Make output of lasclip() a catalog() instead of a list()

[ptompalski]

When processing the point cloud data to create wall-to-wall forest inventory layers, the first step is to calculate point cloud metrics for reference plots. It is a simple task that consist of several steps:

1. create a catalog containing all point cloud data
2. load plot data (e.g. as sf() object)
3. Use lasclip() to clip point clouds to plot borders
4. Calculate metrics using lasmetrics()

The issue here is that the output of lasclip() is a list(). Therefore to calculate metrics we cannot use lasmetrics(clipped_plots, .stdmetrics), but rather calculate the metrics using lapply or a simple for loop. This is not consistent with other functions in lidR, that in most cases accept either a single las file as input, or catalog.

I suggest two changes:

• change the output of lasclip from a list() to a catalog(). This would allow to use functions like catalog_apply(), use multi-core processing etc. (all benefits of catalogs). Such catalog would not be a wall-to-wall object of course, and should processed file-by-file.
• allow lasmetrics() to accept catalog() as input. In that case the function would run on every plot returning an output similar to grid_metrics().

[Jean-Romain]

Your requests makes a lot of sense however your first request already exists.

library(lidR)

# Load a LAScatalog
LASfile <- system.file("extdata", "Megaplot.laz", package = "lidR")
ctg <- readLAScatalog(LASfile)

# Load the plot centers
plots = matrix(c(684800, 684850, 5017850, 5017900), ncol = 2)
plots = SpatialPointsDataFrame(plots, data.frame(NAME = c("P1", "P2")))

# Visual check
plot(ctg)
plot(plots, add = T)

# Set output_files option to write the results on disk
opt_output_files(ctg) <- paste0(tempdir(), "/{NAME}")

# new_ctg is a LAScatalog
new_ctg <- lasclip(ctg, plots, radius = 10)
plot(new_ctg)

Notice that in the specific case of lasclip there is an internal hack to bypass regular catalog_apply behavior. The point cloud is not loaded in memory at all. The points are streamed from input files to output files.

Your second requests is also pertinent however I cannot implement it for consistency reasons. Indeed what you are asking has a meaning only for the specific case of not wall-to-wall catalog of plots. But in the general case there is no meaning at computing lasmetrics on a LAScatalog. This why I won't add this capability to lasmetrics.

However I can modify lasmetrics to allows the following:

opt_chunk_buffer(new_ctg) <- 0
opt_chunk_size(new_ctg) <- 0
opt_filter(new_ctg) <- "-keep_first"
opt_output_files(new_ctg) <- ""
output <- catalog_apply(new_ctg, lasmetrics, func = .stdmetrics)
output <- data.table::rbindlist(output)

plots@data <- cbind(plot@data, output)

This is equivalent to lasmetrics(new_ctg, .stdmetrics) but force the user to use catalog_apply and thus master the LAScatalog processing engine.

[Jean-Romain]

Done in lidR 2.1.1

[ptompalski]

this is great JR! thanks a lot!