Laminaria hyperborea

For Laminaria hyperborea, plant sizes ranged from 0.1 to 3 kg in fresh weight and from 90cm to 300cm in length (Figure A6a) with plants in Scotland reaching much larger maximum sizes than those in England and Wales (Figure A6d). Sub-canopy plants less than 70cm in length were excluded. The weight of the blades (or lamina) was directly proportional to the length of the plants, and typically formed 50-80% of the total weight (Figure A6b), albeit with some variation among sampling locations. Percentage cover and the number of plants per unit area correlated strongly for sparse populations (Figure 8c), reaching 100%

cover at around 7 plants per m². Complete cover (100%) covered a wide range of plant densities (4-17 plants per m²). The overall relationship between cover and density was similar across regions.

Estimation of the biomass of Laminaria hyperborea plants per unit area from the diver-collected data was done in two ways: (1) using regional average plant weights (Figure A6d) multiplied by number of plants in each quadrat (Figure A6c), then averaging the per-quadrat estimates of plant biomass (Figure A8d) to give a site-specific biomass density; and (2) by multiplying location-specific average plant density by location-specific average plant weight (Figure A7a). Uncertainty for the first method was expressed as the standard deviation and standard error of the average total plant weight per quadrat, but this approach did not account for the variation in weight among individual plants at each location. For the second approach, the separate standard deviations and standard errors for the mean plant sizes and mean plant densities per location were combined using the formula for the propagation of uncertainties for products. Thus standard deviation for estimated biomass is a combination of the contributing means and standard deviations:

= + ( ) ,

where b is biomass per m², d is density as number of plants per m², and w is the weight of individual plants in kg; with the same equation giving the standard error of estimated biomass by substituting SE for SD.

(a) (b)

(c) (d)

Figure A6. Laminaria hyperborea collected in UK diver surveys: (a) fresh weight versus total length; (b) weight of lamina as a proportion of total weight; (c) the relationship between plant density and percentage cover in 1m² quadrats; (d) the distribution of plant fresh weights (FW g) in each region.

(a) (b)

(c)

Figure A7. Laminaria hyperborea collected in UK diver surveys: (a) average weight per plant versus average number of plants per quadrat with error bars showing standard deviations;

(b) as (a) but with error bars showing standard errors of estimates for plant weight and density; (c) estimated of biomass density as wet weight per m² with standard error bars from combined averaged plant size and density.

Ranges of estimates for biomass (fresh weight) per m² estimates for each survey location (Figure A8d) values were within the 0 to 25kg/m² range of values reported by Kain (1977).

(a) (b)

(c) (d)

Figure A8. Laminaria hyperborea collected in UK diver surveys: (a) plant density per location; (b) plant density per region; (c) plant biomass obtained by multiplying regional average plant sizes with plant density versus plant density per 1m² quadrat. Midpoint densities for each SACFOR category are shown as dotted vertical lines; (d) Estimated biomass per quadrat across sites.

The next step was to compare observed values for biomass per unit area from diver-surveyed locations with those predicted by the model for the same locations. Biomass scale 1 (Table A1) values for each category were used to convert the predicted likelihood of kelp presence to biomass across the whole UK model domain. Locations surveyed in this study, supplemented by recent data collected by Dan Smale and Pippa Moore since 2015, were used to extract predicted biomass values from the mapped data. It became clear that the 200-m scale of the larger model did not effectively give the depth of diver survey locations, so depth values extracted from maps were replaced with the actual depth of each survey.

Likelihood of kelp presence was then recalculated using map-derived estimates of temperature, chlorophyll concentrations and wave exposure and survey-specific depth.

Combining the revised likelihoods with biomass per abundance category once more gave more exactly comparable estimates.

Given that the two methods of estimating biomass were entirely independently derived, the correspondence between the two estimates was reassuring (Figure A9). With biomass scale 1 (Table A1) using percentage cover associated with each abundance category to scale biomass per category from zero to a maximum of 28kg/m², the kelp suitability model consistently under-predicted kelp biomass: suitability model predictions of biomass using biomass scale 1 were between 2 and 7 kg/m² and were 43% of those estimated from locally determined plant densities and regional average plant sizes.

Figure A9. Laminaria hyperborea collected in UK diver surveys. Comparison between biomass estimated from local average plant size and plant density (x-axis) and biomass predicted by the UK-scale kelp habitat suitability model using biomass scale 1. The no-intercept regression model (dotted line) had an R² of 0.86 (Model estimate = 0.000434 (+/- 0.000050 standard error) x local estimate). Circles show estimates for survey locations and squares show averages across the three locations in each region.

Habitat suitability model predictions for biomass were further produced for locations reported in Kain (1977), using observed biomass data extracted from Kain’s Figure 1. Comparison between modelled and observed biomass per unit area again showed underestimation of kelp biomass on average, especially in shallow areas with over-prediction biomass at greater depths, evident in data from Barra (SE Muldoanich, Figure A10b). Both the comparisons with diver survey data and those from the literature suggested that the biomass estimates used for each category were too small in Scale 1.

This under-fitting suggested that modification was needed to the biomass scaling relationship. Scaling biomass to the plant density instead of percentage cover (Biomass Scale 2, Table A1) did not alter the fit, but increasing the biomass for the “Abundant”

category to 25 kg/m² did substantially improve the fit of the model to both Kain’s data and the diver surveys (Figure A11).

(a) (b)

(c) (d)

Figure A10. Laminaria hyperborea biomass per unit area estimates from Kain (1977).

Biomass Scale1: (a) versus modelled biomass (y-axis) for the same locations and depths, with error bars derived from the standard errors of GAM model estimates, and (b) against depth (squares, observations; circles, model predictions). (c, d) as (a, b) but using Biomass Scale 2 modified.

Figure A11. Laminaria hyperborea collected in UK diver surveys. Model predictions (y-axis) versus estimated biomass (x-axis) for Biomass Scale 2 modified. Error bars show approximate 95% confidence intervals: on the x-axis as 2 x standard error of the mean biomass estimated from site-specific average plant sizes and plant densities (R²=0.758, b=0.443 +/-0.064).