13.4 ED2

Model Information
Home Page http://moorcroftlab.oeb.harvard.edu/
Source Code https://github.com/EDmodel/ED2
Authors Paul Moorcroft, …
PEcAn Integration Michael Dietze, Rob Kooper

13.4.1 Introduction

Introduction about ED model

13.4.2 PEcAn configuration file additions

The following sections of the PEcAn XML are relevant to the ED model:

  • model
    • id – BETY model ID. Each ID corresponds to a different revision of ED (see below)

    • revision – The revision (a.k.a. release) number of ED (e.g. “r82”). “rgit” indicates the latest code on the ED repository.

    • edin – Name of the template ED2IN configuration file. If this is a functional path that points to a specific file, that file is used. If no file is found following the path literally, the workflow will try the path relative to the PEcAn.ED2 package using system.file (recall that files in inst are moved in the package root, but all other directory structure is preserved). If this is omitted, PEcAn.ED2::write.configs.ED2 will look for a file called ED2IN.<revision> (e.g. ED2IN.rgit, ED2IN.r86) in the PEcAn.ED2 package.

      • Example: <edin>ED2IN.rgit</edin> will use the ED2IN.rgit file shipped with PEcAn.ED2 regardless of the revision of ED used. (Note however that if a file called ED2IN.rgit exists in the workflow runtime directory, that file will be used instead).
    • start.date, end.date – Run start and end date, respectively

    • met.start, met.end – Start and end year of meteorology inputs. By default (if omitted), these are set to the years of start.date and end.date, respectively. Setting these values to a shorter interval than start.date and end.date will cause ED to loop the meteorology input over the specified years. This may be useful for, for example, spinning up ED under a set of constant climate conditions.

    • phenol.scheme

    • phenol

    • phenol.start

    • phenol.end

    • all_pfts – (Logical) If false or missing (default), only run ED2 with the PFTs configured via PEcAn (i.e. in the <pfts> section of the XML). If true, run with all 17 of ED2’s PFTs, using ED2’s internal default parameters for all PFTs not configured through PEcAn. See below for more details.

    • ed2in_tags – Named list of additional tags in the ED2IN to be modified. These modifications override any of those set by other parts of the PEcAn workflow. These tags must be in all caps. Any tags that are not already in the ED2IN file will be added; this makes this an effective way to run newer versions of ED2 that have new ED2IN parameters without having to provide an entire new ED2IN. For example:

    • barebones_ed2in – Whether or not to try to annotate the ED2IN file with comments. If “true”, skip all comments and only write the tags themselves. If “false” (default), try to transfer comments from the template file into the target file.

    • jobtemplate

    • prerun – String of commands to be added to the job.sh model execution script before the model is run. Multiple commands should be separated by proper bash syntax – i.e. either with && or ;.

      • One common use of this argument is to load modules on some HPC systems – for instance:

        <prerun>module load git hdf5</prerun>
      • If your particular version of ED is failing early during execution with a mysterious “Segmentation fault”, that may indicate that its process is exceeding its stack limit. In this case, you may need to remove the stack limit restriction with a prerun command like the following:

        <prerun>ulimit -s unlimited</prerun>
    • postrun – Same as <prerun>, but for commands to be run after model execution.

    • binary – The full path to the ED2 binary on the target machine.

    • binary_args – Additional arguments to be passed to the ED2 binary. Some common arguments are:

      • -s – Delay OpenMPI initialization until the last possible moment. This is needed when running ED2 in a Docker container. It is included by default when the host is rabbitmq.
      • -f /path/to/ED2IN – Full path to a specific ED2IN namelist file. Typically, this is not needed because, by default, ED searches for the ED2IN in the current directory and the PEcAn workflow places the ED2IN file and a symbolic link to the ED executable in the same (run) directory for you.
  • run/site
    • lat – Latitude coordinate of site
    • lon – Longitude coordinate of site
  • inputs
    • met/path – Path to ED_MET_DRIVER_HEADER file
    • pss: [required] location of patch file
    • css: [required] location of cohort file
    • site: [optional] location of site file
    • lu: [required] location of land use file
    • thsums: [required] location of thermal sums file
    • veg: [required] location of vegetation data
    • soil: [required] location of soil data

13.4.3 PFT configuration in ED2

ED2 has more detailed PFTs than many models, and a more complex system for configuring these PFTs. ED2 has 17 PFTs, based roughly on growth form (e.g. tree vs. grass), biome (tropical vs. temperate), leaf morphology (broad vs. needleleaf), leaf phenology (evergreen vs. deciduous), and successional status (e.g. early, mid, or late). Each PFT is assigned an integer (1-17), which is used by the ED model to assign default model parameters. The mappings of these integers onto PFT definitions are not absolute, and may change as the ED2 source code evolves. Unfortunately, the only authoritative source for these PFT definitions for any given ED2 version is the Fortran source code of that version. The following is the mapping as of ED2 commit 24e6df6a (October 2018):

  1. C4 grass
  2. Early-successional tropical
  3. Mid-successional tropical
  4. Late-successional tropical
  5. Temperate C3 grass
  6. Northern pine
  7. Southern pine
  8. Late-successional conifer
  9. Early-successional temperate deciduous
  10. Mid-successional temperate deciduous
  11. Late-successional temperate deciduous
  12. Agricultural (crop) 1
  13. Agricultural (crop) 2
  14. Agricultural (crop) 3
  15. Agricultural (crop) 4
  16. Subtropical C3 grass (C4 grass with C3 photosynthesis)
  17. “Araucaria” (non-optimized southern pine), or liana

ED2 parameter defaults are hard-coded in its Fortran source code. However, most parameters can be modified via an XML file (determined by the ED2IN IEDCNFGF field; usually config.xml in the same directory as the ED2IN file). The complete record of all parameters (defaults and user overrides) used by a given ED2 run is stored in a history.xml file (usually in the same directory as the ED2IN) – this is the file to check to make sure that an ED2 run is parameterized as you expect.

As with other models, PEcAn can set ED2 parameters using its built-in trait meta analysis. The function specifically responsible for writing the config.xml is PEcAn.ED2::write.config.xml.ED2 (which is called as part of the more general PEcAn.ED2::write.config.ED2). The configuration process proceeds as follows:

First, the mappings between PEcAn PFT names and ED2 PFT numbers are determined according to the following logic:

  • If the PFT has a <ed2_pft_number> XML tag, that number is used. For example:

  • If <ed2_pft_number> is not provided, the code tries to cross-reference the PFT name against the pftmapping.csv data provided with the PEcAn.ED2 package. If the name is not matched (perfectly!), the function will exit with an error.

Second, the PFT number from the previous step is used to write that PFT’s parameters to the config.xml. The order of precedence for parameters is as follows (from highest to lowest):

  1. Explicit user overrides. These are specified via a <constants> tag in the PFT definition in the pecan.xml. For example:


    Note that these values are passed through PEcAn.ED2::convert.samples.ED, so they should generally be given in PEcAn’s default units rather than ED2’s.

  2. Samples from the PEcAn meta analysis. These are also converted via PEcAn.ED2::convert.samples.ED.

  3. ED2 defaults that PEcAn knows about. These are stored in the edhistory.csv file inside of PEcAn.ED2 This file is re-generated manually whenever there is a new version of ED2, so while we try our best to keep it up to date, there is no guarantee that it is.

  4. (Implicitly) Defaults in the ED2 Fortran source code. In general, our goal is to set all parameters through PEcAn (via steps 1-3), but if you are running PEcAn with new or experimental versions of ED2, you should be extra careful to make sure ED2 is running with the parameters you intend. Again, the best way to know which parameters ED2 is actually using is to check the history.xml file produced once the run starts.

The ED2IN field INCLUDE_THESE_PFT controls which of these PFTs are included in a given ED2 run. By default, PEcAn will set this field to only include the PFTs specified by the user. This is the recommended behavior because it ensures that all PFTs in ED2 were parameterized (one way or another, at least partially) through PEcAn. However, if you would like ED2 to run with all 17 PFTs (NOTE: using ED2’s internal defaults for all PFTs not specified by the user!), you can set the <all_pfts> XML tag (in the <model> section) to true:


13.4.4 Model specific input files

List of inputs required by model, such as met, etc.

13.4.5 Model configuration files

ED2 is configured using 2 files which are placed in the run folder.

  • ED2IN : template for this file is located at models/ed/inst/ED2IN.<revision>. The values in this template that need to be modified are described below and are surrounded with @ symbols.
  • config.xml : this file is generated by PEcAn. Some values are stored in the pecan.xml in <pfts><pft><constants> section as well as in <model> section.

An example of the template can be found in ED2IN.r82

The ED2IN template can contain the following variables. These will be replaced with actual values when the model configuration is written.

  • **@ENSNAME@** : run id of the simulation, used in template for NL%EXPNME

  • **@START_MONTH@** : start of simulation UTC time, from <run><start.date>, used in template for NL%IMONTHA

  • **@START_DAY@** : start of simulation UTC time, from <run><start.date>, used in template for NL%IDATEA

  • **@START_YEAR@** : start of simulation UTC time, from <run><start.date>, used in template for NL%IYEARA

  • **@END_MONTH@** : end of simulation UTC time, from <run><end.date>, used in template for NL%IMONTHZ

  • **@END_DAY@** : end of simulation UTC time, from <run><end.date>, used in template for NL%IDATEZ

  • **@END_YEAR@** : end of simulation UTC time, from <run><end.date>, used in template for NL%IYEARZ

  • **@SITE_LAT@** : site latitude location, from <run><site><lat>, used in template for NL%POI_LAT

  • **@SITE_LON@** : site longitude location, from <run><site><lon>, used in template for NL%POI_LON

  • **@SITE_MET@** : met header location, from <run><site><met>, used in template for NL%ED_MET_DRIVER_DB

  • **@MET_START@** : first year of met data, from <run><site><met.start>, used in template for NL%METCYC1

  • **@MET_END@** : last year of met data, from <run><site><met.end>, used in template for NL%METCYCF

  • **@PHENOL_SCHEME@** : phenology scheme, if this variabe is 1 the following 3 fields will be used, otherwise they will be set to empty strings, from <model><phenol.scheme>, used in template for NL%IPHEN_SCHEME

  • **@PHENOL_START@** : first year for phenology, from <model><phenol.start>, used in template for NL%IPHENYS1 and NL%IPHENYF1

  • @PHENOL_END@** : last year for phenology, from <model><phenol.end>, used in template for NL%IPHENYSF and NL%IPHENYFF
    @PHENOL@** : path and prefix of the prescribed phenology data, from * <model><phenol>, used in template for NL%PHENPATH

  • **@SITE_PSSCSS@** : path and prefix of the previous ecosystem state, from <model><psscss>, used in template for NL%SFILIN

  • **@ED_VEG@** : path and prefix of the vegetation database, used only to determine the land/water mask, from <model><veg>, used in template for NL%VEG_DATABASE

  • **@ED_SOIL@** : path and prefix of the soil database, used to determine the soil type, from <model><soil>, used in template for NL%SOIL_DATABASE

  • **@ED_INPUTS@** : input directory with dataset to initialise chilling degrees and growing degree days, which is used to drive the cold-deciduous phenology, from <model><inputs>, used in template for NL%THSUMS_DATABASE

  • **@FFILOUT@** : path and prefix for analysis files, generated from <run><host><outdir>/run.id/analysis, used in template for NL%FFILOUT

  • **@SFILOUT@** : path and prefix for history files, generated from <run><host><outdir>/run.id/history, used in template for NL%SFILOUT

  • **@CONFIGFILE@** : XML file containing additional parameter settings, this is always “config.xml”, used in template for NL%IEDCNFGF

  • @OUTDIR@** : location where output files are written (without the runid**), from <run><host><outdir>, should not be used.

  • **@SCRATCH@** : local scratch space for outputs, generated /scratch/<username>/run$scratch, should not be used right now since it only works on ebi-cluster

13.4.6 Installation notes

This section contains notes on how to compile the model. The notes for the VM might work on other machines or configurations as well. VM BU geo TACC lonestar

module load hdf5
curl -o ED.r82.tgz http://isda.ncsa.illinois.edu/~kooper/EBI/ED.r82.tgz
tar zxf ED.r82.tgz
rm ED.r82.tgz
cd ED.r82/ED/build/bin
curl -o include.mk.lonestar http://isda.ncsa.illinois.edu/~kooper/EBI/include.mk.lonestar
make OPT=lonestar TACC stampede

module load hdf5
curl -o ED.r82.tgz http://isda.ncsa.illinois.edu/~kooper/EBI/ED.r82.tgz
tar zxf ED.r82.tgz
rm ED.r82.tgz
cd ED.r82/ED/build/bin
curl -o include.mk.stampede http://isda.ncsa.illinois.edu/~kooper/EBI/include.mk.stampede
make OPT=stampede