Modelling water richness in riparian flood plain wetland using bivariate statistics and machine learning algorithms and figuring out the role of damming

Abstract

The present work tried to develop multiparametric statistical and machine learning models of wetland water richness at two phases in the pre-dam period and at one phase in the post-dam period. Eight models four each in statistical and machine learning approaches including six relevant parameters have been developed. The performance of the EBF among statistical models and REPtree among machine learning models are found as the best representative. REPtree model has identified 212.58 km², 136.51 km², and 34.02 km² areas as very high water richness at phase I, II, and III respectively in the core wetland stretches. The decreasing rate has almost doubled after damming (35% to 75%). The Reduction rate of water richness area in-between phase I and II in the pre-dam period is not statistically significant (p < 0.98), but this is significant between pre- and post-dam period figuring out the crucial role of damming on wetland water richness.

Keywords:

• Water richness
• damming effect
• statistical models
• machine learning models and testing role of damming

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by University Grants Commission-India (UGC Ref. No.:3430/(NET-DEC 2018).