“…We did this using a priori geographical data (see Supplementary Methods S3.2). After the experiment, we then directly quantified the level of spatial environmental heterogeneity in each cluster by calculating multivariate dispersion on 14 z-score standardised environmental variables: turbidity (depth of visibility (m), 15 m resolution), relative wave exposure (unitless, 15 m resolution) from geographic data 52 along with variables that we measured in situ during the experiment, namely: distance from panel to the seabed (m), panel depth (m), the average, coefficient of variation, maximum and minimum temperature over the course of the experiment (ºC) and the average, coefficient of variation and maximum light level over the course of the experiment (lux), salinity (ppt), Secchi depth (m) and water movement (mass reduction g hour -1 ) (see Supplementary Methods S3.3 for details). Multivariate dispersion was calculated using the betadisper() function from the vegan package 53 in R v4.1.2 54 .…”