Rasters

CI Codecov Aqua.jl Quality Assurance

Rasters.jl defines common types and methods for reading, writing and manipulating rasterized spatial data.

These currently include raster arrays like GeoTIFF and NetCDF, R grd files, multi-layered stacks, and multi-file series of arrays and stacks.

EarthEnv HabitatHeterogeneity layers trimmed to Australia

A RasterStack of EarthEnv HabitatHeterogeneity layers, trimmed to Australia and plotted with Plots.jl

Lazyness

  • Data is loaded lazily wherever possible using DiskArrays.jl. Indexing a RasterStack by name is always lazy, while view of a Raster is lazy and getindex will load to memory. read can be used on any object to ensure that all data is loaded to memory.
  • Broadcast over disk-based objects is lazy - it will only run when the array is indexed. Always prefer broadcasts to explicit loops - these can be very slow with disk-based data.

Data-source abstraction

Rasters provides a standardised interface that allows many source data types to be used with identical syntax.

  • Scripts and packages building on Rasters.jl can treat AbstractRaster, AbstractRasterStack, and AbstrackRasterSeries as black boxes.
    • The data could hold GeoTiff or NetCDF files, Arrays in memory or CuArrays on the GPU - they will all behave in the same way.
    • RasterStack can be backed by a Netcdf or HDF5 file, or a NamedTuple of Raster holding .tif files, or all Raster in memory.
    • Users do not have to deal with the specifics of spatial file types.
  • Projected lookups with Cylindrical projections can by indexed using other Cylindrical projections by setting the mappedcrs keyword on construction. You don't need to know the underlying projection, the conversion is handled automatically. This means lat/lon EPSG(4326) can be used seamlessly if you need that.
  • Regions and points selected with Between and Contains select the right point or whole interval no matter the order of the index or it's position in the cell.

Named dimensions and index lookups

Rasters.jl extends DimensionalData.jl so that spatial data can be indexed using named dimensions like X, Y and Ti (time) and e.g. spatial coordinates.

Dimensions can also be used in most Base and Statistics methods like mean and reduce where dims arguments are required. Much of the behaviour is covered in the DimensionalData docs.

See the docs for more details and examples for Rasters.jl.