Greenery change and its impact on human-elephant conflict in Sri Lanka: a model-based assessment using Sentinel-2 imagery

被引:1
作者
Gunawansa, Thakshila D. [1 ,2 ]
Perera, Kithsiri [1 ,5 ]
Apan, Armando [1 ]
Hettiarachchi, Nandita K. [3 ]
Bandara, Dananjana Y. [4 ]
机构
[1] Univ Southern Queensland, Fac Hlth Engn & Sci, Sch Surveying & Build Environm, Toowoomba, Qld, Australia
[2] Uva Wellassa Univ, Fac Technol Studies, Dept Engn Technol, Badulla, Sri Lanka
[3] Univ Ruhuna, Fac Engn, Dept Mech & Mfg Engn, Galle, Sri Lanka
[4] Minist Water Supply & Estate Infrastruct Dev, Water Supply & Sanitat Improvement Project, Battaramulla, Sri Lanka
[5] Univ Southern Queensland, Fac Hlth Engn & Sci, Sch Surveying & Built Environm, West St, Toowoomba, Qld 4350, Australia
关键词
Human elephant conflict; remote sensing; random forest; support vector machine; object-based image analysis; greenery changes; SUPPORT VECTOR MACHINE; RANDOM FOREST; TIME-SERIES; COVER CHANGE; CLASSIFICATION; CHALLENGES; SELECTION;
D O I
10.1080/01431161.2023.2244644
中图分类号
TP7 [遥感技术];
学科分类号
081102 ; 0816 ; 081602 ; 083002 ; 1404 ;
摘要
Human-elephant conflict (HEC) is a significant conservation issue for Asian elephants (Elephas maximus) and an environmental and socioeconomic crisis in elephant range countries, including Sri Lanka. Approximately 14,897 HEC incidents were recorded in Sri Lanka between 2015 and 2021. In this study, we present a Sri Lanka-wide analysis to explore the impact of greenery change on HEC. Our sources were official government data, and land use and land cover maps developed using Sentinel-2 satellite imagery. We applied the support vector machine (SVM), random forest (RF), and object-based image analysis classifications to classify land cover into six categories. This classification scheme also considered the differences observed in Sri Lanka's woody vegetation, consisting of forest, open forest, paddy fields, homestead gardens, and other crops. Analysis of the accuracies of the three types of classifiers confirmed that the supervised classification with two machine learning algorithms, RF and SVM, delivered a higher level of precision in land cover classification. RF was the best option, with a 97.34% overall accuracy and a 0.94 kappa coefficient, while SVM recorded a 94.68% overall accuracy and a 0.89 kappa coefficient. According to the findings, most HEC incidences were recorded in open forests (54%), while 62% were recorded within 2 km of the forest edge. Results indicated that HEC coincides with the human-occupied changed landscape adjacent to forest reservations and patches. The findings could be valuable for HEC management by identifying areas where elephants are most likely to conflict with humans, and the government may declare these as protected areas. Also, we propose an early warning system as an effective approach that helps detect and monitor elephant herds' movement. Therefore, implementing long-term land use planning is crucial for protecting the forest and natural habitats, restoring elephant habitats, and mitigating HEC by minimizing human encroachment and promoting sustainable land use practices.
引用
收藏
页码:5121 / 5146
页数:26
相关论文
共 61 条
  • [1] Measuring land cover change in Seremban, Malaysia using NDVI index
    Aburas, Maher Milad
    Abdullah, Sabrina Ho
    Ramli, Mohammad Firuz
    Ash'aari, Zulfa Hanan
    [J]. ENVIRONMENTAL FORENSICS 2015, 2015, 30 : 238 - 243
  • [2] Comparison of Random Forest and Support Vector Machine Classifiers for Regional Land Cover Mapping Using Coarse Resolution FY-3C Images
    Adugna, Tesfaye
    Xu, Wenbo
    Fan, Jinlong
    [J]. REMOTE SENSING, 2022, 14 (03)
  • [3] [Anonymous], 2015, Sentinel-2 User Handbook, V1, P64
  • [4] The Role of Agricultural Land Use Pattern Dynamics on Elephant Habitat Depletion and Human-Elephant Conflict in Sri Lanka
    Anuradha, J. M. P. N.
    Fujimura, Miho
    Inaoka, Tsukasa
    Sakai, Norio
    [J]. SUSTAINABILITY, 2019, 11 (10):
  • [5] Comparison between random forest and support vector machine algorithms for LULC classification
    Avci, Cengiz
    Budak, Muhammed
    Yagmur, Nur
    Balcik, Filiz Bektas
    [J]. INTERNATIONAL JOURNAL OF ENGINEERING AND GEOSCIENCES, 2023, 8 (01): : 1 - 10
  • [6] Random forest in remote sensing: A review of applications and future directions
    Belgiu, Mariana
    Dragut, Lucian
    [J]. ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 2016, 114 : 24 - 31
  • [7] Object based image analysis for remote sensing
    Blaschke, T.
    [J]. ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 2010, 65 (01) : 2 - 16
  • [8] A comprehensive survey on support vector machine classification: Applications, challenges and trends
    Cervantes, Jair
    Garcia-Lamont, Farid
    Rodriguez-Mazahua, Lisbeth
    Lopez, Asdrubal
    [J]. NEUROCOMPUTING, 2020, 408 : 189 - 215
  • [9] Machine Learning Classification of Fused Sentinel-1 and Sentinel-2 Image Data towards Mapping Fruit Plantations in Highly Heterogenous Landscapes
    Chabalala, Yingisani
    Adam, Elhadi
    Ali, Khalid Adem
    [J]. REMOTE SENSING, 2022, 14 (11)
  • [10] Choudhury A., 2008, IUCN RED LIST THREAT