Change detection and monitoring

The earth and its biosphere are constantly changing. It therefore is critical to detect change in order to understand change processes and their impact on terrestrial ecosystems. Satellite sensors provide consistent and repeatable measurements that enable the capturing of effects of many processes that cause change, including natural (e.g. fires, insect attacks) and anthropogenic (e.g. deforestation, urbanization, farming) disturbances. Time series of satellite images provide ways to detect and monitor changes over time and space (e.g. over the last 30 years globally).

A wealth of remotely sensed image time series covering the earth is now available. We are entering the big-data era, especially with the advent of European Sentinel constellation. Efficient ways to detect and understand changes (e.g. deforestation, forest degradation, or climate change) are lacking.

We focus on providing new ways for quantifying and understanding changes:

  • Deforestation, forest degradation, and regrowth from local to global scale
  • Effects of global warming on plant phenology and productivity
  • Impact of drought and fire on vegetation dynamics

Research within the Laboratory of Geo-Information Science and Remote Sensing is focused on:

  • Change detection
  • Near real-time change monitoring
  • Trend analysis and break detections
  • Monitoring of change dynamics (e.g. degradation, regrowth)
  • Time series data fusion (e.g. optical-radar, medium-coarse resolution optical)
  • Early warning and tipping points
  • Phenology

Hereby a few examples of satellite sensors we are using for change monitoring; Sentinel-1, ALOS PALSAR, Landsat, MODIS, AVHRR, Rapid-Eye and SPOT to prepare for Sentinel-2 and 3.

Current study areas include sites in Ethiopia, Brazil, Peru, Fiji, Bolivia, Vietnam, Laos, Sierra Leone, and other locations across the globe.

Available open-source software for trend analysis and change detection:

  • BFAST package for R: http://bfast.r-forge.r-project.org
     
    The BFAST package provides generic functionality for continuous change monitoring, trend analysis, and near-real time disturbance detection of any kind of disturbance or change process (gradual e.g. forest degradation to more abrupt like deforestation, droughts). The methods can be applied to any series of satellite images going from for example MODIS, Landsat, Rapid Eye, or RADAR data (see papers below). Besides that generic change detection functions bfast for time series segmentation, or bfastmonitor for near-real time monitoring can be applied to any kind of time series like for example rainfall, temperature, or dendrochronology series.
     
  • bfastSpatial package for R: http://github.com/loicdtx/bfastSpatial
     
    The bfastSpatial package provides utilities to perform change detection analysis (De Vries et al. 2015, Dutrieux et al. 2015, Verbesselt et al. 2010 and 2012) on time-series of spatial gridded data, such as time-series of remote sensing images (Landsat, MODIS and the likes). The tools provided by bfastSpatial allow user to perform all the steps of the change detection workflow, from pre-processing raw surface reflectance Landsat data, inventorying and preparing them for analysis to the production and formatting of change detection results. The present document aims at providing guidance to the users of bfastSpatial by detailing every steps of the process.
     
    A tutorial is available via this link and and introduction to the package is available via this presentation.
     
  • timeSyncR package for R: http://github.com/bendv/timeSyncR
     
    The timeSyncR package is based on the TimeSync method (Cohen et al., 2010, Remote Sensing of Environment) and provides a tool to aid in the visualization and interpretation of Landsat time series in R for calibration/validation of change detection methods.
     
  • MODIS package for R: http://modis.r-forge.r-project.org
     
    The package contains a collection of the functions downloading, mosaicing, re-sampling, re-projecting, analysing and visualising of MODIS data. Special attention is given to spatio-temporal filtering, change detection and phenological metric extraction.
     
  • MulTiFuse package for R: http://github.com/jreiche/multifuse
     
    The MulTiFuse (Multi-sensor Time series Fusion) package provides functions to fuse optical and SAR time series. The MulTiFuse approach has been published in Reiche et al. 2015, where it was used to fuse Landsat NDVI and ALOS PALSAR L-band SAR backscatter time series. The package provides example data and a detailed description.

Key Publications

Remotely sensed resilience of tropical forests
Verbesselt, J., Umlauf, N., Hirota, M., Holmgren, M., van Nes, E. H., Herold, M., et al. (2016). Remotely sensed resilience of tropical forests. Nature Climate Change, 1–5.