Managing CABLE-POP and ACCESS differences in vegetation distribution

CABLE-POP currently runs with 3 tiles per grid-cell: primary forest, secondary forest, grass. The type of forest and grass is climate (and latitude?) dependent and determined in an initial spin-up phase of the simulation. A grid-cell is either fully vegetated or fully bare-ground/ice/tundra.

ACCESS is flexible on the number of tiles per grid-cell. CMIP6 simulations were set up so that all tiles that would be needed to account for land-use change were simulated across pre-industrial, historical and scenarios. Tiles not required at any given time are given a tile fraction of 1e-5. Any vegetation types that would never exceed a tile fraction of 0.1 were excluded. Most grid-cells would have up to 4 active tiles but some grid-cells may be up to 7 types (from memory).

This file has plots of CABLE-POP ‘S2: no land-use change’ global vegetation distributions from a 1 degree simulations side-by-side with ACCESS-ESM1.5 1850 distributions.
CABLE-POP-S2_vs_ACCESS-1850.pdf (620.4 KB)

To enable POPLUC to work with ACCESS, it will be necessary to determine how best to deal with the different vegetation distributions. Options include

  • a generic re-mapping
  • simplifying the ACCESS vegetation distribution to work in a similar manner to CABLE-POP
  • adding more flexibility to CABLE-POP to allow a vegetation distribution closer to ACCESS
  • some combination of the above.

@inh599 has prepared a document with much more detail, including ideas around how to do a generic remapping.
POP-in-ACCESS-configuration-stuff_v1.pdf (273.2 KB)

@RachelLaw is testing the sensitivity of simplifying the ACCESS vegetation distribution.

  1. Test impact of using a single tree type per grid-cell by taking the dominant tree type from the current distribution. Difference plots for tree types:
    DomTreeDifference.pdf (139.1 KB)

  2. Check if and where c3/c4 grasses overlap in ACCESS. Test if c4 fraction capability in CABLE still works to deal with any overlap (but how are parameters dealt with in this case).

  3. Check whether spatially varying parameters rather than pft-based parameters could be useful to account for different grass-based types e.g. tundra.

  4. Check sensitivity to fractional bare-ground. What would be the impact of replacing with grass either with standard parameters or with parameters set for lower productivity?

  5. Re-visit how wetlands and lakes are treated in ACCESS.

Other questions

  • Could CABLE-POPLUC be adapted to allow for fractional bare ground? @Juergen notes that POPLUC would need major restructuring of the code to implement more than three tiles. Hence, if considered necessary, this needs to be a high priority, early task.
  • How are crops currently treated in each model (ACCESS is just a more productive grass with prescribed N fertiliser) and how would we like to treat them e.g. to account for harvest?

Slides discussed at 22/2/2024 CABLE4 meeting. Showing comparisons of forest and grass types between POPLUC and ACCESS and change across time from 1700-2022 and 1850-2014. Highlights need to check how crop has been processed for POPLUC as not showing any significant change over time. Also some regional differences between POPLUC and ACCESS that are worth checking since they are both derived from LUH2.
POPLUC-ACCESS-compare.pptx (546.6 KB)

1 Like

Aggregating grass tile: breakout discussion
The following 4 screenshots provide content around a breakout discussion on the science of aggregating/disaggregating multiple grass tiles in CABLE-in-ACCESS as necessary provide the single grass tile in POPLUC.

A recording of this discussion was taken - please contact @inh599 if you would like to review this (too large to be uploaded)

Example of stepped algorithm to accommodate multiple grass tiles with ACCESS and POPLUC
The following slides provide a simple numerical example of how a land-use change event results in changes to land area, carbon density and total carbon in ACCESS-ESM1.5, a hypothetical version of ACCESS that can accommodate primary and secondary forests and multiple grass tiles, and the potential ACCESS-with-POP where an aggregation/disaggregation of the grassy tiles is needed.

This is an illustrative example only: importantly these calculations have been done in excel, CABLE code to do the ACCESS3 calculations does not exist.

Take away message - there are notable differences between the current ESM and proposed ACCESS-with-POP approaches, and between cases where net or gross land cover transitions are used. The differences are dependent on the detail (e.g. initial states) of the specific example - nevertheless some of the differences are qualitative (i.e. of differing sign)