Author: Benjamin Stocker
The {grsofun} package provides functions that wrap a call to runread_pmodel_f() from the {rsofun} package. This call is a point-scale simulation of the P-model. Functions of {grsofun} enable spatially distributed simulations on each gridcell and a parallelisation by gridcells for high computational speed.
The implementation follows a tidy data paradigm. All model input and output are organised as tidy data frames where forcing and output time series are nested inside cells and each row represents one gridcell. Because runread_pmodel_f() requires full forcing time series sequences and because reading full time series with global coverage is commonly limited by memory, the forcing data first has to be re-organised as tidy data frames and split into separate files for each longitudinal band. After running the simulations, outputs, written as tidy data frames into separate files for each longitudinal band, have to be collected again before further processing of spatial fields.
Each of these steps is implemented by a separate {grsofun} function. Functions share the settings argument which contains a list of all user-specified and simulation-specific information. Model parameters (which may be calibrated using {rsofun}) are provided as another separate argument.
The workflow of running a global simulation of one year with WATCH-WFDEI climate forcing and MODIS FPAR is demonstrated below.
Load libraries for this vignette.
library(ggplot2)
library(dplyr)
library(rnaturalearth)
library(cowplot)Note: currently, {grsofun} is set up run in demo mode with a prescribed constant CO2 and globally uniform constant cloud cover. This is to be accounted for by reading from data.
settings <- list(
fileprefix = "test", # simulation name defined by the user
model = "pmodel", # in future could also be "biomee", but not yet implemented
year_start = 2018, # xxx not yet handled
year_end = 2018, # xxx not yet handled
grid = list( # a list specifying the grid which must be common to all forcing data
lon_start = -179.75,
dlon = 0.5,
len_ilon = 720,
lat_start = -89.75,
dlat = 0.5,
len_ilat = 360
),
source_climate = "watch-wfdei", # a string specifying climate forcing dataset-specific variables
dir_climate = "~/data/watch_wfdei", # path to where climate forcing data is located
dir_climate_tidy = "~/data/watch_wfdei/tidy/", # path to where tidy climate forcing data is to be written
source_fapar = "modis", # a string specifying fAPAR forcing dataset-specific variables
file_fapar = "~/data/MODIS-C006_MOD15A2_LAI_FPAR_zmaw/MODIS-C006_MOD15A2__LAI_FPAR__LPDAAC__GLOBAL_0.5degree__UHAM-ICDC__2000_2018__MON__fv0.02.nc", # path to where fAPAR forcing data is located
dir_fapar_tidy = "~/data/MODIS-C006_MOD15A2_LAI_FPAR_zmaw/tidy/", # path to where tidy fAPAR forcing data is to be written
file_whc = "~/data/mct_data/cwdx80_forcing_halfdeg.nc", # path to where root zone storage capacity forcing data is located
dir_whc_tidy = "~/data/mct_data/tidy/", # path to where tidy root zone storage capacity forcing data is to be written
file_landmask = "~/data/archive_legacy/landmasks/WFDEI-elevation.nc", # path to where land mask data is located
dir_landmask_tidy = "~/data/archive_legacy/landmasks/WFDEI-elevation_tidy/", # path to where tidy land mask data is to be written
file_elv = "~/data/archive_legacy/landmasks/WFDEI-elevation.nc", # path to where elevation data is located
dir_elv_tidy = "~/data/archive_legacy/landmasks/WFDEI-elevation_tidy/", # path to where tidy elevation data is to be written
save_drivers = TRUE, # whether rsofun driver object is to be saved. Uses additional disk space but substantially speeds up grsofun_run().
dir_drivers = here::here("input/tidy/"), # path where rsofun drivers are to be written
overwrite = FALSE, # whether files with tidy forcing data and drivers are to be overwritten. If false, reads files if available instead of re-creating them.
spinupyears = 10, # model spin-up length
recycle = 1, # climate forcing recycling during the spinup
dir_out = here::here("output/tidy/"), # path for tidy model output
dir_out_nc = "xxx", # xxx not yet handled
save = list( # a named list where names correspond to variable names in rsofun output and the value is a string specifying the temporal resolution to which global output is to be aggregated.
gpp = "mon"
),
nthreads = 1, # distribute to multiple nodes for high performance computing - xxx not yet implemented
ncores_max = 1 # number of parallel jobs, set to 1 for un-parallel run
)par <- list(
kphio = 0.04998, # setup ORG in Stocker et al. 2020 GMD
kphio_par_a = 0.0, # set to zero to disable temperature-dependence of kphio
kphio_par_b = 1.0,
soilm_thetastar = 0.6 * 240, # to recover old setup with soil moisture stress
soilm_betao = 0.0,
beta_unitcostratio = 146.0,
rd_to_vcmax = 0.014, # value from Atkin et al. 2015 for C3 herbaceous
tau_acclim = 30.0,
kc_jmax = 0.41
)Convert forcing data, provided as NetCDF, into a tidy format (time series data frames for each gridcell along rows) and saved for longitudinal bands in a binary format.
tidy_out <- grsofun_tidy(settings)
Read forcing data and construct the driver object for an rsofun run. Run the model and save outputs in a tidy format in .rds files for each longitudinal band.
error <- grsofun_run(par, settings)Read outputs saved in separate files for each longitudinal band. Collect data from all longitudinal bands and apply temporal aggregation steps as specified by the settings.
This can either write the aggregated data again to tidy files by longitudinal band, or return the temporally aggregated data into the R environment.
df <- grsofun_collect(settings, return_data = TRUE)Plot a map.
coast <- rnaturalearth::ne_coastline(
scale = 110,
returnclass = "sf"
)
# January
gg1 <- df |>
filter(month == 1) |>
ggplot() +
geom_raster(
aes(x = lon, y = lat, fill = gpp),
show.legend = TRUE
) +
geom_sf(
data = coast,
colour = 'black',
linewidth = 0.3
) +
coord_sf(
ylim = c(-60, 85),
expand = FALSE
) +
scale_fill_viridis_c(
name = expression(paste("gC m"^-2, "s"^-1)),
option = "cividis",
limits = c(0, 15)
) +
theme_void() +
labs(
subtitle = "January monthly mean"
)
# July
gg2 <- df |>
filter(month == 7) |>
ggplot() +
geom_raster(
aes(x = lon, y = lat, fill = gpp),
show.legend = TRUE
) +
geom_sf(
data = coast,
colour = 'black',
linewidth = 0.3
) +
coord_sf(
ylim = c(-60, 85),
expand = FALSE
) +
scale_fill_viridis_c(
name = expression(paste("gC m"^-2, "s"^-1)),
option = "cividis",
limits = c(0, 15)
) +
theme_void() +
labs(
subtitle = "July monthly mean"
)
cowplot::plot_grid(
gg1,
gg2,
ncol = 1
)