Description

Rocks and minerals do not change by season or by day. However, the environment in which we acquire remote sensing images changes continuously. Major influences are the sun (illumination) and the atmosphere (weather). How difficult is it to ignore such rapidly changing external changing factors and to only monitor slow changes in surface mineralogy?
In this proposed experiment, we would like to review the concept of multi-temporal remote sensing, by performing a long-term series of spectral measurements of a rock or soil surface that does not change in time. The main question in this research is: How stable are the essential mineral spectra in time? Does the key assumption, that the environment changes over time but the “rock signal” does not change, hold?

Objectives and Methodology

The following research questions could be considered:
- Can we improve on existing multi-temporal image analysis methods?
- Can we make our measurements of geological land surface cover independent of time?
- Can we distinguish weather conditions that interfere with mineral spectra most?

Over 300 Sentinel-2 images are directly available for analysis. Besides downloading all this data and processing it locally, it is also possible to use the Google Earth Engine. Spectral indices will be calculated to highlight locations with exposed minerals (that should not change over time) and vegetation (which should change over time).
By analyzing the temporal behavior at known locations in the image, you try to deduce what environmental parameter (sun angle, weather, soil moisture) has influence on mineral spectral indices. Once a source of error has been identified, you can try to invent a method to reduce its influence; this should eventually lead to a more robust measurement technique. In addition, weather databases can be consulted as independent source of information on things that do change in time.

Further reading

http://essay.utwente.nl/85173/