GENIE:Land

Introduction
This scheme is based on Peter Cox's Interactive Vegetation Model (IVM), which is itself a (very) reduced version of the tiled UM land-surface scheme MOSES2 (IVM was developed for inclusion into UVic). More detailed notes on MOSES can be found in Hadley Centre Technial Notes 24 and 56 (TRIFFID) and 30 (MOSES2). TRIFFID is the dynamical vegetation component of MOSES2. (http://www.metoffice.gov.uk/research/hadleycentre/pubs/HCTN/index.html).

GENIE-land was developed by Phil Harris during GENIE stage 1.

Tile structure
In each gridbox, the energy, water and carbon balance is calculated for seven tile types (broadleaf tree, needleleaf tree, C3 grass, C4 grass, shrub, bare soil, land ice) under the same forcing. Gridbox means (GBMs) of a particular quantity are calculated by weighting tile values by the fractional coverage of each tile in the gridbox. e.g. for a quantity $$Q$$ the GBM is calculated as,

$$Q = \sum_{i=1}^{ntiles} f_i Q_i$$

where $$f_i$$ is the fraction of the gridbox covered by tile type $$i$$.

Land surface physics
...will be covered soon.

At present GENIE-land calculates the surface energy balance. In future this may have to change since there are issues with the sea-ice scheme.

Photosynthesis
Photosynthesis is directly calculated from the environmental conditions, using the models developed in Collatz et al (1991) (C3 type photosynthesis) and Collatz et al (1992) (C4 type photosynthesis). Predicted rates of photosynthesis not only determine the carbon fluxes but also predict the stomatal conductance, linking photosynthesis to the surface energy fluxes. This approach is classified as a third generation land surface scheme (Sellers et al 1997). An implication of this methodology is that changes in the atmospheric carbon dioxide concentration can (in C3 plants) alter the moisture requirements. See Cox et al (1998) for details of the integrated photosynthesis model.

Dynamic vegetation structure
The structure of vegetation in GENIE-land can be either set to dynamic or static. In static vegetation patterns can change. This dynamical vegetation option runs TRIFFID (whos role is purely to update the vegetation structure, though TRIFFID and MOSES2 are sometimes confused). The dynamics underlying TRIFFID are lotka-volterra equations for competition (i.e. not the predator-prey equations, these ones don't have chaotic cycles). The setup of TRIFFID also (just about) excludes the possibility of unstable equlibria (see Hadley centre technical note 56). Vegetation in MOSES2 is modelled in terms of a volume of carbon. Plant function types (PFTs) grow through the balance between photosynthesis and respiration. Competition is for the proportion of the grid box (i.e. space). Generally the sucessfull competitor will have the highest carbon assimilation rate. However vegetation is put into a hierarchy of dominance. Trees automaticall displace everything else (shrub and grasses), and shrub automatically displaces grasses. The competition for space actually only occurs between two PFTs on the same level (e.g. between C3 and C4 grasses). The timescales for each PFT vary as a function of photosynthesis rates (see Hughes et al, 2006 for a discussion of the implications of this for earth system modelling), but in general trees have a re-growth timescale of around 100-300 years, shrub of 50 years and grasses of around 1-2 years.

Code structure
The top level routine for GENIE-land physics is tstep_land.f. The start of TRIFFID is in tstep_triffid.f.

Known issues
The land surface scheme is typically run with dynamic vegatation switched off, and then only with the IGCM. The option exists to run it with the EMBM, however this hasn't been fully tested.


 * Naming convention of land_mn files was unable to handle the longer dates (producing ***** names). This can be solved by editing the create_filename_land in land_netcdf.f90, specifically the code defining the cyear variable.


 * Bug appears to cause a crash after 84 years of simulation (when used in IGCM). This is most probably a technical (not scientific) problem and restarting the model from the output of the last month (or a few back) allows the model to be run on.  Suggest running in 50 year sections until this bug is tracked down.  Additional progress - successfully ran when writing to a local disc (sometimes?), try running on your own machine before assuming it will be a problem and consider reducing the frequency of output file writing.


 * At present surface climatology (when run in the IGCM with prescribed SSTs from HadCM3) does not look realistic enough to be able to run genie-land in dynamic vegetation mode. This is currently being explored (by J. Hughes).


 * Preliminary experiments (completed by J. Hughes) suggests an incompatibility between the original land surface exchange scheme and the tiedtke convection scheme. No clear linkage between the two model components exist and this is simply flagged as a warning.