Do you mean hdf5?
I extensively used COGs (cloud optimised geotiffs) and NetCDF4 (based on hdf5) at work over the last 10 years. Both have their pros and cons.
The main limitation with geotiff is its pretty much only usable for layered 2D raster data.
NetCDF4 (hdf5) can set up frames of any dimensionality, you can have datetime axes, time series data, 100d ensemble data, etc.
Yeah. h5 is the typical industry shorthand and file extension.
The h5 saga was NASA saying “we’re going to create a file format that does EVERTHING”, and well… it does… poorly.
Everything that h5 is allegedly better for is better solved by just moving to either sql or postgres. And if the data aren’t that complex, then just send me a geotiff.
If you send me an h5 the first thing I’m doing is moving it over to sqlite or postgres.
Litterally a scientist working with NASA data and it’s scientists and I’m not the intended audience?
I’ve been in the remote sensing game almost 25 years. And a good amount of that at the federal government. I’ve sat at the table and shared beers and dinner with the chief scientists behind the modis and gedi mission. There isn’t a geospatial data type or format I haven’t encountered, and half of them I’ve buried.
So please, spare this old hand any lectures.
The fact is geospatial has been able to explode because we finally got away from these kinds of anachronistic approaches to data. It’s litterally never been a better time to be a geospatial data scientist. Praise be that the age of h5s and local processing is over.
Do you mean hdf5? I extensively used COGs (cloud optimised geotiffs) and NetCDF4 (based on hdf5) at work over the last 10 years. Both have their pros and cons.
The main limitation with geotiff is its pretty much only usable for layered 2D raster data.
NetCDF4 (hdf5) can set up frames of any dimensionality, you can have datetime axes, time series data, 100d ensemble data, etc.
Yeah. h5 is the typical industry shorthand and file extension.
The h5 saga was NASA saying “we’re going to create a file format that does EVERTHING”, and well… it does… poorly.
Everything that h5 is allegedly better for is better solved by just moving to either sql or postgres. And if the data aren’t that complex, then just send me a geotiff.
If you send me an h5 the first thing I’m doing is moving it over to sqlite or postgres.
HDF5 was designed for multidimensional numeric arrays, which are particularly ill-suited to putting in a classic relational DB.
It’s a scientific data format, not an image format and it sounds like you’re not the intended audience.
Litterally a scientist working with NASA data and it’s scientists and I’m not the intended audience?
I’ve been in the remote sensing game almost 25 years. And a good amount of that at the federal government. I’ve sat at the table and shared beers and dinner with the chief scientists behind the modis and gedi mission. There isn’t a geospatial data type or format I haven’t encountered, and half of them I’ve buried.
So please, spare this old hand any lectures.
The fact is geospatial has been able to explode because we finally got away from these kinds of anachronistic approaches to data. It’s litterally never been a better time to be a geospatial data scientist. Praise be that the age of h5s and local processing is over.