A method to reduce sampling bias in multi-level tall-tower eddy covariance systems

Tall-tower eddy covariance (TTEC) systems are increasingly used to monitor land–atmosphere exchanges over complex agricultural and urban landscapes. However, interpreting flux estimates is challenging because the eddy covariance footprint varies significantly with meteorological conditions, which can introduce considerable bias assuming that sources and sinks are not uniformly distributed across the landscape or over the diel cycle. For fluxes with systematic diurnal patterns, such as traffic-related emissions, photosynthesis, or agricultural activities, uneven temporal sampling can prevent capturing a full daily cycle, introducing temporal sampling bias into daily flux estimates. The objective of this study was to evaluate the performance of a multi-level TTEC system in reducing footprint-related sampling bias.

The study site is located in an agricultural landscape west of Copenhagen, Denmark. A 15-month dataset (2023-2024) was collected, representing a heterogeneous landscape dominated by grassland and cropland, with scattered settlements, hedgerows, and forested areas. The TTEC system was installed on a 300 m telecommunication tower and equipped with three measurement levels at 70, 90, and 115 m. These sampling heights were selected a priori based on flux footprint estimates from wind data of a nearby tall tower, ensuring a more uniform footprint at a wider range of atmospheric stability conditions. Each level was equipped with a 3D ultrasonic anemometer (uSonic-3 Class A MP, METEK, Germany). A fast-response gas analyser was connected to the system and configured to sample air from one of the three heights at a time based on criteria related to optimal footprint size and constant flux layer requirements.

The results of the study showed that a greater number of observations were collected at the upper sampling height during daytime whereas nighttime observations were predominantly obtained from the lower level. The intermediate level was primarily used during the transition periods between day and night. The multi-level sampling scheme enabled a substantial reduction in sampling bias by actively controlling the horizontal extent of the flux footprint compared to a single-level TTEC system. Consequently, footprint size and the relative contributions of different land-cover types were more consistent across atmospheric stability regimes. The findings from this study highlight the importance of implementing a multi-level approach, particularly for TTEC systems operating over landscapes with greater heterogeneity than those typically sampled by conventional eddy covariance systems.

 

Acknowledgements

This project is supported by the Independent Research Fund Denmark (DFF-grant 1127-00308B - Observation System of Greenhouse Gas Sources and Sinks at the Landscape Scale for Verification of the Green Transition of Denmark). The authors wish to thank Cibicom A/S for sponsoring access to Hove telecommunication tower.