Segmentation (with a basic workflow)

Data

Download links to data:

Make a directory for this lab, and the next, called susques. Put the data into a data folder. Open QGIS, start a project. Save it as susques.qgz. (You will answer questions later in lab_11_answers.txt)

susques/
├── susques.qgz
├── data/
│   ├── s2_susques_max_2025_clipped.tif
│   └── s2_susques_median_2025_clipped.tif
└── lab_11_answers.txt

For this project, you have two Sentinel-2 images (from Google Earth Engine’s Harmonized Sentinel-2 MSI Level-2A Surface Reflectance collection) to work with. The first is a median composite image and, the second, a max composite image. Both images are of the same scene, near Susques, Argentina, In the Atacama Desert. They are taken over the period from 2025-04-01 to 2025-10-26.

The bands available in the two composite images are listed in Table 2.

Segmentation using a binary mask

In this exercise you will segment the image to find the lithium brine evaporation ponds, using only the raster calculator, and the Polygonize tool (Raster to Vector, from GDAL)

To do this you will need to:

  1. Select an appropriate spectral index (or indices) and calculate it using Raster Calculator
  2. Create a binary mask (pond / not pond) by using a threshold value of your spectral index.
  3. Calculate the total area of the ponds.
  4. Answer these questions in lab_11_answers.txt

q1. What spectral index (or indices) did you use?
q2. What bands are involved in the index and why are those bands useful for this task?
q3. What was the threshold value used to create you mask?
q4. What is the total area of the brine evaporation ponds?
q5. What CRS did you use when measuring the area of the brine evaporation ponds?
q6. What is the difference in mean SWIR1 (band 11) within and outside of evaporation ponds in the image?
q7. What is a median composite image? A maximum composite image?

Submission

Submit zipped project folder. Scoring will be conducted as shown in Table 1

Tables

Table 1: Rubric for scoring
Item Percentage of score
Projects opens and all files are present 40%
Ponds successfully Segmented 30%
Answers to questions 30%
Table 2: Sentinel-2 bands available in the composite images
GeoTIFF band Band name Description Native pixel size
1 B1 Aerosols 60 m
2 B2 Blue 10 m
3 B3 Green 10 m
4 B4 Red 10 m
5 B5 Red edge 1 20 m
6 B6 Red edge 2 20 m
7 B7 Red edge 3 20 m
8 B8 Near infrared (NIR) 10 m
9 B8A Red edge 4 20 m
10 B9 Water vapor 60 m
11 B11 Shortwave infrared 1 (SWIR1) 20 m
12 B12 Shortwave infrared 2 (SWIR2) 20 m
13 AOT Aerosol optical thickness 10 m
14 WVP Water vapor pressure 10 m
15 SCL Scene classification map 20 m
16 TCI_R True color image, red channel 10 m
17 TCI_G True color image, green channel 10 m
18 TCI_B True color image, blue channel 10 m
19 MSK_CLDPRB Cloud probability map 20 m
20 MSK_SNWPRB Snow probability map 10 m
21 QA10 Always empty 10 m
22 QA20 Always empty 20 m
23 QA60 Cloud mask 60 m
24 MSK_CLASSI_OPAQUE Opaque cloud classification 60 m
25 MSK_CLASSI_CIRRUS Cirrus cloud classification 60 m
26 MSK_CLASSI_SNOW_ICE Snow/ice classification 60 m