Contribution of automatically generated radar altimetry water levels from unsupervised classification to study hydrological connectivity within Amazon floodplains

Abstract

Study regionThe Curuaí floodplain in the low Amazon river in the Pará state of Brazil and Juruá basin, a major Solimões tributary. Study focusCharacterizing the hydrological dynamics of Amazon floodplains is essential to better understand and preserve these environments providing important resources to local populations. Radar altimetry is an effective remote sensing tool for monitoring water levels of continental hydrosystems, including floodplains. An unsupervised classification approach on radar echoes to determine hydrological regimes has recently been tested and showed a strong potential on the Congo River basin. This method is adapted to Envisat and Saral satellite radar altimetry data on two study areas in the Amazon Basin. The aim is to improve inland water detection along altimeter tracks to automatically generate water level time series (WLTS) over rivers, lakes, and poorly monitored floodplains and wetlands. New hydrological insightsResults show a good agreement with land cover maps obtained with optical imagery over selected Amazonian wetlands (70–80% accuracies with Envisat data and 50–60% with Saral data). Automatically generated WLTS are strongly correlated to the manually generated WLTS (R² ≈ 0.9; RMSE < 1 m). Compared to the manual method, the automatic method is faster, more efficient and replicable. Densifying the WL network in the floodplains bring crucial information on the connectivity dynamic between lakes and rivers.