Hydrogen (H2) has a key role in helping the world reach net-zero carbon emissions by 2050 by complementing other decarbonization pathways such as direct electrification, biofuels, synthetic fuels (e-fuels), carbon capture and storage (CCS), nuclear fission, wind energy, etc. The interest in the climate impact and behaviour of H2 in the troposphere has been increasing as the research and infrastructure expands to support the clean energy transition. The literature concerning atmospheric H2 has evolved significantly over the last 50 years, from initial observations and measurements to sophisticated modelling studies. The global budget of atmospheric hydrogen is constrained, but the soil uptake has high levels of heterogeneous uncertainties regarding its magnitude and response to environmental changes. Updating the photolysis scheme of formaldehyde to take into account recent experimental studies has an impact on global model estimates of photochemical H2 formation with an 8% reduction of the global burden of H2 from that using previous data. H2 is also a candidate for use in aviation fuel with the intent to lower radiative forcing impacts. However, inevitable leakage of H2 into the atmosphere would have a wider impact on atmospheric composition, which will partially offset some of the climate gains obtained in a switch to H2. A future hydrogen energy economy would have to be carefully monitored and regulated, particularly to minimise fugitive and point source leaks, to avoid further and unintended “greenhouse consequences”.
This study evaluates CMIP6 simulations of aerosol optical depth (AOD) and dust optical depth (DOD) over the Middle East–North Africa (MENA) region during March–May (MAM) and June–August (JJA). Historical AOD from 30 models for 2001–2014 was screened for output availability and broad consistency with observed magnitude, seasonal contrast, and spatial patterns. Twelve retained models formed the equal-weight MMM12 and were evaluated against MODIS and MERRA-2 using bias, monthly climatology, and Taylor-diagram metrics. GFDL-ESM4 showed the most balanced individual performance and was selected for dust-specific evaluation, supported by ERA5 winds and GLDAS soil moisture. MMM12 reproduced broad AOD patterns but underestimated hotspot magnitudes by approximately 0.2–0.3. GFDL-ESM4 better represented spatial patterns, seasonality, and high-AOD monthly occurrence but underestimated DOD and dust dominance. DOD/AOD ratios in the Bodélé Depression were 82.2% and 73.1% in MAM and JJA, respectively, versus 92.4% and 83.4% in MERRA-2; corresponding Arabian Peninsula ratios were 50.9% and 52.9% versus 71.1% and 66.6%. Negative DOD biases were spatially consistent with weaker winds and, regionally, wetter soils, without establishing causality. For 2040–2050, GFDL-ESM4 projected its largest DOD increase and more months with DOD > 0.4 under SSP5-8.5, especially in JJA. Future-minus-historical diagnostics showed only small, spatially limited changes in soil moisture and wind speed, suggesting that neither widespread drying nor stronger mean winds alone explains the projected DOD increase. Four other models with complete future DOD output preserved the main Saharan and Arabian source regions but showed substantial differences in magnitude and scenario sensitivity.
PM2.5 mass concentration alone may not fully represent particle oxidative activity. This study investigated autumn–winter variation in the oxidative potential (OP) of PM2.5, its chemical associations, and source-related predictors in urban Beijing from September 2019 to January 2020. Carbonaceous components, water-soluble ions, and elements were measured, and OP was quantified using an ascorbic acid depletion assay. Although PM2.5 concentrations were comparable in autumn and winter, autumn volume-normalized OP (OPv) and mass-normalized OP (OPm) were 2.3 and 2.6 times the winter means, respectively. Season-stratified analyses showed positive associations of OPv with sulfate, ammonium, and several water-soluble elements, particularly Pb, Cu, Zn, As, Fe, and Ni, in both seasons. Positive matrix factorization resolved five factors interpreted as traffic-related non-exhaust emissions, secondary nitrate, coal and biomass burning, metal-rich secondary-sulfate, and crustal dust. Repeated nested cross-validation with season stratification and temporal blocking showed that the season-adjusted random forest performed best among the tested models, with a mean cross-validated R2 of 0.461 ± 0.024. Cross-fitted Shapley additive explanations (SHAP) and within-season held-out permutation importance identified the metal-rich secondary-sulfate factor as the leading source-related predictor of OPv. The traffic-related non-exhaust emissions factor ranked second in held-out permutation importance, whereas the crustal dust factor showed relatively large SHAP importance but limited incremental predictive value. These findings highlight the value of combining source apportionment with interpretable machine learning to characterize PM2.5 oxidative activity beyond particle mass.
Terrestrial gross primary productivity (GPP) is strongly influenced by atmospheric variability, yet its interannual response remains uncertain in observationally upscaled products. Using the Unified FLUXes (UFLUX) eddy-covariance upscaling framework, we examine how atmospheric drivers, vegetation proxies, temporal constraints, and flux-tower sampling influence global GPP variability. We combine FLUXNET2015 and ICOS observations with multiple vegetation indices, solar-induced chlorophyll fluorescence, and various climate reanalyses to generate daily GPP estimates at 0.25° resolution. At the tower scale, the framework reproduces observed GPP variability reasonably well (R2 ≥ 0.8). Globally, vegetation proxies mainly affect mean GPP magnitude, whereas interannual variability is much more sensitive to temporal constraints and flux-network representativeness. However, replacing the calendar-year predictor with the Oceanic Niño Index (ONI) preserves major interannual anomalies, demonstrating that macroclimate indices offer a physically grounded constraint over arbitrary time trends, even though ENSO teleconnections vary regionally. A European case study shows that dense, continuous flux coverage stabilizes temporal estimates. Empirical regressions across Europe indicate a weakening net carbon uptake coinciding with stimulated ecosystem respiration under rising temperature and atmospheric vapor pressure deficit. However, these patterns reflect statistical associations rather than direct causal attribution, as confounding factors such as land management were not disentangled. Overall, these results indicate that robust assessment of global GPP variability depends not only on satellite and meteorological inputs, but also on physically meaningful treatment of climate variability and improved flux-network representativeness.
