Work

Flood pulse monitoring in wetlands with multi-temporal Sentinel-1 interferometric coherence data: Application to the Okavango Delta (Botswana)

Year: 2025

Type: article

Abstract: Flood-pulsed wetlands are characterized by significant seasonal water fluctuations, which play a critical role in the dynamics of these sensitive ecosystems. Among the growing number of existing remote sensing products, we explore the potential of interferometric (InSAR) coherence time series, derived from Sentinel-1 synthetic-aperture radar images, to characterize the hydrological dynamics of the Okavango Delta, a vast flood-pulsed wetland. Interferometric coherence reflects changes in surface conditions, making it a powerful tool for detecting flood propagation. By fitting harmonic functions, we produce parameters that quantify the seasonality of coherence time series with short isotemporal baselines (12 days). In particular, we developed a normalized seasonal index based on the ratio between the seasonal amplitude and the root-mean-square error of the fitted harmonic function, to map the seasonality of the coherence time series. A multi-annual analysis of coherence time series reveals a strong relationship between their seasonality, land cover, and flood frequency. Unsupervised clustering applied to statistical and seasonal metrics of coherence time series yields consistent classifications that map the variability of flood frequencies across wetland areas and clearly distinguish wetlands from dry zones. Similarly, thresholds applied to normalized seasonal indices delineate the year-to-year extent of flood pulses with accuracy around 79 %. We show that coherence time series in never flooded areas exhibit a pronounced seasonal pattern driven by rainfall cycle, whereas this seasonality is disrupted by flood pulses in wetlands. Building on this, we developed a change-detection approach to map the floods by identifying the date when coherence time series diverge from their seasonal pattern. The resulting flood arrival dates achieve 74–83 % accuracy compared to a reference dataset derived from optical data. Our results highlight the potential of coherence time series as a robust indicator of seasonal variations in inundation extent in flood-pulsed wetlands.

Source: Remote Sensing of Environment

DOI: 10.1016/j.rse.2025.115173