Skip to contents

Annual AET on a grid from MODIS MOD16A3GF v061

Source: R/wet_mod16_aet.R
wet_mod16_aet.Rd

Downloads the MOD16A3GF v061 granules (Running, Mu and Zhao, NASA LP DAAC, doi:10.5067/MODIS/MOD16A3GF.061; gap-filled annual ET, 500 m sinusoidal) for every sinusoidal tile grid touches and every year in years, averages the annual ET per pixel, and puts the mean on grid. MOD16 is a Penman-Monteith model driven by MODIS land cover, LAI and albedo and by GMAO meteorology, so unlike a soil bucket it is not capped by a precipitation field.

Usage

wet_mod16_aet(
  grid,
  years = 2001:2020,
  dir = "data/mod16",
  overwrite = FALSE,
  timeout = 600,
  fact = 4,
  min_years = 10
)

Arguments

grid

Target grid: a SpatRaster or a path to one, in lon/lat (EPSG:4326).

years

Calendar years to average.

dir

Cache directory for the granules and the result.

overwrite

Logical. Rebuild the grid even when cached.

timeout

Seconds allowed for each download (a granule is about 20 MB).

fact

Subdivision of each grid cell for the nearest-neighbour step; 4 makes a 30 arc-second cell about 230 x 130 m at 55 N, finer than 500 m.

min_years

Valid years a pixel needs for its mean to count.

Value

Path to a GeoTIFF with layers et_mod16 (mm per year) and frac_mod16.

Details

The ET_500m layer (kg/m2/yr, i.e. mm/yr; stored as unsigned 16-bit with a 0.1 scale that terra applies) has a valid range of 0-65500. Codes 65529-65535 mark pixels the product does not model (unclassified, urban, permanent wetland, snow and ice, barren, water, fill), so every value above the valid range is a gap. A pixel's mean is taken over its valid years and kept only when it has at least min_years of them, since its land-cover class can change between years.

The tiles are mosaicked in their sinusoidal CRS and put on grid in two steps, as in wet_landcover_nrcan(): nearest onto a grid fact times finer than grid and aligned with it, then block means. A single warp would take each target cell's footprint as an axis-aligned box in the source CRS, and the sinusoidal grid is strongly sheared at BC's longitudes. The result has two layers: et_mod16, the mean over the valid part of each cell, and frac_mod16, the share of the cell that is valid (0-1). Where the share is 0, et_mod16 is NA; filling the gaps is left to the caller, and frac_mod16 lets it weight the fill by area.

Granules are listed through NASA's CMR search (no login) and downloaded from LP DAAC with an Earthdata login read from a netrc file: the option wet.earthdata_netrc, otherwise the first of $NETRC, the file earthdatalogin keeps (tools::R_user_dir("earthdatalogin")/netrc) and ~/.netrc that has a machine urs.earthdata.nasa.gov entry. Downloads are written to .part, renamed on HTTP 200 and opened before they are kept, since a failed login can end on a page served with 200. Granules already in dir are used without a search, and the grid is cached under a key that includes the md5 of every granule read.

Examples

if (FALSE) { # interactive()
# needs an Earthdata login in a netrc; about 20 MB a granule on first use
g <- terra::rast(xmin = -120, xmax = -119, ymin = 50, ymax = 51, res = 1 / 120)
m <- terra::rast(wet_mod16_aet(g, years = 2010:2011, min_years = 1))
terra::global(m, "mean", na.rm = TRUE)
}