Predicting species richness and abundance of tropical post-larval fish using machine learning

Abstract

Post-larval prediction is important, as post-larval supply allows us to understand juvenile fish populations. No previous studies have predicted post-larval fish species richness and abundance combining molecular tools, machine learning, and past-days remotely sensed oceanic conditions (RSOCs) obtained in the days just prior to sampling at different scales. Previous studies aimed at modeling species richness and abundance of marine fishes have mainly used environmental variables recorded locally during sampling and have merely focused on juvenile and adult fishes due to the difficulty of obtaining accurate species richness estimates for post-larvae. The present work predicted post-larval species richness (identified using DNA barcoding) and abundance at 2 coastal sites in SW Madagascar using random forest (RF) models. RFs were fitted using combinations of local variables and RSOCs at a small-scale (8 d prior to fish sampling in a 50 × 120 km2 area), meso-scale (16 d prior; 100 × 200 km2), and large-scale (24 d prior; 200 × 300 km2). RF models combining local and small-scale RSOC variables predicted species richness and abundance best, with accuracy around 70 and 60%, respectively. We observed a small variation of RF model performance in predicting species richness and abundance among all sites, highlighting the consistency of the predictive RF model. Moreover, partial dependence plots showed that high species richness and abundance were predicted for sea surface temperatures <27.0°C and chlorophyll a concentrations <0.22 mg m-3. With respect to temporal changes, these thresholds were solely observed from November to December. Our results suggest that, in SW Madagascar, species richness and abundance of post-larval fish may only be predicted prior to the ecological impacts of tropical storms on larval settlement success.