In emf-creaf/medfateland: Mediterranean Landscape Simulation

knitr::opts_chunk$set(echo = TRUE)

About this vignette

The aim of this vignette is to provide users with a rough estimation of computing times for simulation models included in package medfateland.

The results presented here were obtained using 3 months of simulation with the example watershed on a laptop (16 GiB memory and 11th Gen Inter Core I5 processor @ 2.40 GHz x 8) with Ubuntu Linux OS.

medfate_ver <- packageVersion("medfate")

library(medfateland)
library(kableExtra)
library(knitr)

file_out <- paste0("computing_times/ct_", medfate_ver, ".csv")
if(!file.exists(file_out)) {
  data("example_watershed")
  # Set crop factor 
  example_watershed$crop_factor <- NA
  example_watershed$crop_factor[example_watershed$land_cover_type=="agriculture"] <- 0.75
  # Get bounding box to determine limits
  b <- sf::st_bbox(example_watershed)

  # Define a raster topology, using terra package, 
  # with the same CRS as the watershed. In this example cells have 100 m side.
  # Coordinates in the 'sf' object are assumed to be cell centers
  r <-terra::rast(xmin = 401380, ymin = 4671820, xmax = 402880, ymax = 4672620, 
                nrow = 8, ncol = 15, crs = "epsg:32631")

  data("examplemeteo")
  data("SpParamsMED")
  # Set simulation period
  dates <- seq(as.Date("2001-01-01"), as.Date("2001-03-31"), by="day")
  # Watershed control parameters (TETIS model; Frances et al. 2007)
  ws_control <- default_watershed_control("tetis")

  #spwb
  df_res_spwb <- data.frame(model = "spwb",
                       transpirationMode = c(rep("Granier", 3)),
                       soilDomains = rep(c("buckets", "single", "dual"),1),
                       Elapsed_time_sec = NA)
  for(i in 1:nrow(df_res_spwb)) {
    control <- defaultControl(df_res_spwb$transpirationMode[i], df_res_spwb$soilDomains[i])
    control$verbose <- FALSE
    st <- system.time({spwb_land(r, example_watershed, SpParamsMED, examplemeteo, 
                       dates = dates, summary_frequency = "month",
                       watershed_control = ws_control, local_control = control, 
                       progress = FALSE)})
    df_res_spwb$Elapsed_time_sec[i] <- st[[3]]
  }
  #growth
  df_res_growth <- data.frame(model = "growth",
                       transpirationMode = c(rep("Granier", 3)),
                       soilDomains = rep(c("buckets", "single", "dual"),1),
                       Elapsed_time_sec = NA)
  for(i in 1:nrow(df_res_growth)) {
    control <- defaultControl(df_res_growth$transpirationMode[i], df_res_growth$soilDomains[i])
    control$verbose <- FALSE
    st <- system.time({growth_land(r, example_watershed, SpParamsMED, examplemeteo, 
                       dates = dates, summary_frequency = "month",
                       watershed_control = ws_control, local_control = control, 
                       progress = FALSE)})
    df_res_growth$Elapsed_time_sec[i] <- st[[3]]
  }
  df_res <- dplyr::bind_rows(df_res_spwb, df_res_growth)
  names(df_res)[4] <- as.character(medfate_ver)
  write.csv2(df_res, file = file_out, row.names = FALSE)
}

Table of computational times

Computational times were estimated using system.time(), are in seconds and are shown by medfate package version.

files <- list.files("computing_times/", full.names = TRUE)
df_res <- NULL
for(i in 1:length(files)) {
  df_res_i <- read.csv2(files[i])
  names(df_res_i)[4] <- substr(names(df_res_i)[4], 2, nchar(names(df_res_i)[4]))
  if(is.null(df_res)) {
    df_res <- df_res_i
  } else {
    df_res <- df_res |>
      dplyr::left_join(df_res_i, by = c("model", "transpirationMode", "soilDomains"))
  }
}
df_res |>
  dplyr::rename("function" = "model") |>
  kbl() |>
  kable_styling()