Work in progress

i.landsat.import is a GRASS-GIS module that imports Landsat satellite imagery scenes as native GRASS raster maps in its data base.

Landsat scenes can be acquired in two forms. One is a (packed and compressed) tar.gz file. Another is an (uncompressed and unpacked) directory containing the multispectral band and thermal channel acquisitions in form of GeoTiFF files. Albeit, along with various metadata. The module treats both forms. [option scene, single or multiple inputs]

The MTL metadata file is copied under the target mapset's cell_misc directory. This can be cancelled by using the -c flag. Date (year, month, day) and time (hours, minutes, seconds, timezone) of acquisitions are transferred to each imported band. [see r.timestamp]

Alternatively, the module creates pseudo GRASS raster maps via the -e flag. Instead of creating native GRASS raster maps, it links directly to the original GeoTIFF files. [see r.external]

For whatsoever might be the reason, it is possible to rerun the import process. Existing bands may be retained by skipping the import via the -s flag. At the same time, for bands which might lack of a timestamp, time stamping may be forced via the -f flag.

Multiple scenes are imported in individual Mapsets. That is bands of one scene, are imported in one indepenendet Mapset. If requested, all scenes are imported in one single Mapset. [flag -1 and option mapset]. For the latter, band names are prefixed with each scene's unique identifier. This may ease off building time series via GRASS' temporal t.* modules.

The module has got some handy skills to count the number of scenes inside a given pool directory [flag -n], list basic metadata and bands that pertain to each scene [flag -l] as well as print, or export in a file, a valid TGIS list of timestamps, one to use along with t.register [flag -t].

In the following examples, work is demoed with the LC81840332014146LGN00 (decompressed and unpacked) scene.

First grass-y things first: create a Location for this scene. Remember, a Location is defined by one and only spatial reference system. The scene in question lies inside the WRS2 tile Path 184 and Row 033, which is covered by UTM zone 34N. This is done automagically by using the geo-meta-tags that are part of a GeoTIFF file.

grass78 -c LC08_L1TP_184033_20180403_20180417_01_T1_B1.TIF /grassdb/wrs2_184033/

Let's list basic metadata and bands

i.landsat.import -l LC08_L1TP_184033_20180403_20180417_01_T1

will return

Band Filename
1 LC08_L1TP_184033_20180403_20180417_01_T1_B1.TIF
2 LC08_L1TP_184033_20180403_20180417_01_T1_B2.TIF
3 LC08_L1TP_184033_20180403_20180417_01_T1_B3.TIF
4 LC08_L1TP_184033_20180403_20180417_01_T1_B4.TIF
5 LC08_L1TP_184033_20180403_20180417_01_T1_B5.TIF
6 LC08_L1TP_184033_20180403_20180417_01_T1_B6.TIF
7 LC08_L1TP_184033_20180403_20180417_01_T1_B7.TIF
8 LC08_L1TP_184033_20180403_20180417_01_T1_B8.TIF
9 LC08_L1TP_184033_20180403_20180417_01_T1_B9.TIF
10 LC08_L1TP_184033_20180403_20180417_01_T1_B10.TIF
11 LC08_L1TP_184033_20180403_20180417_01_T1_B11.TIF
BQA LC08_L1TP_184033_20180403_20180417_01_T1_BQA.TIF

Get its timestamp, in form to use with GRASS' TGIS, via:

i.landsat.import -t LC08_L1TP_184033_20180403_20180417_01_T1

LC08_L1TP_184033_20180403_20180417_01_T1|03 Apr 2018 09:10:20.674074 +0000

We can import the scene in its own Mapset and request from the module to not copy the MTL file under cell_misc

i.landsat.import -c LC08_L1TP_184033_20180403_20180417_01_T1

Date Time Timezone
2018-04-03 09:10:20.674074 +0000

Target Mapset
@LC08_L1TP_184033_20180403_20180417_01_T1

Band Filename
1 LC08_L1TP_184033_20180403_20180417_01_T1_B1.TIF
2 LC08_L1TP_184033_20180403_20180417_01_T1_B2.TIF
3 LC08_L1TP_184033_20180403_20180417_01_T1_B3.TIF
4 LC08_L1TP_184033_20180403_20180417_01_T1_B4.TIF
5 LC08_L1TP_184033_20180403_20180417_01_T1_B5.TIF
6 LC08_L1TP_184033_20180403_20180417_01_T1_B6.TIF
7 LC08_L1TP_184033_20180403_20180417_01_T1_B7.TIF
8 LC08_L1TP_184033_20180403_20180417_01_T1_B8.TIF
9 LC08_L1TP_184033_20180403_20180417_01_T1_B9.TIF
10 LC08_L1TP_184033_20180403_20180417_01_T1_B10.TIF
11 LC08_L1TP_184033_20180403_20180417_01_T1_B11.TIF
BQA LC08_L1TP_184033_20180403_20180417_01_T1_BQA.TIF

-------------------------------------------------------------------------------
MTL not transferred to:
/geoyeux/grassdb/wrs2_180034/LC08_L1TP_184033_20180403_20180417_01_T1/cell_misc
-------------------------------------------------------------------------------

By the way, the module will create a new Mapset, named after the directory that contains the requested scene. Expectedly, the directory is named after the scene's unique identifier. Re-running the import command for the same scene, will simply pass the step of recreating the existing Mapset.

To create a copy of the MTL file under cell_misc, at a later stage, we may re-run the import process. The -s flag is therefore useful to skip over existing bands which otherwise would break the execution. As well, it helps in cases where some bands have been removed.

i.landsat.import LC08_L1TP_184033_20180403_20180417_01_T1 -s

Noteworthy is the memory option. It is passed, internally, to r.in.gdal, the actual importer. [see also r.in.gdal]

As usual in GRASS GIS, the --o flag is always handy in case overwriting existing maps is desired.

Using the -e flags, the module calls internally r.external. GeoTIFF files will be linked to GRASS' data base via pseudo GRASS raster maps.

*** Below To Update ***

We import both scenes in one step, each in its own Mapset. For this example, the command

i.landsat.import scene=LC81840332014146LGN00,LC81840332014226LGN00 --v

will fail at first since we have already imported the first scene. We may remove the respective Mapset and and re-import everything. Or, we may use the -s flag to skip existing material

i.landsat.import scene=LC81840332014146LGN00,LC81840332014226LGN00 --v -s

In case of many scenes, we can avoid typing or collecting all scene identifiers to feed the scene option and collect all scenes under one directory. This one directory can be fed to the module's pool option. For testing, we instruct the following and expect to get the list of (already existing band scenes, each in its own Mapset)

i.landsat.import pool=scenes --v -s

The module will fail if, among the scenes, it will detect one with a projection that does not match the current Location! Moving the scene out of the "pool" directory, will let the module (re-)run successfull. If, however, the user knows that the suspicious scene is indeed in the same projection, as the Location's one, the -o flag will override this check.

Assuming we have in addition the scene LC81840332014226LGN00, we can list out their components

i.landsat.import scene=LC81840332014146LGN00,LC81840332014226LGN00 -l

Date            Time

2014-05-26      09:10:26.7368720 +0000


Band    Filename

BQA     LC81840332014146LGN00_BQA.TIF
B11     LC81840332014146LGN00_B11.TIF
B6      LC81840332014146LGN00_B6.TIF
B1      LC81840332014146LGN00_B1.TIF
B8      LC81840332014146LGN00_B8.TIF
B10     LC81840332014146LGN00_B10.TIF
B9      LC81840332014146LGN00_B9.TIF
B7      LC81840332014146LGN00_B7.TIF
B4      LC81840332014146LGN00_B4.TIF
B3      LC81840332014146LGN00_B3.TIF
B2      LC81840332014146LGN00_B2.TIF
B5      LC81840332014146LGN00_B5.TIF
-------------------------------------------------------------------------------
Date            Time

2014-08-14      09:10:58.4714680 +0000


Band    Filename

B4      LC81840332014226LGN00_B4.TIF
B3      LC81840332014226LGN00_B3.TIF
B2      LC81840332014226LGN00_B2.TIF
B5      LC81840332014226LGN00_B5.TIF
B10     LC81840332014226LGN00_B10.TIF
B8      LC81840332014226LGN00_B8.TIF
B6      LC81840332014226LGN00_B6.TIF
B1      LC81840332014226LGN00_B1.TIF
B7      LC81840332014226LGN00_B7.TIF
BQA     LC81840332014226LGN00_BQA.TIF
B9      LC81840332014226LGN00_B9.TIF
B11     LC81840332014226LGN00_B11.TIF
-------------------------------------------------------------------------------

get their timestamps

i.landsat.import scene=LC81840332014146LGN00,LC81840332014226LGN00 -t

LC81840332014146LGN00<Suffix>|2014-05-26 09:10:26.7368720 +0000
LC81840332014226LGN00<Suffix>|2014-08-14 09:10:58.4714680 +0000

or/and import both scenes in a Mapset customly named LC8184033

i.landsat.import scene=LC81840332014146LGN00,LC81840332014226LGN00 --v -1 mapset=LC8184033

To list all bands (and their GeoTiFF filenames),

i.landsat.import pool=landsat_scenes -l

To derive a list of scenes (with a prefix --To Do--) and a valid TGIS timestamp (to use with t.register):

i.landsat.import pool=landsat_scenes -t

And more to add...

[High]

• Test for empty Landsat scene directories and corrupted files. Act upon accordingly. At the moment, a corrupted Landsat GeoTIFF will break the import process. This will, subsequently, brake any scripted workflow in which i.landsat.import is part of.

[Low]

• Update README.md and manual
• Test for range of input date, time, timezone
• What other meta can be transferred from the MTL file? Why not all! As in https://github.com/NikosAlexandris/i.landsat8.swlst/blob/master/landsat8_mtl.py?

• The "first" source for this module was a script published in https://grasswiki.osgeo.org/wiki/LANDSAT#Automated_data_import.
• Microseconds are ignored in the TGIS world

Unsorted

• http://landsathandbook.gsfc.nasa.gov/pdfs/Landsat_Calibration_Summary_RSE.pdf
• http://landsat.usgs.gov/band_designations_landsat_satellites.php
• https://landsat.usgs.gov/what-are-band-designations-landsat-satellites
• https://landsat.usgs.gov/collectionqualityband