Because I’ve been in the forest and unable to upload posts, and because of a short lag-time in getting this blog up and running, my posts have fallen behind a bit. With this entry, I’ll bring us up to date as of 8 May. Our forest work falls into three natural stages: i) surveying the inventory plots, ii) collecting photographs of bark slashes and fallen leaves of each tree, and iii) collecting samples of twig-wood and quality specimens for herbarium matching and DNA analysis. Stages (ii) and (iii) have been running since 8 April, and we have now examined all the trees in plots A-E (five out of six), and made specimens of just over half the morphotypes.
What’s slowing us down the most is matching the morphotypes within and among plots (after stage ii), which involves my sitting at the laptop until late at night, looking at the day’s images of bark slashes and leaves and comparing them to images in our growing ‘library’ of morphotypes (I’ll leave a discussion of the issues involved in defining and matching morphotypes for another post). Yet this step is vital in determining how many morphotypes we have so that specimens may be collected for them. So far, I have finished the matching in four plots (A-D), and from these 752 trees (larger than 10 cm diameter) of 236 morphotypes. Since the plot size is a quarter hectare, it is tempting to announce that this also corresponds to a diversity of 236 species per hectare, among the highest for any rain forest anywhere. However, the addition of plots from two distinct habitats will inflate the estimate over what one would expect for a single, contiguous hectare in either one of the habitats.
A better approach is to ‘rarefy’ the individuals in each habitat down to a standard number, say 100 trees. I’ve run some preliminary analyses on these data, using the comprehensive, ‘free and open source’ statistical tool R. Using a subsampling algorithm (rarefy in the vegan package) to generate a diversity estimate for 100 individuals in each plot (thus correcting for different numbers of trees in each plot), the diversity of the bottomland plots A and C is 68 and 50 morphotypes, and of the hill plots B and D is 54 and 65 morphotypes (mean = 59.5). This still places our site among the most diverse in the world: corresponding estimates for Yasuni (Equador), Lambir (Malaysia), Barro Colorado Island (Panama), and Sinharaja (Sri Lanka) are 67, 61, 35, and 27, respectively (Condit et al. 2004). Using an estimator developed by Taiwanese statistician Anne Chao, the total number of tree species in this patch of forest is predicted to be 390!
That we are dealing with (at least) two different habitat types is clear, based on a simple clustering analysis. Imagine an abstract, multidimensional space with 236 axes, one each for the abundance of each tree species (or at least imagine you can imagine it!). A forest plot is represented by a point in that space (with coordinates of, e.g., 1,0,0,2,0,1,1,…). If two plots are similar in species composition, they will be close together in our abstract “species space.” If they share few species, their points will be far apart. One can simply measure the ordinary distance between plot points, and collapse those distances into a “clustergram” (see figure). What we see with our own data is that alluvial plots A and C are similar in composition, and hill plots B and D are similar in composition, but that these two clusters (of two plots each) are quite dissimilar. This is what any one of our team would expect, having observed how some trees only grow on the hills and some only on the flat land; it’s always nice, though, to see it in black and white!
All in all, I’m pleased with our progress to date. We have a lot more specimens to collect, but the team is at it right now as I type. I’m hoping we’ll be done with the expedition in two weeks.