- That’s right
- We do not provide documentation on this because it is domain, and potentially even site, specific. Visual impact indices may vary for different types of features. The example from that paper is for wind farms, but you might expect something different for aquaculture, landfills, or other large infrastructure. My professional judgement is that your results will likely be more robust if you do not use a visual impact function and stick to visibility. You’ll note that the remainder of that paper uses the framing of visibility without assigning values to it. If you do choose to use an impact function, it should be derived from other studies that investigate the features of interest. Importantly, it needs to be formulated as a function of distance to be used in this model.
The previously linked paper also uses the “weight” feature (an attribute of the point vector layer input) of the model to weight the viewsheds by the number of viewer-days at each point. This can be a useful hybrid approach where you can say something more about the density of viewers across the landscape, yet stopping short of assigning a value. I’d read that paper carefully, to get a sense about how the model can be used from the perspective of viewers, or the viewed, with different interpretations of model outputs.