Generate apsimx file for sensitivity analysis"

In rapsimng: APSIM Next Generation

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(rapsimng)
library(tidyverse)

Sensitivity analysis is conducted to check output variations when a parameter is changed. This post is to show how to conduct sensitivity analysis using rapsimng package using factorial simulations.

We use the base phyllochron as an example to demonstrate how to generate a new apsimx file from a template.

The base phyllochron is a key parameter for wheat phelonogy and leaf appearance rate. The range of base phyllochron is from 60 to 130 degree days.

The data.frame requires three columns (i.e. parameter, value, name) and multiple parameters can be specified here.

phyllochron_para <- tibble(parameter = "[Phenology].Phyllochron.BasePhyllochron.FixedValue", 
                           value = seq(60, 130, by = 1)) %>% 
  mutate(name = paste0("Cultivar", seq_len(n())))
head(phyllochron_para)

The template is an apsimx file setup for the actual experiment or the specified environments (i.e. locations, sowing date or years). I assume there is a factor Cv for culivar in the Permutation model which specified the cultivar by [Sowing].Script.CultivarName.

template <- read_apsimx(system.file("wheat.apsimx", package = "rapsimng"))

update_cultivar can be used to add the list of cultivars in the Replacements. The Specification in the Permutation.Cv can be replace with new values.

template <- update_cultivar(template, phyllochron_para)

node <- search_path(template, "[Permutation].Cv")    
if (length(node) == 0) {
  stop("[Permutation].Cv is not found")
}
new_value <- paste0("\\1", paste(phyllochron_para$name, collapse = ","))
node$node$Specification <- gsub("(.+ *= *)(.+)$", new_value, node$node$Specification)

node$node$Specification

template <- replace_model(template, node$path, node$node)

Finally the new model can be write into file system and run with APSIM NG. Uncomment the sections below, update the path to Models.exe.

# write_apsimx(template, "new-path.apsimx")
# models_path <- "path-to-Models.exe"
# run_models(models_path, sim_name, csv = TRUE, parallel = FALSE)

The example in this post can be modified into parallel codes for sensivity analysis in the large scales.

rapsimng documentation built on Sept. 9, 2021, 9:07 a.m.