The Temperature Range Squeeze hypothesis

• Description
• Usage
• Climate data
• Dependencies

Description

Repository to reproduce the analyses in the paper “Diurnal temperature range as a key predictor of plants’ elevation ranges globally”.

Usage

Simply use the main.R file to run all the analyses performed in the manuscript.

You can run multiple models automatically using a set of for loops. For example:

# global-scale analyses

model <- load_model(scope = "global")

for (expr in c(~dtr, ~ts, ~past_dmat)) {
  for (elevation_span in ELEV_SPANS) {
    for (exclusion_zone in EXCLS) {
      mdl_data <- compile_mdl_data(
        trs,
        clim_data = trs_bioclim,
        elevation_span = elevation_span,
        exclusion_zone = exclusion_zone,
        singleton_thr = SINGLETON_THR,
        std_elev_grad = TRUE,
        average = TRUE,
        std_from = "top",
        cols = c("location", "sp_range", "land_type"),
        expr = expr
      )
      
      run_jags(
        mdl_data,
        model = model,
        n.iter = JAGS_ITER,
        n.thin = JAGS_THIN,
        n.chains = JAGS_CHAINS,
        n.burnin = JAGS_BURN_IN,
        save = TRUE,
        path = glue("data/jags/{model$scope}-scale/")
      )
      
      gc() # to free up RAM
    }
  }
}

Note that the expr parameter in compile_mdl_data() takes a formula for the global-scale analyses and a character string for the local-scale analyses.

A list of the different models ran in the manuscript can be found below.

Once the models have finished running, use:

• results.R to plot model estimates and get model statistics
• mdl_diagnostics.R to perform model diagnostics

Models

• Global-scale analyses:
  • main models: ~dtr, ~ts, ~past_dmat
  • land type: ~dtr * land_type, ~ts * land_type, ~past_dmat * land_type
  • climate-related interactions: ~dtr * ap, ~dtr * mat, ~past_dmat * past_map
• Local-scale analyses: "dtr", "ts"

Gilchrist’s hypothesis

Run the following line to make the required data set to reproduce the analyses on Gilchrist’s hypothesis.

trs_dtr_lower_third <- filter(trs, bio2 <= max(bio2, na.rm = TRUE) / 3)

Run the ~past_dmat model using the trs_dtr_lower_third data set.

model <- load_model(scope = "global")

for (elevation_span in ELEV_SPANS) {
  mdl_data <- compile_mdl_data(
    trs_dtr_lower_third,
    clim_data = trs_bioclim,
    elevation_span = elevation_span,
    exclusion_zone = EXCL_DEFAULT,
    singleton_thr = SINGLETON_THR,
    std_elev_grad = TRUE,
    average = TRUE,
    std_from = "top",
    cols = c("location", "sp_range", "land_type"),
    expr = ~past_dmat
  )
  
  run_jags(
    mdl_data,
    model = model,
    n.iter = JAGS_ITER,
    n.thin = JAGS_THIN,
    n.chains = JAGS_CHAINS,
    n.burnin = JAGS_BURN_IN,
    save = TRUE,
    path = "data/jags/global-scale/gilchrist hypothesis/"
  )
}

Use the GlobalAnalyses class to plot model estimates:

gh <- GlobalAnalyses$new("data/jags/global-scale/gilchrist hypothesis")

gh$regressions(
  vars = "past_dmat",
  facet_cols = "elevation_span",
  labels = "∆ mean annual temperature (0-1980) (°C)"
)

gh$posterior_distributions(
  vars = "past_dmat",
  yvar = "elevation_span",
  facet = FALSE,
  reverse = TRUE,
  scales = .6
)

Climate data

Climate data are already provided in the trs.csv data set.

If you would like to perform GIS analyses again, proceed as follows:

1. Open TRS.Rproj and execute the following function in the console: TRS.utilities::setup_gis(). The function will create all the required directories to store SRTM and climate-related data.
2. Download bioclim and digital elevation data (as .tif) as well as PaleoView. See here for instructions to generate past climate data.
3. Save each file in the appropriate folder (see file tree below).
4. Once all the files are in their respective folders, use the gis.R script to extract present and past climate data in each location. Output data will be stored in the gis/clim/extracted/ subfolders.

gis
├── clim
│   ├── data
│   │   ├── past                        <- PaleoView-related folder
│   │   │   ├── base                    <- past climate data go here
│   │   │   └── var
│   │   │       ├── mean precipitation  <- generated mean precipitation files go here
│   │   │       └── mean temperature    <- generated mean temperature files go here
│   │   └── present                     <- present bioclim files go here
│   └── extracted
│       ├── past
│       └── present
└── srtm                                <- save SRTM tiles in the corresponding subfolders
    ├── Afghanistan
    ├── Alborz Mountains
    ├── Azores
    ├── Baekdudaegan Mountains
    ├── Bioko
    ├── Canary
    ├── Cantabria
    ├── Cape Verde
    ├── Chicauma
    ├── Colombian Andes
    ├── Crete
    ├── Cyprus
    ├── Denali
    ├── Drakensberg
    ├── Euboea
    ├── Golestan
    ├── Hawaii
    ├── Hengduan
    ├── Jamaica
    ├── Jaya
    ├── Kenya
    ├── La Amistad
    ├── Lazio
    ├── Mt Ararat
    ├── Mt Etna
    ├── Mt Kilimanjaro
    ├── Nanga Parbat
    ├── Nepal
    ├── Nevada Test Site
    ├── Owens Peak
    ├── Reunion
    ├── Santa Rosa Mountains
    ├── Sierra Nevada
    ├── Sierra San Pedro Martir
    ├── Socotra
    ├── South-Eastern Pyrenees
    ├── Swiss Alps
    ├── Tahoe
    ├── Taiwan
    ├── Tajikistan
    ├── Tasmania
    ├── Utah
    ├── Venezuelan Andes
    └── Wind River Mountains

Generating past climate data

Open PaleoView and download the climate data (see chapter 2 of PaleoView’s user manual) for mean temperature and precipitation. Configure the input data location (see chapter 3 in the user manual). Run PaleoView with the same settings as shown in the image below for temperature and precipitation:

Dependencies

Below are the different R packages required to run the code in this repository:

• tidyverse
• TRS.utilities
• glue
• bayesplot (only for model diagnostics)

To install TRS.utilities, use:

# install.packages("pak")
pak::pak("arnaudgallou/TRS.utilities")

In addition, you will need to have JAGS installed on your machine.