Estimating the biomass density of macroalgae in land-based cultivation systems using spectral reflectance imagery

Sub-sampling large-scale, land-based macroalgal cultures to estimate stocking densities is time-consuming, labour-intensive, and inaccurate. Therefore, the development of innovative methods to monitor stocking densities are required to maximise macroalgal productivity. In this study the spectral reflectance of a range of biomass densities (0.5-12.0 g L-1) of Ulva ohnoi was measured using a spectroradiometer (ASD) and a multispectral camera (REMX). A two-band normalised difference vegetation index (NDVI), which was standardised by depth of the culture system, was quantified from both sets of data, and used to establish biomass density sensing models. The models developed using the ASD and REMX data had strong piecewise linear relationships between the standardised NDVI and biomass density (R-2 > 0.85, residual mean standard error < 1.31, and mean standard error < 0.99), but differed in breakpoint and slope of the regression lines. Subsequently, both models were validated using data collected in 4000 L high-rate algal ponds, representing practical conditions of the commercial land-based production of macroalgae. Notably, the REMX model had a better fit with the validation data, and the mean difference in the actual versus predicted densities was <= 0.76 g L-1, with a maximum difference of 1.73 g L-1. Therefore, the REMX biomass density model is a useful management tool for differentiating between low (< 2 g L-1), medium (2-4 g L-1), and high densities (> 4 g L-1) using the standardised NDVI. This study has demonstrated that there is a predictive relationship between spectral reflectance and the biomass density of U. ohnoi. Notably, the REMX model will inform decision making around the frequency of harvest, as well as the quantity of biomass to be harvested, to maintain an optimal stocking density and, therefore, high productivities at a commercial scale.