My primary research interests are:

  • Planet boundary layer and convection, and their parameterizations
    • Infrared radiation parameterization
    • Climate modeling and diagnostics
    • History of atmospheric science
    • Research projects:

      Boundary layer and convection

      GFDL AM4 underestimates marine stratocumulus amount on the west coasts of North and South America and South Africa, leading to excessive shortwave absorption in these regions. To address this issue, we implement the Mellor-Yamada-Nakanishi-Niino Eddy-Diffusivity/Mass-Flux (MYNN-EDMF) scheme into the AM4. The major implementation challenges include (1) incompatibility of the MYNN-EDMF cloud scheme and AM4 cloud scheme, and (2) coupling the MYNN-EDMF with other schemes. The performance of the MYNN-EDMF in AM4 is evaluated using AMIP simulation. AM4 with MYNN ED shows moderate improvements in marine stratocumulus biases. However, AM4 with MYNN-EDMF worsens the already large marine stratocumulus biases, partly due to coupling with the AM4 stratiform cloud scheme.

      Planet boundary layer (PBL) and moist convection closely couple with each other. Turbulence in the PBL effectively transports heat and moisture from surface to the atmosphere, thus helping convective clouds to form. These convective clouds can in turn affect the PBL turbulence and structure. Here we explore two coupling strategies of PBL and convection schemes in GFDL AM4, namely, (1) PBL_then_Conv, in which the convection scheme sees the state updated by the PBL scheme, and (2) PBL_and_Conv, in which both PBL and convection schemes see the same state. The AMIP results show that these coupling strategies have the strongest impact on marine shallow cumulus regime. PBL_and_Conv has weaker convection, stronger PBL activities, and more low cloud than those in the PBL_then_Conv. We hypothesize that these are because the convection scheme in PBL_and_Conv “sees” a less unstable state, leading to weaker convection.

      For further details, please refer to our poster at 2022 AGU Fall Meeting.

      In order to better represent cloud microphysical processes and aerosol-cloud interactions in cumulus parameterization schemes, I incorporated a two-moment (mass and number), two hydrometeor species (cloud water and rain) warm-rain scheme into the Zhang-McFarlane deep cumulus parameterization. The unique feature of this new scheme was that raindrops can move upward or downward depending on their terminal velocity and the updraft velocity. This treatment was more physical than other current schemes which assumes raindrops either leave the updraft immediately or only move upward with the updraft. Sensitivity tests of this new scheme using the NCAR CAM5 single column model showed that the portion of convective precipitation decreased while the total precipitation remains unchanged, suggesting a way to address the overestimation of convective precipitation in CAM5 and other climate models.

      For further details, please refer to my Master thesis.

      Longwave radiation

      Cloud longwave scattering has never been deemed as a necessity in climate models. Out of all climate models in the IPCC fifth and sixth assessments, only three modeling centers have longwave scattering included in their models. Our study explained why the traditional wisdom of neglecting longwave scattering breaks down for the simulation of high‐latitude climate in the fully coupled models. We showed the critical importance of atmosphere‐surface radiative coupling for correctly assessing the role of cloud longwave scattering in the model simulation of climate mean state as well as climate changes, an issue overlooked by all previous studies. We argued that the cloud longwave scattering is a necessity in climate models, not an option.

      For further details, please refer to our paper published in 2020 in the Geophysical Research Letters.

      I investigate how changes of surface longwave (LW) emissivity in the Sahara and Sahel influence simulated regional climate and beyond. Surface emissivity is a function of surface types and usually varies with wavelength. However, most general circulation models (GCMs) neglect this spectral dependency and assume surface emissivity as a constant over the all LW bands (usually close or equals to unity). This assumption can lead to biases in climate simulations over the Sahara and Sahel, where the surface LW emissivity can be as low as 0.6-0.7 over the atmospheric infrared window band due to the features of sand surface. In order to assess the extent to which the emissivity treatment in the Sahara and Sahel influences the simulated regional climate, my colleagues and I have implemented a realistic surface emissivity dataset into the NCAR CESM, carried out climate simulations, and analyzed the simulation results.

      Compared to the simulation that treats surface as blackbody, the simulation with realistic surface emissivity treatment shows that surface air temperature increases over the Sahara and Sahel. This is mainly because the inclusion of realistic, non-blackbody surface emissivity decreases the surface emission and keeps more LW energy at the surface, which in turn increases the surface temperature. In addition to that, the planetary boundary layer over the Sahara and Sahel becomes warmer and wetter, favoring more convection and produce more convective rainfall, especially in the Sahara. The inclusion of surface emissivity in the Sahara and Sahel also leads to changes in moisture flux convergence in the adjacent regions, resulting in precipitation changes, in particular south of the Sahel.

      Our study highlights the need to treat surface emissivity realistically in the GCMs, in order to represent the surface-atmosphere LW coupling processes in a more physical way.

      For further details, please refer to our paper published in 2019 in the Journal of Climate.

      I compared two widely-used longwave and shortwave radiation schemes in climate models, namely CLIRAD (CLImate and RADiation) developed by the NASA Goddard Space Flight Center, and RRTMG (Rapid Radiative Transfer Model for GCMs) developed by Atmospheric & Environmental Research. This comparison focused on their gaseous absorption treatment, radiative transfer solver, cloud overlap assumption, and simulation results with different standard atmospheric profiles, such as mid-latitude summer and tropics. I also implemented the CLIRAD into the NCAR CESM climate model (the default is RRTMG) as a part of the project for Taiwan Earth System Model (TaiESM).