Read ESM netCDF data for arbitrary variables

The variable data should be a netCDF file of annual grid values. The data is assumed to have its dimensions of time, lat, lon, where lon is the most rapidly varying index, and time is the least.

read.general(
  filename,
  len = NULL,
  tag = basename(filename),
  varname = "tas",
  latvar = "lat",
  lonvar = "lon",
  timevar = "time"
)

Arguments

filename	Name of the input netCDF file
len	Maximum length of the time series to read. If the data read is longer, it will be trimmed. (Default: read entire time series, regardless of length.)
tag	A string identifying the name of the scenario. If omitted, the tag will default to the filename.
varname	Name of the variable to read from the netcdf file.
latvar	Name of the latitude dimension variable in the input file.
lonvar	Name of the longitude dimension variable in the input file.
timevar	Name of the time dimension variable in the input file.

Value

A griddata list (see details).

Details

The output will be a list with six fields:

vardata

Matrix of variable data.

globalop

Vector operator for global mean value of the variable.

lat

Vector of latitude values. These are only replicated once, so their length is nlat, not ngrid. You have to use rep(lat, nlon) to get latitudes for each grid cell.

lon

Vector of longitude values. These are only replicated once, so their lingth is nlon, not ngrid. Getting longitudes for each grid cell is tricksy. Try as.vector(matrix(rep(lon, nlat), nrow=nlat, byrow=TRUE)). Fortunately, you don't need them very often.

time

Vector of time values, given as years since the base year of the dataset.

tags

A list of datasets that were concatenated to get this structure. The names of the list are the tags given to the dataset, and the values are a vector of start-row, end-row pairs.

The data at each time is represented as a flattened vector of grid cells. The flattening is performed by transposing to lat, lon ordering, so that lat will be the most rapidly varying index (because R uses Fortran-style indexing). Then the spatial dimensions are discarded, resulting in a 1D vector. The time dimension is kept, resulting in a matrix with years in rows and grid cells in columns. The dimensions of the matrix will be Nyear x Ngrid, where Nyear is the number of years in the data set, and Ngrid is the number of grid cells.

The globalop vector is a vector whose dot product with a flattened grid is the area-weighted global mean for the grid. It is stored as a column vector, so vardata %*% globalop is the time series of global mean values.

The lat and lon dimension variables from the input file are also stored in the structure. These are primarily useful for writing out generated grids as netCDF files. The time dimension variable is converted to integer years, starting at 0.

Conventionally, we refer to the output list as griddata. Notably, any other function with a griddata argument is expecting one of these structures.