Earth & Climate Systems

Overview

Earth & Climate Systems is built around a conviction that the atmosphere, the land surface and the hydrological cycle are best understood as a coupled dynamical system — not as independent objects of study, but as components whose behavior at any scale is shaped by feedbacks with the others. ECS researchers maintain their own observation infrastructure because they regard it as inseparable from the modeling work: sensors constrain models, models identify what the sensors should measure next.

The Atmospheric Dynamics Group studies how air masses organize into storms, how terrain shapes precipitation and how the planetary boundary layer governs the initiation of deep convection. The group maintains a network of X-band dual-polarization radars and conducts intensive field campaigns in collaboration with national meteorological services. The Hydrology & Earth Surface Lab works at the catchment scale, building process-based models of runoff generation, sediment transport and groundwater recharge, and testing them against long-term gauge records and satellite-derived soil-moisture products. The Climate Informatics Group develops machine-learning tools for a different part of the problem: emulating full climate models at a fraction of the computational cost, downscaling coarse-resolution projections to local scales and extracting attribution signals from noisy observations.

ECS is one of the largest users of the Meridian HPC Cluster, running regional climate simulations and ensemble weather prediction at 1 km resolution. The division contributes atmospheric and land-surface data to the Scientific & Analytical Services portfolio and collaborates with the Computational & Data Systems division on surrogate modeling and uncertainty quantification.

Research themes

• Mesoscale atmospheric dynamics — extratropical cyclone structure, orographic precipitation enhancement and convective organization under varying thermodynamic environments.
• Catchment hydrology and land surface — rainfall-runoff modeling, soil moisture dynamics, groundwater-surface water exchange and sediment connectivity under changing land use.
• Climate-model emulation — neural network surrogates trained on CMIP archives, with quantified uncertainty and physically interpretable internal representations.
• Downscaling and bias correction — statistical and dynamical methods for producing local-scale climate projections from global or regional models.
• Detection, attribution and extremes — attributing observed changes in precipitation, drought and heat to anthropogenic forcing; characterizing return periods of compound events.
• Paleoclimate reconstruction — Bayesian inversion of proxy records (speleothems, pollen, tree rings) for pre-instrumental hydroclimate.