A Cloud-Based Post-Processing Pipeline for Eddy Covariance Flux Datasets: From Actual Evapotranspiration Measurements to Spatial Water Balance

Eddy covariance stations collect in situ high frequency (10 Hz) measurements of wind speed and direction in three dimensions alongside high frequency (10 Hz) measurements of water vapor concentration. Raw high frequency data undergo processing onboard field-deployed sensor systems to provide flux measurements at 30-minute averaging intervals. The processed flux data are then transmitted via message queue telemetry transport (MQTT) to a cloud-based platform where post-processing can occur. This presentation provides an overview of a post-processing pipeline built on 30-minute fluxes and ancillary meteorological data inputs to arrive at cleaned and gap-filled evapotranspiration fluxes over multiple timeframes. The steps to arrive at these intermediate data products include physically plausible threshold detection, quality-control based on error condition for the flux averaging interval, statistical outlier detection, and gap filling using the marginal distribution sampling (MDS) method. Drivers for MDS gap filling shown include vapor pressure deficit (VPD), incoming shortwave radiation (SWin), and air temperature (Tair). Accumulated fluxes are then spatialized based on the flux footprint associated with the accumulation period, using inputs from flux sensors and global weather models as well as ancillary remote sensing multispectral imagery from the European Space Agency (ESA) Sentinel2 constellation. The provenance associated with this pipeline is available to promote scientific reproducibility.