Upscaling terrestrial carbon dioxide fluxes in Alaska with satellite remote sensing and support vector regression

Abstract

[1] Carbon dioxide (CO 2 ) fluxes from a network of 21 eddy covariance towers were upscaled to estimate the Alaskan CO 2 budget from 2000 to 2011 by combining satellite remote sensing data, disturbance information, and a support vector regression model. Data were compared with the CO 2 budget from an inverse model (CarbonTracker). Observed gross primary productivity (GPP), ecosystem respiration (RE), and net ecosystem exchange (NEE) were each well reproduced by the model on the site scale; root-mean-square errors (RMSEs) for GPP, RE, and NEE were 0.52, 0.23, and 0.48 g C m À2 d À1 , respectively. Landcover classification was the most important input for predicting GPP, whereas visible reflectance index of green ratio was the most important input for predicting RE. During the period of 2000-2011, predicted GPP and RE were 369 ± 22 and 362 ± 12 Tg C yr À1 (mean ± interannual variability) for Alaska, respectively, indicating an approximately neutral CO 2 budget for the decade. CarbonTracker also showed an approximately neutral CO 2 budget during 2000-2011 (growing season RMSE = 14 g C m À2 season À1 ; annual RMSE = 13 g C m À2 yr À1 ). Interannual CO 2 flux variability was positively correlated with air temperature anomalies from June to August, with Alaska acting as a greater CO 2 sink in warmer years. CO 2 flux trends for the decade were clear in disturbed ecosystems; positive trends in GPP and CO 2 sink were observed in areas where vegetation recovered for about 20 years after fire.

Upscaling CH4 Fluxes Using High-Resolution Imagery in Arctic Tundra Ecosystems

“…Eddy covariance (EC) and chambers are the most common methods to measure CH 4 fluxes in the field, and both give useful information [5,9,10]), but can also result in inaccurate emission predictions [29,30]. EC towers measure trace gas fluxes with a high temporal resolution across a footprint with radii of roughly 100-300 m for a tower 1-3 m tall [31].…”

Mapping Forest Health Using Spectral and Textural Information Extracted from SPOT-5 Satellite Images

“…The most frequently used approaches in forestry include regression and decision trees [102,103], artificial neutral networks [70,104] random forests [105][106][107], and support vectors [108,109]. The size of the training data set for machine learning greatly influences the stability and accuracy of the trained model [110].…”

Assessing the dynamics of vegetation productivity in circumpolar regions with different satellite indicators of greenness and photosynthesis

“…Heterogeneous respiration and carbon emissions to the atmosphere are stimulated in warmer soils at lowered water table depth (Billings et al, 1982;Yi et al, 2014;Oechel et al, 1993;Huemmrich et al, 2010b; phenological observations, spectral reflectance as well as gas flux measurements are conducted in-situ, both under natural conditions and in manipulative experiments. Many studies can be found in the literature that evaluate eddy-covariance or chamber gas flux measurements with respect to spatial patterns of NEE at a fixed point in time, or in-situ NEE integrated over the growing season and its variations between years (López-Blanco et al, 2017;Lund et al, 2010;Ueyama et al, 2013a;McFadden et al, 2003;Williams and Rastetter, 1999;Marushchak et al, 2013;Kross et al, 2014). However, only few sites exist com-10 pared to temperate regions and observations are usually not done in a continuous manner over the complete year but during individual measurement campaigns or dedicated periods during the growing season.…”

“…Further, the heterogeneity of the landscape poses limits to the spatial representativeness of the relationships between the carbon fluxes and meteorological and soil conditions 20 that have been identified in-situ (Pirk et al, 2017;Tuovinen et al, 2018). Therefore, in spatial up-scaling exercises (Ueyama et al, 2013a;Marushchak et al, 2013;Tramontana et al, 2016) strong extrapolations are necessary.…”