The initialization of vegetation structure in the ED-2.2 model requires that the user define, for each polygon, the attributes of the hierarchy of sites, patches, and cohorts. It is usually relatively easy to convert plot census information into vegetation files by assuming each plot corresponds to a patch and each tree in a patch corresponds to a cohort. It may be necessary to break large plots or mapped stands into subplots. The following sections explain the structure on how to build these files. In order to use these files to initialize the model, the user must set NL%IED_INIT_MODE to 3 (in case site level information is available) or 6 (in case it is not) in ED2IN.

In case you would like to retrieve the forest structure from a previous ED-2.1 or ED-2.2 simulation, the user may set NL%IED_INIT_MODE=5 in ED2IN. Important: this will only set the forest structure from the previous simulation, but otherwise it will start a new simulation (e.g. soil moisture, vegetation temperature will use the typical initial conditions).

In case you would like to resume a previous simulation, use NL%RUNTYPE='HISTORY' in ED2IN. This will read every variable in the history file, and continue the simulation. Important: this option is intended to be used within the same version of the code (i.e. same Github commit): for example, you may want to use this option to resume a simulation interrupted due to a power outage. We strongly discourage setting NL%RUNTYPE='HISTORY' with a new commit and providing history files from an older commit.

The files described below are used by options NL%IED_INIT_MODE=3 or NL%IED_INIT_MODE=6.

The site file is only used when running with NL%IED_INIT_MODE = 3. This file comes in two formats and file format is specified in the file header:

nsites  <num> file_format <ff>

For file format #2 there is then an N x (N+1) adjacency matrix that is used in the runoff calculation. For file format 1 the adjacency matrix is approximated based on the proportional abundance of each site and assuming a random distribution of sites. File format #3 is like #1 but a soil class is specified for each soil layer rather than one class for all layers. The adjacency matrix has N+1 columns because the last column is the transition to the water patch.

The patch file is a space delimited ASCII file, that describe patch-level properties such are the are and the soil carbon and nitrogen stocks. The format of the pss file depends on the NL%IED_INIT_MODE.

• NL%IED_INIT_MODE=3. The file should contain the following elements:

• NL%IED_INIT_MODE=6. The file should contain the following elements:

Dummy variables are legacy variables that are no longer needed by ED-2. You can set them to zero, for example.

Dummy variables are legacy variables that are no longer needed by ED-2. You may want to include the values to keep things consistent, but otherwise, you may set them to zero.

A sample set of scripts to convert forest inventory plots to pss/css files (NL%IED_INIT_MODE=6) is available here. Currently this script works only for tropical forest PFTs.

To use it, you must download all the three files, and place them in the same directory.

• allometry.r. This subroutine computes the dummy variables in the css files, so they are consistent.
• pft_allom_table.csv. List of default parameters by allometry (consistent as of Jan 2019).
• make_bioinit.r. The main script, which calls/reads the other files.

• plots — Unique plot identification. Refrain from using very large plots. The definition of "gap" in ED-2 is typically associated with the crown area of a large tree (about 175 m² in the tropics). In case the plot is much larger than this, split the original plots and provide the smaller plots in the CSV file.
• tag — Unique ID for tree. Do not repeat tags for trees in the same plot.
• scientific — Scientific name. Currently not used by the script, but it could be useful for non-tropical PFTs.
• wood.dens — Wood density (g cm^-3). In case the wood density is not known, use some wood density data base (e.g. Chave et al.,2009) and the scientific names to provide estimates.
• dbh — Stem diameter at breat height (cm).

• place — Name of the site, for output directory.
• census.csv — Name of the CSV file with the forest inventory data (no paths).
• iata – Short code used to identify the site.
• identity – Tag for forest inventory data (it may be useful for multiple data sets in the same site).
• lon – Site longitude. Longitude goes from -180 to 180. Make sure to use the correct coordinates, because ED-2 reads this information.
• lat – Site latitude. Make sure to use the correct coordinates, because ED-2 reads this information.
• year – Year of measurements, used to populate dummy variable year.
• subplot.area – Area of sub-plots (m²), in case of differential sampling efforts. Put the full size in case all trees the same sampling effort was used for all measured trees.
• allplot.area – Total plot area (m²).
• nplots — Number of plots.
• min.dbh.subplot — Minimum DBH (cm) for trees to be measured in the sub-plot. In case the same sampling effort was used for all measured trees, set this to the minimum DBH measured in the plot.
• min.dbh.allplot — Minimum DBH (cm) for trees to be measured in the entire plot.
• fast.soil.c — Carbon stocks in the fast (metabolic litter + microbial) soil carbon (kg_C m^-2). Currently the same value is used for all plots.
• struct.soil.c — Carbon stocks in the structural (woody debris) soil carbon (kg_C m^-2). Currently the same value is used for all plots.
• struct.soil.l — Lignin stocks in the structural (woody debris) soil carbon (kg_C m^-2). Currently the same value is used for all plots.
• fast.soil.c — Carbon stocks in the slow (humified + passive) soil carbon (kg_C m^-2). Currently the same value is used for all plots.
• min.soil.n — Nitrogen stocks in the mineralised pool (kg_N m^-2). Currently the same value is used for all plots.
• fast.soil.c — Nitrogen stocks in the fast (metabolic litter + microbial) soil pool (kg_N m^-2). Currently the same value is used for all plots.
• iallom — Allometric setting to fill in the dummy variables in the css files. This is consistent with NL%IALLOM% in ED2IN.
• pft.idx — List of PFTs to use. Currently the script works only for tropical forest PFTs (2,3,4).

There are a number of ways for the user to prescribe the initial soil temperature, and soil moisture content. These are controlled by the ED2IN setting ISTWFLG:

• 0 – Interpolate from the coarser nested grid
• 1 – Read values from file (ISTWFLN)
• 2 – Set values prescribed values

Land cover setting is specified by the ED2IN setting IVEGTFLG. This setting is used only for regional runs, and the idea is to eliminate polygons that would fall over a water body (ocean, lakes, rivers). In case of regional run, the first grid should always have IVEGTFLG set to 1. The standard land cover setting comes from the Olson Global Ecosystem data with roughly 1 km resolution. Even though the data itself contains information on which type of vegetation (or water) exists at each point, all that ED2 uses is the information on whether the point has land or water.

The subroutines that control the land cover mapping are adaptations of the ones available at RAMS and BRAMS.