The stakes of this measurement problem are enormous. Terrestrial ecosystems absorb roughly one quarter of the carbon dioxide released by human activities, a service that significantly slows the rise of atmospheric CO2 and the pace of global warming. NEE, defined as the difference between the CO2 released by ecosystem respiration and the CO2 sequestered by photosynthesis, is the quantity that tells us whether a landscape is a net carbon sink or source. Yet NEE can only be measured directly at the plot scale, using eddy covariance towers that monitor the vertical exchange of CO2 between vegetation and air. Everything beyond the tower footprint must be estimated, and that is where the disagreements begin.

Scientists have traditionally taken two routes to continental-scale carbon budgets. Top-down atmospheric inversions combine measurements of atmospheric CO2 concentrations with transport models to infer surface fluxes, offering an independent perspective but at coarse spatial resolutions limited by sparse sampling. Bottom-up approaches use process-based terrestrial biosphere models or data-driven machine learning to extrapolate tower observations across landscapes. The trouble is that bottom-up estimates have often diverged dramatically from inversions, partly because conventional machine learning treats each time step independently and fails to capture the complex, nonlinear ways ecosystems respond to disturbances and climate extremes. Ecosystems remember: a drought can suppress carbon uptake for months or years after the rains return, and most models simply do not carry that memory.

The LSTM architecture is what sets MemoryFlux apart. Developed originally to solve long-term dependency problems in sequence modelling, the network uses a six-month look-back window of historical climate and vegetation data to predict current carbon exchange, explicitly encoding the legacy effects of antecedent conditions. The team trained separate ecosystem-specific models for ten major plant functional types, from evergreen needleleaf forests to permanent wetlands, using tower-measured meteorology combined with a rich suite of satellite observations. Those predictors included the normalized difference vegetation index, leaf area index, the fraction of absorbed photosynthetically active radiation, and solar-induced chlorophyll fluorescence, a direct signal of photosynthetic activity, alongside ERA5-Land climate reanalysis variables such as temperature, radiation, precipitation, soil water content, and vapour pressure deficit.

The validation results are striking. MemoryFlux’s seasonal cycle of NEE correlated at r=0.96 with the ensemble mean of the OCO-2 v10 Model Intercomparison Project inversion and at r=0.97 with CarbonTracker2022, both top-down estimates derived from fundamentally different methodology. Perhaps more importantly, the mean annual carbon sink of −1.27 ± 0.12 petagrams of carbon per year sits far closer to the inversion range of −0.83 to −0.70 petagrams than existing machine-learning upscaling products, which have suggested sinks as large as −3.30 petagrams per year. That gap between bottom-up and top-down estimates has long been one of the most stubborn discrepancies in regional carbon cycle science, and narrowing it builds confidence in both approaches.

One of the dataset’s most celebrated achievements is its reproduction of the Midwest Corn Belt phenomenon. Atmospheric inversions have consistently shown that the croplands of the Upper Midwest pull in enormous amounts of carbon during the peak growing season in July and August, a continental-scale signal recognized as a model benchmark since the early days of the CarbonTracker project. Yet most existing global flux upscaling datasets fail to capture it, instead placing their strongest seasonal uptake in the southeastern United States. MemoryFlux correctly identifies the Corn Belt as the dominant carbon-uptake region during the peak growing season, matching the empirical evidence and the inversion picture, even though the uptake is transitory because most of the absorbed carbon ends up in harvested crops consumed by humans and livestock.

The spatial patterns across the continent also align with ecological understanding. The strongest annual carbon uptake appears in the humid southeastern United States, followed by the northern Pacific Coast and tropical regions, while major carbon sources concentrate in the tundra of northern and central Canada and the arid central United States, where warming could release soil carbon and accelerate climate change. Forests and wooded ecosystems emerge as the continent’s dominant carbon sink, contributing a mean uptake of −0.76 petagrams of carbon per year, with MemoryFlux attributing more than 85 percent of the total sink to forests, a share higher than most other products suggest.

To test whether the memory mechanism actually matters, the researchers ran a controlled experiment. They built a parallel dataset, dubbed nonMemoryFlux, using the same predictors, architecture, and validation scheme but feeding the model only contemporaneous variables, with no look-back window. The differences were substantial. At the plant functional type level, the coefficient of determination rose from 0.63 to 0.79 when historical information was included, and error metrics dropped accordingly. The memory-free version also estimated stronger carbon uptake in recent years, precisely the period when the 2015/16 El Niño and repeated droughts and heatwaves struck North America, suggesting it failed to capture the lingering legacy of those extremes. The gap between the two versions reached 186 teragrams of carbon in a single year, about 14.6 percent of the mean continental uptake.

MemoryFlux also proved its worth during climate disasters. The researchers examined six major events, including the 2011 drought in Mexico and the southern United States, the 2012 North American drought, the 2013 California drought, the 2017 flash drought across the northern plains, the 2019 Midwest floods, and the southwestern drought and wildfires of 2020 to 2021. In each case, the dataset’s spatial anomaly patterns coincided with negative soil moisture and fluorescence anomalies and with the documented extent of the events, showing anomalous carbon release where ecosystems were stressed. The memory-equipped version represented these regional anomaly patterns more clearly than its memory-free counterpart, particularly for the 2011 and 2012 droughts and the 2019 flood.

The authors are careful about limitations. Flux tower coverage remains uneven, with some ecosystem types represented by only two sites, and the dataset lacks a fully spatially explicit uncertainty framework. A mismatch between tower-scale training data and gridded reanalysis predictors introduces additional uncertainty, and the static land-cover map cannot track real transitions such as cropland expansion or fire-induced vegetation change. Differences with inversions persist in the managed Midwest croplands, hinting that human management information could further improve the models. Still, with the dataset and code openly available, MemoryFlux offers climate scientists a powerful new lens on how a continent’s ecosystems respond to a rapidly changing climate, and a template for bringing memory into the next generation of carbon cycle models worldwide.

Subject of Research: LSTM-based upscaling of net ecosystem CO2 exchange across North America from flux tower and satellite observations

Article Title: Improved estimation of net ecosystem CO2 exchange over North America using LSTM-based flux upscaling (2001–2021)

Article References: Improved estimation of net ecosystem CO2 exchange over North America using LSTM-based flux upscaling (2001–2021). (n.d.). https://doi.org/10.5194/essd-18-7367-2026

Image Credits: AI Generated

DOI: 10.5194/essd-18-7367-2026

Keywords: carbon cycle, net ecosystem exchange, LSTM, deep learning, MemoryFlux, eddy covariance, atmospheric inversion, drought, Corn Belt, remote sensing, North America, carbon sink

The research team, led by Charuni Jayasekara of RMIT University together with colleagues at the University of Melbourne, Queensland University of Technology, the Queensland Herbarium and La Trobe University, conducted a systematic quantitative literature review following established PRISMA-style protocols. From an initial pool of 341 articles identified through databases including Web of Science, Scopus and Google Scholar, the authors applied strict inclusion criteria: only field-based studies reporting annual net ecosystem exchange values, meaning full-year measurements or repeated seasonal measurements extrapolated to a complete year, qualified for analysis. Plot-scale studies smaller than ten square meters, laboratory experiments and short-term snapshots were excluded to ensure that every data point represented the carbon balance of a functioning ecosystem at policy-relevant temporal scales. The final dataset of 58 articles and 120 sites spans both hemispheres, every continent except Antarctica, and measurement records ranging from a single year to an exceptional 17-year series.

The quantitative picture that emerges is one of dramatic divergence between land uses. Peatlands under their original, undrained land use were, on average, net carbon sinks absorbing about 30 grams of carbon per square meter per year, while agricultural peatlands emitted a mean of 490 grams of carbon per square meter per year, a statistically significant difference. Among agricultural categories, tropical plantations were the worst offenders, averaging 1,132 grams of carbon emitted per square meter annually, with the single highest recorded emission of 3,758 grams coming from an oil palm plantation established on drained tropical peat. Pastures and croplands each emitted roughly 420 grams per square meter per year. Peat extraction sites were also net sources, though less extreme, at around 137 grams. The underlying mechanism is well understood: drainage lowers the water table, exposing previously saturated peat to oxygen and accelerating aerobic microbial decomposition of organic matter that took centuries to accumulate, while the removal of natural vegetation simultaneously eliminates the photosynthetic uptake that once offset respiration.

Peatland condition told a parallel story. Degraded sites, about 55 percent of the dataset, averaged net emissions of 306 grams of carbon per square meter per year, significantly higher than intact sites, which averaged a net uptake of 124 grams. But the most consequential finding concerns restoration. Of the 18 restored sites in the analysis, the category as a whole remained a net carbon source, averaging 110 grams of emissions per square meter per year, and 39 percent of individual restored sites were still emitting carbon dioxide at the time of measurement. The authors attribute this to insufficient recovery time, incomplete hydrological recovery, legacy effects of severe degradation, and in some cases restoration methods that prioritize biodiversity or hydrological objectives over immediate carbon benefits. Severely degraded sites that have lost substantial peat may require many decades to regain sink function, and some may never return to it.

The single most powerful predictor of carbon exchange across all peatland types and climatic zones was water table depth. Across 98 sites with paired hydrological data, the regression analysis revealed that every centimeter of water table drawdown increases carbon dioxide emissions by approximately 7.2 grams of carbon per square meter per year. The relationship explained about 19 percent of the variance in net ecosystem exchange and, critically, identified a transition point: peatlands shift from net sink to net source when the water table drops to roughly 7.8 centimeters below the peat surface. This figure aligns closely with hydrological tipping points reported in other experimental studies, which place the threshold between about 14 and 24 centimeters depending on peatland type. The finding carries direct policy weight, because it converts an abstract ecological principle into a concrete management target: keep the water table within a few centimeters of the surface, and the peatland keeps functioning as a carbon bank; drain it, and the stored carbon begins flowing into the atmosphere.

Climatic geography added further nuance. Tropical peatlands were the strongest mean net sources, emitting on average 270 grams of carbon per square meter per year, and showed by far the greatest variability, a reflection of the intense land-use pressure they face, particularly in Southeast Asia where peat swamp forests are being rapidly converted to plantations. Alpine and sub-polar peatlands, by contrast, were the strongest sinks, absorbing roughly 34 to 37 grams per square meter per year. Temperate sites, which dominated the dataset at 77 percent of all locations, averaged as net sources overall, largely because so many European temperate peatlands have been drained for agriculture. Vegetation type mattered independently of land use: native shrub and grass systems and moss-dominated sites were mean sinks, while tree-dominated natural sites averaged as sources, and every agricultural vegetation category was a net emitter.

Perhaps the most sobering pattern in the data concerns the reliability of simple classification schemes. Nineteen percent of sites classified as intact were actually net carbon sources, while 21 percent of degraded sites were net sinks. This means that categorical labels based on land-use history or visual assessment cannot reliably capture the functional carbon status of a peatland. The authors point to several explanations: natural interannual variability means one or two years of measurement may misrepresent long-term trajectories; cryptic degradation through nitrogen deposition, invasive species or subtle hydrological shifts can erode sink function without visible structural change; and in some counterintuitive cases, vegetation changes following disturbance, such as dense tree establishment on a drained bog, can increase photosynthetic uptake enough to maintain or even strengthen net carbon sequestration despite lowered water tables.

The review also exposed a glaring geographic bias in the science itself. Eighty-six percent of the 120 sites lie in the Northern Hemisphere, and 60 percent are in Europe, with Germany, Denmark, Ireland and Latvia alone hosting dozens of studies. Only 14 percent of sites are in the Southern Hemisphere. Africa and South America contributed just three sites each, and Oceania six, despite these regions containing substantial, carbon-dense and highly threatened peatlands. The authors attribute the gap to technical, financial and political constraints on establishing long-term flux monitoring infrastructure, seasonal logistics in cold climates, and the likelihood that research from underrepresented regions exists in non-English or gray literature inaccessible to global syntheses. This underrepresentation is particularly troubling for tropical peatlands, which hold the largest share of global peatland carbon, accumulate carbon faster than their boreal counterparts, at rates of 59 to 145 grams per square meter per year, and face the fastest rates of conversion for agriculture and forestry.

Methodological fragmentation compounds the problem. The review identified 22 different software programs used to process carbon flux data, 15 for eddy covariance measurements alone, alongside fundamentally different measurement approaches: eddy covariance systems integrate fluxes over footprints of 100 to 1,000 square meters continuously, while chamber methods sample patches of less than one square meter and require spatial upscaling and temporal interpolation to estimate annual budgets. Different gap-filling algorithms, quality control protocols and chamber placement strategies introduce inconsistencies that may partly explain the enormous variability observed in restored peatland fluxes, which ranged from a sink of 397 to emissions of 1,170 grams of carbon per square meter per year. The authors argue that methodological standardization, combined with longer monitoring periods, since 43 of the 120 sites reported only a single year of data, is essential if net ecosystem exchange is to serve as a dependable restoration monitoring tool.

The implications extend from national greenhouse gas inventories to the growing portfolio of nature-based climate solutions. With approximately 65 million hectares, 16 percent of the world’s peatlands, already drained, and degraded peatlands responsible for roughly three percent of anthropogenic greenhouse gas emissions in 2020, the evidence assembled here supports a clear hierarchy of action: protect intact peatlands before they are lost, rewet drained ones with water table targets anchored to reference conditions from functioning sink systems, and commit to long-term monitoring and adaptive management rather than expecting rapid carbon payback from restoration. The authors also caution that raising water tables promotes methane production, meaning that the net climate benefit of any restoration project should be judged against the full greenhouse gas balance, not carbon dioxide alone. What the synthesis makes unmistakable is that the fate of hundreds of gigatonnes of soil carbon now hinges on a hydrological variable measurable with a ruler: centimeters of water above or below the peat surface.

Subject of Research: Global patterns and drivers of net ecosystem exchange in peatlands across land use and environmental gradients

Article Title: Global Patterns of Net Ecosystem Exchange in peatlands: A Systematic Review and Meta-analysis of Drivers Across Land Use and Environmental Gradients

Article References: Jayasekara, C., Leigh, C., Shimeta, J., Silvester, E., & Grover, S. (2026). Global Patterns of Net Ecosystem Exchange in peatlands: A Systematic Review and Meta-analysis of Drivers Across Land Use and Environmental Gradients. Environmental Management, 76(9), Article 310. https://doi.org/10.1007/s00267-026-02621-y

Image Credits: AI Generated

DOI: 10.1007/s00267-026-02621-y

Keywords: peatlands, net ecosystem exchange, carbon cycle, water table depth, peatland restoration, land use change, CO2 emissions, carbon sink, drainage, tropical peatlands, meta-analysis, nature-based climate solutions