Wetland Monitoring Using Dual-Polarized X-Band and C-Band Data of Lac Bam, West-Africa

Monitoring of wetlands is a very important task for water management as well as for wetland ecology. Water bodies and wetlands in semi-arid African areas show strong seasonal dynamics throughout the year that require time series imagery with narrow temporal intervals for satellite-based monitoring. The study site for this work is Lac Bam, which is the largest natural lake in Burkina Faso, and as such of high importance for the local inhabitants for irrigated farming, animal watering, and extraction of water for drinking and sanitation. Semi-arid African wetlands often contain areas of floating vegetation or flooded vegetation, which causes difficulties for water body detection and monitoring over time when using synthetic aperture radar (SAR) intensity data or medium to high resolution optical data. The water of Lac Bam is very turbid and sediment-rich, particularly during the rainy season when also strong cloud-cover is a limiting factor for the use of optical data. SAR intensity data has been frequently used to detect open water surfaces. Multi-polarized SAR data, in addition, have proven to be able to detect flooded vegetation, thanks to their ability to distinguish between different scattering mechanisms in different polarizations. Two different repeat-pass time series have been acquired in X-band and C-band dual-polarimetric mode: TerraSAR-X StripMap data with HH (transmitted and received horizontally)–VV (transmitted and received vertically) polarization with an interval of 11 days have been acquired from August 2013 until present (January 2015). Radarsat-2 fine beam data with HH (transmitted and received horizontally)–HV (transmitted horizontally, received vertically) polarization with an interval of 22 days have been acquired since October 2014 until present. The Kennaugh decomposition produces four normalized Kennaugh elements from the complex HH and VV, or HH and HV input channels, respectively. A novel image enhancement algorithm called Multi-scale Multi-looking stabilizes the radiometric measurements without loss of geometric resolution. Results demonstrate that open water could be well detected in the Kennaugh element representing the total intensity from both co-polarized HH-VV (X-band) as well as cross-polarized HH-HV (C-band) data. The three remaining polarimetric Kennaugh elements deduced from X-band clearly enable the detection of vegetation in or on water, and can discriminate it from vegetation on land. Applied onto the full time series surface water fluctuations and changes in flooded vegetation can be monitored throughout the year. Seasonal curves extracted from time series of the Kennaugh elements allow concluding to different types of flooded vegetation and the duration of flooding. Results of the analysis are validated based on optical very high resolution data from WorldView-2 with a resolution of 0.5 m, and RapidEye with a resolution of about 6 m for selected dates throughout the time series. Moreover, the performance of TerraSAR-X and Radarsat-2 is compared using data acquired on the same date in October 2014. Despite the great importance of water management and ecology in African wetlands, regular monitoring is often not carried out and applications of polarimetric SAR imagery in Africa are very scarce. This study contributes to understanding of the dynamics of surface water including flooded vegetation in or floating vegetation on water. It demonstrates the capability of dual-polarimetric SAR imagery for wetland delineation, and concludes to the applicability of Sentinel-1 dual-polarimetric data for almost continuous wetland monitoring