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 Priscila Belen Cano of Purdue University’s Department of Agronomy together with Gustavo Angel Maddonni of the Universidad de Buenos Aires and Ignacio Antonio Ciampitti, developed what they call a yield-based environmental regionalization framework. Rather than starting with climate maps or soil surveys, the researchers began with the crop itself. They compiled median annual maize yields for each of Indiana’s 92 counties from 2000 to 2024, drawing on insured production records from the United States Department of Agriculture Risk Management Agency. Because long-term yield integrates the cumulative effects of climate, soils, topography, and management, spatially persistent differences in productivity can be read as realized environmental differences, the authors explain.

To transform those yield patterns into coherent geographic zones, the team applied a fuzzy c-means clustering algorithm, a machine learning technique that assigns each county a degree of membership in multiple clusters rather than forcing hard boundaries. The feature matrix combined each county’s median yield with the latitude and longitude of its geographic centroid, with the coordinates weighted by a spatial factor of 1.5 to encourage geographic contiguity while preserving yield homogeneity. Selecting the optimal number of clusters required a multi-criteria evaluation using the Xie–Beni index, the Fukuyama–Sugeno index, and the Fuzzy Partition Coefficient. Because these indices favored different solutions, the researchers settled on five clusters as a balanced compromise between partition quality, model complexity, and agronomic interpretability. The result was five yield-based environmental regions spanning the state: northwest, northeast, west-central, southwest, and southeast.

The regionalization exposed a striking productivity gradient. The west-central region emerged as Indiana’s maize powerhouse, with median yields of approximately 11.2 megagrams per hectare, while the northwest region lagged at roughly 8.5 megagrams per hectare, accompanied by the greatest spatial dispersion in county yields, with a standard deviation of 2.6 megagrams per hectare compared to less than 0.7 in the other regions. The northeast, southwest, and southeast occupied intermediate positions with relatively similar median yields. Crucially, climate variables were deliberately excluded from the clustering procedure so they could later serve as an independent test of whether the yield-defined zones corresponded to genuinely distinct climatic environments.

That test came in the form of a principal component analysis of long-term climatic means, built from the Daymet gridded meteorological dataset at one-kilometer resolution. The first two principal components together explained nearly 84 percent of the total variance, and the dominant variables separating the regions were vapor pressure deficit, reference evapotranspiration, and a heat-stress index capturing cumulative degrees of daily maximum temperatures above 35 degrees Celsius. The first component discriminated between environments with longer growing seasons, higher atmospheric demand, and greater heat stress, exemplified by the southwest, and cooler environments with shorter frost-free periods and more grain-filling precipitation, exemplified by the northeast and northwest. The second component captured contrasts in diurnal temperature amplitude and solar radiation, particularly during the critical period, cleanly separating the northeast from the northwest and the southwest from the southeast.

Yields rose everywhere over the study period, but not at the same relative pace. Absolute gains ranged narrowly from 0.172 to 0.185 megagrams per hectare per year, yet in relative terms the lower-yielding northwest improved at about 2 percent per year while the high-yielding west-central region gained about 1.6 percent annually. This pattern suggests a tendency toward relative convergence in maize performance across the state, even though absolute yield gaps between regions may persist. Interannual variability also differed, with coefficients of variation around 16 to 18 percent in the higher-yielding regions and approximately 21 percent in the southern zones, hinting that southern Indiana farmers face a rockier road to consistent harvests.

To isolate the climatic drivers of year-to-year swings, the researchers first removed long-term technological trends by fitting linear regressions of yield against year for each county, retaining the residuals as a measure of pure interannual variability. They then correlated these residuals with climatic anomalies, defined as the deviation of each county-year observation from the county’s 2000 to 2024 long-term mean for each variable. The results were unambiguous. During the vegetative period, relationships were weaker and more region-dependent, though elevated vapor pressure deficit and heat-stress indices consistently tracked negative yield residuals. During grain filling, higher solar radiation anomalies were reliably beneficial. But it was the critical period, defined as fifteen days before to fifteen days after silking, that exhibited the strongest and most consistent relationships across all five regions.

Vapor pressure deficit around flowering showed the most powerful association with interannual yield variability in the northeast, southeast, and southwest regions, with correlation coefficients ranging from minus 0.44 in the northwest to minus 0.78 in the southeast. Slope values were equally telling, spanning from minus 3.24 megagrams per hectare per unit of VPD anomaly in the northwest to minus 9.53 in the southeast. In the northwest and west-central regions, the heat-stress index during the same window took precedence, though with somewhat weaker relationships, a pattern the authors attribute to the more continental thermal regime of western Indiana, where episodic extreme heat exerts a proportionally larger influence than background evaporative demand. Across every region, the worst yield outcomes clustered where high VPD and extreme heat anomalies coincided during pollination.

The physiological story behind these statistics is well grounded. Maize is most vulnerable around silking because this stage determines kernel number per plant, the foundation of final yield. Elevated vapor pressure deficit during this window reduces pollen viability, accelerates silk desiccation, and impairs pollination success, leading directly to kernel set failure. In parallel, high atmospheric demand intensifies plant water stress and triggers stomatal closure, cutting carbon assimilation and increasing kernel abortion. Extreme temperatures above 35 degrees Celsius compound these effects, while heat during grain filling reduces kernel weight and quality. The authors caution that because many of the evaluated variables are physically interrelated, their correlation analysis identifies the variables most strongly associated with yield variability rather than proving independent causal effects, and that disentangling those contributions would require multivariate approaches beyond the study’s scope.

The practical implications reach from the seed bag to the planting calendar. In the southern and eastern regions, where the frost-free period is longer, delayed sowing could shift flowering toward windows of lower atmospheric demand, though any adjustment must be coordinated with hybrid maturity selection to avoid truncating grain filling or raising late-season frost risk. In the northwest and west-central regions, where the frost-free window is tighter, choosing hybrids with enhanced drought and heat tolerance during pollination and grain filling becomes the primary management lever, tempered by the recognition that stress-tolerant hybrids may carry yield penalties in favorable years. The framework itself, built entirely from publicly available yield records and gridded climate data, is potentially transferable to the wider Corn Belt and other maize-producing regions, offering a scalable empirical route to matching agronomic strategy and breeding priorities to the environments where they will actually matter. As atmospheric drying intensifies with rising temperatures, knowing precisely where and when vapor pressure deficit strikes hardest may prove one of the most valuable tools Indiana’s farmers can hold.

Subject of Research: Environmental characterization of rainfed maize yield variability in Indiana using yield-based regionalization and climatic analysis

Article Title: Environmental characterization of maize yield variability in a major Corn Belt state

Article References: Environmental characterization of maize yield variability in a major Corn Belt state. (n.d.). https://doi.org/10.1007/s00704-026-06582-4

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06582-4

Keywords: maize, Indiana, Corn Belt, vapor pressure deficit, heat stress, yield variability, environmental regionalization, fuzzy c-means clustering, rainfed agriculture, climate variability, flowering stage, agricultural climatology