One quadrat or two?

Sometimes a 30m x 30m quadrat is not sufficient to cover the variation and patchiness inside an area of bushland. The more open the overstorey becomes, the harder it is to set up a quadrat which is relatively homogeneous in structure. For example, if you picked a spot at random in an open red-gum woodland, you might end up in an open area with no trees, which is not representative of the whole the area, or an area with one large redgum, which would also not be representative of the area as a whole. What are the options in such a case?

One option might be to have a bigger quadrat, say 100 x 100 m, which would be sure to take in a representative proportion of each area (open and tree'd). This is a problem for the method, however, since a number of the benchmarks are specific to a 30 x 30 m area, and adjusting for size is not simply a matter scaling up linearly (e.g. the species - area curve). The other problem with a bigger quadrat is that it takes longer to do.

The other option (and the one I am recommending here) is to do 2 quadrats, each 30 x 30m, in homogenous sections of each type of area (e.g. open v. tree'd). If a single condition indicator is required, then a composite indicator could be derived from an estimate of the proportion of the patch comprising each type (e.g. if open areas comprise approximately 60% of the patch, and tree'd 40%, then Species Diversity score would be 60% x Species Diversity Score in the open + 40% x Species Diversity Score in the tree'd area).

It means more work, but there is really no alternative. Trying to line up a quadrat so that it gets a representative bit of open ground and a representative bit of tree'd area in the approximate proportions of the whole patch is difficult to gauge, and I suspect it ends up with a bias towards tree'd areas, especially as the vegetation associations get more open.

What do you think?