This will depend on the level of mixing in the water body, which may vary seasonally.
For water bodies where there is a lot of mixing (i.e. rivers) eDNA is more homogeneously distributed.
In still water (i.e. ponds) then there is much more spatial heterogeneity and so the probability of detection is lower if water is taken from a single point, and appropriate (sub)sampling design is key to cover all microhabitats.
However, see Lawson-Handley et al. (2019) for a comprehensive study of spatial dynamics of eDNA in large lakes. This study concluded that shoreline sampling was sufficient to detect all species in Lake Windermere during the winter when more mixing occurred, and only missed one species (Arctic Charr, which lives deep in the middle of the lake) during summer when there was less mixing.