A6 Elevation Class Proposal for ACME V3

Owned by Steve Ghan (Unlicensed)
Last updated: May 23, 2017
4 min read

Previous work by Leung and Ghan (1995, 1998) and Ghan et al. (2002, 2996a,b) showed that applying all atmospheric and land physics to multiple elevation classes within each grid cell and adding an orographic forcing term to the heat and tracer budgets within each elevation class can yield greatly improved simulations of local climate in regions with complex terrain. The spatial distributions of simulated surface temperature, precipitation, and snow water in particular, after mapping from elevation classes according to a high-resolution (order 5 km) distribution of surface elevation, agree much better with data mapped to the same resolution from station measurements. The computational cost of the scheme depends on the width of the elevation bands selected for the classification, but Ghan found impressive results in CAM3 with a maximum of twelve elevation classes that yield on average about two classes per grid cell. A load balancing scheme (Ghan and Shippert, 2005) limited the computational burden to about a factor of two.

However, because the scheme (originally developed and applied to a regional model MM5 by Leung and Ghan,1995, 1998) does not distinguish between precipitation on the windward and lee sides of subgrid topography, it consistently under-simulates rainshadow effects of topography when applied at resolutions too coarse to resolve rainshadows.

The ACME model is designed to run at 25 km resolution, which should be sufficiently fine to resolve rainshadows explicitly, leaving the elevation class scheme to improve fidelity at finer scales. Moreover, since more topographic variability is explicitly resolved at finer resolution, the average number of elevation classes is smaller than at coarser resolution, so the computational burden will be less than a factor of two.

We propose to apply the elevation class scheme to the ACME model in three phases. In the first phase, targeting ACME V2.0, elevation classes are added to the land model but not the atmosphere model. Land forcing fields (temperature, precipitation, downward longwave) from the atmosphere are lapsed with surface elevation at prescribed rates. Adjustments to ensure conservation of the grid cell mean are applied. Fluxes returned from land to atmosphere are aggregated to form grid cell mean. No changes in the coupler or atmosphere are needed. The elevation classes defined in the land model will be the same as that used in the second stage to facilitate comparison of results from the first and second stages and to reduce efforts for the second stage.

The second phase, targeting ACME V3.0, will assume the atmosphere and land models operate on the same grid and adopt the same elevation classification algorithm. This results in a one-to-one mapping between elevation classes in the land and atmosphere, so that no interpolation is required to couple the land and atmosphere for each elevation class. Although we expect this configuration to yield spectacular results, it does not allow the land model to distinguish subgrid watersheds.

In phase three, targeting ACME V4.0, the land model will represent watersheds, so that interpolation will be required to couple elevation classes in the land and atmosphere. Conservation will require normalizing fields to preserve grid cell means.

There are numerous challenges with implementing elevation classes in ACME that did not exist in the first implementations in MM5 and CAM3, mostly because a coupler was not used to couple the atmosphere and land in MM5 or CAM3.

A design document that identifies the challenges, describes methods to address them, and specifies tests to determine whether the scheme is working properly in the ACME model, has been prepared. While some of the work has been done as May 2017, staffing limitations have been a bottleneck and need to be resolved. Completing development for each phase will require involvement of ACME staff from the land, atmosphere, and coupling teams.

An estimate of staffing for completing tasks in phase one is listed here.

An estimate of staffing for completing tasks in phase two is listed here.

Staffing for phase three is expected to be more involved. An estimate of staffing will wait until the first two phases are nearly complete.



References

Ghan, S. J., X. Bian, A. G. Hunt, and A. Coleman, 2002: The thermodynamic influence of subgrid orography in a global climate model, Climate Dynamics, 20, 31-44, 10.1007/s00382-002-0257-5.

Ghan, S. J., and T. Shippert, 2005: Load balancing and scalability of a subgrid orography scheme in a global climate model. Int. J. High Performance Comput. Appl., 19, 237-245, doi: 10.1177/1094342005056112

Ghan, S. J., T. Shippert, and J. Fox, 2006: Physically-based global downscaling: Regional evaluation. J. Climate, 19, 429–445, http://dx.doi.org/10.1175/JCLI3622.1.

Ghan, S. J., and T. Shippert, 2006: Physically-based global downscaling: Climate change projections for a full century. J. Climate, 19, 1589–1604, http://dx.doi.org/10.1175/JCLI3701.1.

Leung, L. R., and S. J. Ghan, 1995: A subgrid parameterization of orographic precipitation. Theoretical and Applied Climatology, 52, 95-118. http://link.springer.com/article/10.1007/BF00865510

Leung, L. R., and S. J. Ghan, 1998: Parameterizing subgrid orographic precipitation and surface cover in climate models. Mon. Wea. Rev., 126, 3271-3291. http://dx.doi.org/10.1175/1520-0493(1998)126<3271:PSOPAS>2.0.CO;2

Leung, L.R., J.G. Michalakes, and X. Bian, 2001: Parallelization of a Subgrid Orographic Precipitation Scheme in an MM5-based Regional Climate Model. Lecture Notes on Computer Science – ICCS 2001. Springer Verlag, New York, pp195-203. http://link.springer.com/chapter/10.1007/3-540-45545-0_28