Kim et al. (2026) Comment on Golkar and de Oliveira (2026): Missing eddy covariance observations undermine reported OpenET biases
Identification
- Journal: Agricultural and Forest Meteorology
- Year: 2026
- Date: 2026-11-11
- Authors: Yeonuk Kim, Sayantan Majumdar, John Volk, Peter ReVelle, Todd G. Caldwell, Justin Huntington
- DOI: 10.1016/j.agrformet.2026.111586
Research Groups
Division of Hydrologic Sciences, Desert Research Institute, Reno, NV, USA
Short Summary
This comment identifies substantial inconsistencies in the eddy covariance (EC) reference data used by Golkar and de Oliveira (2026), demonstrating that the apparent treatment of missing observations as zero and omission of energy balance correction led to inflated OpenET overestimation and questionable machine learning bias correction effectiveness.
Objective
- To identify and quantify inconsistencies in the eddy covariance (EC) reference data used by Golkar and de Oliveira (2026) that challenge their conclusions regarding OpenET biases and machine learning bias correction.
- To demonstrate how errors in processing observational reference data can fundamentally alter the evaluation of satellite evapotranspiration (ET) performance.
Study Configuration
- Spatial Scale: 22 AmeriFlux sites across the United States.
- Temporal Scale: Daily resolution over unspecified extended periods (data from a prior study).
Methodology and Data
- Models used: N/A (This paper is a comment focused on evaluating the data processing of a prior study's reference observations, not on running predictive models.)
- Data sources: Original AmeriFlux datasets.
Main Results
- Unusually low daily mean observed ET and frequent exact zero-valued observations were identified in the reference data used by Golkar and de Oliveira (2026).
- Independent reconstruction of daily ET from original AmeriFlux datasets reproduced the reported site-mean ET within 0.04 mm d⁻¹ at 20 of 22 sites when missing observations were treated as zero.
- The reported ET statistics by Golkar and de Oliveira (2026) are reproducible only when omitting the energy balance correction described in their analyses.
- The apparent treatment of missing EC observations as zero and the omission of the stated energy balance correction substantially inflated apparent OpenET overestimation.
- The conclusions of Golkar and de Oliveira (2026) regarding seasonal and climate-dependent correction mechanisms likely reflect artifacts in the reference observations rather than genuine deficiencies in OpenET.
Contributions
- Provides compelling quantitative evidence that missing eddy covariance observations were effectively incorporated as zero ET in a prior study, systematically lowering reference measurements.
- Demonstrates that errors in processing observational reference data (e.g., treating missing data as zero, omitting stated corrections) can fundamentally alter the evaluation of satellite ET performance.
- Underscores the critical importance of rigorous quality control and reproducible analysis in distinguishing genuine model biases from observational artifacts.
- Challenges the reported effectiveness of a machine learning bias correction due to issues in the reference data.
Funding
N/A
Citation
@article{Kim2026Comment,
author = {Kim, Yeonuk and Majumdar, Sayantan and Volk, John and ReVelle, Peter and Caldwell, Todd G. and Huntington, Justin},
title = {Comment on Golkar and de Oliveira (2026): Missing eddy covariance observations undermine reported OpenET biases},
journal = {Agricultural and Forest Meteorology},
year = {2026},
doi = {10.1016/j.agrformet.2026.111586},
url = {https://doi.org/10.1016/j.agrformet.2026.111586}
}
Original Source: https://doi.org/10.1016/j.agrformet.2026.111586