Problem
Vietnam required more detailed and consistent national-scale agricultural land-use information to support agricultural monitoring, resource management and food-security research.
Context
The work integrated multitemporal Earth-observation and elevation data across heterogeneous agricultural landscapes and examined whether a trained model could transfer across years.
Approach
A UNet++ semantic-segmentation framework was applied using multitemporal Sentinel-1, Sentinel-2 and DEM inputs. The published study used an adaptive weighted combined loss function to address class imbalance.
My role
I contributed to the national-scale framework, cross-year transferability evaluation and resulting peer-reviewed publication.
Result
National agricultural land-use maps were produced for 2020 and 2024. The published study reports 15 land-cover categories and overall classification accuracies of 83.01 ± 1.37% for 2020 and 80.09 ± 0.76% for 2024.
Evidence
The method and results are documented in Remote Sensing (2026), DOI: 10.3390/rs18030430.
Potential applications
Agricultural monitoring, land-use planning, food-security research, crop-system analysis and national environmental reporting.