My main research goal is to understand the spatial and temporal structure of climate feedback processes, how this structure will determine future climate change, and how it can be constrained from historical patterns of both anthropogenic change and natural variability. This goal entails a complex series of problems that require a structured, hierarchical approach. I favor a research philosophy of building conceptual analytical models, testing them in large, complex numerical simulations, and developing the statistical tools required to constrain the relevant physics in observational records and palaeoclimate proxies.

Peter Huybers, Kyle Armour, Gerard Roe, David Battisti, LuAnne Thompson, Cecilia Bitz, Aaron Donohoe, Malte Stuecker, Nick Lutsko, Nick Siler, Steve Po-Chedley, Robb Jnglin Wills, Jerry Mitrovica, Adam Maloof, Frederik Simons, Andy Rhines.

  • Climate Sensitivity is arguably the single most used metric of the earth's response to anthropogenic greenhouse gases. I try to understand how spatial patterns of warming interact with radiative feedbacks to set both the time-evolution and equilibrium value of climate sensitivity.
  • Natural variability is often treated as a nuisance in attempts to identify anthropogenic components of climate change. However, the structure of internal variability encodes a wealth of information about the underlying physical processes. I am particularly interested in the scale-dependence of the dominant radiative and air-sea feedback and forcing terms.
  • Palaeoclimate archives offer a unique opportunity to stress-test our theories about how the climate operates under markedly different conditions than the present. I am particularly interested in using the palaeoclimate record to constrain low-frequency climate variability and time-scale dependent feedbacks. I also have a long-standing interest in understanding the interplay between orbital forcing of past climates, and the coupling of climate, cryosphere, sea level, volcanism, and the carbon cycle.
  • Ultimately, I am very interested in understanding and improving forecasts for impacts of climatic changes. How do we separate forced changes to the hydrological cycle from natural variability? How do we improve forecasts of local changes in temperature, humidity, and heat stress? When are we committed to passing critical thresholds with policy implications?
  • I use General Circulation Models (GCMs) to test theoretical and statistical models under idealized "perfect information" scenarios. Given my interest in conceptual models and natural variability, I have an affinity for model hierarchies, model ensembles, and idealized simulations.
  • I often examine the Dynamics of the Climate System by treating Climate as a Dynamical System. Ideas from the control theory of dynamical systems make appearances.
  • Working on the physics of natural variability means I deal with stochastic dynamics, which I examine by extensively using spectral analysis.
  • The problems I approach typically involve several sources of evidence - such as data, proxies, and model output, and several layers of uncertainty - due to natural variability, observational error, and model error. Such problems are often best approached in Bayesian frameworks.


To understand how much the Earth will warm in response to changes in greenhouse gases, we need to understand how much the Earth needs to warm in order to come back into energy balance. At equlibrium: F=R(T).

They key quantity is the net radiative feedback,  or the efficiency of radiative damping, λ=R/T. The stronger the (outgoing) radiative feedback, the less the Earth has to warm to come back into equilibrium

We show that warming in regions of deep convection (like the West Pacific) leads to very efficient radiation to space (large, negative radiative feedback), while warming in other regions leads to either little radiation to space, or to more radiation being trapped (Fig. 1a). As future warming patterns will have relatively more warming in less radiatively effective regions such as the East Pacific and Southern Ocean (Fig. 1b), the efficiency which which the Earth radiates to space will become smaller.

When underlain by the observed SSTs, atmospheric models yield radiative feedbacks consistent with estimates from present day energetic imbalances. The strong radiative feedback - and low climate sensitivity - inferred from observations is attributable to the fact that the West Pacific has historically experienced more warming relative to the global mean that it is expected to in the future.

Proistosescu & Huybers 2017 (pdf) (supp).
Dong, Proistosescu, Armour, & Battisti, in review (preprint).


Over the satellite record, joint observations of top-of-atmosphere radiation and sea surface temperatures offers a unique opportunity to empirically constrain radiative feedbacks. However, the satellite record is dominated by “noise” or, formally, stochastic variability. Efforts to extract information about feedbacks from this noise, requires an appropriate stochastic physical model.

We offer a physical model that can explain the dominant features of the stochastic relationship between near surface air temperature and TOA anomalies. We find that there are there modes of variability, each with their own radiative feedback, variance, and time-scales. 
-The first mode is associated with variability driven by weather anomalies and strong air-sea feedbacks. 
-The second mode is associated with joint atmosphere-mixed layer variability driven by radiative anomalies. 
-The third mode is associated with ENSO-like quasi-periodic variability arising from coupled atmosphere-ocean dynamics. 

Proistosescu, Donohoe, Armour, Bitz, Stuecker, & Roe  2018 (pdf)


Fig. 1 Raw and dynamically adjusted trends in 1-April snowpack (as measured by snow water equivalent) at Snowfall Telemetry sites across the western United States between 1984 and 2018 (water years)


Fig 2: Top: The observed trend in winter-mean SST∗, defined as the local SST minus the average SST within the region shown. Bottom: The observed trend in winter-mean U500 (colors) and Z500 (contours), normalized by the interannual standard deviation (𝜎) at each grid point. The Z500 trend was computed after subtracting the global-mean value in each winter.

Melting snowpack is a vital source of water in the western United States during the summer, when rainfall is usually scarce. Although the amount of water contained in the snowpack has declined over the past century, it has been surprisingly stable since the 1980s, despite 1 ∘C of warming over the same period. 

We show that the contribution of global warming to western U.S. snowpack loss has in reality been large and widespread since the 1980s, but mostly offset by natural variability in the climate system, driven by the anomalous recent pattern of Pacific Sea Surface Temperatures and associated anomalies in atmospheric circulation. 

This result points to a faster rate of snowpack loss in coming decades, when the impact of global warming is more likely to be amplified, rather than offset, by natural variability.

Siler, Proistosescu, & Po-Chedley, 2019 (pdf) (supp)


The pattern of zonal mean warming associated with increased CO2 in the tropics only (TROP: 7S-7N), mid-latitudes (MLAT: 7N-70N), and polar regions (POLAR: 70N-90N). The linear sum of the response to localized forcing is a good approximation to the response to globally uniform CO2 forcing (GLOBAL).

One of the most salient features of global warming is the strongly amplified warming in polar regions. Polar warming arises from radiative forcing associated with increases in atmospheric CO2 concentrations across all latitudes. However, only the local CO2 forcing leads to polar amplified warming.

The excess energy input due to CO2 has be balanced either by radiating it to space, or exporting it to lower latitudes. The high latitudes are inefficient at radiative to space, i.e. the radiative feedbacks are more positive in polar regions. The lower moister content in the cold high latitudes makes it hard to export energy up the moisture gradient into the lower latitudes. 

Due to the inefficiencies in exporting energy either to space or to lower latitudes, a large polar temperature increase is required to balance the local radiative forcing.

Stuecker et al 2018 (pdf).


The SST patterns associated with the classical Principal Component based Pacific Decadal Oscillation (PC-PDO) and the low frequency component of the PDO


Gain and phase relation between the North Pacific Index (NPI) of Alleutian Low variability (a.b), and the Low Frequency component of the PDO. A reconustruction of the LFC-PDO time-series assuming a simple AR(1) integration of NPI variability in two models. The explained variance is shown for the full time-series and for decadally filtered time-series.

The Pacific Decadal Oscillation is one of the primary modes of interannual variability in the climate system. The traditional description of the PDO in terms of Principal Components (PC-PDO), subsumes variability induced by ENSO, gyre dynamics and extratropical atmospheric variability. 

Here we isolate the low-frequency component of the PDO (LFC-PDO), and show that it primarily arises due to slow gyres dynamics integrating variability in the Aleutian low. The LFC-PDO is adequately modeled as red noise integration of the North Pacific Index (NPI) (Fig. 2).

Wills, Battisti, Proistosescu, Thompson, Armour, & Hartmann (pdf) (supp).