The rainforest in the Congo Basin is under threat of disturbance by the expansion of agriculture, selective logging, and mining (Tyukavina et al., 2018). In the future, this threat is likely to increase due to an expected fivefold population growth by 2100. Land use change simulation models are used to assess the likely locations of future land use change and the effect of potential measures to mitigate unwanted changes, like deforestation. Given the above-described challenges in the Congo Basin, a land use change simulation model could help to protect the rainforest. However, such a model needs to be set up and calibrated first.
Bayesian data assimilation can update a simulation model during runtime at time steps when observations are available, and thereby increase the model accuracy. Hereto, not only the observations themselves are required, but also an estimate of the uncertainty in these observations.
In 2014, Verstegen et al. used Bayesian data assimilation to calibrate a land use simulation model. For their case study, the method could reduce the model projection uncertainty by a factor 3. However, they noticed that obtaining a time series of land use observations was already challenging, but obtaining cell-based uncertainty information was impossible; they had to synthetically create the uncertainty.
The new dataset of forest disturbance in the Congo Basin by Reiche et al. is unique in its provision of forest disturbance probabilities per cell, i.e. not the classes forest (undisturbed) or non-forest (disturbed), but the (un)certainty about the presence of these classes. The aims of this thesis are 1) to set up a land use change model for (part of) the Congo Basin to simulate deforestation based on the drivers identified in literature, 2) to assimilate the dataset of Reiche et al. into this model to calibrate it, and 3) to run the calibrated and uncalibrated model towards the future to assess the effect of the calibration on the projection uncertainties.
- Set up a land use change simulation model for (part of) the Congo Basin.
- Use Bayesian data assimilation and the dataset of forest disturbance by Reiche et al. to calibrate this model.
- To run the calibrated and uncalibrated model towards the future to assess the effect of the calibration on the projection uncertainties
- Tyukavina A., Hansen M.C., Potapov P., Parker D., Okpa C., Stehman S.V., Kommareddy I. & Turubanova (2018). Congo Basin forest loss dominated by increasing smallholder clearing. Science Advances 4 (11). DOI: 10.1126/sciadv.aat2993.
- Verstegen, J.A., Karssenberg, D.J., van der Hilst, F. & Faaij, A.P.C. (2014). Identifying a land use change cellular automaton by Bayesian data assimilation. Environmental Modelling & Software 53, 121-136. DOI: 10.1016/j.envsoft.2013.11.009.
- Reiche, J., et al. (2021). Forest disturbance alerts for the Congo Basin using Sentinel-1. Environmental Research Letters 16 (024005). DOI: 10.1088/1748-9326/abd0a8.
- Experience with R or Python or both
Theme(s): Modelling & visualisation; Human – space interaction