fluxtower-sites

The goal of fluxtower-sites is to estimate total above ground biomass for the "footprints" of flux towers from a variety or raster data products. This is a collaboration between CCT Data Science and the Moore lab at University of Arizona.

Files

• data/ contains raw data and metadata
• R/ contains R scripts with functions to be source()ed
• notes/ contains R scripts with work-in-progress workflows

Reproducibility

[!NOTE] External data is required for this code to be reproducible. See EcosystemEcologyLab/AGB-products for code and metadata on how this external data was created.

There are two options for running the workflow. I recommend the "targets" version, but the "manual" version may be easier to edit and understand.

Manual

Run extract_agb_fluxtowers.R. This extracts data for all the data products and combines the results into a single .csv file. This workflow is very slow because of some high resolution data products. There is an option to un-comment some code to run the extraction function on a single product at a time.

Targets

Simply run targets::tar_make(as_job = TRUE) to kick off the pipeline. as_job = TRUE runs this as a background process so your R console is still free. You can monitor the progress of the pipeline by running targets::tar_visnetwork() periodically or by using tar_watch() to launch a Shiny app. When the pipeline completes, a .csv file for each data product will be saved to output/. You can read them all in as a single data frame with:

library(readr)
agb <- read_csv(list.files("output", full.names = TRUE))

This is also quite slow for some of the products (first run of the pipeline took over a day with 2 workers and 25GB of RAM). Depending on your computer's specifications you may improve performance by increasing the number of workers used by editing the crew_controller_local() function arguments in _targets.R. You might also speed things up by increasing the value for the max_cells_in_memory in the extract_mean_agb() function call in _targets.R---this allows more pixels to be loaded into memory. Keep in mind that this value is per worker, so if you have many parallel workers you probably need to decrease max_cells_in_memory. Also note that any edits to functions in the pipeline will invalidate all downstream targets and cause them to be re-run when you next run tar_make().

Developed in collaboration with the University of Arizona CCT Data Science team