Satellite-based monitoring systems are the primary tools for providing timely information on newly deforested areas in vast and inaccessible tropical forests. Their potential to empower governments and communities to enact timely actions against illegal and unsustainable forest activities, and to respond to natural disasters is increasingly recognised. Monitoring of changes in tropical forest cover has relied predominantly on optical satellite sensors, however persistent cloud cover limits its use for near real-time forest change detection in many tropical regions. Spaceborne SAR data have the advantage of providing cloud-free observations. With Sentinel-1 (launch 2014) for the first time, dense and regular SAR time series data are provided over tropical forest areas free and openly.
This thesis aims to explore the potential of dense Sentinel-1 time series to detected deforestation using Google Earth Engine (GEE) platform. The study site will be Sumatra, Indonesia. The thesis will focus on exploring machine learning approaches (e.g. Random Forest) to detect new deforestation in the Sentinel-1 data. This will include, data handling in GEE, design and collection of samples to train the machine learning algorithm, parameter testing and validating of the results. A potential combination with optical data (Landsat, Sentinel-2) may be explored.
- To detect new deforestation in Sumatra, Indonesia using Sentinel 1, machine learning methods and the Google Earth Engine platform.
- Testing and evaluating different Sentinel-1 metrics for detecting deforestation
- Testing and evaluating machine learning algorithms (e.g. random forest) and parameters
- Reiche, J.; Verhoeven, R.; Verbesselt, J.; Hamunyela, E.; Wielaard, N.; Herold, M. Characterizing tropical forest cover loss using dense Sentinel-1 data and active fire alerts. Remote Sens. 2018, 10.
- Margono, B., & Turubanova, S. (2012). Mapping and monitoring deforestation and forest degradation in Sumatra (Indonesia) using Landsat time series data sets from 1990 to 2010. Environmental Research Letters, 7(3), 34010.
- Advanced Earth Observation course (GRS-32306)
- Geo-scripting course (GRS-33806)(Good knowledge in scripting is an asset; e.g. R, python, java script)
- Affinity to work with optical and radar remote sensing data
Theme(s): Modelling & visualisation, Integrated Land Monitoring