How_to_run_LPX.md

To run the model, you will have to have the following installed:

1) fortran compiler - to compile the main, science bit of the model. At Macquarie, we use gfortran

2) c++ compiler - to compile the driver (I/O) bit. We use GCC c++

3) Netcdf for c++. Specifically, -lnetcdf_c++ -lnetcdf libraries

4) A little file that helps fortran and c++ talk to each other, called something along the lines of

lstdc++

Before compiling, you need to alter the "makefile" compilers and paths. . If you are running the

model at Bristol on the holocene, lgm or eocene servers, you might be able to get away with

skipping this bit. Look for and alter the following lines:

1) NETCDF =

2) FC =

3) FCOPTIONS = <relevant fortran options, including option for pre-compiling>. For gfortran, these

are typically -x -f77 -cpp-input

4) CPPOPTIONS= <you c++ libraries>

5) cxx = <you c++ compiler>

6) cxxoptions = <relevant c++ options> Typically -o3

7) you will finally need to change the 2nd path listed on the line "LDLIBS" to point to your talky

c++ Fortran file (lstdc++)

Once you've done this, it’s time to compile. LPX has 3 compiling options. Type either:

1) > make one_step makes a file for running the model in one go. Useful for equilibrium

climate runs.

2) > make two_step makes two running options: a spin-up; and a run. Spin-up is important for

dynamic runs, as you need to get the plants growing and various carbon pools filled before

you can do a proper run. This is the option most people should use

3) > make three_step: makes three options: a spin-up; a ramp-up; and a run. This is useful for

anyone doing futures runs, or for anyone unsure about how long their spin-up should take.

Spin up gets the plants and carbon pools going, ramp-up can either continue the spin-up (i.e, if

the model is not in equilibrium – see spinup protocol on wiki); or run the model dynamically to a

common point (e.g. to the present day), where all data is saved, and you can start new runs

(e.g. each future run) from that one point.

There is also an option call "make all", but I don’t think you'll need this unless you’re doing some

funny development stuff. This makes all the options from all the other make commands.

Now, before running the model, you need your climate inputs. The historical CRU inputs most of us

use are on terrafirma server at Macquarie Uni. Server address is sceince.terrafirma.mq.edu.au.

Hopefully at some point, we will make some inputs public on www.bcd.mq.edu.au , but for now,

you'll need a terrafirma login to get data. Email Rhys on rhys.whitely@mq.edu.au for more info.

LPX takes inputs in netcdf format. If you want to provide you own, then visit our wiki for

information about what you need.

For most runs, you will need spin-up (typically detrended) data for the spin up, and then real (or

transient) data from the actual run. See configuration help file for full list of input data.

The configuration file (lpj.cfg) lists all the information to set up the model. To set the model

properly, change each file path in the config file (paths ending in .nc) to the relevant inputs

(examples given in the file already):

TEMP: monthly gridded mean temperature over the period of the run

PREC: monthly gridded precipitation over the period of the run

WETDAYS: monthly gridded number of wetdays over the period of the run

SUN: monthly gridded cloud over the period of the run

TMIN: monthly gridded mean minimum temperature over the period of the run

TMAX: monthly gridded mean maximum temperature over the period of the run

WINDSPEED: monthly gridded windspeed over the period of the run

MASK: A landmask (no time dimension) defining which cells should be simulated. All other

inputs must be on the same sized grid as this (i.e, if this is 0.5 degree, so should

everything else). If there is data in the other files that is not in the defined mask from here

(i.e, if any of the other inputs have ocean values), they will simply be ignored.

MASK_VAR: name of the variable in the landmask file that corresponds to the mask.

SOIL_TYPE: gridded soil type (no time dimension). Soils should be listed as of Sitch et al

(2003).

SOIL_TYPE_VAR: name of variable in soiltype file corresponding to the soil.

POPDENS: annual gridded population density. Not currently used by the model, but we

haven’t taken it off the input list cos we might use it again one day

Crop: annual gridded cropland fraction in each cell

Pasture: annual gridded paster for each cell. Only used on some runs, but still there’s a line

for it in the file.

Lightn: gridded lighting climatology (no time dimension…. yet).

A_ND: no longer used, but still listed as an input. Maybe I’ll find time to remove this at some

point. Annual, gridded input of whatever you want, cos it makes no difference.

CO2_FILE: Annual (not gridded) atmospheric CO2 concentration

C13_FILE: Annual (not gridded) atmospheric delta C13 values.

C14_NORTH_FILE: Annual (not gridded) atmospheric delta C14 values for the northern

hemisphere.

C14_EQUATOR_FILE: Annual (not gridded) atmospheric delta C14 values for the equator

C14_SOUTH_FILE: Annual (not gridded) atmospheric delta C14 values for the southern

hemisphere.

CO2_VAR: CO2 variable in CO2 file

C13_VAR: c13 variable in delta C13 file

C14_VAR: c14 variable in delat C14 file

For the spin-up, these line should point to your detrended data. Once the spin-up is complete, this should be

your real data.

Then, set the following:

If you are using cloud data as inputs, set SUN_IS_CLOUD: to “TRUE”.

SPINUP_YEARS: Set to the number of years you wish to do a spin-up. Different applications

have different protocols for the length of the spin-up. There will be some listed on

the wiki at some point.

RUN_YEARS: The number of years the model needs to run past spin-up. E.g, for spin ups

ending in 1850, set to 150 to run the model onto 2000.

RUN_OUT_YEARS: The year from which you want LPX to start giving you outputs. Ie., staying

with the same example, if you want LPX to give you outputs from 1991, set to 141.

RUN_OUT_FREQ: How often you want the model to output. The model averages the

variables over this time. i.e, if you want the average output over the period 1991-

2000, set to 10.

SPINUP_FILE: This is the file where the state of the model after the spin-up will be stored.

RAMPUP_FILE: This is the file where the state of the model after the ramp-up will be stored.

OUTPUT_FILE: This is the netcdf file where defined outputs will be stored.

OUTPUT_VAR: A variable you’d like to output. You will need an OUPUT_VAR line for each variable

your outputting. You’ll have to look in the lpjmain code and search through the available

variables for now. At some point, I will put an OUPUT key together.

Once this is all set up, you’re ready to run. Run the model with the following commands:

1) > motif-lpj : obtained by the make one_step command. Runs the model in 1 step, useful for

equilibrium runs

2) > motif-lpj-step1a : obtained by the make two_step & make three_step commands. Runs the

spinup component and outputs the model state at the end of the spin-up into the file listed

after "SPINUP_FILE:" in lpj.cfg

3) > motif-lpj-step1b : obtained by the make three_step command Runs the rampup component.

Reads state of model from SPINUP_FILE: , runs through rampup, and outputs state of

model into RAMPUP_FILE:

4) > motif-lpj-step2 : obtained by the make two_step & make three_step commands. Runs the

final step of the model. Read state of model form SPINUP_FILE: and outputs defined

inputs into netcdf file defined by OUTPUT_FILE: .

The model outputs the defined variables (listed after OUTPUT_VAR in the config file) into a netcdf

file. The output file will be named after the OUTPUT_FILE: line, and will have the year the

output related to after it. This year is in model years, so if you run a 7000 year spin up to pre-

industrial 1850, and you want to find your output for the year 1990, it will be referred to as 7000

(number of spin-up years) +140 (number of years after pre-industrial) = 7140. So the output will be

called -7140.nc . If you have asked for an output freqancy of e.g. 10 years, the output file in

this example, 7140, will cover 1981-1990 (i.e, the previus 9 years, plus the year its outputting on).

And thats how to setup and run the model.

Goodluck!

Doug (douglas.kelley@students.mq.edu.au)

Sitch, S., Smith, B., & Prentice, I. (2003). Evaluation of ecosystem dynamics, plant geography

and terrestrial carbon cycling in the LPJ dynamic global vegetation model. Global Change Biology