Johannes Reiche, Bart Slagter & Johannes Balling 23-03-2020
- Software
- Google Earth Engine (GEE)
- Dataset provided via GEE
- Sentinel-1 SAR
- Introduction to GEE
- Visualisation and RGB generation of Sentinel-1 images in GEE
- Basic time series analysis of Sentinel-1 images in GEE
- Understanding SAR backscatter characteristics over forest landscapes
-
Google Earth Engine Introduction
-
Sentinel-1 SAR data
-
Study area (Riau, Indonsia)
-
Google Earth Engine Exercises
E1. Sentinel-1 visualisation and RGB generation
Google Earth Engine (GEE) provides vast amounts of pre-processed satellite imagery and geospatial datasets ready to use for researcher, scientists and developers (Gorelick et al., 2017). Satellite images include ESA’s Sentinel missions providing both dense optical (Sentinel-2) and radar (Sentinel-1) data streams; and also annual 50 m (ALOS PALSAR-1/2 mosaics). For a brief introduction on how to use GEE you may watch the videos of this playlist. Suggested videos in particular helpful to accomplish this exercise are:
- General introduction of GEE
- Difference between a server and a client. It is important to notice that GEE is as server based platform and certain operations usually used in R or other programming software are either not working or not advisable.
- General functions on filtering and displaying data in GEE.
The rest of the videos are not mandatory for this practical, but give a more in-depth view on possibilities in GEE.
To start using GEE the user needs to sign up for an GEE account (if not done beforehand). Visit the following website and click on the sign-up button on the top of the page (Fig. 3). The confirmation of a GEE account might take up to three days according to the website (usually it is faster).
Figure 3. GEE website - for using GEE, please sign-up on the top right (red box).
With an account you can login via the GEE code editor and start coding (Watch the videos mentioned above to understand the structure of the GEE code editor).
The Sentinel-1 mission comprises a constellation of two polar-orbiting satellites, operating day and night performing C-band synthetic aperture radar (SAR) imaging, enabling them to acquire imagery regardless of the weather. The C-band radar satellites provide - depending on the acquisition mode - co- and cross-poralized images, a spatial resolution of 10 m and a temporal resolution of up to 6 days in the tropics. In this practical we will use Ground Range Detected (GRD) products. GRD products consist of focused SAR data that has been detected, multi-looked and projected to ground range using an Earth ellipsoid model. Phase information is lost. However, the polarization information is kept in single co-polarization (VV) or dual-band cross-polarization (VV + VH) depending on the instrument's polarization settings. For a more in-depth description of Sentinel-1 data please visit ESA's website.
Following pre-processing steps were applied to GRD data within GEE (as implemented in the Sentinel-1 toolbox):
- Apply orbit file
- Updates orbit metadata with a restituted orbit file.
- GRD border noise removal
- Removes low intensity noise and invalid data on scene edges.
- Thermal noise removal
- Removes additive noise in sub-swaths to help reduce discontinuities between sub-swaths for scenes in multi-swath acquisition modes.
- Radiometric calibration
- Computes backscatter intensity using sensor calibration parameters in the GRD metadata.
- Terrain correction (orthorectification)
- Converts data from ground range geometry, which does not take terrain into account, to σ°.
Note: Slope correction and speckle filtering are not included the standard Sentinel-1 GEE pre-processing.
The study area is located in the province of Riau (Indonesia). Riau is located on the east coast of central Sumatra and experiences tropical equatorial climate with persistent cloud cover throughout the entire year. Primary and secondary dryland-, swamp- and mangrove forests dominate the natural forest areas. Riau has the highest forest cover loss rates in Indonesia (Margono et al 2014) mainly driven by expansion and conversion to acacia, coconut, rubber plantations and oil palm (Fig.1) (Uryu et al 2008). Oil palm, representing the largest plantation area by species covers about 3.08 Mha of all land area (Fig.1) (GlobalForestWatch).
Figure 1. Forest conversion (left) and oil palm plantation (right) in Riau (Indonesia)
The study area extends approximately 80 x 60 km (0°24’ N; 102°40’ E) and is part of the Pelalawan regency with several plantations and peatlands. The latter are diminishing in their quantity due to drainage for land conversion making them more sensible to fires. Although forbidden by law, fire use for forest removal is still wide-spread in Riau (Gaveau et al 2014). These fire activities are man-made and can be traced back to fire-related land management practices on forest plantations and can spread to natural forest (Reiche et al 2018). These fire impacts and also logging activities are visible in Sentinel-1 images (Fig. 2).
Figure 2. Sentinel-1 RGB (Red: VV, Green: VH, Blue: VV-VH) (left) of an area in Riau and the corresponding photos of the land cover (right – top: water body; middle: secondary peatland forest; bottom: plantation logging and secondary peatland forest)
Please click on the different .md files given in this github project and follow the instructions and tasks within them. It is important to carry out the different parts in the intended order of:
E.1. Sentinel-1 visualisation and RGB generation