01-05 December 2025
INCOIS, Hyderabad, India.
| Abstract Submission No. | ABS-09-0136 |
| Title of Abstract | A Machine Learning Framework for long-term monitoring of mangroves along the Ernakulam Coast, Kerala, India |
| Authors | Farzana Haris*, Ranith Rajamohanan Pillai, Nandini Menon N, Roshin P Raj |
| Organisation | Nansen Environmental Research Center (India) |
| Address | Khadiriyya manzil, Prakkulam PO Kollam- 691602 Kollam, Kerala, India Pincode: 691602 E-mail: harisfarzana107@gmail.com |
| Country | India |
| Presentation | Oral |
| Abstract | Mangroves are ecologically significant coastal habitats that play a critical role in shoreline stability, carbon sequestration, and the maintenance of biodiversity. Accurate, continuous mapping and monitoring of mangrove cover and health are essential for understanding the ecosystem changes under anthropogenic and climatic pressures. This study presents a standardized machine learning based framework for mangrove mapping using medium resolution satellite imagery along the Ernakulam district of Kerala coast, and to monitor the relative change in mangrove cover over a 25-year period (2000⿿2025). Multispectral datasets from Landsat, LISS-3/4, and Sentinel-2 were collected as post-monsoon median composites and classified using multiple ML algorithms, including Random Forest, Gradient Boosting, and Artificial Neural Network, to identify the best model and an efficient ensemble that can be used to understand the change in mangrove cover using satellite data. Extensive field-validated mangrove datasets were used for supervised training and independent validation. Classification performance was assessed using multiple accuracy metrics such as confusion matrix, overall accuracy, Kappa coefficient, precision, specificity, sensitivity, and estimated mangrove area in Ernakulam (sq.km) to ensure mapping accuracy. Among the evaluated ML models, XGBoost and CatBoost emerged as the best-performing classifiers, achieving high Kappa coefficients of 0.925 and 0.908 respectively. This study presents a ML based framework to standardize mangrove cover mapping across diverse multi-spectral sensors and temporal scales, enabling consistent and comparable monitoring efforts of coastal mangrove ecosystems. This framework also enabled us to determine the relative change in mangrove cover, which in turn helped in identifying hotspots of mangrove degradation that need specific mitigation measures, as well as areas of mangrove replenishment. Applied to the Ernakulam coast, the approach demonstrates its potential for integration in regions with fragmented datasets, contributing to scalable frameworks for environmental assessment, conservation planning, and marine spatial management. |
| Are you part of IIOE-2 endorsed project | no |
| Keywords | Mangrove Mapping, Machine Learning, Remote Sensing, Coastal Ecosystem, Change Detection |
| For Awards | yes |
| Date Of Birth | 30-10-1997 |
| ECSN Registration Number | IIOE2-ECSN-0112 |