Pal S, Debanshi S. Machine learning models for wetland habitat vulnerability in mature Ganges delta.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH INTERNATIONAL 2021;
28:19121-19146. [PMID:
33398756 DOI:
10.1007/s11356-020-11413-8]
[Citation(s) in RCA: 5] [Impact Index Per Article: 1.7] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Received: 05/22/2020] [Accepted: 10/26/2020] [Indexed: 06/12/2023]
Abstract
The present study attempts to measure wetland habitat vulnerability (WHV) in the Indian part of mature Ganges delta. Predictive algorithms belonging to bivariate statistics and machine learning (ML) algorithms were applied for fulfilling the data mining and generating the models. Results show that 60% of the wetland areas are covered by moderate to very high WHV, out of which > 300 km2 belong to very high WHV followed by a high vulnerability in almost 150 km2. This areal coverage increases by 10-15% from phase II to phase III. On the other hand, a relatively safe situation is confined to < 200 km2. The receiver operating characteristic curve, root-mean-square error, and correlation coefficient are used to assess the accuracy of these models and categorization of habitat vulnerability. Ensemble modeling is done using the individual models having a greater accuracy level in order to increase accuracy. A field-based model of the same is prepared by gathering information directly from the field which also exhibits similar results with the algorithm-based models. Analysis of residuals in standard regression strongly supports the relevance of the selected parameters and multi-parametric models.
Collapse