MSc thesis topic: Above-ground Biomass Change (ΔAGB) mapping using stacked machine learning models

Changes in above-ground biomass are useful for national carbon accounting i.e., UN-SEEA framework (Hein et al., 2020). Though there have been recent data sources usable for carbon accounting such as the ESA-CCI maps (Santoro and Cartus, 2021) and the WRI-Flux model (Harris et al., 2021), current analysis revealed that their ΔAGB have a level of disagreement (Araza et al., under preparation).  An option to make good use of these recent ΔAGB maps is to make an ensemble ΔAGB i.e., through model stacking approach. Such approach involves two machine learning models: (1) “base learners” where different ML models are used to predict ΔAGB using ΔAGB from the map products and other ecological variables; and (2) “meta learners” where the predictions from each base learner serve as inputs. Predictions from this “stacked model” should be cross-validated using an appropriate method. The final ΔAGB map will be the basis for creating carbon accounting tables.

Objectives

• Generate an ΔAGB stacked model
• Apply appropriate model hyper-parameter tuning, modelling and cross-validation of ΔAGB predictions
• Map ΔAGB and create carbon accounting tables

Literature

• Healey, S. P., Cohen, W. B., Yang, Z., Brewer, C. K., Brooks, E. B., Gorelick, N., ... & Zhu, Z. (2018). Mapping forest change using stacked generalization: An ensemble approach.Remote Sensing of Environment,204, 717-728.
• Harris, N. L., Gibbs, D. A., Baccini, A., Birdsey, R. A., De Bruin, S., Farina, M., ... & Tyukavina, A. (2021). Global maps of twenty-first century forest carbon fluxes.Nature Climate Change,11(3), 234-240.
• Hein, L., Remme, R. P., Schenau, S., Bogaart, P. W., Lof, M. E., & Horlings, E. (2020). Ecosystem accounting in the Netherlands. Ecosystem Services, 44, 101118.

Requirements

• R geo-scripting and machine learning

Theme(s): Integrated Land Monitoring