Unless the .68 ha was sampled on a random (or other probability) basis from the full 1.33, you can't get a confidence interval. (If it was, then I suppose you could, as long as you apply a Finite Population Correction.) It sounds like it was not randomly selected, so about all you can do is inflate the count of trees by a multiplication factor of 1.33/.68, which of course assumes the part of the forest not counted had about the same tree density as the part included in the count. But maybe they only did the inventory where there were the most trees (worse yet, maybe the rest was paved over), which would invalidate that inflation factor. As the adage goes, it's always dangerous to extrapolate beyond the observed data.

Thanks for the answer Burke.

So, the 0.68 area was fully measured, not sampled.

I've checked satellite images of the area and It seems to be the same tree density.

The variable of interest we want asses is the biomass. But we are doing it with the mean size of the tree. That's why I need a tree density.

The most honest thing to do is assume that the 1.33 area has the same density of the 0.68 area?

I understand the .68 was fully measured. My point was, whether the fraction of the total area that was to be fully measured was selected as a random subarea of the total area, or just because the counters stopped when they got tired of counting, or they intentionally ignored non-forested areas, or whatever. It sounds like the area to be fully counted was not randomly selected, but appears from the satellite images to be similar in tree density to the area as a whole. In that case, then yes, assuming a roughly uniform density seems like the way to proceed. And the honest thing is to acknowledge you are making that assumption, with your justification being the sat. images.

Thanks, Burke D. Grandjean you really helped me a lot, I'll proceed that way.

Take care.