network that can represent a complex set of data on a two-dimensional “map.” When SOMs are used to analyze texts, documents full of words are projected onto a two-dimensional “feature map” based on the original locations and importance of the words from the input documents. At the end of the process, a SOM can help researchers visually explore the importance of and relationships among words in the input documents. The SOM preserves the value and connectedness and context between words in documents. This paper uses meeting notes recorded at 38 SNAMP public meetings in 2005-2012.
Four different SOM maps were created: 1) the general landscape of discussion topics at SNAMP public meetings, 2) the popular topics among them, 3) how these topics have changed over time, and 4) the relationships among these topics. The topics that emerge from the discussions had three main groupings: 1) scientific, 2) administrative, and 3) the adaptive management process. The most consistently discussed topics relate to the impact of forest treatments on fire and water conditions, on the funding and process of the project, on the use of remote sensing technology called light detection and ranging (lidar), and on issues surrounding the Pacific fisher. In the early years of the project, discussion focused on planning the science dominated the discussions. In later years, topics related to wildlife science and to the treatments begin to dominate discussion. Topics that were usually discussed in isolation belonged to the six science research categories: “lidar/return,” “water/snow,” “species/tree,” and “bobcat/predation.” In contrast, there were topics commonly discussed together, such as planning and participation (“plan/resilience,” “moup/ucst,” and “participation/demographics”).
We think that natural resource managers might take advantage of this new technique SOM in environmental management in a number of ways.
- The representations of large sets of unstructured meeting notes can provide indicators about the state of management and collaboration in natural resource management.
- The visualizations of discussion topics can also quickly inform mangers about what issues concern environmental stakeholders the most and how these issues relate to one another.
- By seeing a map of issues, including how they change over time, managers might develop a management scheme for prioritizing and tackling the issues they are confronted with.
- Managers might also use these maps as an education tool to show environmental stakeholders the issues that they have been discussing, the progress they are making in the discussions, and how much their issues may be related to one another. This may give stakeholders a sense of progress in collaboration.
Lei, S. and M. Kelly. 2015. Evaluating adaptive collaborative management in Sierra Nevada forests by exploring public meeting dialogues using Self-Organizing Maps. Society and Natural Resources (28)8: 873-890
The full paper is available here.
For more information about the SNAMP project and the Public Participation Team, please see: http://snamp.cnr.berkeley.edu and http://snamp.cnr.berkeley.edu/teams/public-participation.