IIASA is responsible for this WP, with Michael Obersteiner in the leadership role. 

In this WP the resilience of landscapes mapped by WP 1 are integrated with the drivers identified by WP 2. The platform thus integrates drivers and biophysical, ecological and socio-economic processes. It examines trends in policies of land distribution, forest conservation and exploitation, as well as the local dynamics of land use practices in the forestry, agro-forestry and agriculture sectors. WP 3 develops tools allowing inference calculation, prediction and mapping of dynamic processes at regional and national level, taking into account the uncertainties of the input data and models.

Three specific tasks were defined under WP 3: 

Task 3.1. Interlinking global models.

Taking genera distribution data from the WP 1, here we have integrated three models: GLOBIOM (Biosphere Global Model Management), G4M (Global Model forestry), and the Vegetation model. GLOBIOM is a global dynamic model incorporating the agricultural sectors, bioenergy and forestry, which aims to provide strategic advice on the extent of land use. G4M is an agent-based, spatially explicit model that simulates decisions by landowners on deforestation, afforestation and forest management taking into account the profitability of forestry and agriculture. The Vegetation model is the result of the CoForChange project and allows the prediction of spatio-temporal evolution of functional forest types according to geophysical variables and management options. Once fully integrated, the Vegetation / G4M / GLOBIOM model will allow predicting the impact of land use practices and politics on the dynamics of functional diversity in the forest of the Congo Basin, given the uncertainty on the parameters of the models.

Task 3.2. Managing uncertainties through stochastic frameworks. 

The goal here is to develop unified inference approaches allowing the propagation of various sources of uncertainty. The integrated Vegetation / G4M / GLOBIOM model is defined using a hierarchical Bayesian framework, mixing in a probabilistic approach the uncertainty associated with an incomplete knowledge of a parameter, to the random uncertainty of fluctuating states of nature.

Procedures based on simulations, using Bayesian algorithms among others, are used to quantify the model prediction errors and build confidence intervals around each scenario. These methods provide an envelope of uncertainty to help as a decision support tool when different scenarios will be explored.

Finally, calibration and simulation work will allow to feed global sensitivity studies, for prioritising sources of uncertainty that affect the process of decision making. These results will highlight the input variables, which require further study.

Task 3.3. Prediction of biodiversity outcomes of various governance options.

Policymakers, the industry, NGOs and other stakeholders all need estimates of biodiversity with the associated uncertainties to make decisions related to land use and development. We want to apply a matrix model of species areas to explore the consequences of land use and land cover changes projected in the WP2. This new model could predict the amount of remaining forest in the landscape but also the quality of land uses that form the matrix. Not only can this model can produce more accurate predictions of the loss of biodiversity due to changes in land use, but can also predict the potential recovery of biodiversity following the improvement of the quality of the matrix in the area of the Congo Basin.