Multivariate generalized linear model‐based discrimination of striped dwarf catfish Mystus vittatus stocks from the River Ganga using body morphometrics

Abstract

Understanding spatial population structure is essential for the effective management of Mystus vittatus due to its declining population in the River Ganga. This study aims to investigate potential stocks of M. vittatus using comprehensive body morphometrics with both distance‐based and model‐based analyses. A total of 313 fish were collected from four sampling sites (Narora, Kanpur, Varanasi and Bhagalpur) along the River Ganga. In total, 36 truss distances were created by joining 16 morphometric landmarks on the fish body. Distance‐based analysis (PERMANOVA and CAP) revealed significant morphological variation among sites; however, heterogeneous dispersion violated key assumptions, necessitating model‐based approaches. Consequently, model‐based analyses that explicitly accounted for the mean–variance relationship inherent in morphometric data were applied to avoid potentially misleading inferences. A multivariate generalized linear model (MGLM) fitted with a Tweedie error distribution detected significant differences in body morphology among the four sites. Further, univariate GLMs indicated that most head, trunk and caudal measurements varied significantly among sites. Random forest classification achieved an overall accuracy of 85.30%, with the highest allocation success for Narora (95%), supporting its distinctiveness. Collectively, the results suggest the presence of heterogeneous stocks of M. vittatus in the River Ganga and highlight the importance of adopting model‐based multivariate frameworks for robust stock discrimination in fisheries research. These findings highlight the need for site‐specific management and conservation strategies to ensure the long‐term sustainability of the population of the target fish species.

Bibliographic Information