39 papers
The paper introduces Task-Aware Modulation with Representation Learning (TAM-RL), a novel framework that combines spatio-temporal representation learning with physically grounded constraints to significantly improve the accuracy and generalizability of global terrestrial carbon flux estimates compared to existing state-of-the-art methods.
The paper introduces CarbonBench, the first standardized benchmark comprising over 1.3 million global observations from 567 sites, designed to rigorously evaluate and compare zero-shot spatial transfer learning methods for upscaling terrestrial carbon fluxes across diverse, unseen ecosystems and climate regimes.
This study validates a method for estimating circumsolar sky brightness by demonstrating quantitative agreement between direct coronagraphic measurements and aerosol-inferred radiance at Mauna Loa, thereby enabling multi-decadal analysis of daytime coronal observing conditions using AERONET data.
This paper employs a global Vapor Kinetic Energy (VKE) budget analysis to reveal that atmospheric rivers evolve through consistent physical mechanisms worldwide, primarily intensifying via potential-to-kinetic energy conversion and decaying through condensation and turbulent dissipation, thereby establishing the VKE framework as a powerful diagnostic tool for understanding AR dynamics and regional variability.
This study utilizes ERA5 data and CESM1.2 simulations enhanced by a rare events algorithm to demonstrate that persistent double jet states over Eurasia are dynamically linked to positive Northern Annular Mode and quasi-wave-3 patterns, which collectively drive extreme heatwaves across three specific centers in the Northern Hemisphere.
This paper identifies and characterizes a distinct advective boundary layer (ABL) regime that emerges during South Asian monsoon onset when cross-equatorial flow intensifies, causing a dynamical transition from frictional to advection-controlled momentum balance that is governed by a linear diagnostic relation between meridional wind and geopotential gradient scaled by the inertial timescale.
This study combines field measurements and high-resolution WRF simulations to demonstrate that while the model accurately captures the structure of the Bolzano basin's valley-exit wind regardless of the planetary boundary-layer scheme, the scheme's ability to correctly simulate basin temperature stratification is critical for predicting the wind's onset, duration, and surface impact.
This study utilizes dry idealized simulations to demonstrate that poleward-shifted, broader, and deeper jet streams accelerate cyclone intensification, promote anticyclonic Rossby wave breaking, increase the frequency of cyclone merging, and dynamically precondition the atmosphere for persistent stationary anticyclones, thereby modulating midlatitude weather extremes.
While Kossin et al. (2020) attributed a rising ratio of major tropical cyclones to a decline in weaker Category 1 storms, extending the data through 2023 reveals a genuine intensification trend driven by both fewer weak storms and an actual increase in strong Category 3 and 4 storms.
This study demonstrates that machine learning can effectively parameterize unresolved midlatitude mesoscale vertical fluxes in Earth system models by leveraging vertically non-local information from temperature, moisture, and meridional wind, thereby offering a targeted approach to improve the representation of complex processes like slantwise convection and frontal dynamics.
This study employs Bayesian structural time-series causal inference on multi-decadal satellite data to demonstrate that the rapid expansion of nickel processing facilities at Indonesia's Morowali Industrial Park has caused a significant, quantifiable degradation of coastal water clarity, posing potential risks to the region's high-biodiversity marine ecosystems.
Using Bayesian inference and generalised extreme value regression on five CMIP6 models, this study projects that extreme temperatures in Earth's desert regions will significantly increase over the next century, with the most pronounced rises in annual maximums under severe forcing scenarios.
This study demonstrates that a scalable, dispatcher-led contrail avoidance workflow integrated into standard airline operations significantly reduces contrail formation rates without increasing fuel consumption, as validated by a randomized control trial using satellite imagery.
This paper demonstrates that a $50,000 Kaggle competition successfully crowdsourced diverse machine learning architectures for subgrid parameterization, which, when coupled with a full-physics climate model, achieved reproducible online stability and state-of-the-art performance, marking a significant milestone in advancing hybrid physics-ML climate simulations.
This paper presents a decision-theory framework and a blended AI-statistical forecasting system that successfully delivered skillful, tailored monsoon onset predictions to 38 million Indian farmers in 2025, enabling better agricultural decision-making under uncertainty.
To address the critical data scarcity of extreme rare weather events that hinders robust machine learning models, this paper proposes a physics-informed diffusion model based on Context-UNet that generates physically consistent, multi-spectral synthetic satellite imagery conditioned on key atmospheric parameters, thereby effectively mitigating extreme class imbalance and enhancing operational weather detection algorithms.
This paper introduces a data-driven integration kernel framework that enhances the interpretability and efficiency of nonlocal operator learning in climate modeling by separating nonlocal information aggregation via learnable weighting functions from local nonlinear prediction, thereby achieving competitive performance with fewer parameters and clearer physical insights.
Using an intermediate-complexity climate model, this study reveals that the Atlantic Meridional Overturning Circulation (AMOC) undergoes a boundary crisis under rising CO levels, where the collapse of its stable state and the resulting long chaotic transients governed by edge states explain large ensemble variances and apparent stochastic bifurcations in Earth system models.
This paper presents an improved mechanistic model for wave energy attenuation in drifting sea ice that accounts for ice-water drag, successfully replicating observed non-exponential decay profiles and spatial variations in attenuation rates that traditional exponential models fail to capture.
This study evaluates Aurora, an AI weather model, and finds that while it achieves strong short-range (1–7 day) forecast skill for various weather extremes comparable to traditional methods, its ability to predict the intensity of these events degrades significantly beyond 7–10 days as predictions regress toward climatology, indicating that intrinsic atmospheric dynamics still limit the practical predictability horizon for deterministic AI extreme-event forecasting.