Home » LCCMcons


euflagThis project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 703862.


Land use/cover change (LULCC) driven by rapid human population growth and increasing demand for agricultural is a huge environmental risk. Tools that allow policy-makers and landscape managers to visualize and understand the impact of their decisions, prior to its implementation, are of the utmost importance.

Addressing two real-world case studies this project will demonstrate that LULCC models can be excellent supports to the implementation and assessment of conservation policies; as well as to help landscape managers understand and visualize the consequences of their decisions.

  1. Road development in the Serengeti-Mara ecosystem, Tanzania: model simulations will be run with the proposed (government and scientific community) road configurations, in order to provide policy makers the first set of spatially-explicit maps of future LULCC in the Serengeti.
  2. Economic incentives towards forest restoration in the Atlantic Forest, Brazil: model simulations will be run in the state of Santa Catarina, where the great majority of the land is privately owned, to help directing those benefits, and evaluate their outcome.

In both regions, collaboration with representatives of non- governmental organisations with extensive work in the region, and a good connection with local/regional/national governments (EarthWatch Brazil and Frankfurt Zoological Society, respectively) was established, thus ensuring that the outcomes of this project are both relevant and used.


Introduction and State-of-the-art:
Land use/land cover change (LULCC) driven by rapid human population growth and increasing demand for agricultural and forest products1 is main driver of ecosystem change. It is now one of the major causes of biodiversity loss2, the second largest source of carbon emissions to the atmosphere3, and has a major impact on the water cycle4. However, there are still large uncertainties about where and at what rate these changes are occurring and their impacts. Much of our planet’s future will depend on how LULCC impacts the success of conservation policies and interventions. Therefore, knowing where, at what rate, and the consequences of future LULCC will allow us to anticipate its impacts and develop preventive and/or adaptive measures, much as climate change models form the basis for climate adaptation and mitigation measures5. Spatially-explicit LULCC models can be helpful tools to achieve such a goal6, however, modelling LULCC processes is still a great challenge. Not only is the physical environment itself highly variable from one region to another and constantly changing, but the underlying processes that drive LULCC are usually very complex, combining socio-economic, cultural, political and environmental factors7. To understand the complex interactions among human and biophysical factors that drive LULCC, numerous spatial models have been developed8,9, but their use to address real-world conservation issues is often undermined by poor clarity on the calibration and validation methods, weak communication of model outputs, and lack of involvement of stakeholders, and decision makers in the modelling process8,9.

Objectives and Overview
Improving LULCC models, and involving key stakeholders in the modelling process, would increase their potential to help solving real-world problems. Thus, the main objective of this proposal is to develop a next- generation LULCC model that can be confidently used as a tool to address pressing conservation questions at regional scale (Work Package 2, WP2). The novelty and innovative nature of this model is in the way that it statistically tackles the main problem with all LULCC models: dealing with non-stationary aspects of LULCC. Because the relationships between proximate causes and drivers (e.g. road expansion, socio-economic dynamics) and LULCC are spatially and temporally heterogeneous, the model will explicitly incorporate region- and land-use transition-specific parameter estimates, which will be allowed to vary through time as well.

Two real-world case studies on two different biomes – WP3: Serengeti-Mara Ecosystem; WP4: Atlantic Forest, Brazil (Figure 1) – were identified to demonstrate the applicability of the model. Further, a collaboration was established with non-governmental organisations (NGOs) with extensive presence and work in these two regions (WP3: Frankfurt Zoological Society (FZS); WP4: EarthWatch Brazil (EWB)) to ensure the outcomes of this project are relevant and used. These two case studies address particularly important conservation issues that can be found in many ecosystems and regions of the world: WP3 – conflict between infrastructure development and conservation – infrastructure development is an ongoing major cause of LULCC worldwide, with severe ecological and biological consequences14,15; WP4 – implementation of financial incentives to promote conservation/restoration – with very large areas of abandoned land worldwide16, financially compensating private landowners is seen as a promising tool to boost ecosystem restoration.

study areas

Figure 1 – Location and recent land cover maps of: (A) Serengeti National Park and surrounding areas (the Kenya part will not be covered in this project), data from GlobCover 2009; (B) state of Santa Catarina in the Atlantic Forest of Brazil, data from SOS Mata Atlântica.

In particular, WP3 will examine the road development planned by the Tanzania government to cross the northern part of the Serengeti National Park (SNP), and the alternatives suggested by the scientific community17,18. The Tanzania government has proposed building several new infrastructures around SNP to improve the wellbeing of the population. A road crossing the northern part of the park has been proposed, but a court order as recently prevented it, or at least postponed it. This northern road was heavily criticized both by researchers and the international public. Dobson et al.17, more recently, supported also by Hopcraft et al.18, suggested an alternative southern route, which would never cross the park’s boundaries, and would have (they argue) much less environmental impacts. Alternative views on Dobson et al.’s 17 remarks were presented by Homewood et al.19, Fyumagwa et al.20 and Kideghesho et al.21. They all agree, though, that whichever path these roads follow they will have significant environmental impacts as well as socio-economic costs and benefits that require rigorous analysis. Using the improved LULCC model, calibrated to best represent the observed relationships between proximate causes and drivers of LULCC in the region (SNP and surrounding areas, hereafter Serengeti), I will quantify and compare potential LULCC impacts until 2020 of the construction of such road(s). Model simulations will be run with different configurations of the roads proposed, in order to provide policy makers the first set of spatially- explicit maps of future LULCC in the Serengeti. This will help them visualize the consequences of the planned infrastructure development, before its implementation.

roads2(Hopcraft et al. 2015)

With regards to WP4, the main goal is to use the improved model, calibrated to best represent the observed relationships between forest cover dynamics and its proximate causes and drivers in the region, to project future forest restoration, and deforestation (until 2050). In particular, two scenarios, representing the implementation or absence of financial incentives towards conservation, will be simulated in the state of Santa Catarina (hereafter SC), in Brazil, located in the Atlantic Forest. The Atlantic Forest in Brazil, one of the world’s biodiversity ‘hotspots’22 has been severely degraded and deforested since colonial times23. The Atlantic Forest Restoration Pact (AFRP) aims to restore 15 million ha of degraded areas by 205020. Since the great majority of the land is now privately owned24, one way to potentially increase forest restoration is through the provision of economic incentives and payments of ecosystem services (PES) is one of the options being investigated. In the state of São Paulo, for example, between 2005-2010 there were municipalities with more than 4% gains of forest cover, which shows that even without these benefits, there are regions already recovering. However, our understanding of why this is happening is still unclear. WP2 of this research proposal will shed some light on possible factors driving these changes. Importantly, this shows that some regions might need more incentives than other regions where regeneration is occurring per se. A tool that could help directing those benefits, such as the one that will be developed here, is timely needed. The main objective is to quantify the impacts on the rate and location of future forest restoration, and deforestation, in SC of the implementation or absence of economic incentives.

cristina(Banks-Leite et al. 2015)

Research Methodology and Approach
After the initial data collection and processing (step one), for both study areas (Serengeti and SC), I will analyse the spatial and temporal trends in land cover change and associated proximate causes and drivers (WP2) in both study areas (step two). With regards to the analysis of historical LULCC change in SC, I will focus only on two transitions, from forest to non-forest (deforestation) and vice versa (regeneration); and in Serengeti, I will analyse agricultural gain and loss (2000-2012). Then, for both study areas, I will use the socioeconomic, geographical and biophysical data collected (Table 1), together with multivariate techniques, to identify sub- regions within the landscapes (Serengeti in WP3, and SC in WP4) that share the same LULCC history and socio- economic trajectories, thus making clear the spatial dependency of LULCC and associated proximate causes and drivers. These relationships will be analysed through time, whenever data availability allows, thus detailing the temporal fluctuations in the association between drivers of LULCC and observed rates and location of change.


Figure 2 – Diagram of the modelling approach at the sub- regional level. The outputs of each sub-regional model will be combined to produce the region-wide LULCC final map.

Afterwards, using the data collected and analysed before, I will improve the LULCC model that I developed during my PhD9 (step 3), to better represent the spatial dependency of LUCC dynamics (WP2). A statistical model will be calibrated with the data collected in order to estimate the probability of occurrence, over time, of agricultural gain/loss in the Serengeti or deforestation/regeneration in SC, on each sub-region identified before. The outputs of these sub-models will be combined in the end to produce a region-wide map (Figure 2). To incorporate temporal variability in the dependency between LULCC and its drivers, both in Serengeti and SC, a set of model projections will be produced using several starting points (e.g. randomly selected from the pool of available LULCC data), thus creating an “ensemble” of model outcomes. The outputs of the model will be validated against independent data (step 4). Finally, as described above, two real-world case studies were identified to demonstrate how useful the model outputs will be to address pressing conservation issues (step 5). With my strong background in LULCC modelling and a well-established network with experienced researchers with a strong knowledge of the two areas covered by this research proposal (Dr. Giorgi from EWB (WP4) and Dr. Rentsch from FZS (WP3)), I am confident that the outcomes of this proposal will be used, and will have significant impacts in the regional landscape management.

datatableTable 1 – Data sources for this project

— — — — — —


  1. Foley et al. 2012 Science, 309:570-574.
  2. Pereira et al. 2012 Annual Review Environemtal Research, 37: 25–50.
  3. Le Queré et al. 2009 Nature Geosciences, 2:831-836.
  4. Sterling et al. 2013 Nature Climate Change, 3: 385–390.
  5. Surampalli et al. 2013 American Society for Civil Eng. 728 pp.
  6. Verburg and Overmars 2009. Landscape Ecology, 24(9): 1167-1181.
  7. Geist and Lambin. 2002. Bioscience, 52:143-150.
  8. Brown et al. 2013. Current Opinion on Environmental Sustainability, 5(5):452-457.
  9. Rosa et al. 2014a. Global Change Biology, 20(6): 1707–1722.
  10. Rosa et al. 2013. PLOS One 8(10): e77231.
  11. Rosa et al. 2015. Regional Environmental Change.
  12. Dalla-Nora et al. 2014. Land Use Policy, 39: 403–411
  13. Hurtt et al. 2011. Climate Change, 109(1-2): 117-161.
  14. Laurance and Balmford. 2013. Nature, 495: 308–309.
  15. Laurance et al. 2009. Trends in Ecology Evolution, 24(12):659-69
  16. Cramer et al. 2008. Trends in Ecology Evolution, 23(2): 104–112.
  17. Dobson et al. 2012. Journal of Applied Ecology, 49:1325-1333
  18. Hopcraft et al. 2014. Conservation Biology 00, 0, 1-5.
  19. Homewood et al. 2010. Nature 467:788–789;
  20. Fyumagwa et al. 2013. Conservation Biology, 27: 1122–1125;
  21. Kideghesho et al. 2013. Nature Conservation, 6: 1-29;
  22. Baptista and Rudel 2006. Environmental Conservation, 33: 195–202
  23. Gilroy et al. 2014. Nature Climate Change, 4: 503–507;
  24. Rodrigues et al. 2009. Biological Conservation, 142: 1141–1153.