In order to harvest seaweed in a sustainable manner it is essential to understand the distribution and the status of the seaweed. Hafrannsóknastofnun (The Icelandic Marine Research Institute) conducted a project in which the total distribution of seaweed at Breiðafjörður (West Iceland) had to be assessed. Svarmi’s task was to design a remote sensing approach in order to assess the final extension. The approach was to use UAV imagery to ground-truth coarser satellite image classification (Landsat). We flew over predefined areas along Breiðafjörður, with both NIR and RGB cameras.  The satellite classification was then ground-truthed with the UAV imagery, giving a final error estimation that was used to estimate the classification accuracy of the Landsat images.


Seaweed mapping and classification




Fixed wing drone, multispectral and a high resolution RGB camera. Reflectance target for calibration. Satellite images (Landsat 8 OLI-TIRS).


Satellite image atmospheric correction with Python. UAV and satellite image classification and post-classification conducted with ENVI. 

Thousands of high resolution RGB images

A high resolution RGB camera was used to deliver high resolution calibrated images during the worst conditions of the Icelandic winter

Multisprectral imagery

Both a high resolution true color camera along with a multispectral camera were used to be able to identify all the characteristics of the seaweed by using radiometric calibrated imagery

Click and drag the slider to reveal multispectral image.

Click and drag the slider to reveal multispectral image.

Satellite classification ground-truthing

Satellite imagery (Landsat 8 OLI-TIRS) was classified and ground truthed with UAV imagery. The red pixels at the image are the final classified seaweed extent.