Scholarlink

<p>Mining activities have long been associated with significant environmental impacts, including deforestation, habitat degradation, and biodiversity loss, necessitating targeted strategies like afforestation to mitigate ecological damage. Tata Steel’s afforestation initiative near its Noamundi iron ore mining site in Jharkhand, India, spanning 165.5 hectares with over 1.1 million saplings planted, is a critical case study for evaluating such restoration efforts. However, assessing the success of these initiatives requires robust, scalable methods to monitor land use changes over time, a challenge compounded by the need for accurate, cost-effective tools to validate ecological recovery and support environmental governance frameworks. This study introduces a novel approach by integrating multiple machine learning (ML) algorithms, classification and regression tree (CART), random forest, minimum distance, gradient tree boost, and Naive Bayes, with multi-temporal, multi-resolution satellite imagery (Landsat, Sentinel-2A, PlanetScope) on Google Earth Engine (GEE) to analyze land use dynamics in 1987, 2016, and 2022. In a novel application to such contexts, high-resolution PlanetScope data (3&#xa0;m) and drone imagery were leveraged to validate classification accuracy using an 80:20 training–testing data split. The comparison of ML methods across varying spatial resolutions and temporal scales provides a methodological advancement for monitoring afforestation in mining landscapes, emphasizing reproducibility and precision. Results identified CART and Naive Bayes classifier classifiers as the most accurate (83% accuracy with PlanetScope 2022 data), effectively mapping afforestation progress and land use changes. These findings highlight the utility of ML-driven remote sensing in offering spatially explicit, cost-effective monitoring of restoration initiatives, directly supporting Environmental, Social, and Governance (ESG) reporting by enhancing transparency in ecological management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging machine learning for monitoring afforestation in mining areas: evaluating Tata Steel’s restoration efforts in Noamundi, India

  • 期刊论文

  • 发表日期：

    2025年6月27日

查看全文
  • Wang Xiuqing,
  • Saied Pirasteh,
  • Hishmi Jamil Husain,
  • Bhavesh Chauhan,
  • Vidhya Lakshmi Sivakumar,
  • Mahdieh Shirmohammadi,
  • Davood Mafi-Gholami

摘要

Mining activities have long been associated with significant environmental impacts, including deforestation, habitat degradation, and biodiversity loss, necessitating targeted strategies like afforestation to mitigate ecological damage. Tata Steel’s afforestation initiative near its Noamundi iron ore mining site in Jharkhand, India, spanning 165.5 hectares with over 1.1 million saplings planted, is a critical case study for evaluating such restoration efforts. However, assessing the success of these initiatives requires robust, scalable methods to monitor land use changes over time, a challenge compounded by the need for accurate, cost-effective tools to validate ecological recovery and support environmental governance frameworks. This study introduces a novel approach by integrating multiple machine learning (ML) algorithms, classification and regression tree (CART), random forest, minimum distance, gradient tree boost, and Naive Bayes, with multi-temporal, multi-resolution satellite imagery (Landsat, Sentinel-2A, PlanetScope) on Google Earth Engine (GEE) to analyze land use dynamics in 1987, 2016, and 2022. In a novel application to such contexts, high-resolution PlanetScope data (3 m) and drone imagery were leveraged to validate classification accuracy using an 80:20 training–testing data split. The comparison of ML methods across varying spatial resolutions and temporal scales provides a methodological advancement for monitoring afforestation in mining landscapes, emphasizing reproducibility and precision. Results identified CART and Naive Bayes classifier classifiers as the most accurate (83% accuracy with PlanetScope 2022 data), effectively mapping afforestation progress and land use changes. These findings highlight the utility of ML-driven remote sensing in offering spatially explicit, cost-effective monitoring of restoration initiatives, directly supporting Environmental, Social, and Governance (ESG) reporting by enhancing transparency in ecological management.