IIRSpy

Toolkit for working with ISRO's Chandrayaan-2 IIRS dataset. This is a work in progress - comes with no guarantees, is not yet tested, and is subject to change without notice!

Installation

1. Fork / clone this repository.
2. Get uv package manager installation
3. From the main iirspy directory run uv run make test
   • This will make a new virtual environment and run the iirspy tests
   • You can now use this venv from an IDE or uv run ipython for an interactive prompt

Current Workflow

1. Download IIRS data from the PRADAN / ISSDC system (requires an account).

2. Use iirspy to read data by giving the datetime basename and a local directory. Use L0 for raw or L1 for calibrated. The package unzips the downloaded files if necessary.

   python import iirspy extent = (5, 245, 4000, 8000) # Optional (xmin, xmax, ymin, ymax), defaults to full image iirs_l0_raw = iirspy.L0("20210720T2333026105", "/path/to/data/", extent) iirs_l1_rad = iirspy.L1("20210720T2333026105", "/path/to/data/", extent)

3. (Optional): Check that files were downloaded and unzipped correctly (slow!). Useful when files are added or changed. Prints nothing if successful.

   python iirs_l0_raw.checksum() iirs_l1_raw.checksum()

4. Plot images. Images are handled with xarray and can be accessed directly with L0.img or L1.img.

   ```python ax = iirsl0raw.plot() # Plot band 12 by default

   Plot desired band from raw and calibrated image side by side

   fig, axs = plt.subplots(1, 2, figsize=(10, 4)) iirsl0raw.plot(band=79, ax=axs[0]) iirsl1rad.plot(band=79, ax=axs[1])

   Plot the median spectrum of a subset of the image using xarray

   iirsl0raw.img.sel(y=slice(6000, 6250)).median(dims=('x', 'y')).plot() ```

5. Calibration (coming soon)

   python radiance = iirs_l0_raw.calibrate_to_rad() reflectance = iirs_l1_rad.calibrate_to_refl()

Other features

The IIRS classes L0 and L1 provide access to image metadata.

```python l0 = iirspy.L0("20210720T2333026105", "/path/to/data/")

Print paths to the image files in this image bundle

print("Qub path", l0.qub) print("Hdr path", l0.hdr) print("Xml path", l0.xml)

...etc for lbr, oat, oath, spm. L1 files also have csv and csv-xml

The projection, start and stop time, exposure and gain are stored in metadata

print("Metadata", l0.metadata)

Not all metadata is loaded in by default, but you can access other values in the XML file with .metaget (thanks to pdr!)

print("Solar inc (deg):", l0.metaget("isda:solar_incidence")) ```

Works in progress

Calibrate from L0 raw to L1 radiance

Note: Raw data from ISSDCC is already dark subtracted.

1. (Optional): Remove bad pixels from gain and offset array
2. Apply gain and offset from lookup table to convert DN to radiance [mW cm^-2 sr^-1 um^-1]
3. (Optional): Destripe the result.
4. (Not implemented) Postprocessing (keystone correction, radiance adjustment at order-sorting-filters and at sensor edges).
5. Write to file as 32-bit floating point binary BSQ.

Georeferencing to LOLA DEM

Using the ldem_80s_20m from LOLA.

1. Read image with lat/lon extent in python and find nearest latitudes given in the geom .csv file
2. Crop to desired lines using gdal
3. Import into QGIS
4. Select tie-points from ldem to IIRS image
5. Warp DEM to image coords (bilinear interpolation, thin plate spline)

Manual method:

• Read reflectance image into QGIS
• Choose tie points (or use the previous ones for dem, reversed)
• Project to Spole stereo (IAU_2015:30135)

Reflectance correction

1. Read IIRS L1 data from .qub
2. Convert radiance units to SI (x0.01) [1000 mW/cm^2/sr/um] -> [W/m^2/sr/um]
   • (Optional) Find bad bands (iirs_bad_bands.csv) for this img's exposure and gain (.xml) and set those bands to nodata
   • (Optional) Run the fourier destriper on the IIRS image to smooth artefacts
3. Read solar spectrum (ch2_iirs_solar_flux.txt) and scale by solar distance (spice) at image collection time
4. Get local solar incidence and azimuth for each line of the image from .spm
   • (Optional) Adjust solar incidence by dem
     • Resample the ldem to the resolution of the IIRS image
     • Get slope from gradient of ldem
     • Dot product between solar inc/az vector and surface normal vector
5. Do I/F refl correction (radiance / (cosinc * solarflux)) on all pixels
   • (Optional): apply spectral polish (don't - makes the spectra worse)
   • (Optional): apply spectral smoothing (usually do the 3-band boxcar)
6. Write to geotif or envi img/hdr format
   • (Optional): Edit envi hdr / geotiff metadata to add wavelength band labels