A year-long eddy covariance dataset over an Alpine Steppe on the central Tibetan Plateau: a landscape perspective on carbon and energy fluxes

Abstract. The Tibetan Plateau (TP) is warming rapidly, with future projections suggesting continued warming that may amplify climate–carbon feedbacks. However, sparse in-situ observations and pronounced spatial heterogeneity in vegetation, soil moisture, and climate have limited our understanding of these ecosystem responses. Here, we present a continuous record of carbon and energy fluxes measured at a landscape scale (∼30 ha (0.3 km2)) in an alpine steppe ecosystem on the TP from July 2018 to June 2019. The dataset yields a cumulative net ecosystem exchange (NEE) of -12.6±16.4 g C m−2, indicating a near carbon-neutral budget. Flux measurements were quality-filtered, retaining 33 %–44 % of high-quality data across fluxes, and gap-filled to produce a complete seasonal record using two complementary approaches, marginal distribution sampling (MDS) and random forest (RF), with RF outperforming MDS for all the fluxes. Eddy covariance measurements represent integrated fluxes over their footprint area, which are often much smaller than most model grids or remote sensing pixels, particularly in grassland ecosystems. Owing to the higher measurement height (19 m) at this site, the footprint climatology (90 % source area ≈ 280 000 m2) closely aligns with the spatial coverage of a Moderate Resolution Imaging Spectroradiometer (MODIS) 500 m pixel (∼250 000 m2), making this dataset particularly suitable for future landscape-scale comparisons and satellite product validation. The data include quality flags indicating observed or gap-filled values, uncertainty estimates, footprint diagnostics, and auxiliary meteorological variables (air and soil temperature, soil moisture, radiation, relative humidity, vapour pressure deficit). The data described in this manuscript provides a robust foundation for examining carbon–climate interactions in alpine environments and supporting ecosystem modeling (https://doi.org/10.5880/GFZ.TKVR.2026.001, Pillai et al., 2026).