Terrain characteristics and Land cover of Fogo Island

Sentinel 2 data was used to derive landcover raster data showing different landcover classes which included built-up, vegetation, water, bare land, agricultural land, and lava.

SRTM data was used to depict the elevation of the study area and also to generate slope and aspect raster datasets.

With the application of QGIS and R software, the data was processed to achieve the objective.

After pre-processing, the respective datasets were then overlaid, in order to extract observations for the values of landcover class, elevation, slope, and aspect.

These observations were then used to test for the relationship between terrain characteristics and landcover class.

Due to the difference in spatial resolution of the two datasets (SRTM and Sentinel), the sentinel data was downscaled to 30m and re-projected to suitably overlay.

The aspect and slope were calculated from SRTM data using the Terra library in R. The result was a raster dataset showing different directions and degrees of slope.

Machine learning, specifically Random Forest algorithm was used to predict the class of each pixel in the satellite image.

lava and water were correctly classified, this could be attributed to the big areas covered by the two land cover classes, which yielded more training points.

The other land cover classes were, however, incorrectly classified, probably due to the similarity in particular cover classes e.g., agriculture and vegetation which facilitated the confusion.

Most landcover classes e.g. agriculture, vegetation, and built environment are in the south-facing slopes due to more exposure to sunlight. 

Steep slopes deter the establishment of most landcover types.

Certain landcover types are associated with high elevation such as lava, while some are associated with low elevation such as water.

The relationship can also be visualized with density plots and histograms.

Certain landcover classes are closely associated with specific terrain characteristics.

There was a positive correlation between elevation and lava, while water was found to be at the lowest elevation.

This suggests that terrain plays a crucial role in shaping the spatial patterns on landcover across our study area.