Spatial autocorrelation and scales of variability in litter density: implications for survey designs

Abstract (English)

A lack of standardisation of survey methods makes it difficult to assess and compare macrolitter densities within and across the different environmental compartments (e.g., beaches, seafloor and surface) in time and space. A recent systematic review of research on marine macrolitter densities published from 2015–2020 highlights a considerable range in survey designs. For seafloor surveys, for example, scuba dive surveys covering relatively small square or rectangular plots were most commonly used for sites within a km from shore, and generally reported the highest densities of macrolitter. Contrastingly, trawl surveys utilising much larger plots which are long and narrow (width of trawl, length of haul) were commonly used further from shore, along with image-based ROV surveys. The latter dominated exclusively the most remote surveys, and also yielded plots of similar length or longer than trawl surveys, but with an extremely narrow field of view. Plot size and shape has been shown to influence survey results in ecological surveys and may also do so for marine litter surveys. The distribution of litter is not very well understood, but it is clearly heterogeneous and patchy in both space and time. Image-based ROV surveys reported clean seafloor plots more frequently than trawl surveys, possibly because the former's narrow field of view makes the interception of a patch less likely, although this is confounded somewhat by tendencies to use the two methods on different substrates and potential differences in detection probability of litter intercepted. The scales of spatial heterogeneity and patchiness in macrolitter distribution are also not well understood and potential spatial autocorrelation has implications for sample independence and statistical analyses. The review points to long-established considerations from ecological research pertaining to independence of data points, spatial autocorrelation, sampling scale, and plot size and shape and highlights how these are highly relevant also for marine litter research. These considerations are relevant both for global standardisation efforts and for independent studies. In addition, the author will introduce the results of simulations on the effect of sample size on precision and accuracy of beach litter surveys based on a negative binomial distribution of beach litter density data from nearly 300 beaches along the coast of Norway. 

Notes