Numerical investigation of observational flux partitioning methods for water vapor and carbon dioxide

Authors: ZAHN, E - Princeton University; GHANNAM, K - Princeton University; CHAMECKECKI, M - University Of California; MOENE, A - Wageningen University And Research Center; Kustas, William; GOOD, S - Oregon State University; BOU-ZEILD, E - Princeton University

Submitted to: Biogeosciences
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/28/2024
Publication Date: 6/21/2024
Citation: Zahn, E., Ghannam, K., Chameckecki, M., Moene, A., Kustas, W.P., Good, S., Bou-Zeild, E. 2024. Numerical investigation of observational flux partitioning methods for water vapor and carbon dioxide. Biogeosciences. Article e2024JG008025. https://doi.org/10.1029/2024JG008025.
DOI: https://doi.org/10.1029/2024JG008025

Interpretive Summary: The partitioning of evapotranspiration (ET ) into plant transpiration and soil evaporation is important in agriculture because knowledge of actual water use by the crop or transpiration determines actual crop water use efficiency and stress. The most promising methods to separate ET into surface evaporation € and canopy transpiration (T ) use flux tower measurements of high frequency (turbulent) observations of water vapor and carbon dioxide (eddy covariance). The problem is that validating the partitioning methods under different cropping and environmental conditions would require eddy covariance measurements across a wide variety of crop types and climatic conditions. To overcome this limitation, numerical simulations of canopy flows relying on the Large-Eddy Simulation (LES) technique are used to simulate eddy covariance flux tower measurements. The partitioning methods using eddy covariance show great utility in estimating E and T for most cropping systems but will be most challenging for sparse row crops such as vineyards unless the measurements are close to the surface. This analysis serves as a benchmark indicating that the eddy covariance methods being used in flux tower networks globally offer the possibility to determine E and T for a wide variety of agroecosystems leading to improvements in crop water use and stress management.

Technical Abstract: This paper investigates the partitioning of evapotranspiration (ET ) and CO2 fluxes (Fc) into plant and soil components. Two previously proposed partitioning approaches based on the turbulent behavior of CO2 and water vapor perturbations are investigated: the Flux-Variance method (FVS) and the Conditional Eddy Covariance (CEC). In addition, we also propose and test two additional methods: the Conditional Eddy Accumulation (CEA) method and an approach combining CEC with the water-use efficiency (W ), labeled CECw. Since experimental tests of these methods are always fraught with the uncertainty of the “correct standard” to compare to, Large-Eddy Simulations (LES), where the four flux components can be imposed individually, are used in this study. Simulations over domains with different plant canopy geometries are performed to generate high-frequency time series that mimic tower observations at various heights. Our results indicate that canopies with large exposed soil patches increase the turbulence mixing of scalars, negatively impacting the performance of the partitioning methods, all of which require some degree of uncorrelatedness between CO2 and water vapor. The CEC and CEA methods were found to perform well when data are collected at the canopy top and when all components are non- negligible, with CEA slightly outperforming CEC. In the present tests, the FVS method displayed the best performance under all conditions, with the caveat that this method also requires, and is sensitive to, the water-use efficiency. While W is imposed and known in our LES, its estimates for observational analyses have significant uncertainty. When a known W is ingested into the CEC method, the resulting CECw performs better than CEC at different heights and under more combinations of flux components. CECw is also found to be less accurate but more robust that the FVS method. The overall findings highlight (i) the importance of selecting an appropriate measurement height, (ii) the benefits of a simultaneous application of all methods to increase confidence in the results, and (iii) the improved performance of the methods when the water use efficiency is accurately known (for FVS and CECw) or when all fluxes are non-negligible (for CEC and CEA). Regarding this last criterion, we investigate the connection between the water-use efficiency and the correlation coefficient between water vapor and CO2, indicating the possibility of quick screening of the magnitudes of the fluxes and of recovering biophysical information from simple high-frequency data measurements.