Amazon's 2023 Drought: Sentinel-1 Reveals Extreme Rio Negro River Contraction

TL;DR

This study demonstrates near-real-time mapping of the Rio Negro River's water surface at 10 m resolution using Sentinel-1 SAR and a U-Net segmentation model, producing 12-day mosaics for 2022–2023. The approach achieves a high F1-score (~0.93) and is validated against JRC Global Surface Water and MapBiomas 2022 datasets, capturing more fine-scale rivers than 30 m products. During the 2023 drought, water surface contracted to 68.1% of its maximum and 81.0% of the period median, with strong alignment to concurrent water-level measurements at Manaus (r = 0.887). The work highlights the practical potential of cloud-penetrating SAR combined with deep learning for near-real-time hydrological monitoring in tropical regions and discusses pathways and limitations for scaling to larger areas and future SAR missions.

Abstract

The Amazon, the world's largest rainforest, faces a severe historic drought. The Rio Negro River, one of the major Amazon River tributaries, reaches its lowest level in a century in October 2023. Here, we used a U-net deep learning model to map water surfaces in the Rio Negro River basin every 12 days in 2022 and 2023 using 10 m spatial resolution Sentinel-1 satellite radar images. The accuracy of the water surface model was high with an F1-score of 0.93. The 12 days mosaic time series of water surface was generated from the Sentinel-1 prediction. The water surface mask demonstrated relatively consistent agreement with the Global Surface Water (GSW) product from Joint Research Centre (F1-score: 0.708) and with the Brazilian Mapbiomas Water initiative (F1-score: 0.686). The main errors of the map were omission errors in flooded woodland, in flooded shrub and because of clouds. Rio Negro water surfaces reached their lowest level around the 25th of November 2023 and were reduced to 68.1\% (9,559.9 km$^2$) of the maximum water surfaces observed in the period 2022-2023 (14,036.3 km$^2$). Synthetic Aperture Radar (SAR) data, in conjunction with deep learning techniques, can significantly improve near real-time mapping of water surface in tropical regions.

Figures (6)

• Figure 1: Geographical location in blue of the entire Rio Negro River and its tributaries before it forms the Amazon with the Solimões river at Manaus, adapted from FAO2022. The study area is represented with a dashed black line, and the approximate extents of the 24 Sentinel-1 images taken for each orbit are in magenta.
• Figure 2: U-Net model architecture used for water surface estimation from Sentinel-1 images, adapted from Ronneberger2015. The number of channels is indicated above the cuboids, and the vertical numbers indicate the row and column size in pixels. The operations (convolutions, skip connections, max pooling and upsampling) performed in each layer and their sizes are indicated by the colored arrows.
• Figure 3: Observation and prediction for 18 images of the validation dataset with a F1-score above 0.97.
• Figure 4: Observations and predictions for six images of the validation dataset with an F1-score below 0.5 or cloud artifacts.
• Figure 5: Daily water level of the Rio Negro measured at the Port of Manaus and water surface estimated every 12 days with our Sentinel-1 based model for the Rio Negro study area, for the period 2022-2023.