Tutorial

From EnMAP Level-1B to Level-2A

In this tutorial, we generate EnMAP Level-2A from Level-1B data using the EnMAP processing tool (EnPT). Please refer to the overview about EnMAP data products here to learn more about the differences between EnMAP processing levels.

As described in the Usage section of this documentation, the input parameters of EnPT can be set using the Python application programming interface (API), the command line interface (CLI) or the graphical userface (GUI) within the EnMAP-Box in QGIS. In this tutorial, we will use the GUI for simplicity. More information about the GUI can be found here.

Requirements

Software

This tutorial requires some software to be installed on your machine:

Please see here for a description to install all together into a single MambaForge Conda environment.

Data

You can download a subset of an EnMAP Level-1B dataset from here, which can be used to reproduce this tutorial. However, the actual EnMAP dataset that was used for this tutorial is not contained in the EnPT repository due to its file size.

Inspect the Level-1B data

The EnMAP L1B zip-archive used for this tutorial contains an acquisition covering Berlin and the surrounding area.

Extract the zip-archive on your local disk and inspect the extracted files. An explanation of all the contained files is given here.

_images/tut__contents_test_dataset.png

Note

The contents of this zip-archive may change in future.

Depending on the data format in which the Level-1B data was ordered from the ground segment, the raster data may also have different file extensions such as *.BSQ.

Open QGIS, start the EnMAP-Box and load the ENMAP*L1B*-SPECTRAL_IMAGE_VNIR.TIF and ENMAP*L1B*-SPECTRAL_IMAGE_SWIR.TIF image datasets. Some general handling instructions regarding the EnMAP-Box are given in Exercise A of the EnMAP-Box tutorials.

Unfold the respective entries in the Data sources panel to explore some metadata of both images. Both images have 1000 x 1024 pixels in x- and y-direction. The VNIR image contains 91 and the SWIR contains 133 spectral bands. Both images have no coordinate reference system (Arbitrary CRS). This is because EnMAP Level-1B data are provided in sensor geometry.

To visualize the first spectral band of the VNIR hyperspectral image, unfold the Bands entry of the VNIR image in the Data sources tab, right-click on Band 01:Band 1 (418.24) and press Open in new map. Then click the Spectral Library icon in the EnMAP-Box toolbar to open a new Spectral Library window and click somewhere into the visualized image in the Map window. This makes the corresponding spectral signature show up in the Spectral Library window.

_images/tut__enmapbox_l1b_vnir.png

Open the first band of the SWIR image and visualize any SWIR spectral signature as described above.

_images/tut__enmapbox_l1b_swir.png

With regard to the spectral signature plot, it is noted that EnMAP Level-1B data contain top-of-atmosphere radiance spectra, with each spectral band scaled according to specific gain and offset values given in the image metadata. The visualized spectra can therefore NOT directly be interpreted as top-of-atmosphere radiance but must first be converted. Even if all bands were equally scaled, the spectra would not be comparable to other hyperspectral EO data because they are still affected by numerous spatio-temporal varying effects, e.g., originating from the atmospheric state at the acquisition time.

Now open the VNIR and SWIR images in two different map views. Right-click into Map #1 and press Link with other maps. Then zoom into the image and compare the VNIR and SWIR images. At this point, a spatial offset between both images (around 20 pixels along-track and <1 pixel across-track) is clearly visible which means that the spectral information of both EnMAP detectors cannot be interpreted together. Panning through the image also reveals some vertical striping effects.

_images/tut__enmapbox_l1b_vnir_swir_shift_striping.png

To sum up, EnMAP Level-1B data cannot be directly used for downstream remote sensing applications due to the following reasons:

  • separate VNIR and SWIR detector images

  • images provided in sensor geometry, i.e., without coordinate reference system (CRS)

  • spectral information represents top-of-atmosphere radiance, separately scaled for each band

  • VNIR and SWIR images do not show the same spatial extent (offsets in along- and across-track direction)

  • image artefacts such as vertical striping

  • spectral information is affected by various spatio-temporal effects, e.g., originating from the atmospheric state

  • some sensor-induced effects that are not directly visible (spectral smile, keystone, …)

Note

The EnMAP-Box features a tool to import EnMAP Level-1B data (EnMAP-Box ‣ Project Menu ‣ Add product ‣ EnMAP L1B) which directly applies scales and offsets to each band and adds some metadata such as central wavelength positions. This is convenient as the data is directly shown in radiance units. However, users should keep in mind that the actual (scaled) L1B pixel values are different as they are stretched to the unsigned integer 16-bit value range.

Generate Level-2A data using EnPT

The EnMAP processing tool (EnPT) generates Level-2A from Level-1B data, i.e., processes the EnMAP data from the raw format (top-of-atmosphere radiance in sensor geometry) to geometrically and atmospherically corrected bottom-of-atmosphere reflectance. Please refer to the Algorithm descriptions section of this documentation for more information on the underlying algorithms.

Open the EnPT GUI to run the processing chain. You can find it here: QGIS 3.xx ‣ EnMAP-Box ‣ Applications Menu ‣ EnPT (EnMAP Processing Tool) ‣ Start EnPT GUI.

_images/screenshot_enpt_enmapboxapp_v0.7.4.png

Select the EnMAP L1B zip-archive (ENMAP01-____L1B*.ZIP) at the parameter L1B EnMAP image. All other parameters are set to a default or not strictly needed to generate Level-2A data. However, they may improve the output quality. Documentation about the parameters is provided here.

Press the Run button to start the processing. The current status is shown in the Log panel and the QGIS Python console. After processing, the output directory is indicated in the Log panel.

Inspect the Level-2A data and compare them with Level-1B

The output folder contains the EnMAP Level-2A image data ENMAP*L2A*-SPECTRAL_IMAGE.TIF and a couple of other files, such as metadata, quality layers, etc.:

_images/tut__contents_l2a_output.png

For descriptions of the individual files, see here.

Note

The output directory will contain additional layers in future.

Load the ENMAP*L2A*-SPECTRAL_IMAGE.TIF into the EnMAP-Box. Unfold the entry in the Data sources panel to explore some metadata of the Level-2A EnMAP-image. The image has a dimension of 1266 x 1204 pixels in x- and y-direction and contains 206 bands. The different image dimensions compared with Level-1B data are due to the geometric correction / orthorectification applied by EnPT. Furthermore, the image is now projected in WGS84 / UTM zone 33N.

Right-click on the image in the Data Sources panel and select Open in new map ‣ True Color to visualize an RGB band combination and open a Spectral Library window as described above. The spectral information now contains atmospherically corrected bottom-of-atmosphere reflectance data (scaled between 0 and 10000 by default). The two detector images have been merged so that their spectral information can now be used together. Spatio-temporal varying interferences on the spectra have been corrected as far as possible.

_images/tut__enmapbox_l2a_output.png

The generated EnMAP Level-2A data can now be used for subsequent remote sensing applications.

Note

Also for EnMAP Level 2A data, the EnMAP-Box provides a convenient data import dialog (EnMAP-Box ‣ Project Menu ‣ Add product ‣ EnMAP L2A) which scales the EnPT L2A output between 0 and 1.