A novel framework for assessing the spatiotemporal dyamics of soil water vapor adsorption using a global observation network

Quantifying liquid water inputs to the Earth’s surface is crucial for understanding water and energy exchange between the surface and the atmosphere. While rainfall is the main source, small non-rainfall inputs, fog, dew, and soil water vapor adsorption (SVA), are ecologically significant, but remain poorly quantified due to scarce and inconsistent field data. This dissertation examines whether the eddy covariance (EC) method can detect and quantify SVA at the ecosystem scale. Specifically, it investigates (1) the seasonal dynamics of non-rainfall inputs in a Mediterranean ecosystem, (2) whether downward movements of water vapor measured by EC can be interpreted as SVA, and (3) if SVA can be identified across the global FLUXNET network, and what factors control its detectability. We combine high-precision lysimeter measurements with EC fluxes, supported by independent data streams and models. Lysimeter data show that non-rainfall inputs occur on 84% of nights at a Mediterranean site, with dew and fog during wet seasons, while SVA dominates dry summers, serving as the only water input for several weeks annually. EC measurements showed periods when water vapor moved from the atmosphere back into the soil, coinciding with lysimeter-based SVA, confirming detection despite underestimation of flux magnitude. Extending the analysis globally, we identified SVA using EC-based signals of vapor direction, atmospheric humidity, and soil water content. Adsorption was found to be widespread, particularly in arid, sparsely vegetated ecosystems, averaging ~4 hours per day and up to 150 days per year. Overall, this work demonstrates that EC can provide long-term, ecosystem-scale information on SVA, bridging the gap between laboratory studies and large-scale observations. This enables future research on seasonal variability, climate sensitivity, and management impacts, and supports the integration of SVA into process and remote sensing models.