Riparian Wetland Mapping and Inundation Monitoring Using Amplitude and Bistatic Coherence Data From the TanDEM-X Mission

This article focuses on bistatic coherence as an additional feature complementing amplitudes in classification space, permitting to monitor temporal changes in water extent on the wetland comprising surface water and inundated vegetation. The research was conducted on a herbaceous wetland. The TanDEM-X images were acquired during the science phase in bistatic mode with long perpendicular baselines. Two different sets of observations were computed: polarimetric amplitudes (PAs) and interferometric coherences in single-pass mode. Next, the datasets composed of a multitemporal stack of images were classified using object-based image analysis. The main outcome of the experiment is that bistatic coherences increased greatly the overall accuracy (OA) of expected thematic classes. The OA shows that thematic categories were classified with higher accuracy when the bistatic coherence complemented PAs. The OA is greater than 85% for all analyzed datatakes. The accuracy achieved using amplitudes only was higher than 70% but varied overtime. The bistatic coherence at X-band turned out to be really helpful in mapping high vegetation, which can be an indicator that this methodology can be directly used in the monitoring of common reed mowing or mapping highly invasive vegetation. Additionally, we could observe that short inundated vegetation was also mapped correctly, allowing flooded areas in this floodplain to be mapped with great precision throughout the growing season.

Automated summary

This article focuses on bistatic coherence as an additional feature complementing amplitudes in classification space, allowing to monitor temporal changes in water extent on the wetland. The experiment showed that bistatic coherence significantly improved the accuracy of wetland classification, especially in mapping high vegetation.