The project was done to compare the accuracy of unsupervised classification with supervised classification based on Sentinel 2 satellite imagery. The area that was selected for the project is the Amvrakikos Wetlands National Park in Greece and its surroundings. The area has diverse terrain consisting of lagoons, marshes, swamps, reed beds, forests and sand dunes. In supervised classification, many decisions are left to the operator. He or she must, based on vision and analysis of satellite imagery, categorize and mark selected fields that characterize the chosen land class. In unsupervised classification, on the other hand, class patterns are developed automatically, without the operator's involvement. To facilitate supervised classification, I additionally made images for two indices: NDVI and NDWI. To get more results, I selected images for two dates:
June 27, 2022 and March 19, 2023.
Sentinel 2 satellite imagery of Amvrakikos Wetlands National Park, Greece:
27/06/2022 vs 19/03/2023
NDVI
NDVI (Normalized Difference Vegetation Index) is an indicator used in remote sensing in satellite image analysis, to assess the amount and condition of vegetation in an area. The higher the index value, the more healthy lush vegetation.
NDWI
NDWI (Normalized Difference Water Index) is widely used for analyses related to water resources. It can be used to monitor water levels in rivers and lakes, identify wetlands, analyse changes in water cover and assess areas at risk of flooding and drought. The higher the value of the indicator, the greater the presence of water in a given image.
Supervised vs unsupervised classification
The colouring of selected areas was classified as follows:
deep water - dark blue
high vegetation - dark green
cultivated fields - light green
urban areas - white
coastal waters - light blue
marshy waters - dark brown
marshland - olive
bare soil - orange
high vegetation - dark green
cultivated fields - light green
urban areas - white
coastal waters - light blue
marshy waters - dark brown
marshland - olive
bare soil - orange
Next to the colour classification, the percentage coverage of the satellite image with each class, which was calculated automatically in the SNAP programme, is shown.
27/06/2022
19/03/2023
