Surface mass balance is the net result of accumulation and ablation at a glacier’s surface. Direct observations remain sparse and unevenly distributed, making it difficult to apply one calibrated model across large numbers of glaciers. Temperature-index models are widely used because they estimate accumulation from precipitation and melt from near-surface air temperature, often with separate degree-day factors for snow and ice. Their small parameter sets are interpretable and practical, but their accuracy can depend strongly on glacier-specific calibration. That dependence limits transfer to glaciers with different geometries and climate regimes, particularly because global geodetic mass-change products can be noisy at the scale of an individual glacier and do not represent exactly the same quantity as surface mass balance.

GlaUnTI retains the temperature-index structure but adds a shallow convolutional neural corrector. The underlying model represents solid precipitation, snow accumulation, fractional snow cover and temperature-driven melt using smooth functions that support automatic differentiation. The corrector learns spatial and temporal adjustments to five key quantities: precipitation sensitivity, the snow degree-day factor, the relative ice degree-day factor, the centre of the rain-to-snow transition and a temperature offset. Because these adjustments modify interpretable parameters rather than replacing the mass-balance equations, the model keeps a physical backbone while allowing data-driven compensation for systematic errors. In the comparison with the data-driven baseline, the temperature-index backbone used seven trainable parameters, whereas the gated recurrent unit had 961.

The neural component has separate spatial and temporal branches. The spatial branch uses an initial mass-balance estimate, sensitivities of that estimate to small temperature and precipitation perturbations, glacier outlines, surface elevation, elevation uncertainty and, in one variant, satellite-derived glacier facies. The temporal branch processes monthly temperature and precipitation totals. Their outputs are combined and resampled to generate glacier- and grid-cell-specific parameter corrections. The facies maps classify surface conditions such as ice, snow, debris, firn and refrozen-like material at the end of the ablation season when available; the study assembled 163 such maps for 68 glaciers. The two GlaUnTI variants used identical structural changes in this respect, with the facies-enabled version adding classified surfaces and classification confidence as predictors where the maps existed.

The study compiled 10,483 annual, 15,421 winter and 3,482 summer point mass-balance measurements, alongside 1,637 annual, 1,278 winter and 1,279 summer glacier-wide estimates. These records were combined with daily ERA5-Land temperature and precipitation, annual elevation estimates and time-varying glacier outlines. The data covered 78 glaciers, with 38 in the Alps, 21 in Scandinavia, nine in Svalbard, eight in Iceland and two in the Pyrenees. A glacier-level stratified split assigned 52 glaciers to training, 13 to validation and 13 to an entirely independent test set spanning contrasting climates and geometries. The test set contained 793 annual, 756 winter and 314 summer point measurements, together with 312 annual, 235 winter and 233 summer glacier-wide estimates. The authors used the glacier-wide estimates as weaker constraints than direct point observations because monitoring programmes derive them by extrapolating measurements across each glacier.

At point locations, the conventional temperature-index model produced an annual correlation of 0.854, a root mean square error of 1.707 metres water equivalent and a mean bias of +0.298 metres water equivalent. The purely data-driven gated recurrent unit performed worse, with a correlation of 0.798 and root mean square error of 2.066 metres. GlaUnTI without facies maps raised the correlation to 0.933 and reduced error to 1.095 metres, while the facies-enabled version achieved a correlation of 0.940, an error of 1.068 metres and a mean bias of just +0.009 metres. Seasonal results generally followed the same ranking, although the data-driven model retained the lowest winter bias. Per-glacier distributions showed generally lower errors and narrower spreads for the GlaUnTI models, indicating that the aggregate improvement was not confined to one glacier.

At the glacier-wide scale, the differences were also pronounced. The temperature-index baseline achieved an annual correlation of 0.606 and root mean square error of 0.805 metres water equivalent, while the data-driven recurrent network fell to a correlation of 0.331 and an error of 1.009 metres. The two GlaUnTI variants improved the correlation to 0.684 and 0.700 and reduced error to 0.651 and 0.627 metres, respectively. Including facies information produced moderate benefits for summer and annual glacier-wide estimates, although the evidence was limited: the facies-available evaluation contained 27 annual glacier-wide estimates and 32 annual point measurements. The researchers therefore treated the maps as useful auxiliary information rather than evidence of a large general gain. Their results are also consistent with the physical expectation that facies information can act as a proxy for processes such as surface-albedo contrasts, firn retention and debris cover, rather than serving as a causal driver of mass balance in its own right.

The hybrid model also approached the performance obtained when the conventional model was calibrated separately for each test glacier. In that deliberately favourable comparison, each temperature-index model used the target glacier’s observations from 1995–2012, was validated on 2013–2018 and was evaluated on 2019–2024. The locally calibrated model achieved an annual point-level correlation of 0.948, a mean bias of −0.308 metres water equivalent and a root mean square error of 0.905 metres. GlaUnTI, evaluated without glacier-specific fine-tuning, achieved a correlation of 0.944, a bias of +0.107 metres and an error of 0.883 metres. It also had lower absolute bias in winter and summer and comparable seasonal errors. The comparison shows that GlaUnTI approached local calibration in this test while its original spatial training used no observations from the target glaciers.

GlaUnTI nevertheless has important limitations. Its temperature-index backbone does not explicitly resolve radiation, debris effects, snow ageing, terrain shading or the transport of snow by ice flow, so some physical processes remain embedded in learned corrections rather than represented mechanistically. Point stakes may also fail to represent conditions across an entire 100-metre grid cell, particularly on glaciers with strong microclimate or topographic variation. Elevation observations contain gaps and uncertainty, and climate forcing has its own errors. The study’s sensitivity analysis addressed the choice of how elevations were extended outside the 2000–2019 observation window: replacing quadratic temporal extrapolation with backfilled values changed mass-balance estimates by less than ±2.5%, indicating that this extrapolation strategy was secondary to other error sources. The European scope and limited facies observations also mean that transfer beyond the sampled climates and glacier types remains to be tested.

Beyond prediction, the model’s differentiability enables analytical applications demonstrated in the study. It supports gradient-based calibration, input attribution, transfer learning, inverse optimisation and the propagation of climate-forcing uncertainty through long mass-balance trajectories. In one example on Grosser Aletsch in the Swiss Alps, the researchers optimized a temperature offset for 2015–2024 that made the model’s average annual glacier-wide surface mass balance for 2020–2024 equal zero. The resulting effective offset was −2.42°C. The authors emphasized that this was a reduced inverse experiment with fixed glacier geometry, not a complete climate-change scenario, and that the inferred offset could compensate for both forcing discrepancies and residual model biases. In another example, integrated gradients separated the temperature- and precipitation-associated contributions to annual anomalies for Schwarzberg. A separate uncertainty calculation propagated temporally correlated temperature and precipitation errors through cumulative mass-balance trajectories using first-order automatic differentiation.

The study’s broader implication is methodological rather than a claim that all local glacier processes have been resolved. The fully data-driven recurrent network had the weakest spatial transfer, while the physically structured temperature-index model generalized more reliably and benefited from a learned corrector that could adjust its key parameters. End-to-end differentiability also makes it possible to connect mass balance with differentiable ice-flow models, perform input attribution and ask what effective forcing perturbation would be needed to reach a specified balance target. The authors identify possible extensions involving albedo products, transient snowlines, satellite indicators of melt, more explicit energy-balance forcing and regional geodetic mass-change records. They caution that geodetic observations integrate processes beyond surface mass balance, including dynamic thinning, frontal ablation and basal melt, and that climate-forcing uncertainty should be represented explicitly rather than hidden within learned structural corrections.

Subject of Research: Transferable estimation of glacier surface mass balance using a hybrid physics–machine learning model

Article Title: GlaUnTI: a hybrid physics–machine learning model enables transferable glacier surface mass balance estimation

Article References: Maslov, K. A., Schellenberger, T., Persello, C., & Stein, A. (2026). GlaUnTI: a hybrid physics–machine learning model enables transferable glacier surface mass balance estimation. The Cryosphere, 20(10), 6045-6070. https://doi.org/10.5194/tc-20-6045-2026

Image Credits: AI Generated

DOI: 10.5194/tc-20-6045-2026

Keywords: glacier surface mass balance, hybrid modelling, machine learning, temperature-index model, European glaciers, climate forcing, remote sensing, neural corrector, snow accumulation, glacier melt, GlaUnTI, hybrid

Firn models are numerical tools that simulate how snow densifies into glacial ice, how heat moves through the column, and how meltwater percolates and refreezes. They are used to interpret satellite observations, to understand firn physics, and to project future surface mass balance of the ice sheets. Every such model integrates coupled differential equations in time, and the boundary conditions at the surface, such as snowfall, melt, temperature, and wind speed, must be supplied at some discrete interval. In the published literature, that climate forcing time step varies enormously, from one hour to days, months, and even years. The reasons for these choices are usually practical rather than physical: forcing data may not exist at finer resolution, files become unwieldy, or computational budgets run out. Until now, the implications of this choice had barely been evaluated.

The research team tested the sensitivity of the IMAU firn densification model, or IMAU-FDM, by forcing it with output from the regional climate model RACMO2.3p2, itself downscaled from the ERA5 reanalysis. They ran simulations for the Antarctic Peninsula Ice Sheet and the southern Greenland Ice Sheet, two regions spanning a wide range of accumulation and melt conditions, using four different forcing time steps: 3 hours, 6 hours, 1 day, and 1 month. The model itself integrates at a 15-minute step, but the climate inputs were averaged to the coarser intervals and held constant between updates. The team focused on firn air content, the vertically integrated pore space of the firn column, a key quantity that determines how much meltwater the firn can absorb before runoff begins.

The results were striking. Across both regions, larger forcing time steps systematically produced more firn air in the final simulated firn layer, and the size of the discrepancy grew with the time step. On the Antarctic Peninsula, the effect was largest on the low-lying ice shelves, where firn air content is naturally small: switching from 3-hourly to daily forcing inflated firn air content by about 15 percent, and monthly forcing by up to 44 percent. In the southern Greenland Ice Sheet, the largest differences appeared in a band at intermediate elevations, where melt is substantial but the firn is not yet fully depleted. There, daily forcing added on average 2.1 meters of firn air compared to 3-hourly forcing, and monthly forcing added 4.0 meters, differences that are far from trivial for meltwater retention studies.

The root cause, the authors show, is the presence or absence of a diurnal cycle in the input data. With 3-hourly or 6-hourly forcing, surface temperatures swing through the daily cycle, reaching the melting point during melt events and dropping below freezing at night. Melt and refreezing therefore alternate in time. With daily or monthly averages, however, temperature and melt become decoupled: the model can simultaneously apply a melt flux and hold the surface below zero degrees, a physically impossible combination. In that unphysical state, any meltwater that the pore space can retain refreezes immediately near the surface, forming a dense, ice-like crust. Subsequent melt events then strip away this dense refrozen material rather than the porous firn beneath it, so the model loses less firn air than it should. The result is an overestimate of the firn’s water-buffering capacity, an artifact born purely of the forcing resolution.

Deeper in the column, a second mechanism amplifies the differences. With 3-hourly forcing, meltwater percolates farther down before refreezing, releasing its latent heat at depth. Because firn conducts heat poorly, that trapped warmth makes the deeper firn warmer, which accelerates densification. Simulations with coarse forcing, by contrast, refreeze water near the surface and keep the deeper column colder, slowing compaction. The team also uncovered a subtler artifact tied to the fresh snow density parameterization used for Antarctica, which depends on instantaneous surface temperature and 10-meter wind speed. Averaging the wind speed over longer intervals changes the snowfall-weighted wind speed, altering the modeled density of newly fallen snow. In nearly melt-free zones this effect can even reverse the sign of the bias, producing slightly less firn air at fine resolution, on the order of half a meter.

The stakes become clearest in Greenland, where melt rates above roughly 500 kilograms per square meter per year can sustain firn aquifers, perennial reservoirs of liquid water buried in the firn. In the team’s simulations, aquifers formed during the spin-up period under 3-hourly forcing but never developed under daily forcing, because the deeper refreezing associated with fine resolution leaves the firn column warmer and less able to freeze all incoming water. Where aquifers did form under both resolutions, the timing and evolution diverged: under fine forcing the aquifer existed from the start of the historical simulation, while under daily forcing it emerged only around 2005. Because liquid water pins the firn temperature to the freezing point, the densification equations respond differently in each case, and the simulated firn air content trends can even acquire opposite signs. The modeled runoff limit, the elevation above which meltwater no longer escapes to the sea, shifted more than 10 kilometers inland between 3-hourly and daily forcing, and up to 35 kilometers compared to monthly forcing.

These discrepancies ripple directly into satellite altimetry, one of the primary tools for tracking ice sheet mass balance. Altimeters measure changes in surface height, but part of that signal reflects firn compaction rather than genuine ice loss, so firn models supply the correction. The study found that surface height changes differed by 0.2 to 0.4 meters on the Antarctic Peninsula depending on forcing resolution, and by 8 to 30 percent of the firn air content at a Greenland aquifer site. Corrections based on coarsely forced simulations would overestimate surface height gains, which means that during periods of thinning, ice loss would be underestimated, and during thickening, ice gain would be underestimated. In a warming Arctic where Greenland’s contribution to sea level rise is closely watched, a bias of this kind matters for projections and for the satellite records that policymakers rely on.

The authors distill two lessons for the modeling community. First, firn models that are prescribed melt fluxes and surface temperatures, as IMAU-FDM is, need a forcing time step fine enough to resolve at least the diurnal cycle; otherwise melt and subfreezing temperatures coexist and the physics breaks down. Models that compute melt internally from a surface energy balance may be less vulnerable to this specific artifact, though they face their own accuracy challenges. Second, parameterizations should be applied only under the conditions and resolutions for which they were developed. The Antarctic fresh snow density formula was tuned for 3-hourly forcing, and the densification equation assumes dry, steady-state snow, yet both are routinely stretched beyond those limits. As climate change accelerates melt across the polar regions, the study argues, the humble choice of a time step deserves the same scrutiny as the equations themselves, because it quietly shapes how much water the firn can hold, where runoff begins, and how vulnerable ice shelves are to collapse.

Subject of Research: Effect of climate forcing time step resolution on firn air content simulated by an ice-sheet firn densification model

Article Title: Impact of climate forcing time step in an ice-sheet firn model

Article References: van den Aker, T. E. A., Kuipers Munneke, P., van de Berg, W. J., Immerzeel, W. W., & van den Broeke, M. R. (2026). Impact of climate forcing time step in an ice-sheet firn model. The Cryosphere, 20(9), 5475-5490. https://doi.org/10.5194/tc-20-5475-2026

Image Credits: AI Generated

DOI: 10.5194/tc-20-5475-2026

Keywords: firn, ice sheet, Greenland, Antarctic Peninsula, firn air content, firn model, climate forcing, time step, meltwater retention, firn aquifer, satellite altimetry, The Cryosphere

Brown carbon, often abbreviated BrC, is the enigmatic cousin of black carbon. While black carbon is the familiar soot that absorbs sunlight across the entire visible spectrum, brown carbon is a complex brew of organic molecules that absorbs light selectively, with its absorbance climbing steeply toward ultraviolet wavelengths. This wavelength dependence makes brown carbon both scientifically fascinating and difficult to model. Recent modeling studies suggest that brown carbon accounts for more than twenty percent of the direct radiative effect caused by carbonaceous aerosols, and in some regions, such as the remote tropical upper troposphere, it may even dominate that effect. Beyond its climatic role, brown carbon can suppress the ultraviolet sunlight that drives tropospheric photochemistry, slowing the photolysis of ozone and radicals, and some of its chromophores, including polycyclic aromatic and nitro-heterocyclic compounds, carry significant oxidative potential with consequences for human health.

The new study, led by Can Wu and Gehui Wang of East China Normal University together with colleagues, set out to resolve a puzzle. Over the past decade, China has implemented stringent emission controls in the North China Plain, shifting residential heating from coal to natural gas and electricity, eliminating small coal-fired industrial boilers, and banning straw burning. Given that primary emissions from incomplete combustion are a major source of brown carbon, its levels should have plummeted. Yet strongly light-absorbing brown carbon kept appearing in high concentrations across the region. To understand why, the team conducted simultaneous measurements at five sites during the winter of 2023, from mid-November through the end of December: the cities of Beijing, Tianjin, Handan and Jinan, plus a rural station near Luancheng free from significant industrial influence. At each site, fine particulate matter smaller than 2.5 micrometers was collected every twelve hours onto prebaked quartz filters mounted on rooftops fifteen to twenty meters above the ground.

The spatial patterns the team uncovered were revealing. The average light absorption coefficient at 365 nanometers, a standard metric for water-soluble brown carbon, was highest at the rural Luancheng site, reaching 8.0 plus or minus 4.7 megapercubic megameter, roughly 1.1 to 3.5 times higher than at the urban sites. Beijing recorded the lowest value at just 2.3 plus or minus 1.9. Across the four urban sites, average light absorption had declined by about forty-five percent between 2018 and 2023, and in Beijing the value was only sixteen percent of what it had been a decade earlier, when the city’s average fine particle concentration stood at 158 micrograms per cubic meter compared with roughly a quarter of that during this campaign. The clean air actions, in other words, are visibly working. Yet multi-day haze episodes with hourly peaks reaching 203 micrograms per cubic meter, nearly three times the national grade-II standard of 75, still struck all five sites simultaneously, confirming that regional pollution remains a persistent winter challenge, particularly in the south.

When the researchers normalized absorption by the amount of water-soluble organic carbon, producing the mass absorption efficiency, a distinctly different spatial picture emerged. Jinan recorded the strongest light absorptivity of any site, at 1.40 plus or minus 0.02 square meters per gram, a value on par with severely polluted regions such as Xingtai, Xi’an and Delhi, India, and squarely within the range documented for brown carbon emitted from residential bituminous coal combustion. Tianjin, Luancheng and Handan shared values near 1.1 square meters per gram, closely matching biomass-burning-derived brown carbon, and concentration-weighted trajectory analysis linked their high absorption loadings to regions dense with satellite-detected fire hotspots. Beijing, by contrast, showed the weakest absorptivity at 0.71 square meters per gram, likely a combination of vehicle emissions, which typically produce weakly absorbing brown carbon, and atmospheric aging, since the ratio of the stable benzo[e]pyrene to its easily degraded counterpart benzo[a]pyrene indicated Beijing’s aerosol was the most chemically aged, and aging is known to bleach brown carbon’s color over time.

The pivotal insight came when the team compared clean periods, when fine particle concentrations stayed below 75 micrograms per cubic meter, with haze periods above that threshold. At most sites, the light absorptivity of water-soluble brown carbon increased as haze developed, and this enhancement tracked a parallel rise in the ratio of nitrogen to carbon within the water-soluble organic fraction. A positive correlation between that nitrogen-to-carbon ratio and light absorption held across the entire region, and a high nitrogen-to-carbon ratio also coincided with an elevated ratio of water-soluble to total organic carbon, a fingerprint of secondary formation. Together, these patterns pointed to nitrogenous organic compounds, or NOCs, as the crucial chromophores modulating brown carbon’s optical properties, and to secondary atmospheric production, rather than direct emissions, as their dominant source. A tracer-based apportionment confirmed the point: secondary formation explained roughly 64 plus or minus 21 percent of the water-soluble organic nitrogen in Beijing, comparable to other urban sites and about 1.2 times the rural fraction, and the secondary share grew as pollution intensified.

To identify what drives the formation of this secondary organic nitrogen, the researchers turned to machine learning. A random forest model, trained on seventy percent of the dataset and validated with ten-fold cross-correlation, attributed roughly forty-eight percent of the variance in secondary water-soluble organic nitrogen to ammonium and about twenty-eight percent to aerosol liquid water content, with a generalized additive model independently confirming both factors. The statistical evidence aligned with chemistry: reactions of ammonia or ammonium with atmospheric dicarbonyls such as glyoxal and methylglyoxal are a well-established pathway to light-absorbing nitrogen heterocycles like imidazoles, and laboratory work has shown that the uptake of these dicarbonyls, and the resulting yields of nitrogen-containing products and their oligomers, are significantly higher on neutral ammonium sulfate seeds than on acidic ammonium bisulfate. Thermodynamic calculations indicated the region’s aerosols were indeed dominated by ammonium sulfate and ammonium nitrate. During a humid haze episode in late December, offline aerosol mass spectrometry detected imidazole-related fragment ions that correlated strongly with ammonium, with a coefficient of determination of 0.93, but not with combustion tracers, and the ratio of these fragments to organic nitrogen climbed dramatically as the episode evolved, all pointing to ammonia-driven aqueous chemistry as the dominant formation route under moist conditions.

Aerosol acidity added a further layer of nuance. The mass fraction of secondary organic nitrogen correlated negatively with aerosol pH, suggesting that mildly acidic conditions favor nitrogenous chromophore formation, plausibly because ammonia partitions more readily into acidic particles, and because carbonyl-ammonia reactions are generally acid catalyzed. The team’s partitioning calculations showed that the fraction of ammonia present in the gas phase at pH 3.0 was an order of magnitude higher than at pH 5.0, consistent with this picture. Gas-phase photochemistry contributed too: nitrogen dioxide and ozone together explained over fourteen percent of the variance in secondary organic nitrogen, and nine quantified nitro-aromatic compounds, dominated by 4-nitrophenol and 4-nitrocatechol, were formed roughly forty-six to sixty-four percent by secondary oxidation, correlating with nitrogen dioxide rather than with biomass-burning tracers and displaying a temperature dependence consistent with gas-to-particle partitioning theory.

The broader implications stretch well beyond the North China Plain. A recent modeling study cited by the authors suggests that nitrogen-containing compounds dominate global organic aerosol absorption, accounting for eighteen percent of the direct radiative effect induced by carbonaceous aerosols, and previous work by the same group has shown that ammonia-driven aqueous reactions can generate brown carbon even in lifting air masses aloft. If ammonia is one of the key factors sustaining high loadings of strongly absorbing brown carbon across China’s boundary layer, then the persistently high ammonia loadings from insufficiently controlled agricultural emissions represent a blind spot in current air quality strategy. As the authors conclude, targeted ammonia emission control may prove indispensable for further mitigating wintertime haze and brown carbon pollution in ammonia-rich regions, especially during moist haze episodes when abundant aerosol liquid water opens the door to the aqueous chemistry that turns invisible organic vapors into climate-warming, light-absorbing particles.

Subject of Research: Light absorption and formation mechanisms of water-soluble brown carbon aerosols in the North China Plain

Article Title: Light absorption properties and composition of water-soluble brown carbon in North China Plain: implication for an enhancing role of nitrogenous organic compounds

Article References: Wu, C., Zhang, H., Sun, K., Li, R., Yan, Z., Chen, Y., Li, Z., Xiao, B., Ren, Y., & Wang, G. (2026). Light absorption properties and composition of water-soluble brown carbon in North China Plain: implication for an enhancing role of nitrogenous organic compounds. Atmospheric Chemistry and Physics, 26(19), 14229-14243. https://doi.org/10.5194/acp-26-14229-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-14229-2026

Keywords: brown carbon, ammonia, nitrogenous organic compounds, haze, North China Plain, aerosol optics, secondary organic aerosol, aqueous-phase chemistry, imidazoles, nitro-aromatic compounds, air quality, climate forcing

The methodology behind the study is straightforward but rigorous. The authors searched Scopus, the multidisciplinary citation database launched by Elsevier in 2004, using keywords such as climate change, paleoclimate, Holocene, climate variability and Quaternary, combined with Boolean operators and explicit inclusion and exclusion criteria. After removing duplicates and screening titles and abstracts for relevance, they retained 1,141 articles published across the 22-year window. The dataset was then analyzed and visualized with VOSviewer, the bibliometric mapping software developed at Leiden University’s Centre for Science and Technology Studies, alongside Microsoft Excel. This allowed the team to track annual publication trends, identify the most productive countries, institutions and authors, map keyword co-occurrences, and construct citation networks linking the field’s core journals. The authors are careful to note that their findings depend entirely on the coverage, accuracy and classification practices of Scopus itself, a caveat that shapes how the results should be read.

The temporal trends reveal a field in sustained expansion, though not without turbulence. In 2002, the earliest year of the analysis, only 26 documents appeared in the search results. Output climbed irregularly thereafter, with a sharp jump between 2012 and 2013, when the count leapt from 42 to 73 papers, and another surge between 2016 and 2018, culminating in a peak of 84 publications in 2018. The trajectory was not monotonic: 2014 saw a steep fall back to 42 documents, and 2022 dropped to 58 after hovering between 65 and 73 in the preceding years, before recovering slightly to 68 in 2023. These oscillations, the authors suggest, may reflect irregularities in publication patterns or in how the database collects and indexes records. The overall message, however, is unambiguous: scientific interest in Holocene climate variability has grown substantially over the long term, mirroring broader concern about understanding natural climate baseline conditions against which modern anthropogenic warming can be compared.

The geography of the field is strikingly lopsided. The United States leads with more than 300 publications, followed by Germany with roughly 250 and the United Kingdom with around 225. China ranks fourth with approximately 175 papers, and France contributes about 150. Canada, Australia, Spain, Switzerland and Sweden round out the top ten, each producing between 75 and 125 documents. At the institutional level, the Chinese Academy of Sciences stands out as the single most prolific affiliation, exceeding 100 papers, with France’s CNRS close behind. The University of Bern in Switzerland, Spain’s Consejo Superior de Investigaciones Científicas and Germany’s Alfred-Wegener-Institut Helmholtz-Zentrum occupy an intermediate tier, while Lanzhou University, the Deutsches GeoForschungsZentrum GFZ, China’s Ministry of Education and the University of Minnesota Twin Cities contribute more modest volumes. The map that emerges is one of dense research capability in North America, Western Europe and China, and comparatively thin coverage elsewhere.

Individual researchers show similarly uneven productivity. Melanie J. Leng tops the author ranking with around 18 papers in the dataset, well ahead of R. Lawrence Edwards at roughly 10. Achim Brauer, Hai Cheng and Feng Sheng Hu each published between six and eight papers, while most other highly ranked authors produced fewer than five. In the journal landscape, Quaternary Science Reviews dominates both in output and influence, typically publishing between 5 and 15 relevant papers per year and reaching approximately 15 in 2023. The Holocene follows a fluctuating path of 3 to 10 papers annually, Quaternary International varies between 2 and 8 with growth after 2010, and Palaeogeography, Palaeoclimatology, Palaeoecology records 2 to 7 per year. The Journal of Quaternary Science trails with 1 to 3. Citation network analysis confirms Quaternary Science Reviews as the most influential source, followed by Quaternary International, with related outlets such as Climate of the Past, the Journal of Paleolimnology and Marine Geology anchoring the multidisciplinary edges of the network.

Perhaps the most revealing finding concerns disciplinary breadth. Earth and Planetary Sciences accounted for about 37.7 percent of the publications, the largest single share, but the remainder is scattered across a remarkable range of fields: Environmental Sciences at 18.4 percent, Agricultural and Biological Sciences at 15.9 percent, and Arts and Humanities at nearly the same 15.8 percent, with Social Sciences contributing 8.1 percent. Medicine, chemistry and energy each registered only fractions of a percent. Keyword density mapping reinforces this picture of interdisciplinarity. The word Holocene dominates the center of the co-occurrence network, surrounded by clusters referencing paleoenvironment, precipitation, monsoon and glaciology, alongside oceanographic and isotopic terms such as Atlantic Ocean, oxygen isotope and sea surface temperature. Palynological and sedimentary vocabulary, including lacustrine deposit, pollen and fire history, appears prominently, and terms like genetics, mitochondrial DNA and genetic variability hint at links to biological evolution and species adaptation. Holocene climate research, in other words, is not the province of climatologists alone but a genuinely hybrid enterprise spanning geology, ecology, archaeology and even the humanities.

The study contextualizes this bibliometric portrait within the scientific substance of the field. Holocene climate variability has been reconstructed from a diverse suite of paleoclimate archives, each with distinct strengths. Ice cores from polar regions deliver annual-resolution records of oxygen isotopes with precise dating; speleothems from caves capture seasonal-scale signals in isotope ratios and trace elements; marine sediment cores, rich in foraminiferal assemblages and magnesium-to-calcium ratios, offer centennial-scale resolution with moderate dating accuracy. Pollen assemblages document past vegetation and hydroclimate, charcoal reflects fire activity, and stable isotopes such as δ18O and δ13C are widely used to infer temperature, precipitation sources and moisture balance. The driving mechanisms span multiple timescales: slow orbital forcing from Milankovitch cycles shaped the early to mid-Holocene climatic optimum, while solar variability, volcanic eruptions and internal ocean-atmosphere dynamics, including the Atlantic Meridional Overturning Circulation, the North Atlantic Oscillation and the El Niño-Southern Oscillation, generated abrupt and regionally heterogeneous responses such as the 8.2 ka cooling event. Transient model simulations from the Paleoclimate Modelling Intercomparison Project broadly reproduce orbitally driven trends, though discrepancies persist in the magnitude and timing of regional hydroclimatic shifts, particularly in the tropics and Southern Hemisphere.

The bibliometric data also expose structural weaknesses that the authors argue the field must confront. Regional disparities are pronounced: Europe, North America and China dominate the literature, while Africa and South America remain markedly underrepresented, a gap that echoes the spatial bias already known in proxy data coverage itself. Temporal resolution varies enormously across archives, from annually layered tree rings and ice cores to coarser marine sediments, complicating direct comparison of short-lived climatic events. Chronological uncertainties arising from radiocarbon calibration and reservoir effects further limit the precision of regional syntheses, and proxy sensitivity can differ depending on local environmental conditions. The authors also flag the limitations of their own approach, noting that English-language dominance in the indexed literature, keyword selection and Scopus’s disciplinary classifications all introduce potential biases into the analysis. They call for stronger integration of climate models with high-resolution proxy data, improved dating techniques, and expanded interdisciplinary and global collaboration to fill the geographic and methodological gaps.

Why does this matter beyond the census-taking? Because the Holocene is the stage on which human civilization developed, and its climate history provides the natural baseline against which modern, human-driven warming must be judged. Since the Industrial Revolution, emissions of carbon dioxide and methane have intensified the greenhouse effect to a degree that now overwhelms the natural forcings, orbital, solar and volcanic, that governed Holocene variability for millennia. Understanding how climate behaved under purely natural conditions, and how societies responded to events like the African Humid Period’s abrupt termination or the sapropel S1 interval in the eastern Mediterranean, informs climate resilience strategies, resource management and policy development. The authors expect research on the topic to keep growing in the coming years, and their analysis suggests where that growth is most needed: in the under-sampled regions of the global South, in the integration of models with proxy records, and in the high-resolution chronologies that would let paleoclimatologists pin down exactly when and how fast past climates changed. A field that can map its own blind spots, this study shows, is far better equipped to close them.

Subject of Research: Bibliometric analysis of Holocene climate variability research publications from 2002 to 2023

Article Title: Bibliometric analysis of Holocene climate variability research papers from 2002 to 2023 identifying trends leading contributors and thematic developments

Article References: Bouabid, N., Mnasri, H., Srarfi, F., & Tagorti, M. A. (2026). Bibliometric analysis of Holocene climate variability research papers from 2002 to 2023 identifying trends leading contributors and thematic developments. Discover Geoscience, 4(1), Article 325. https://doi.org/10.1007/s44288-026-00682-4

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00682-4

Keywords: Holocene, climate variability, bibliometrics, Scopus, paleoclimate, VOSviewer, proxy records, Quaternary Science Reviews, research trends, climate forcing, interdisciplinarity, research gaps