TROLL forest simulator relies on climate tables with half-hourly variations
of a typical day and monthly variations of a typical year which are recycled
through simulation days and years. Initially, TROLL climate tables were
computed from the Nouraflux dataset. Variations in quantities of interests
(temperatures, ...) were averaged to the target resolution (half-hour for
daily variation or month for monthly variation). The purpose of climate
generation functions is to compute equivalent climate tables from the ERA5
land reanalysis dataset (Muñoz-Sabater et al. 2021). With these functions,
rcontroll users only need inventories and associated functional traits to
run TROLL simulations. See the corresponding vignette
vignette("climate", package = "rcontroll") for further details.
generate_climate(
x,
y,
tz,
era5land_hour,
era5land_month,
daytime_start = 7,
daytime_end = 19
)num. Longitude in UTM. Can be obtained from the location name with
nominatimlite::geo_lite_sf().
num. Latitude in UTM. Can be obtained from the location name with
nominatimlite::geo_lite_sf().
num. Time zone. Can be obtained from the coordinates with
lutz::tz_lookup_coords().
str. Path to ERA5 land data monthly averaged reanalysis
by hour of day in netCDF. See the corresponding vignette
vignette("climate", package = "rcontroll") to download corresponding
data from Copernicus in R.
str. Path to ERA5 land data monthly averaged reanalysis
in netCDF. See the corresponding vignette vignette("climate", package
= "rcontroll") to download corresponding data from Copernicus in R.
int. Daytime starting hour to compute nigh and day variables (default 7).
int. Daytime ending hour to compute nigh and day variables (default 19).
A list with two data.frame(): daytimevar and climatedaytime12.
The TROLL forest model simulates tree growth based on ecophysiological
processes, with an external climate forcing. Input climatic conditions are
provided in the form of climate tables with (i) half-hourly standardised
variation of a typical day, and (ii) monthly average values of a typical
year, which are currently recycled through simulation. Initially, TROLL
climate tables were computed from the Nouraflux dataset (Poncy et al., 1998).
The variation in quantities of interest (irradiance, temperature, vapour
pressure, rainfall, and wind speed) were averaged to the target resolution
(half-hour for daily variation or month for monthly variation).
The purpose of the climate generation function is to compute equivalent
climate tables from a global climatic reanalysis dataset. With
generate_climate, rcontroll users no longer need to format complex
climate input fields, but can generate them from global and carefully
documented climate distributions to run TROLL simulations. The selected
input climate product for this version of rcontroll is ERA5-Land
(Muñoz-Sabater et al. 2021). The ERA5-Land climate reanalysis has two main
advantages over other climate reanalysis products: (1) the data are at a
spatial resolution of 9km and have been available at hourly temporal
resolution since 1950, and (2) daily or monthly averages are available and
their uncertainties are reported.
The following assumptions are made in the generation of climate data:
The temperature at 2m and its derivatives (d2m) from ERA5-land
corresponds to air temperature measurement used as an input in TROLL;
We can calculate the vapour pressure deficit using the Buck equation;
The extraction of standardised half-hourly values of an average day and of monthly average values of a year for the climate variables of interest is based on the decomposition of the raw time series into: (i) an overall trend over the study period, (ii) seasonal or daily variation across months or hours depending on the study level, and (iii) the remaining variation;
Half-hourly values are not available for ERA5-land data. Spline functions are used to interpolate hourly values for downscaling to half-hourly resolution.
TROLL variablesTROLL climate tables summarise temporal variation of quantities of interest. These variations are called seasonal pattern and are computed from time series under the additive assumption :
\(X(t) = Trend_X(t)+Seasonal_{x,Period}(t \mod p [mod Period]) + Irregular(t)\)
With :
\(Trend_X(t)\) a moving average covering one period;
\(Seasonal_{x,Period}(t \mod p [mod Period])\) the seasonal pattern contribution at t time modulo the period;
\(Irregular(t)\) the remainder part.
The de-seasonally average of X is defined as :
\(mean(X)=mean(Trend_X(t))\)
These values of seasonal pattern and de-seasonally average are used to compute the climate table TROLLv3_climatedaytime12 and TROLLv3_daytimevar.
There is a restricted set of variables needed to generate the TROLL climate files:
t2m: Temperature at 2m in K
d2m: Dew point temperature at 2m in K
tp: Cumulative rainfall in m
sp: Atmospheric pressure at the surface in Pa
ssrd: Cumulative Net solar irradiation in J/m2
u10: Zonal wind component at 10m in m/s
v10: Meridional wind component at 10m in m/s
The transition from ERA-Land data to TROLL data requires several
transformations. The TROLL climate files correspond to seasonal components,
either daily from 7am to 7pm, or monthly. The extraction of these seasonal
components is possible by analysing the time series of the data. An additive
decomposition of the variables allows one to obtain the pattern of interest
at the original resolution (hourly or monthly). Interpolation of the pattern
using spline functions of the periodic type, ensuring the boundary conditions
(i.e. value at 0 am is the same as 12 pm), except for the ssrd which is not
a continuous periodic function and which requires natural type spline
interpolation. An evaluation of the quality of the pattern extraction is
possible by measuring the standard deviation of the error to the original
time series. This error can be calculated for each unit of the pattern (for
each hour for example).
The wind speed is the norm of the vector generated by the u10 and v10 components of the wind. We can therefore deduce that the wind speed corresponds to:
\(WindSpeed= \sqrt{u10^2+v10^2}\)
The calculation of the vapour pressure deficit can be done according to three
variables (t2m, d2m and sp) using the formula of (Buck 1981):
\(VPD =e_{sat}(d2m,sp)-e_{obst}(2m,sp)\)
\(e_*(t2m|d2m,sp)=611.21xf(t2m|d2m,sp)x(1)\)
\((1)=18.678-(t2m|d2m-273.15)/234.5x(t2m|d2m-273.15)/ (240.97+t2m|d2m-273.15)\)
\(f(t2m|d2m,sp)=1.0007+10^{-7}xspx0.032+5.9x10^{-6}x(t2m|d2m-273.15)^2\)
In conclusion, despite some discrepancies between climate input generated
from local meteorological station data and the generate_climate function
that should be investigated further, the generate_climate function allows
rcontroll users to easily obtain relevant climate data for their study. The
discrepancies may be partly due to the unconventional situation of the
Nouraflux station, which should not be considered as the true climate.
Warning: As TROLL is under constant development, some of the variables
presented here may not be used in the current version (v 3.1.7) and may be
left over from previous versions or may be intended for future versions.
Furthermore, this supplementary information corresponds to the version 3.1.7
of TROLL and the climate variables used by the model may change as new
versions of TROLL are released. We plan to include future major
developments of TROLL in rcontroll to keep the advances of the model
accessible to the community, including the development of the
generate_climate function. We thus invite the reader to check the
corresponding updated vignette on GitHub
(https://sylvainschmitt.github.io/rcontroll/articles/climate.html) according
to the version of TROLL they are using in rcontroll (check with
TROLL.version()).
Buck, Arden L. (1981) New equations for computing vapor pressure and enhancement factor. Journal of Applied Meteorology and Climatology, 1981, vol. 20, no 12, p. 1527-1532.
Muñoz-Sabater, J., Dutra, E., Agustí-Panareda, A., Albergel, C., Arduini, G., Balsamo, G., … Thépaut, J. N. (2021). ERA5-Land: A state-of-the-art global reanalysis dataset for land applications. Earth System Science Data, 13(9), 4349–4383. https://doi.org/10.5194/essd-13-4349-2021
Poncy, O., Riéra, B., Larpin, D., Belbenoit, P., Jullien, M., Hoff, M., & Charles-Dominique, P. (1998). The permanent field research station “Les Nouragues” in the tropical rainforest of French Guiana: current projects and preliminary results on tree diversity, structure, and dynamics. Forest Biodiversity in North, Central and South America, and the Caribbean: Research and Monitoring., 385–410.