[1] We upscaled FLUXNET observations of carbon dioxide, water, and energy fluxes to the global scale using the machine learning technique, model tree ensembles (MTE). We trained MTE to predict site-level gross primary productivity (GPP), terrestrial ecosystem respiration (TER), net ecosystem exchange (NEE), latent energy (LE), and sensible heat (H) based on remote sensing indices, climate and meteorological data, and information on land use. We applied the trained MTEs to generate global flux fields at a 0.5°× 0.5°spatial resolution and a monthly temporal resolution from 1982 to 2008. Cross-validation analyses revealed good performance of MTE in predicting among-site flux variability with modeling efficiencies (MEf) between 0.64 and 0.84, except for NEE (MEf = 0.32). Performance was also good for predicting seasonal patterns (MEf between 0.84 and 0.89, except for NEE (0.64)). By comparison, predictions of monthly anomalies were not as strong (MEf between 0.29 and 0.52). Improved accounting of disturbance and lagged environmental effects, along with improved characterization of errors in the training data set, would contribute most to further reducing uncertainties. Our global estimates of LE (158 ± 7 J × 10 18 yr −1 ), H (164 ± 15 J × 10 18 yr −1), and GPP (119 ± 6 Pg C yr ) were similar to independent estimates. Our global TER estimate (96 ± 6 Pg C yr −1 ) was likely underestimated by 5-10%. Hot spot regions of interannual variability in carbon fluxes occurred in semiarid to semihumid regions and were controlled by moisture supply. Overall, GPP was more important to interannual variability in NEE than TER. Our empirically derived fluxes may be used for calibration and evaluation of land surface process models and for exploratory and diagnostic assessments of the biosphere.

The FLUXNET2015 dataset provides ecosystem-scale data on CO 2 , water, and energy exchange between the biosphere and the atmosphere, and other meteorological and biological measurements, from 212 sites around the globe (over 1500 site-years, up to and including year 2014). These sites, independently managed and operated, voluntarily contributed their data to create global datasets. Data were quality controlled and processed using uniform methods, to improve consistency and intercomparability across sites. The dataset is already being used in a number of applications, including ecophysiology studies, remote sensing studies, and development of ecosystem and Earth system models. FLUXNET2015 includes derived-data products, such as gap-filled time series, ecosystem respiration and photosynthetic uptake estimates, estimation of uncertainties, and metadata about the measurements, presented for the first time in this paper. In addition, 206 of these sites are for the first time distributed under a Creative Commons (CC-BY 4.0) license. This paper details this enhanced dataset and the processing methods, now made available as open-source codes, making the dataset more accessible, transparent, and reproducible.

There were systematic differences in performance among the fluxes, with the following ascending order: net ecosystem exchange (R 2 < 0.5), ecosystem respiration (R 2 > 0.6), gross primary production (R 2 > 0.7), latent heat (R 2 > 0.7), sensible heat (R 2 > 0.7), and net radiation (R 2 > 0.8). The ML methods predicted the across-site variability and the mean seasonal cycle of the observed fluxes very well (R 2 > 0.7), while the 8-day deviations from the mean seasonal cycle were not well predicted (R 2 < 0.5). Fluxes were better predicted at forested and temperate climate sites than at sites in extreme climates or less represented by training data (e.g., the tropics). The evaluated large ensemble of ML-based models will be the basis of new global flux products.Published by Copernicus Publications on behalf of the European Geosciences Union.

The energy balance at most surface-atmosphere flux research sites remains unclosed. The mechanisms underlying the discrepancy between measured energy inputs and outputs across the global FLUXNET tower network are still under debate. Recent reviews have identified exchange processes and turbulent motions at large spatial and temporal scales in heterogeneous landscapes as the primary cause of the lack of energy balance closure at some intensively-researched sites, while unmeasured storage terms cannot be ruled out as a dominant contributor to the lack of energy balance closure at many other sites. We analyzed energy balance closure across 173 ecosystems in the FLUXNET database and explored the relationship between energy balance closure and landscape heterogeneity using MODIS products and GLOBEstat elevation data. Energy balance closure per research site (CEB,s) averaged 0.84 ± 0.20, with best average closures in evergreen broadleaf forests and savannas (0.91–0.94) and worst average closures in crops, deciduous broadleaf forests, mixed forests and wetlands (0.70–0.78). Half-hourly or hourly energy balance closure on a percent basis increased with friction velocity (u*) and was highest on average under near-neutral atmospheric conditions. CEB,s was significantly related to mean precipitation, gross primary productivity and landscape-level enhanced vegetation index (EVI) from MODIS, and the variability in elevation, MODIS plant functional type, and MODIS EVI. A linear model including landscape-level variability in both EVI and elevation, mean precipitation, and an interaction term between EVI variability and precipitation had the lowest Akaike’s information criterion value. CEB,s in landscapes with uniform plant functional type approached 0.9 and CEB,s in landscapes with uniform EVI approached 1. These results suggest that landscape-level heterogeneity in vegetation and topography cannot be ignored as a contributor to incomplete energy balance closure at the flux network level, although net radiation measurements, biological energy assimilation, unmeasured storage terms, and the importance of good practice including site selection when making flux measurements should not be discounted. Our results suggest that future research should focus on the quantitative mechanistic relationships between energy balance closure and landscape-scale heterogeneity, and the consequences of mesoscale circulations for surface-atmosphere exchange measurements

&lt;p&gt;&lt;strong&gt;Abstract.&lt;/strong&gt; Spatial-temporal fields of land-atmosphere fluxes derived from data-driven models can complement simulations by process-based Land Surface Models. While a number of strategies for empirical models with eddy covariance flux data have been applied, a systematic intercomparison of these methods has been missing so far. In this study, we perform a cross-validation experiment for predicting carbon dioxide (CO&lt;sub&gt;2&lt;/sub&gt;), latent heat, sensible heat and net radiation fluxes, in different ecosystem types with eleven machine learning (ML) methods from four different classes (kernel methods, neural network, tree methods, and regression splines). We employ two complementary setups: (1) eight days average fluxes based on remotely sensed data, and (2) daily mean fluxes based on meteorological data and mean seasonal cycle of remotely sensed variables. The pattern of predictions from different ML and setups were very consistent. There were systematic differences in performance among the fluxes, with the following ascending order: net ecosystem exchange (R&lt;sub&gt;2&lt;/sub&gt; &lt; 0.5), ecosystem respiration (R&lt;sub&gt;2&lt;/sub&gt; &gt; 0.6), gross primary production (R&lt;sub&gt;2&lt;/sub&gt; &gt; 0.7), latent heat (R&lt;sub&gt;2&lt;/sub&gt; &gt; 0.7), sensible heat (R&lt;sub&gt;2&lt;/sub&gt; &gt; 0.7), net radiation (R&lt;sub&gt;2&lt;/sub&gt; &gt; 0.8). ML methods predicted very well the across sites variability and the seasonal cycle (R&lt;sub&gt;2&lt;/sub&gt; &gt; 0.7) of the observed fluxes, while the eight days deviations from the mean seasonal cycle were not well predicted (R&lt;sub&gt;2&lt;/sub&gt; &lt; 0.5). Fluxes were better predicted at forested and temperate climate sites than at ones growing in extreme climates or less representated in training data (e.g. the tropics). The large ensemble of ML based models evaluated will be the basis of new global flux products.&lt;/p&gt;

Terrestrial gross primary productivity (GPP) varies greatly over time and space. A better understanding of this variability is necessary for more accurate predictions of the future climate-carbon cycle feedback. Recent studies have suggested that variability in GPP is driven by a broad range of biotic and abiotic factors operating mainly through changes in vegetation phenology and physiological processes. However, it is still unclear how plant phenology and physiology can be integrated to explain the spatiotemporal variability of terrestrial GPP. Based on analyses of eddy-covariance and satellite-derived data, we decomposed annual terrestrial GPP into the length of the CO 2 uptake period (CUP) and the seasonal maximal capacity of CO 2 uptake (GPP max ). The product of CUP and GPP max explained >90% of the temporal GPP variability in most areas of North America during 2000-2010 and the spatial GPP variation among globally distributed eddy flux tower sites. It also explained GPP response to the European heatwave in 2003 (r 2 = 0.90) and GPP recovery after a fire disturbance in South Dakota (r 2 = 0.88). Additional analysis of the eddy-covariance flux data shows that the interbiome variation in annual GPP is better explained by that in GPP max than CUP. These findings indicate that terrestrial GPP is jointly controlled by ecosystem-level plant phenology and photosynthetic capacity, and greater understanding of GPP max and CUP responses to environmental and biological variations will, thus, improve predictions of GPP over time and space. ecosystem carbon uptake | growing season length | photosynthetic capacity | spatiotemporal variability | climate extreme L arge variability exists among estimates of terrestrial carbon sequestration, resulting in substantial uncertainty in modeled dynamics of atmospheric CO 2 concentration and predicted future climate change (1). The variability in carbon sequestration is partially caused by variation in terrestrial gross primary productivity (GPP) (2), which is the cumulative rate over time of gross plant Significance Terrestrial gross primary productivity (GPP), the total photosynthetic CO 2 fixation at ecosystem level, fuels all life on land. However, its spatiotemporal variability is poorly understood, because GPP is determined by many processes related to plant phenology and physiological activities. In this study, we find that plant phenological and physiological properties can be integrated in a robust index-the product of the length of CO 2 uptake period and the seasonal maximal photosynthesis-to explain the GPP variability over space and time in response to climate extremes and during recovery after disturbance.

Abstract. This paper, developed under the framework of the RECCAP initiative, aims at providing improved estimates of the carbon and GHG (CO 2 , CH 4 and N 2 O) balance of continental Africa. The various components and processes of the African carbon and GHG budget are considered, existing data reviewed, and new data from different methodologies (inventories, ecosystem flux measurements, models, and atmospheric inversions) presented. Uncertainties are quantified and current gaps and weaknesses in knowledge and monitoring systems described in order to guide future requirements. The majority of results agree that Africa is a small sink of carbon on an annual scale, with an average value of Published by Copernicus Publications on behalf of the European Geosciences Union. R. Valentini et al.: A full greenhouse gases budget of Africa−0.61 ± 0.58 Pg C yr −1 . Nevertheless, the emissions of CH 4 and N 2 O may turn Africa into a net source of radiative forcing in CO 2 equivalent terms. At sub-regional level, there is significant spatial variability in both sources and sinks, due to the diversity of biomes represented and differences in the degree of anthropic impacts. Southern Africa is the main source region; while central Africa, with its evergreen tropical forests, is the main sink. Emissions from land-use change in Africa are significant (around 0.32 ± 0.05 Pg C yr −1 ), even higher than the fossil fuel emissions: this is a unique feature among all the continents. There could be significant carbon losses from forest land even without deforestation, resulting from the impact of selective logging. Fires play a significant role in the African carbon cycle, with 1.03 ± 0.22 Pg C yr −1 of carbon emissions, and 90 % originating in savannas and dry woodlands. A large portion of the wild fire emissions are compensated by CO 2 uptake during the growing season, but an uncertain fraction of the emission from wood harvested for domestic use is not. Most of these fluxes have large interannual variability, on the order of ±0.5 Pg C yr −1 in standard deviation, accounting for around 25 % of the year-toyear variation in the global carbon budget.Despite the high uncertainty, the estimates provided in this paper show the important role that Africa plays in the global carbon cycle, both in terms of absolute contribution, and as a key source of interannual variability.

No abstract