Datensatz

Zusammenfassung

Natural wetlands constitute the largest and most uncertain source of methane (CH4) to the atmosphere and a large
fraction of them are in the northern latitudes. These emissions are typically estimated using process (bottom-up) or inversion
(top-down) models, yet the two are not independent of each other since the top-down estimates rely on the a priori estimation
of these emissions coming from the process models. Hence, independent validation data of the large-scale emissions would be
10 needed.
Here we utilize random forest (RF) machine learning technique to upscale CH4 eddy covariance flux measurements from 25
sites to estimate CH4 wetland emissions from the northern latitudes (north of 45 °N) during years 2013 and 2014. The predictive
performance of the RF model is evaluated using the leave-one-site-out cross-validation scheme and the performance (Nash-
Sutcliffe model efficiency = 0.47) is comparable to previous studies upscaling net ecosystem exchange of carbon dioxide or
15 studies where process models are compared against site-level CH4 emission data. Three wetland maps are utilized in the
upscaling and the annual emissions for the northern wetlands yield 31.7 (22.3-41.2, 95 % confidence interval), 30.6 (21.4-
39.9) or 37.6 (25.9-49.5) Tg(CH4) yr-1, depending on the map used. To evaluate the uncertainties of the upscaled product it is
also compared against two process models (LPX-Bern and WetCHARTs) and methodological issues related to CH4 flux
upscaling are discussed. The monthly upscaled CH4 flux data product is available for further usage at: https://doi.org/
20 10.5281/zenodo.2560164.