Extracting ERA5-Land data with extract_clima()

[jonathanvonoppen]

This concerns the compatibility of extract_clima() with ERA5-Land for extraction of gridded data. When trying to do that, I get the error:

# specify file 
era5land_nc <- "era5land_alps_2011_1.nc"
# unzip
unzip(gsub(".nc", ".zip", era5land_nc), exdir = gsub(".nc", "", era5land_nc))
# extract data
mcera5::extract_clima(nc = era5land_nc, 
                      long_min = 10.0870, long_max = 10.1445,
                      lat_min = 46.6347, lat_max = 46.6720,
                      start_time = as.POSIXct("2011-01-01 00:00:00", tz = "UTC"), 
                      end_time = as.POSIXct("2011-01-31 23:00:00", tz = "UTC"),
                      format = "microclimf")

Error: [rast] tcc not found. Choose one of: t2m, d2m, sp, u10, v10, tp, ssrd, str, strd

Looking into the code, the varname_list object defined within the function does not yet account for the different sets of variables available in ERA5 and ERA5-Land, respectively. The function thus fails at the following loop trying to rast() the layer.

Might be a quick fix with providing alternative varname_lists conditional to a dataset argument (e.g. "era5" vs "era5land"), or perhaps to allow users to specify a list themselves?

Thanks a lot!

link to zipped file - save in wd

[jonathanvonoppen]

So, since not all variables needed for downstream calculations are available from ERA5-Land anyway, I have found a solution integrating both ERA5 (nc_era5) and ERA5-Land files (era5l) from corresponding time periods. The respective ERA5 layers tcc, msnlwrf, msdwlwrf, fdir and lsm are resampled to the ERA5-Land grid when importing data layers:

var_list <- lapply(varname_list, function(v) {
    if (v == "lsm") {
      # only need one timestep for land-sea mask
      # "lsm" variable is not in ERA5-Land: load from ERA5 data and resample and crop to ERA5-Land grid
      r <- terra::rast(nc_era5[nc_era5 == gsub("land", "", nc_era5l)], subds = v)  # assuming file naming schemes that only differ by "land"
      rref <- terra::rast(nc_era5l, subds = "t2m")
      r <- terra::resample(r, rref, method = "bilinear")
      r <- terra::crop(r, rref)
    } else if(v %in% c("tcc", "msnlwrf", "msdwlwrf", "fdir")) {
      # variables not in ERA5-Land: load from ERA5 data, and resample and crop to ERA5-Land grid
      r <- terra::rast(nc_era5[nc_era5 == gsub("land", "", nc_era5l)], subds = v)
      rref <- terra::rast(nc_era5l, subds = "t2m")
      r <- terra::resample(r, rref, method = "bilinear")
      r <- terra::crop(r, rref)
      # subset down to desired time period
      r <- r[[as.POSIXct(nc_datetimes, tz = "UTC") %in% tme]]
      # Name layers as timesteps
      names(r) <- tme
    } else {
      # For all others, subset down to desired time period
      # terra::time() not identifying time data of ERA5 data from new CDS, so
      # use nc_datetimes
      r <- terra::rast(nc_era5l, subds = v)
      r <- r[[as.POSIXct(nc_datetimes, tz = "UTC") %in% tme]]
      # Name layers as timesteps
      names(r) <- tme
    }
    
    # Subset down to desired spatial extent
    r <- terra::crop(r, terra::ext(long_min, long_max, lat_min, lat_max))
    return(r)
  })

In addition, when extracting coordinates, we need to account for the fact that the t2m template layer may now include NAs, hence adding na.rm = TRUE (using a resampled ERA5 layers as the template instead would work as well).

coords <- as.data.frame(terra::crds(t2m[[1]]))

Some other additions might be useful such as an index-based selection of corresponding ERA5 files, or checking that both nc_era5 and nc_era5l inputs cover the same time period.

Hope this helps!

P.S. unfortunately the corresponding ERA5-Land file is too big to attach here..

[dklinges9]

Hi Jonathan, I'm a bit backed up this week but I'll follow up here shortly!

[jonathanvonoppen]

Just flagging here that apparently and very annoyingly, CDS seem to have changed ERA5-Land .nc file structures.
Among other changes, when loading data with rast(nc_file), variables are now contained in names(rast) instead of the previous varnames(rast), for instance in the form
[1] "SFC (Ground or water surface); 2 metre temperature [C]"
, so my above solution looping through varnames_list won't work anymore for ERA5-Land data.
I'm also looking into it and will let you know if I come up with a solution.