The Amazon forests have adapted to a climate gradient ranging from areas with no dry season to those with up to six consecutive dry months. Ecohydrological studies covering this gradient are rare but essential for understanding forest responses to climate change. The western Amazon, in particular, is the least studied region in the Amazon Basin, despite models predicting it will be the most perniciously impacted by climate change.

To address the lack of continuous ecohydrological measurements across this gradient, a consortium led by the Catholic University of Peru created AndesFlux, which comprises six eddy flux towers and their associated 1-hectare permanent plots. Another tower is scheduled for installation by year-end. Tower locations across the western Amazon's climatic gradient are shown in Figure 1. Four of the towers are registered with AmeriFlux. AndesFlux is a collaborative platform welcoming participation ranging from proposal development to project execution.

In this presentation, we show preliminary results of carbon dioxide (CO₂) and methane (CH₄) fluxes from three towers in southern (Tambopata) and central (Panguana) Peru and southern Ecuador (San Francisco). The Tambopata tower (PE-TNR), located in primary forest, has been operational since 2017. Data analyses from December 2023 to April 2024 (local wet season) show the ecosystem acted as a carbon sink (Net Ecosystem Production, NEP = 1.13 g C m^-2 d^-1) but was a CH₄ source with a median flux of 1.12 mg CH₄ m^-2 d^-1. The average Bowen ratio at solar noon is approximately 0.25. The Panguana tower (PE-PAN), operational since December 2023, indicated that the surrounding secondary rainforest acted as a small net CO₂ source (mean flux = 0.02 g C m^-2 d^-1) for the same period (Dec 2023-April 2024). The estimated Bowen ratio at solar noon was 0.33. At Panguana, the change in CH₄ storage, measured from the vertical profile, was three orders of magnitude less than that of CO₂. The network's northernmost site at San Francisco, Ecuador, located in primary montane forest, displayed a mean NEP for the period 2019–2023 of 2.1 g C m^-2day^-1. AndesFlux data were used to validate and improve carbon and water flux simulations with the trait-based Dynamic Global Vegetation Model LPJmL-FIT, enabling more robust predictions of western Amazon ecosystem stability under climate change.