What is the easiest way to create map using HDF-5 data subset?

Question

I need to visualize some area using values provided in HDF-EOS version 5 file and create map using those values and other GIS layers. Unfortunately, my QGIS do not recognize the HDF5 file.

Do I really need to compile GDAL manually as described at: SciGeo?

Maybe there's an easier way to do that? I thought about conversion to the GeoTIFF. However HEG tool crashes during conversion.

I opened my product in HDFView, but I don't see any option to generate GeoTIFF in this application.

Last, but not least, I've tried R! I read appropriate data set, but I don't know what to do next. Is there any possibility to save this data in different format? Or maybe I am able to (how? I can't find this in h5 package documentation)

My current code is:

pacman::p_load(h5)

# download HDF-5 file
destfile <- 'SMAP.h5'
if(!file.exists(destfile))
{
  download.file('ftp://n5eil01u.ecs.nsidc.org/SAN/SMAP/SPL2SMAP.003/2015.07.07/SMAP_L2_SM_AP_02293_D_20150707T084111_R13080_001.h5',destfile)
}

# read file
file <- h5file(destfile ,'r')
dataset <- file["Soil_Moisture_Retrieval_Data"]["soil_moisture"]

# @Spacedman: Does your R have a GDAL that supports these files?
subset(rgdal::gdalDrivers(),grepl("HDF",long_name))
# it outputs:
#         name    long_name create  copy
# 51 HDF4Image HDF4 Dataset   TRUE FALSE
# 53 HDF5Image HDF5 Dataset  FALSE FALSE

Answer 1

That particular file, or at least that particular dataset within that file, is not a regular grid, and so can't be converted (easily) to a GeoTIFF file or read in as a raster data source.

> long <- file["Soil_Moisture_Retrieval_Data"]["longitude"]
> str(long)
Formal class 'DataSet' [package "h5"] with 7 slots
  ..@ name       : chr "longitude"
  ..@ datatype   : chr "d"
  ..@ dim        : num 297120
  ..@ maxdim     : num 297120
  ..@ chunksize  : num 297120
  ..@ compression: chr "H5Z_FILTER_DEFLATE"
  ..@ pointer    :<externalptr> 

You can use ReadDataSet to get the data:

> lat <- readDataSet(file["Soil_Moisture_Retrieval_Data"]["latitude"])
> long <- readDataSet(file["Soil_Moisture_Retrieval_Data"]["longitude"])
> moisture <- readDataSet(file["Soil_Moisture_Retrieval_Data"]["soil_moisture"])

What you now have is three long vectors of long, lat, moisture values. If you plot(long,lat) you'll see how these points look like they come from a pass of the satellite over a certain part of the globe, and so can't be made into a regular grid without it being 90% missing data.

I think that if you wanted a small square of data within the satellite pass then that might be convertible to a raster - if the above works for you and you can't work out how to subset it then open a new question - this answer should suffice for how to extract data in general from these files.

Note most of my info on this I got from running h5dump, h5ls and h5stat on the file from the Linux command line.

Answer 2

You can find a solution to read the data at Opening data from NSIDC in QGIS?.

SMAP L2 files work with Panoply, while SMAP L3 is readable with GDAL.