Process-based models (PBMs, see table 1 for a list of abbreviations) are essential tools to assess ecosystem response to climate change, land use changes, extreme weather patterns, or other environmental disturbances. PBMs allow to deal with the high level of interactions and feedbacks which are intrinsic to ecological processes, but their complexity comes at the cost of computation time and memory. Because of that, there is a trade-off between the resolution satisfactory to describe ecological processes (e.g. organism, population, species, vegetation type …), and the computing constraints when applying PBMs at a large spatial extent and/or long time periods.
This script was used to construct metamodel and to perform scale analysis
Software : R; Packages : lattice, polspline, lhs, ade4, caret, neuralnet, nnet, MASS, Hmisc, Cairo; functions are loaded from the library.R file.
script_analysis_article.R
R library where the functions are described
Packages used : lattice, polspline, lhs, ade4, caret, neuralnet, nnet, MASS, Hmisc, Cairo.
library.R
Latin hypercube design simulations
Each ".res" file is a one day simulation results of noTG model. There are three design of simulation : construct LHS for the construction of the meta-models; test LHS for neural network metamodel and polynomial regression spline optimizations; and Valid LHS for intercomparaison of different metamodel techniques. A readme file is attached to describe variables of ".res" files.
LHS_9_inputs.rar
Constructed metamodels
A rar file with all constructed meta-models. Meta-model files can be loaded in to R script with the function "load".
metamodels.rar
50 and 100 years simulations with noTG model
Data used to provide a multi-scale validation test. Each simulation is stocked in one repertory. The name of the repertory was generated as follow : "meta" or "mod" for simulation made with or without meta-model; TC is the name of the plot; the number of the sub-plot; the simulated period. Into each repertory, there are five types of output file: - "finals_plant.s" : where dimensions, carbon and nitrogen pools of each tree are recorded at the end of the simulation. - "finals_sol.s" : where carbon, nitrogen, ligin and water content of each pixel are recorded at the end of the simulation. - "XXX_TC_X_XXX.Day1D.res" : where species and soil state variables are recorded for each species and for each day. - "XXX_TC_X_XXXTotal.res" : where cumulated species and soil state variables are recorded for each species and for entire period. - "XXX_TC_X_XXXTree_X.res" : where tree state variables are recorded for each day.
Simulations.rar