By T. Trevor Caughlin, Literature Coordinator
The success of tropical forest restoration projects depends on many interrelated factors, ranging from institutional support to local participation to appropriate tree species selection. Project coordination that accounts for all these complex factors is difficult, so it is perhaps not surprising that many publicly-funded tropical forest restoration projects fail. Because interactions between seemingly-disparate drivers can determine restoration success, scientific research to understand the drivers of successful restoration projects may be hindered by disciplinary boundaries For example, tree mortality due to insect herbivory (an ecological driver) could influence local perception of restoration success (a social driver) leading to project abandonment. Predicting the success of restoration projects will require novel approaches that transcend disciplines and account for feedbacks and interactions between drivers. A new study conducted by Hai Dinh Le and co-authors illuminates the critical role of interactions between socioeconomic, institutional and biophysical drivers for restoration project success.
Le’s study evaluated 43 restoration projects on Leyte Island in the Philippines. Decades of logging and agricultural conversion have resulted in the loss of nearly all of Leyte Island’s original forest cover, and more subsequent agricultural abandonment has led to a large area of the island covered by degraded Imperata grassland. This abandoned land presents an opportunity for restoration in one of the world’s biodiversity hotspots, as well as major challenges for tree establishment due to the legacy of soil degradation and erosion. Le and his team collected data on reforestation success indicators and drivers, including data on project governance, socioeconomic impacts, technical aspects of site management and forest growth performance. The authors then applied Bayesian network analysis, a statistical technique that uses a graphical map of dependencies between variables (i.e. a flowchart) to produce probabilistic estimates of how variables determine restoration outcomes. The key advantage of Bayesian network models for this study was the ability to quantify interactions between a wide range of predictor variables that affect various metrics of success of restoration projects (e.g. tree survival rate, area of trees planted relative to target area and tree biodiversity).
The Bayesian network model identified several surprising interactions related to restoration project success. The interaction with the largest effect size was a positive interaction between short-term tree survival rate, soil depth, and government funding on area planted. Another interesting three-way interaction is a decrease in landslide frequency when planted area compared to target area was >50%, education and awareness programs were implemented, and mixed introduced species were used. Overall, the model indicated that interactions between various predictor variables were far more common than additive effects of predictor variables. In other words, considering single variables independently was insufficient to predict whether a project was successful or not. Although more mechanistic, hypothesis-driven research will be necessary to better understand these interactions and prescribe specific interventions, Le’s study is a valuable starting point for generating ideas and discussion on the complex drivers of restoration success. The study also illustrates the potential value of using existing restoration projects as replicates for studying reforestation—including failed or poorly implemented projects as useful cases.
Photo credit: A typical landscape in the Philippines, including secondary forest, agricultural land and abandoned grassland. Photo by Sharif A. Mukul
