Integrated observations and atmospheric modeling to bridge the scaling gap from local to landscape

Many natural ecosystems are subject to fine scale variability in biogeophysical and biogeochemical properties, consisting of a mosaic of patches with individual characteristics in e.g. vegetation, hydrology, or microclimate. Carbon cycle fingerprints between patch types may exhibit strong differences, and reactions to current variability in external forcing as well as to future climate change may substantially differ across spatial gradients of often just a few meters or less. Capturing a representative carbon budget for such landscapes is highly challenging, since footprints of common observation techniques are either rather small with limited representativeness (e.g. flux chambers), or rather large and therefore aggregating signals across multiple patch types (e.g. eddy covariance).

This study is based on a 2025 field campaign at Stordalen Mire in Northern Sweden, a highly structured wetland consisting of a patchwork of fens, bogs, palsas and open water areas. Observational platforms included 2 eddy covariance towers with different instrument heights but nested footprints, stationary (fixed collars) and mobile chamber flux measurements within the tower footprints, a floating mobile auto-chamber system for distributed observations across different lakes and lake zones, and a drone equipped with in-situ greenhouse gas analyzers and meteorological sensors for landscape-integrating surveys using grid, curtain and profile flights. Since all platforms focused their observations on the same wetland section (about 500x500m), our dataset allows to merge detailed process information for individual ecosystem patches (e.g. from flux chamber data) with the landscape-scale integrative products (e.g. by eddy towers or drone).

We present results from different scaling approaches for deriving ecosystem-scale CO₂ and CH₄ budgets and variability, including e.g. data-driven upscaling, decomposition of eddy-covariance observations into patch-level fluxes, and local scale inversion of drone observations, each focusing on different subsets of the observational database. Through combining all data streams we aim at reducing uncertainties in wetland-scale carbon budgets as well as in the assessment of flux representativeness for the larger region. Comparing upscaled fluxes reveals strengths and weaknesses of individual data streams for constraining net carbon budgets and identifying functional controls, and delivers guidelines towards optimum upscaling strategies.