Earth Observation for tropical wetland management
to support food security in Rwanda
Acknowledgements:
We thank the German Federal Ministry of Economic Affairs and Energy for financial support
(FKZ: 50EE1537) as well as the German Aerospace Center
Introduction
Against the background of population growth and climate change, Sub-Saharan
African (SSA) wetlands are increasingly considered as a possibility to improve food security. Although wetlands are already part of national and regional agricultural
policies, there is still an information gap on their location, physical and land use
characteristics, which can be closed through remote sensing technology. In order to
reflect wetland variability in space and time, the use of time series is imperative. The goals of this study were:
• The development of a mainly automatic EO-based tropical wetland delineation
and characterization scheme which reflects wetland variability
• The creation of national-scale baseline data for Rwanda at a high spatial resolution (10-30 m)
Stefanie Steinbach1,2, Konrad Hentze2, Jonas Franke3, Natalie Cornish3, Adrian Strauch2,4, Frank Thonfeld2,4, Sander Zwart5, Andrew Nelson1
(1) Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, Enschede, Netherlands; (2) Department of Geography, University of Bonn, Germany; (3) Remote Sensing Solutions GmbH, Munich, Germany; (4) Center for Remote Sensing of Land Surfaces (ZFL), University of Bonn; Germany; (5) International Water Management Institute (IWMI), Accra, Ghana
Results
The developed approach for tropical
wetland characterization provides
nationwide data for agricultural planning
and monitoring at different scales.
Key Findings
• Clusters of wetland occurrence and use intensity could be detected using the
multi-layered wetland characterization approach; intensively used areas are
located in the north-east of Rwanda and south of Kigali
• WUI is highest for wide wetland sections under officially restricted use, whereas the often small sections without legal restrictions only show intermediate WUI
Conclusion & Outlook
• The advanced degree of automation at the high spatial resolution of Sentinel-1
and -2 imagery allow the incorporation of this wetland characterization scheme into agricultural planning and wetland monitoring schemes
• The created data could furthermore contribute to agricultural productivity
assessments or ecosystem integrity assessments
Stefanie Steinbach
Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente
s.steinbach@utwente.nl
Figure 2: Wetland Delineation (a), Surface Water Occurrence (SWO) (b), Wetland Use Intensity (WUI) (c) and RGB display of NDVI as Vegetation Response (d) for a wetland complex of the Nyabarongo River south of Kigali. Semi-natural vegetation, extensively used area and a large rice irrigation scheme are clearly distinguishable.
1Otsu, N., 1979. A Threshold Selection Method from Gray-Level Histograms. IEEE Transactions on Systems, Man, and Cybernetics, 9(1), 62-66. ²Franke, J.; Keuck, V.; Siegert, F., 2012. Assessment of grassland use intensity by remote sensing to support conservation schemes. Journal for Nature Conservation, 20, 125–134.
Wetland Delineation
- An object-based approach was applied to a previously developed potential wetlands layer and NDVI and NDWI from a 2017 Sentinel-2 composite; manual editing improved the result
- Validation: 255 randomly selected delineation slices were compared to wetland area digitized from RapidEye imagery with a median agreement of 84.71 %
Surface Water Occurrence (SWO)
- 60 Sentinel-1 SAR images (IW, VV polarization) from 2014 to 2017 were speckle-filtered and each scene classified by applying an Otsu threshold1 in
Google Earth Engine; water frequency was calculated across this time series - Permanent water bodies were derived from the 5th percentile
- Validation: 2 individual classifications were validated against manually digitized water in 4 RapidEye tiles resulting in a 94.77 % overlap
Wetland Use Intensity (WUI)
- The Mean Absolute Spectral Dynamics (MASD²) algorithm was adapted to wetland ecosystems and created from multiple cloud and cloud shadow masked Sentinel-2 images from 2017
- WUI was compared to the legal use status of to-date mapped wetlands as obtained from the Rwandan Environment Management Authority (REMA)
Vegetation Response
- The NDVI was calculated from a cloud-masked time series of Sentinel-2 images in Google Earth Engine from 2016 to 2019
- The 20th, 50th and 90th percentiles were derived to capture vegetation
response to different flooding regimes and wetland use types
NDVI 90th percentile 50th percentile 20th percentile SWO 100 % 0 % Wetland Wetland WUI Low: 0 High:1532
Figure 1: Total wetland area (a) and mean Wetland Use Intensity per ~9x9km unit (b) across Rwanda.
a) a) b) b) c) d) Wetland
Methodology
Mean Wetland Use Intensity High: 121 Low: 0 Country boundaries Country boundaries 81 0
