Artificial Intelligence Reveals a Weaker CMIP6 Terrestrial Carbon Sink with Reduced Uncertainty

2026-03-14

Terrestrial ecosystems have cumulatively sequestered 24% of anthropogenic carbon dioxide (CO2) emissions since 1850 and are critical for mitigating future climate change. However, current Earth System Models (ESMs) remain highly uncertain in projecting future trajectories of this carbon sink capacity, hampering our predictive understanding of climate mitigation potential and impeding effective climate and carbon management policies. This study develops a novel framework that harnesses deep-learning (DL) to constrain uncertainties of ESM-projected Gross Primary Production (GPP) and Net Ecosystem Production (NEP) through 2100. Specifically, we apply DL to characterize the “offset” between ESM-simulated output (using CMIP6 models) and best-available observational products (top-down, bottom-up). This offset is treated as unresolved processes by current ESMs that could be effectively resolved by DL, which, once trained during historical periods, can be applied to adjust CMIP6 projections of the future. We find that DL significantly reduces the inter-model spread of GPP by ~56% and NEP by ~66% across the CMIP6 ESM ensemble . Under the medium emission scenario (SSP 245), the ensemble mean for NEP in 2100 is much weaker, 2.42 ± 1.16 PgC yr⁻¹ compared to 5.52 ± 3.45 PgC yr⁻¹ in the raw CMIP6 projections, suggesting a current overestimation of future carbon sequestration capability. Interestingly, DL revealed a slower trajectory of NEP growth compared to the raw CMIP6 projection. Beyond curbing the uncertainties of CMIP6 projections, DL also captures key environmental sensitivities of carbon cycle processes such as CO2 fertilization and sensitivity to warming. These findings demonstrate the power of DL in effectively curbing ESMs projection uncertainties and suggest that relying solely on natural terrestrial carbon sinks for climate mitigation is unlikely to slow down climate warming.

Anticipatory detection and characterization of urban stormwater system flood risk: a model-based fault diagnosis approach

Zenodo (CERN European Organization for Nuclear Research) · 2026-04-04

File Descriptions lab_system.inp - EPA SWMM model input file. lab_system_config.yml - Configuration file used by UrbanSurge to set up the EPA SWMM model. phy_healthy_impulse_database.csv - Simulated fault-free stormwater system depth and velocity measurements using survey impulse. phy_impulse_2.5yr_fault_database.csv - Simulated stormwater system depth and velocity measurements using design storm impulse under fault conditions. phy_impuslse_fault_database.csv - Simulated stormwater system depth and velocity measurements using survey impulse under fault conditions. pipeline_diagnose.ipynb - Python Jupyter Notebook used to perform fault detection and diagnosis and produce results figures found in the paper. single_fault_diagnose.ipynb - Diagnose a single fault from a single flow measurement. Used to test the accuracy of compound fault diagnosis. create_fault_database.ipynb - Functions to simulate faults in the EPA SWMM model. example_system.inp - SWMM model used to simulate compound faults. compound_fault_<6,4,2,2>_<8,5,10,4>.csv - Simulated flow measurements for the compound fault scenarios.

Local cloud enhancement associated with urban morphology: evidence from observations and idealized large-eddy simulations

Nature Communications · 2026-02-05

Previous studies have noted that cities enhance cloud cover, but the mechanisms of urban morphological types on cloud formation remain elusive. Observations of cloud climatology from 44 major U.S. cities show that cloud enhancement increases with the street-canyon aspect ratio and decreases with building density. Here, to explain these observations, we conducted numerical experiments using urban morphology-resolving large-eddy simulations. In these simulations, urban and rural surfaces retain their respective heat-flux differences, while the moisture sources and background atmospheric water vapor are prescribed to be identical, allowing us to isolate the morphological controls on moist convection. Results show that urban morphology influences cloud formation through two mechanisms: taller buildings intensify urban-breeze circulations at the urban-rural interface, while denser buildings, acting as momentum sinks, reduce vertical turbulent transport at the urban core. These vertical motions modify the transport of moisture in the urban atmospheric boundary layer, causing different cloud amounts across different urban morphology. This study highlights the mechanistic link between urban form, vertical motions, and cloud enhancement, thus providing a basis for city-specific boundary-layer convective parameterizations in large-scale weather and climate models. Using cloud observations from 44 US cities and large-eddy simulations, this study shows that taller buildings strengthen urban-breeze updrafts, while denser buildings suppress vertical turbulence, together reshaping moisture transport and cloud formation in urban environments.

Anticipatory detection and characterization of urban stormwater system flood risk: a model-based fault diagnosis approach

Zenodo (CERN European Organization for Nuclear Research) · 2026-04-24

File Descriptions lab_system.inp - EPA SWMM model input file. phy_healthy_impulse_database.csv - Simulated fault-free stormwater system depth and velocity measurements using survey impulse. phy_impulse_2.5yr_fault_database.csv - Simulated stormwater system depth and velocity measurements using design storm impulse under fault conditions. phy_impuslse_fault_database.csv - Simulated stormwater system depth and velocity measurements using survey impulse under fault conditions. pipeline_diagnose.ipynb - Python Jupyter Notebook used to perform fault detection and diagnosis and produce results figures found in the paper. single_fault_diagnose.ipynb - Diagnose a single fault from a single flow measurement. Used to test the accuracy of compound fault diagnosis. create_fault_database.ipynb - Functions to simulate faults in the EPA SWMM model. example_system.inp - SWMM model used to simulate compound faults. compound_fault_<6,4,2,2>_<8,5,10,4>.csv - Simulated flow measurements for the compound fault scenarios. misspec_M<Manning's n value>_L<fault node>.csv - Simulated flow measurements for the misspecification cases. The Manning's n values is the maximum value used to in the uniform sampling to generate random roughness offsets. Pimpulse_M1_L30_R1.dat - Impulse input file for the SWMM model. M1 denotes the maximum flow rate of 1 cfs, L30 denotes a 30 min impulse length, and R1 is an obsolete index. fault_database.csv - Simulated stormwater system depth and velocity measurements using survey impulse under fault conditions used for misspecification studies. supplemental_materials.pdf - Supplemental materials for the manuscript.

Anticipatory detection and characterization of urban stormwater system flood risk: a model-based fault diagnosis approach

Zenodo (CERN European Organization for Nuclear Research) · 2026-04-04

File Descriptions lab_system.inp - EPA SWMM model input file. lab_system_config.yml - Configuration file used by UrbanSurge to set up the EPA SWMM model. phy_healthy_impulse_database.csv - Simulated fault-free stormwater system depth and velocity measurements using survey impulse. phy_impulse_2.5yr_fault_database.csv - Simulated stormwater system depth and velocity measurements using design storm impulse under fault conditions. phy_impuslse_fault_database.csv - Simulated stormwater system depth and velocity measurements using survey impulse under fault conditions. pipeline_diagnose.ipynb - Python Jupyter Notebook used to perform fault detection and diagnosis and produce results figures found in the paper. single_fault_diagnose.ipynb - Diagnose a single fault from a single flow measurement. Used to test the accuracy of compound fault diagnosis. create_fault_database.ipynb - Functions to simulate faults in the EPA SWMM model. example_system.inp - SWMM model used to simulate compound faults. compound_fault_<6,4,2,2>_<8,5,10,4>.csv - Simulated flow measurements for the compound fault scenarios.

Spectral distortions to momentum and scalar exchanges by non-turbulent motion and patchy landscape variability

Agricultural and Forest Meteorology · 2025-05-26

Modifications to the spectra of turbulent velocity and scalars and co-spectra of vertical fluxes of momentum and scalars due to patchy landscape heterogeneity and non-stationarity are explored for a Mediterranean ecosystem. About 9 months of high frequency measurements of the three velocity components, water vapor concentration, carbon dioxide concentration, and air temperature were analyzed for different seasons (spring/summer) and prevalent wind directions (southeast/northwest). The two wind directions sampled a contrast of clumped and patchy landscape comprised of olive trees (southeast) and wall bounded flow disturbed by the presence of few upwind trees (northwest). The measured spectra and co-spectra were also compared to theoretical scaling forms from stationary, planar homogeneous flow, in the absence of subsidence as derived from the Kansas experiment. To assess the role of low frequency non-turbulent motion on the spectral and co-spectral content, a 5-min Fourier cutoff was introduced and the analysis was limited to near-neutral conditions where the boundary layer depth is shallow compared to its unstable counterpart. It was shown that the velocity statistics were not appreciably impacted by the low-frequency motion causing non-stationarity. Moreover, the turbulent scalar fluxes were also shown not to be significantly impacted by such low frequency motion. The scalar variances were impacted, especially the water vapor variance and its concomitant spectral shape. When the non-turbulent motion was filtered, the scalar spectra at low wavenumbers followed expectations from the so-called attached eddy hypothesis (i.e. exhibited a scaling with defining the longitudinal wavenumber) applicable for near-neutral conditions. For momentum co-spectra, the canonical shapes from the Kansas experiment appear to describe well the measurements here and in both dominant directions and seasons with some adjustment to the integral time scales based on wind direction. For the scalar co-spectra, deviations from the Kansas experiment were prevalent. The most noticeable and surprising deviations were their slow decay with increased sampling frequency at inertial subrange scales. This slow decay was shown not to contribute appreciably to the overall scalar fluxes. At those fine scales, predictions from local isotropy were expected to hold. The scalar co-spectral deviations from local isotropy were then discussed using a simplified co-spectral budget model where scalar–scalar co-spectra naturally emerged and the interplay between landscape heterogeneity and a scale-dependent pressure-scalar de-correlation time was postulated. It is also envisaged that the findings here offer a preliminary template for analyzing eddy-covariance data in situations that deviate from ideal conditions, especially regarding low-frequency modulations of scalar spectra and vertical scalar flux co-spectra.

Real‐Time Flood Inundation Modeling With Flow Resistance Parameter Learning

Water Resources Research · 2025-01-01 · 6 citations

Abstract Emergency response to flood plain inundations requires real‐time forecasts of flow depth, velocity, and arrival time. Detailed and rapid flood inundation forecasts can be obtained from numerical solution of 2D unsteady flow equations based on high‐resolution topographic data and geomorphologically informed unstructured meshes. However, flow resistance parameters representing the effects of land surface topography unresolved by digital terrain model data remain uncertain. In the present study, flow resistance parameters representing the effects of roughness, vegetation, and buildings are determined hydraulically in real‐time using flow depth observations. A detailed numerical reproduction of a real flood has been largely corroborated by observations and subsequently used as a surrogate of the ground truth target. In synthetic numerical experiments, flow depth observations are obtained from a network of in‐situ flow depth sensors assigned to hydraulically relevant locations in the flood plain. Starting from a generic resistance parameter set, the capability of a tandem 2D surface flow model and Bayesian optimization technique to achieve convergence to the target resistance parameter set is tested. Convergence to the target resistance parameter set was obtained with 50 or fewer tandem flow + optimization iterations for each forecasting cycle in which the difference between simulated and observed flow depths is minimized. The flood arrival time errors across a 52 flood plain inundation area were reduced by 3.13 hr with respect to results obtained without optimization from a fixed range of flow resistance parameters. Performance metrics like critical success index and probability of detection reach values above 90% across the flood plain.

Streamflow Forecasts Enhance Reservoir Drought Resilience: A Preference-Preserving Framework for Robustness Evaluation

2025-04-01

Hydropower utilities face increasing threats from climate change, with recent drought-induced generation shortfalls pushing global carbon emissions to record highs. While streamflow forecasts offer promising solutions, translating analytical insights into actionable policies remains challenging. This study introduces a preference-preserving framework that maintains stakeholders’ operational trade-off priorities while assessing reservoir performance under altered climatic scenarios. The framework is demonstrated using the upper Yellow River basin (UYRB) reservoir system in China. Our approach first selects representative policies from a pre-optimized pool based on their proximity to actual operations in multi-objective space. We then quantify the potential value of forecast information by evaluating performance gaps between policies informed by perfect versus binary streamflow forecasts under increasingly drier conditions. Time-varying sensitivity analysis reveals the mechanisms driving performance differences between forecast granularities. Results demonstrate that when stakeholders adhere to their persistent prioritization of firm power output, high-quality forecasts can reduce failure cases by approximately 50% under extremely dry climate scenarios. Policy robustness significantly improves through explicit feedback mechanisms that adapt to shifts in streamflow distributions, even without precise predictions of future climate changes. This framework provides decision-makers with actionable guidance for incorporating forecasts into reservoir operations while respecting existing management priorities.

Estimating anthropogenic heat flux by assimilating meteorological observations with a Kalman filter approach

2025-05-21 · 1 citations

Anthropogenic heat (AH) emissions in urban environments alter the surface energy budget and significantly influence urban climates. However, these emissions vary greatly in both time and space, leading to considerable uncertainty in their estimation. As remote sensing in the urban environment advances, where the remotely sensed urban surface temperatures are becoming increasingly available, such as those retrieved from satellite observations and thermal cameras. Yet, assimilating these observations into surface energy modeling for AH estimation has not been fully explored. In this study, a model for AH estimation based on the Kalman filter and surface energy balance is developed (KF-SEB model). Urban meteorological data, including air temperature and building surface temperature, are assimilated into the Kalman filter, yielding sensible heat flux and building heat storage. AH is subsequently calculated using the SEB equation. The KF-SEB model is evaluated using a forward model with predefined AH emissions. The forward model employs a simple SEB approach at the building exterior surface and adopts a 1-D heat conduction equation for the wall. The results show that the KF-SEB model accurately captures the magnitude and temporal variation of AH, with reduced uncertainties compared to previous studies. This study offers a novel approach of AH estimation based on urban meteorological data and provides important insights into the feedback between urban microclimates and anthropogenic energy use.

Quantification of Hydrogen Emission Rates Using Downwind Plume Characterization Techniques

Environmental Science & Technology · 2025-03-20 · 8 citations

emission rates at an accuracy comparable to that of other well-studied gases.