Modelling the seasonal distribution of macrozoobenthos in intertidal areas of the Dtuch Wadden Sea using novel rmeote sensing and deep learning methods

Abstract

Macrozoobenthos are an essential constituent of intertidal systems and their food webs. Changes in their distributions affect higher levels in the food web, such as fish and coastal birds, and can also can serve as indicators for shifts in boundary conditions of the system. It is therefore essential to regularly monitor macrozoobenthic distributions. Field sampling is a labor-intensive and time-consuming method and is therefore not suitable for frequent monitoring. Remote-sensing methods are also challenging due to the homogeneous spectral signatures of intertidal areas. The present study proposes a novel method to predict the seasonal distribution of macrozoobenthic species presence, biomass and abundance, based on feature extraction from a variational autoencoder (VAE) model. Seasonal field observations were gathered in September 2021, April 2022 and July 2022 and the data was used to test the prediction models’ performances.