Wednesday 7
Community-based management
Madhu Sarin
› 11:30 - 12:30 (1h)
› JOFFRE B
Analysing complex social-ecological systems: A model of sustainability in community-based forest management
Ulrich Frey  1, *@  , Carl Salk  2@  , Hannes Rusch  1@  
1 : University of Giessen
2 : University of Colorado
* : Corresponding author

Motivation:

For many social-ecological systems it is still unclear why some groups thrive while others fail to achieve sustainable outcomes. In particular, it remains debated which factors improve ecological success and resilience. Moreover, it is not clear how potential factors interact. Suggestions for potential success factors include participation of communities, predictable resource flows and many others.

Our approach tries to find robust patterns of more than 20 factors potentially related to success in a large number of case studies (irrigation, fisheries, forestry). However, there are a number of obstacles to overcome. First, the method has to be able to handle a comprehensive set of such factors simultaneously. Second, there is a lack of large-N-studies with data on more than a few factors. Third, variables from different studies do not overlap. Fourth, which success factors should be included, which left out?

 

Methods:

For the last problem, our synthesis uses Ostrom (2009) and Agrawal (2001) as a starting point. The other problems can be overcome with a new methodology – artificial neural networks (ANN) combined with large data sets. Neural networks are especially suited for such tasks because they are capable of analysing the complex interplay of factors since they allow non-linear statistic modelling of complex systems if the interactions between factors are not known.

Here, we use the IFRI data set (community-based forestry) with around 400 codeable cases.

Each case is represented by a vector of values for each success factor (e. g. capability to adapt to change). These vectors serve as input for the ANN. The output of our model is the ecological success of each system. Once the ANN has learned how to relate the input to the output on a training set of vectors, it can be evaluated by predicting the success of cases whose success (output) is known, but has not been part of the training set.

 

Results:

The results of the ANNs show that ANNs are indeed capable of modelling the complexity of SES-systems, which rarely has been done up to now. However, the results on other data sets (irrigation and fishery) yield much more precise results (the error rate is small: the absolute average error is 6.5%) than the models for forestry. We will discuss why this is the case. One suggestion is the complexity of forests and their slower rate in adapting governance measures.

 

Conclusion

Such an ANN-generated model might be a convenient tool to analyse, predict and optimize performance for communities world-wide facing SES-challenges. It may also be able to cope with the real-life complexity of SES-problems. Today many attempts of SES management fail without a clear understanding of the reasons, which makes a quantitative and precise analysis a pressing cause.

 

Literature cited

 Agrawal, A. 2001. Common Property Institutions and Sustainable Governance of Resources. World Development 29(10):1649-1672.

Ostrom, E. 2009. A General Framework for Analyzing Sustainability of Social-Ecological Systems. Science 325:419-422.


Online user: 3 RSS Feed