fastbioclim: Scalable Derivation of Climate Variables

fastbioclim is an R package for efficiently deriving standard bioclimatic and custom summary variables from large-scale climate raster data. It is designed to overcome the memory limitations of traditional approaches by intelligently switching between processing backends.

Overview

Working with large climate datasets often presents a major challenge: the data is too large to fit into memory. fastbioclim solves this problem by providing a powerful and unified interface with a dual-backend architecture:

1. In-Memory ("terra"): For datasets that fit comfortably in RAM, fastbioclim uses the highly optimized terra package for maximum speed.
2. In-Disk ("tiled"): For datasets that exceed available RAM, it automatically switches to a memory-safe tiled workflow. This backend uses exactextractr and Rfast to process the data chunk by chunk, ensuring that even continent-scale analyses can run on a standard computer.

The core of the package is a set of "smart" wrapper functions—derive_bioclim() and derive_statistics()—that automatically select the best backend, providing a seamless experience for the user.

Key Features

• Smart auto Method: Automatically detects data size and chooses the optimal processing workflow (in-memory vs. tiled).
• Scalability: Robustly handles raster datasets of any size, from small study areas to global extents.
• Comprehensive Bioclimatic Variables: Calculates the full set of 35 standard bioclimatic variables, including those based on temperature, precipitation, solar radiation, and moisture.
• Custom Statistics: Provides a flexible derive_statistics() function to compute custom summaries (mean, max, min, standard deviation, etc.) for any climate variable.
• Flexible Time Periods: Easily define custom periods (e.g., weeks, semesters) instead of being limited to standard quarters.
• Advanced Control: Supports static index rasters for advanced time-series analysis and defining specific seasonal periods.

Installation

You can install the development version of fastbioclim from GitHub with:

# install.packages("remotes")
remotes::install_github("gepinillab/fastbioclim")
# Install to get the package example data 
remotes::install_github("gepinillab/egdata.fastbioclim")

Usage

The package provides two primary, user-facing functions:

• derive_bioclim(): For calculating the standard set of 19-35 bioclimatic variables.
• derive_statistics(): For calculating custom summary statistics on any variable.

Example: A Quick Demonstration

This example demonstrates the core functionality using simple, self-contained mock data.

library(fastbioclim)
library(terra)
library(future.apply)
library(progressr)

tmin_ecu <- system.file("extdata/ecuador/", package = "egdata.fastbioclim") |>
  list.files("tmin", full.names = TRUE) |> rast()
tmax_ecu <- system.file("extdata/ecuador/", package = "egdata.fastbioclim") |>
  list.files("tmax", full.names = TRUE) |> rast()
prcp_ecu <- system.file("extdata/ecuador/", package = "egdata.fastbioclim") |>
  list.files("prcp", full.names = TRUE) |> rast()

# The function will automatically use the fast "terra" method for this small dataset
output_dir_bioclim <- file.path(tempdir(), "bioclim_ecuador")

bioclim_vars <- derive_bioclim(
  bios = 1:19,
  tmin = tmin_ecu,
  tmax = tmax_ecu,
  prcp = prcp_ecu,
  output_dir = output_dir_bioclim,
  overwrite = TRUE
)

plot(bioclim_vars[[c("bio01", "bio12")]])

# 3. Derive custom summary statistics for a different variable (e.g., wind speed)
wind_rast <- system.file("extdata/ecuador/", package = "egdata.fastbioclim") |>
  list.files("wind", full.names = TRUE) |> rast()
output_dir_custom <- file.path(tempdir(), "wind_ecuador")

custom_stats <- derive_statistics(
  variable = wind_rast,
  stats = c("mean", "max", "stdev"),
  prefix_variable = "wind",
  output_dir = output_dir_custom,
  overwrite = TRUE
)

plot(custom_stats)

Handling Large-Scale Data (The Tiled Workflow)

The real power of fastbioclim shines with large datasets. The method = "auto" setting in derive_bioclim() and derive_statistics() handles this automatically.

When the wrapper function detects that the input rasters are too large to fit in memory, it seamlessly switches to the tiled workflow.

Important Requirement: For the tiled workflow to function, your input SpatRaster objects must be pointing to files on disk, not held entirely in memory.

# Conceptual example for large, file-based rasters
tmin_neo <- system.file("extdata/neotropics/", package = "egdata.fastbioclim") |>
  list.files("tmin", full.names = TRUE) |> rast()
tmax_neo <- system.file("extdata/neotropics/", package = "egdata.fastbioclim") |>
  list.files("tmax", full.names = TRUE) |> rast()
prcp_neo <- system.file("extdata/neotropics/", package = "egdata.fastbioclim") |>
  list.files("prcp", full.names = TRUE) |> rast()
output_dir_bios <- file.path(tempdir(), "bioclim_neotropics")

# Optional: ACTIVATE PROGRESS BAR
progressr::handlers(global = TRUE)
# Optional: DEFINE PARALLEL PLAN FOR EVEN FASTER PROCESSING
future::plan("multisession", workers = 4)

# The call is identical. `derive_bioclim` will detect the large file size
# and automatically use the memory-safe tiled method.
large_scale_vars <- derive_bioclim(
  bios = 1:19,
  tmin = tmin_neo,
  tmax = tmax_neo,
  prcp = prcp_neo,
  output_dir = output_dir_bios,
  tile_degrees = 20,
  overwrite = TRUE
)
print(large_scale_vars)
plot(large_scale_vars[["bio11"]])

Under Active Development

This R package is currently under active development. While it is functional, it may contain bugs or undergo changes to the API.

Contributions, bug reports, and feature requests are highly encouraged. Please open an issue on our GitHub repository to provide feedback.