Post-Fire Environmental Monitoring in Greece  

Project Focus

Development of a Graphical User Interface (GUI) in eCognition for conducting indices analysis and change detection in areas affected by wildfires.

Case Study

For this project, two Sentinel-2 MSI L2A images from the area of “Schinos” in Corinthos, Greece were utilized, to study a wildfire that broke out in the area on May 19th 2021. The images were pre-processed and they are provided to the user. However, the user can also import their own data, as long as the bands match.

​

​​​​​​​​​​​The application and step-by-step analysis are described in the video below:

​

​

​

​

​

​

​

​

​

​​

GUI Sections 

​

Start 

The data needs to be loaded carefully: the image from Timestep 1 must be loaded first and the one from Timestep 2 second, so that the image bands (layers in eCognition) match those set in the rule set.                   

This section allows users to rename the eCognition image layers as the Sentinel-2 band names for ease of use. The layers are later used in the rule set to calculate indices etc. It also creates the starting layout for the project which consists of two linked maps which have been split horizontally for timestep 1 and timestep 2.

 

Segmentation 

This section allows users to define their own segmentation parameters (scale parameter, shape and compactness) and conduct multiresolution segmentation for both timesteps.

 

Timestep 1 Indices 

This section allows users to calculate the MNDWI, NBR, and NDVI for timestep 1. Users can also define a threshold for each index in order to classify image objects as water, burned area, or vegetation. Included in this section is the option to remove classification so that users can change the classification threshold multiple times to receive different results.  

​​

Timestep 2 Indices 

This section allows users to calculate the MNDWI, NBR, and NDVI for timestep 2. Users can also define a threshold for each index in order to classify image objects as water, burned area, or vegetation. Included in this section is the option to remove classification so that users can change the classification threshold multiple times to receive different results.

 

Export Classification Results 

This section allows users to export the results of the classifications based on the indices mentioned above as vector files in a shapefile format. The users can choose to export the total burned area, area covered by vegetation and water area for each timestep. 

​

Change Detection 

In this part, the user can perform change detection using different Maps. The purpose is to define how much of the burned area has recovered (vegetation re-growth) after two years. For the change detection, we will have two data sets at different time steps. After loading them we will create a Map within one project for each time step and apply a classification on each single time step separately. This is what was performed above. In a final step, we synchronized the two classifications into one Map (main) and compared them to reveal areas of changes and also no changes. 

​

Statistics  

In this section users can calculate different statistics derived from features, like the average (mean) burned area (km2), the mean NDVI for the whole scene as well as the changed and unchanged area (km2) after performing the change detection. 

​

Reset 

This section resets the project to the starting layout with two linked, horizontally split maps for timestep 1 and timestep 2. The alias names assigned in the start section remain, but the previously created classifications, levels, and image objects are all deleted.