Abstract
<strong class="journal-contentHeaderColor">Abstract.</strong> Gap-filling eddy covariance <span class="inline-formula">CO<sub>2</sub></span> fluxes is challenging at dryland sites due to small <span class="inline-formula">CO<sub>2</sub></span> fluxes. Here, four machine learning (ML) algorithms including artificial neural network (ANN), <span class="inline-formula"><i>k</i></span>-nearest neighbors (KNNs), random forest (RF), and support vector machine (SVM) are employed and evaluated for gap-filling <span class="inline-formula">CO<sub>2</sub></span> fluxes over a semiarid sagebrush ecosystem with different lengths of artificial gaps. The ANN and RF algorithms outperform the KNN and SVM in filling gaps ranging from hours to days, with the RF being more time efficient than the ANN. Performances of the ANN and RF are largely degraded for extremely long gaps of 2 months. In addition, our results suggest that there is no need to fill the daytime and nighttime net ecosystem exchange (NEE) gaps separately when using the ANN and RF. With the ANN and RF, the gap-filling-induced uncertainties in the annual NEE at this site are estimated to be within 16âgâCâm<span class="inline-formula"><sup>â2</sup></span>, whereas the uncertainties by the KNN and SVM can be as large as 27âgâCâm<span class="inline-formula"><sup>â2</sup></span>. To better fill extremely long gaps of a few months, we test a two-layer gap-filling framework based on the RF. With this framework, the model performance is improved significantly, especially for the nighttime data. Therefore, this approach provides an alternative in filling extremely long gaps to characterize annual carbon budgets and interannual variability in dryland ecosystems.
Highlights
The eddy covariance (EC) technique has been widely applied for monitoring energy and water fluxes as well as net ecosystem exchanges (NEEs) of carbon dioxide and other trace gases between land and the atmosphere (Baldocchi, 2003; Oncley et al, 2007; Stoy et al, 2013)
We evaluate the performance of four commonly used machine learning (ML) algorithms (ANN, k-nearest neighbors (KNNs), random forest (RF), and support vector machine (SVM)) in filling the extremely long gaps in the NEE data collected at an EC site over a semiarid sagebrush ecosystem in central Washington, USA, from 2016 to 2019, and we assess the uncertainties in the annual NEE introduced by gap-filling methods
The performance of the four ML algorithms in filling the NEE data gaps is evaluated at a semiarid sagebrush ecosystem site
Summary
The eddy covariance (EC) technique has been widely applied for monitoring energy and water fluxes as well as net ecosystem exchanges (NEEs) of carbon dioxide and other trace gases between land and the atmosphere (Baldocchi, 2003; Oncley et al, 2007; Stoy et al, 2013). Gap-filling usually accounts for one large source of uncertainties in the annual NEE (Soloway et al, 2017), together with other sources of uncertainties such as measurement errors and bias related to non-closure of the surface energy balance (Gao et al, 2019; Wilson et al, 2002). Robust NEE gap-filling approaches are critical for quantifying the annual and interannual variability of carbon budgets Yao et al.: Technical note: Uncertainties in eddy covariance CO2 fluxes
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
