Parameterizations that can cause nonlocal interactions - ESCOMP/CTSM GitHub Wiki

Parameterizations that can cause non-local interactions

Background

For the most part, grid cells in CTSM operate independently of each other. If you couple it to an active atmosphere, then changes in one grid cell can of course have non-local effects via interactions with the atmosphere. But if you're running a land-only configuration, it is often the case that changes in one grid cell should not affect any other grid cells. However, there are some exceptions to this, and this page attempts to gather these exceptions.

Moreover, even at the subgrid level, columns operate mostly independently of each other. However, there are some exceptions to this, and this page also attempts to capture these exceptions.

Options that can cause interactions between grid cells in a land-only configuration

When running with a data atmosphere model (i.e., non-interactive atmosphere), interactions between grid cells can arise via the following options:

  • Feedbacks with an ice sheet model: Within ice sheet-covered regions, there can be interactions between grid cells if running coupled to an evolving ice sheet model. This is not the typical configuration: most compsets use CISM2%NOEVOLVE (a non-evolving, diagnostic ice sheet model) or SGLC (a stub glacier model); this is only the case for compsets that use CISM2%EVOLVE.

  • Feedbacks with a river model: If running with an active river model (which is the default for global runs), interactions between land grid cells can arise if any of the following options are enabled, because changes in one grid cell can affect river volumes, which in turn affect other grid cells in the river basin:

    • Methane when finundation_method is TWS_inversion, though this may be a bug (https://github.com/ESCOMP/CTSM/issues/658)

    • Irrigation with runoff-based limitation - i.e., limit_irrigation_if_rof_enabled set to .true.. This is currently .false. by default.

    • Flooding - i.e., RTM_FLOOD_MODE="ACTIVE". This is currently off by default, and only is available when running with RTM, not MOSART.

Interactions between land units within a grid cell

Few (if any) of CTSM's parameterizations operate at the land unit level. Interactions between land units can occur via the interactions between columns discussed below.

Interactions between columns within a grid cell

For the most part, the columns in a grid cell operate independently of each other. However, the following parameterizations can lead to interactions between columns within a grid cell:

  • Methane depends on grid cell total water storage (TWS), though this may be a bug (https://github.com/ESCOMP/CTSM/issues/658)

  • Downscaling of incoming longwave radiation in glacier-covered regions, if glcmec_downscale_longwave is .true. (which is the default): This downscaling is done in a way to conserve energy across the grid cell, which means that the downscaling in one column (which depends on the column's topographic height) can affect other columns after the final normalization is performed. Furthermore, the normalization of longwave radiation can mean that answers are sensitive to roundoff-level changes in column weights on the gridcell.

  • Dynamic column areas, via transient crops or dynamic glaciers: Some state variables in growing columns depend on the state in shrinking columns in the grid cell. In addition, the state in newly-initiating crop columns depends partly on the state in existing columns in that grid cell.

  • Hillslope hydrology (not yet available on master)

  • Urban columns that make up a given urban landunit interact non-linearly with each other through longwave and shortwave radiation reflection and absorption. Consequently, the landunit-level variables are oftentimes not simple combinations of the column quantities. Hence, there is a need for special scaling or special code (e.g., "if urban") to compute landunit-level fluxes and states from the column-level quantities (e.g., the urbanf and urbans scaling in subgridAveMod.F90 needed to handle extensive and intensive state and flux variables). Fluxes from urban columns are also combined in the code itself to solve for landunit-level urban canopy air temperature and humidity for example.

  • Others? The above may not be a complete list at this point.

Interactions between patches within a grid cell

Patches do interact strongly with each other. All of the patches on a given column share column-level (soil and snow) state variables, and so interact via changes in those state variables. So, in typical configurations, all natural vegetated patches interact with each other.

However, a patch should only interact with patches on a different column via the interactions described in the above sections.