Distribution of fish larvae and juveniles on salinity in an estuary predicted from remote sensing and fuzzy logic approach

Abstract

Salinity is one of the main factors influencing the early life stages of fish along an estuary, which shows great temporal and spatial changes. Recently, remote sensing has been widely applied to map salinity changes and fuzzy logic is identified as a suitable and strong tool for modeling complex systems. Based on collections of fish larvae and juveniles and water parameters in May, September, November, and December during 2019 along the Ba Lat estuary of the Red River, northern Vietnam, the present study attempts to predict the mapping and monitoring of the salinity using multispectral satellite imagery from Landsat 8 OLI satellite. The study determined that the NFS machine learning model, when improved by PCA, achieved a higher performance in displaying different salinity levels. The present study also confirms that using high spatial resolution or hyperspectral images would have increased the accuracy of spatial variation in similar modeling and mapping. Fuzzy rule–based modelling suggests that the occurrence of fish larvae and juveniles depended on salinity levels, with an R² > 0.9, compared to the field surveys. This finding further confirms the accuracy data obtained by artificial neural network models.