Explicit physics of boundary layer and low clouds. 2015-present.
Currently I am at UC Irvine working with Mike Pritchard on developing a novel global climate model (GCM) to explicitly simulate the low-clouds. This is a DOE funded multi-institutional research involving outstanding collaborators from University of Washington (Chris Bretherton and Matt Wyant), Stony Brook (Marat Khairoutdinov) and PNNL (Balwinder Singh).
My baby (read the climate model I’m developing) is based on “Ultra-parameterization” and is a high-resolution version of cloud “Super-parameterization”. Both terms in simple words are: a convection-permitting model able to capture the cloud physics. Ultra-parameterization will allow us to address the long-standing problem of low-cloud climate feedback which is the main shortcoming of today’s GCMs. Next plot is a snapshot of what this model is capable of doing. This is a high vertically resolved stratocumulus cloud developing off the coast of Peru, showing just one grid point out of hundreds in which low-clouds develop in the model.
Here the CRM has a 250m horizontal and 20m vertical resolution in the inversion mandating a 1 second time-step size or smaller.
This model is a computational grand challenge and requires a Peta-scale platform to operate on. I think this sort of convection-permitting models will be standard practice in the near future. Today, if you have a state-of-the-art supercomputer in your backyard and looking for some fun, 15,000 cores in 1 hour can give you a 1-day global simulation in Ultra-parameterized framework :).
Ultra-parameterization captures various types of clouds
Stratocumulus (Sc), shallow cumulus (Cu) and Cu-under-Sc clouds are realistically captured in ultra-parameterized framework. In the next plot, simulated height-time evolution of Sc clouds is shown off the coasts of California and Peru, along with shallow Cu clouds in Barbados and also Cu-under-Sc in South/Central Pacific; all of which are validated against NASA C3M satellite data (article submission imminent). The 125 level vertical cloud-resolving grid is shown on the left.
Explicit simulation of boundary layer and low clouds, a comparison.
Two panels of the figure bellow show the one-day zonal mean of the cloud liquid water content simulated for the day of 2008-10-01 using Ultra-parameterization (top panel) and Super-parameterization (bottom panel).
The unrealistically-low clouds of SP can be seen on the bottom panel, while the top panel shows the significantly more realistic zonal mean clouds calculated in UP. Note the clearly defined realistic boundary layer and cloud heights in all latitudes.
Physics of low clouds.
Low clouds are intricate balance of physical/transport processes like turbulence and entrainment that depend sensitively on small scale physics. Boundary layer plays a key role in creating and sustaining the low clouds. A picture is worth a thousand words, so here is Rob Wood‘s beautiful schematic (used with author’s permission) of the main players in the life of marine Stratocumulus (a major type of low clouds) from his excellent review paper: Stratocumulus Clouds, Monthly Weather Review 40, 2012.
Turbulent mixing feeds the cloud from bottom providing moisture and energy. The longwave cooling and entrainment work together at the cloud-top to sustain the deck. Various other weird and wonderful processes are involved which are tediously discussed in the paper.
Why low clouds are so important?
Climate scientists talk about low clouds all the time. The reason is that the low clouds are the most important yet uncertain elements on the planet to regulate the global temperature by reflecting the solar radiation back to space. You have seen the top of the low clouds when flying in an airplane. They are so bright (just like the Sun itself) that you will find it hard to look at them. Enormous areas of such low clouds exist off the coasts of California, Peru and Namibia (marked red in the plot bellow). These are almost permanent type of clouds and they sit there and hang around for a long time. For instance, this is how Peruvian low clouds look like from NASA Earth Observatory:
Indeed IPCC-AR5 shows that the majority of uncertainty in climate feedback is caused by the huge uncertainty of low-clouds in the GCMs. The next plot shows the bias of the short-wave absorbed radiation between the NASA satellite observation and the model. The three red boxes are the areas in which the current GCMs are not able to simulate the low-clouds and therefor a huge bias exists in absorbed radiation in those areas. This bias could be ~100 Watt per square meter or larger. Bottomline, the long-standing question is how and to what extent the low-cloud feedback is important to the climate?
The next plot shows the height time evolution of low-clouds from Ultra-parameterization over the Peruvian coast on various lon/lats. The cloud liquid mixing-ratio is plotted with the max values of 0.3 g/kg. The top right corner point is over land and the rest of the panels are over ocean.
Updates.
12/17/2018, UCLA: Invited talk on prospects of high performance computing for geophysical fluid dynamics.
09/05/2018, UT Austin – UTIG: Invited talk to present UP to an audience of various fascinating backgrounds in the Earth System. The talk mostly focused on the computational aspects of Ultra-parameterization and other HPC problems.
07/13/2018, Vancouver – Cloud Physics Conference: Presented UP results on low cloud feedback.
12/14/2017, New Orleans: Presented UP climate change results in AGU.
08/25/2017, Caltech-JPL: Attending the summer school on “Using Satellite Observations to Advance Climate Models”.
05/01/2017, UCI: Now joined the ACME (Accelerated Climate Modeling for Energy) team in developing/testing one of the leading climate models.
01/03/2017, UCI: Collaborating with Bill Collins and Ben Fildier in Berkeley Lab. on extreme rainfall under an idealized climate change scenario. See the paper here: http://onlinelibrary.wiley.com/doi/10.1002/2017MS001033/full
08/01/2016, Seattle: Visiting UW Atmospheric Science Dept. to work with Chris Bretherton, Matt Wyant and Peter Blossey on the low LWP bias in UPCAM.
06/17/2016, NCAR: First hands-on experience with TEMPEST (a novel atmospheric model under development from Paul Ullrich’s group in UCD). Using this model, I simulated an idealized tropical cyclone and analyzed the evolution of the TC. Here is a short summary of the result.
06/10/2016, NCAR: First hands-on experience with UZIM (a novel atmospheric model from Dave Randall’s group in CSU). This model is quite unique in that it uses Arakawa’s unified system of equations combined with Z grid in vertical direction and icosahedral cells in horizontal direction. Using this model, I am working on an idealized tropical cyclone to evaluate the model performance.
06/06/2016, NCAR: In DCMIP-2016 workshop/school to explore various dycores. Presented and discussed the details of UPCAM with NCAR scientists in a poster session.
10/12/2016, UCI: UPCAM paper is submitted to JAMES. Accepted (05/20/2017). Now published.
02/17/2016, UCI: SL dycore seems to be superior to FV in representing low cloud LWP. Closer examintaion debunks the role of dycore in the low LWP bias.
01/31/2016, UCI: Coerced the Semi-Lagrangian dycore to accept UP vertical grid.
01/05/2016, Boulder: Introduced UPCAM at CMMAP meeting to a more expert audience.
12/18/2015, San Francisco: The Numer8or introduces UPCAM at AGU.