Application of deep learning in assessing the impact of flooding on the endangered freshwater fish Neolissochilus benasi (Cyprinidae) in a northern province of Vietnam

Abstract

Flooding, a sudden disturbance, is considered to affect negatively the survival of fish by causing shock and growth, especially for species living in headwaters of a river. Neolissochilus benasi is a freshwater fish that prefers living in clean, flowing water and rocky bottoms with sands and gravels. Based on a segment in mtDNA obtained from eight specimens collected from northern Vietnam, the present study applied a hybrid novel, genetic algorithm (GA)–artificial neural network (ANN) to understand impacts of floods on N. benasi. The GA–ANN hybrid model was successful in mapping flood susceptibility, which correlates with river density, altitude, and rainfall, being typical in lowlands, along rivers and streams. Strong correlations were found between fish and urban density, agriculture, and land use/land cover, which contribute to the decrease of N. benasi. Habitat destruction, hydropower dams, pollution, overfishing, and using destructive gears are probably the main causes of the N. benasi decline. Importantly, based on GA–ANN model, flooding had a significant impact on N. benasi, which performs a low genetic diversity in the studied regions. Thus, this endangered freshwater fish species would have been easily affected by flooding since very high and high susceptibility of N. benasi was abundant in the province, particularly along the Red River and urban areas. This is the first study to examine the link between flooding and genetic diversity of an aquatic organism in Vietnam applying deep learning models. Accordingly, these results recommend significant suggestions to protect N. benasi in its habitats from northern Vietnam under flooding.