EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Mapping bamboo land cover in Mizoram, India using integrated high-resolution satellite data and deep learning

Muhammad Nazrul Islam, Omkar Shashikant Ghatage, Subham Roy, Suraj Kumar Singh, Giribabu Dandabathula, Apurba Kumar Bera

Discover Environment · 12.07.2026 · 4 (1)

DOI: 10.1007/s44274-026-00896-w ↗

Kurz gefasst

Fachartikel in der Zeitschrift Discover Environment, online erschienen am 12.07.2026. Verfasst von Muhammad Nazrul Islam und 5 weiteren Autor:innen. Die Arbeit nennt 35 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate spatial mapping of bamboo land cover is essential for resource management, ecological monitoring, and livelihood planning in Northeast India. However, the detection of fragmented bamboo-dominated patches in complex mountainous terrain remains challenging. This study develops and evaluates a deep learning-based semantic segmentation framework for mapping bamboo land cover in Mizoram using a nine-channel multi-sensor composite. The composite integrates high-resolution LISS-IV (5.8 m) surface reflectance data with Sentinel-2 Level-2A red-edge and narrow near-infrared bands (B5, B6, B7, B8A), as well as derived vegetation indices (NDVI, EVI2). The integration of high spatial resolution and red-edge spectral information was designed to enhance discrimination of bamboo-dominated vegetation within a multi-source mapping framework. A U-Net architecture with a ResNet-34 encoder was trained on 813 labelled patches acquired across multiple states of Northeast India. The model achieved a test Intersection over Union (IoU) of 0.8456, an F1-score of 0.9163, a precision of 0.9200, and a recall of 0.9127, evaluated at the patch level on a held-out test subset. Independent-site assessment across Mizoram, Manipur, and Assam indicates consistent model performance across diverse landscape conditions. Deployment across nine LISS-IV tiles covering Mizoram produced a bamboo land cover map at 5.8 m spatial resolution, with an estimated extent of approximately 6262 km2 (29.7% of the state’s geographical area). The results demonstrate the potential of integrating high-resolution imagery, red-edge spectral information, and deep learning for bamboo land cover mapping in complex mountainous environments and provide a potentially transferable framework for regional-scale bamboo resource assessment.

Zitieren

APA

Islam, M. N., Ghatage, O. S., Roy, S., Singh, S. K., Dandabathula, G., & Bera, A. K. (2026). Mapping bamboo land cover in Mizoram, India using integrated high-resolution satellite data and deep learning. Discover Environment, 4(1). https://doi.org/10.1007/s44274-026-00896-w

Vancouver

Islam MN, Ghatage OS, Roy S, Singh SK, Dandabathula G, Bera AK. Mapping bamboo land cover in Mizoram, India using integrated high-resolution satellite data and deep learning. Discover Environment. 2026;4(1). doi:10.1007/s44274-026-00896-w

BibTeX
@article{Islam2026,
  author = {Islam, Muhammad Nazrul and Ghatage, Omkar Shashikant and Roy, Subham and Singh, Suraj Kumar and Dandabathula, Giribabu and Bera, Apurba Kumar},
  title = {Mapping bamboo land cover in Mizoram, India using integrated high-resolution satellite data and deep learning},
  journal = {Discover Environment},
  year = {2026},
  volume = {4},
  number = {1},
  doi = {10.1007/s44274-026-00896-w}
}
Kostenlos starten →