Please login first
Broad-leaved and Coniferous Forest Classification in Google Earth Engine using Sentinel Imagery
1  Eskisehir Technical University, Institute of Earth and Space Sciences


Knowledge of forest structure is key to understanding, managing, and preserving forest biodiversity and function. With the well-established need within the remote sensing community for better understanding of canopy structure, in this paper, the effectiveness of Sentinel-2 imagery for broad-leaved and coniferous forest classification within the Google Earth Engine (GEE) platform has been assessed. Here we used Sentinel-2 image collection from the summer period over North Macedonia when the canopy is fully developed. For the sample collection of the coniferous areas and the accuracy assessment of the classification, we used imagery from the spring period when the broad-leaved forests are in early green stage. Support Vector Machines (SVM) classifier has been used for discriminating forest cover groups, namely broadleaved and coniferous forests. According to the results more than 90% of the canopy in North Macedonia are broad-leaved, while less than 10% are conifers. The results in this study showed that with the use of Google Earth Engine, Sentinel-2 data alone can be effectively used to obtain rapid and accurate mapping of main forest types (conifers-broadleaved) with fine resolution.

Keywords: Broad-leaved forest; Coniferous forest, Remote Sensing; Google Earth Engine; Sentinel;
Comments on this paper
Arthur Novikov
Effectiveness of actions
Dear researchers,

I hope you have a few minutes to discuss the following:
1. Is it possible to classify the species of a single tree using Your algorithm?
2. Is it possible to estimate the annual biomass gain with the help of your algorithm?

Gordana Kaplan
Dear Arthur,

Thank you for your interest in the paper.

1. The algorithm hasn't been tested for single tree species, but that is a great idea for future studies!
However, in my experience, taking in mind the area of the study area, I don't think the results would be satisfactory.
But, I am sure good results can be achieved in smaller study areas (Persson, Magnus, Eva Lindberg, and Heather Reese. "Tree species classification with multi-temporal Sentinel-2 data." Remote Sensing)

2. The study was actually inspired by an Agriculture, Forestry, and Other Land Use project, where it was needed to classify the forest into two classes, confers, and broad-leaved forests, so yes, the biomass can be estimated.

Thank you for your kind questions,

Have a nice day!
Arthur Novikov
Dear Gordana,

Thanks a lot!

Have a nice day,

Svetlana Illarionova
reference data
Dear Gordana,

Thank you for sharing interesting and important research results! Could you please comment on reference data for your study? Were labels for sample points obtained from the field-based observation, or were satellite images from the early spring months interpreted to achieve reference labels for four classes?

Best regards,

Gordana Kaplan
Dear Svetlana,

I am so happy that you showed interest in my paper. First of all, I want to note that the study was made with a minimum sample collection for all classes, not more than 20 per class. The samples for the forest classes, Confer and Broad Leaved were selected with coordination from the early spring season when the broad-leaved forests are in the early development stage, which made it easy to detect the confers. The other classes were selected from the summer period.

No filed data were conducted for this research. However, the accuracy assessment of the result gives good confidence.

I hope this answers your question, and please, don't hesitate to contact me again.

Kind regards, wishing you healthy days, Gordana