Recent Research Projects

Global aquaplanet simulation at ultra-high resolutions

Global atmospheric models with horizontal resolution at the order of kilometers provides new opportunities to study the complex coupling of the climate system across scales, which is not possible using the conventional General Circulation Models at the resolutions of O(100km) or Large Eddy Simulations/Cloud Resolving Models over limited domain. We perform a series of global aquaplanet simulations at the horizontal resolution from 50km to 6km using the GFDL AM4-MG2 model without additional tuning. We also perform simulations with and without the cumulus parameterization. Interesting (and sometime conterintuitive) resolution dependency is seen in the global mean precipitation, latitudinal distribution of clouds, ITCZ width and strength, convective organization over the deep tropics as well as water vapor concentration in the stratosphere. We seek a physical understanding of how such resolution dependency arises. More...

Simulate ozone depletion in climate models

The effect of stratospheric ozone depletion is simulated in GFDL AM4 model with three ozone schemes: prescribing monthly zonal mean ozone concentration, full interactive stratospheric chemistry, and a simplified linear ozone chemistry scheme but with full dynamical interactions. While similar amounts of ozone loss are simulated by the three schemes, the two interactive ozone schemes produce significantly stronger stratospheric cooling than the prescribed one. We find that this temperature difference is driven by the dynamical responses to ozone depletion. In particular, the existence of ozone hole leads to strong ozone eddies that are in-phase with the temperature eddies. The coherence between ozone and temperature anomalies leads to a weaker radiative damping as ozone absorbs shortwave radiation that compensates for the longwave cooling. As a result, less wave dissipates at the lower stratosphere, leading to a weaker descending and dynamical heating over the polar lower stratosphere, and hence a stronger net cooling there. The covariance between ozone and temperature is largely suppressed when ozone is prescribed as monthly zonal mean time series, as is the reduction in the radiative damping following ozone depletion. With much lower computational cost, the simplified ozone scheme is capable of producing similar magnitude of ozone loss and the consequent dynamical responses to those simulated by the full chemistry. More...

climo diff — Difference in (a) ozone concentration, (b) temperature, (c) shortwave (SW) heating rate, and (d) dynamical heating rate between the 2010O3 and the 1960O3 experiments. Results are shown at 100 hPa averaged over the southern polar cap.

temp diff — Difference in the polar cap temperature at 100 hPa between the 2000ozone and 1960ozone simulations from CAM3, AM3 and AM3C (AM3 with zonal wind climatology perturbed to be CAM3-like). Shading indicates the 95% confidence interval based on the Student's t test.

We contrast the responses to ozone depletion in two climate models: CAM3 and GFDL AM3. Although both models are forced with identical ozone concentration changes, the stratospheric cooling simulated in CAM3 is 30% stronger than in AM3 in annual mean, and twice as strong in December. We find that this difference originates from the dynamical response to ozone depletion, and its strength can be linked to the timing of the climatological springtime polar vortex breakdown. This mechanism is further supported by a variant of the AM3 simulation in which the Southern stratospheric zonal wind climatology is nudged to be CAM3-like. Given that the delayed breakdown of the Southern polar vortex is a common bias among many climate models, previous model-based assessments of the forced responses to ozone depletion may have been somewhat overestimated. More...