Combining a deep learning model with an optimization algorithm to detect the dispersal of the early stages of spotted butterfish in northern Vietnam under global warming

• Published in: Ecological informatics 2023-12, Vol.78, p.102380, Article 102380
• Main Authors: Do, Anh Ngoc Thi, Tran, Hau Duc
• Format: Article
• Language: English
• Subjects: algorithms; autocorrelation; Climate change; corals; estuaries; GA-ADAM algorithm; GA-XGBoost; habitats; humans; juveniles; larvae; mangrove forests; marine fish; momentum; Scatophagus argus; Spatial distribution; species; temperature; tropical fish; urbanization; Vietnam
• ISSN: 1574-9541
• DOI: 10.1016/j.ecoinf.2023.102380

Climate change and anthropogenic disturbances are growing more severe and have a substantial impact on fish larval and juvenile dispersals, especially in an estuarine system, where many marine fish utilize as their nursery grounds. To estimate distribution patterns of larvae and juveniles of spotted butterfish, Scatophagus argus, a reef-associated bony fish and an economically important species, eleven key natural and human factors in tropical estuaries, northern Vietnam were chosen as input variables. Environmental variables in estuarine habitats frequently exhibit direct relationships, which can contribute to autocorrelation problem and over-modeling. Consequently, the combined genetic-adaptive momentum algorithm (GA-ADAM) model was utilized to map the distribution of S. argus larvae and juveniles in two estuaries from northern Vietnam. The two most crucial factors influencing the fish occurrence, according to GA algorithms, are temperature and mangroves. Computed results suggest that the Catch Per Unit Effort (CPUE) of S. argus in Ba Lat estuary is 11.78 times higher than that in Tien Yen one (more northward zone). The GA based on extreme gradient boosting (XGBoost) model is then applied to examine projected changes in the fish spatial distribution by 2032. In the Tien Yen and Ba Lat estuaries, the temperature in 2032 will rise by 0.32 °C and 0.31 °C, respectively. Simultaneously, rising urbanization and loss of mangrove forests would have diminished S. argus's habitat. Due to the foregoing alteration, the CPUE reduced substantially, with a maximum of 2.39 and 32.76 individuals/haul in Tien Yen and Ba Lat estuaries, respectively. These results show how useful the hybrid model can be as a management tool for safeguarding and conserving tropical fish species from both natural and artificial threats. This work additionally emphasizes the ecological connectivity importance of the mangrove forest for aquatic organisms, including coral reef-associated fish, who utilize this habitat during their early life stages. •Climate change and landscape fragmentation are combined for the first time in this study to examine species distribution.•S. argus larvae and juvenile's occurrence is largely reliant on mangroves and temperature.•The hybrid GA-ADAM model is extremely competent in mapping fish distribution.•Warming temperatures and a decline in mangrove area by 2032 diminish the CPUE of S. argus.•Early stages of S. argus prefer the Ba Lat estuary over the Tien Yen es