Counting using deep learning regression gives value to ecological surveys

Many ecological studies rely on count data and involve manual counting of objects of interest, which is time-consuming and especially disadvantageous when time in the field or lab is limited. However, an increasing number of works uses digital imagery, which opens opportunities to automatise counting tasks. In this study, we use machine learning to automate counting objects of interest without the need to label individual objects. By leveraging already existing image-level annotations, this approach can also give value to historical data that were collected and annotated over longer time series (typical for many ecological studies), without the aim of deep learning applications. We demonstrate deep learning regression on two fundamentally different counting tasks: (i) daily growth rings from microscopic images of fish otolith (i.e., hearing stone) and (ii) hauled out seals from highly variable aerial imagery. In the otolith images, our deep learning-based regressor yields an RMSE of 3.40 day-rings and an [Formula: see text] of 0.92. Initial performance in the seal images is lower (RMSE of 23.46 seals and [Formula: see text] of 0.72), which can be attributed to a lack of images with a high number of seals in the initial training set, compared to the test set. We then show how to improve performance substantially (RMSE of 19.03 seals and [Formula: see text] of 0.77) by carefully selecting and relabelling just 100 additional training images based on initial model prediction discrepancy. The regression-based approach used here returns accurate counts ([Formula: see text] of 0.92 and 0.77 for the rings and seals, respectively), directly usable in ecological research.

No evidence for reduced growth in resident fish species in the era of de-eutrophication in a coastal area in NW Europe

Coastal areas in north-western Europe have been influenced by elevated nutrient levels starting in the 1960s. Due to efficient measures, both nitrate and phosphate levels decreased since the mid-1980s. The co-occurring declines in nutrient loadings and fish productivity are often presumed to be causally linked. We investigated whether four resident fish species (twaite shad, bull-rout, thick-lipped grey mullet and eelpout), that spend the majority of their life in the vicinity of the coast, differed in growth between the historic eutrophication period compared to the recent lower nutrient-level period. Based on Von Bertalanffy growth models of length at age, and the analysis of annual otolith increments, we investigated the difference in sex-specific growth patterns and related these to temperature, eutrophication level (Chlorophyll a), growth window and fish density. In all four species, annual otolith growth rates during the early life stages differed between the two periods, mostly resulting in larger lengths at age in the recent period. All species showed significant correlations between increment size and temperature, explaining the observed period differences. The lack of an effect of total fish biomass provided no evidence for density dependent growth. A correlation with chlorophyll was found in bull-rout, but the relationship was negative, thus not supporting the idea of growth enhanced by high nutrient levels. In conclusion, we found no evidence for reduced growth related to de-eutrophication. Our results indicate that temperature rise due to climate change had a greater impact on growth than reduced food availability due to de-eutrophication. We discuss potential consequences of growth changes for length-based indicators used in management.