NeFut Logo NeFut
Admin Login

[CS.AI] Systematic Evaluation of the COTQ Provincial Land Cover Product: Structural Consistency, Spectral Separability, and Relative Positioning to ESA, ESRI, and Google Datasets

Published at: 2026-09-17 22:00 Last updated: 2026-09-18 00:46
#algorithm #Machine Learning

High‑resolution land‑use/cover (LULC) products derived from Sentinel‑2 imagery are widely used for environmental monitoring and land management, yet their performance varies in regions with complex ecological gradients and heterogeneous surfaces. To address Quebec's need for annual monitoring of land occupation and soil artificialisation, a provincial 10 m land‑cover product, COTQ, was released. This paper does not propose a new mapping method; instead it systematically analyses COTQ's behaviour and consistency using complementary evaluation approaches.

All products are harmonised under a common legend and compared with structural indicators (object‑size distributions, shape complexity, Adjusted Rand Index, Intersection over Union), spectral separability metrics derived from Sentinel‑2 reflectance, and targeted photo‑interpretation of disagreement areas. The analysis covers eight Sentinel‑2 tiles representing Quebec's main bioclimatic domains, from temperate and boreal forests to northern tundra.

The results show that COTQ’s structural and spectral characteristics are most similar to ESA WorldCover, while systematic differences appear for urban areas, wetlands, and rocky or cryptogamic surfaces due to class definitions and thematic priorities. This multi‑criteria evaluation provides an objective characterization of COTQ and clarifies its relative positioning for operational land monitoring in Quebec.

Review

Original Source: https://arxiv.org/abs/2609.17731

[h] Back to Home