We sampled soils from 216 profiles representing 24 sites in the El Yunque National Forest to determine amounts C, N and neutral-salt-extractable Ca++, Mg++ and K+. Following the classic paradigm, we assessed the influence of climate (modeled precipitation, modeled temperature and/or elevation as a surrogate variable for both), forest type (tabonuco, colorado, palm), parent material (quartz diorite, volcaniclastics), and topography (catena positions ridge, slope, valley and % slope) on the distribution of these nutrients. To separate the effects of vegetation from those of climate, half of the sites were located between 500 and 700 m in the three forest types where rainfall and temperature were not significantly different. Using a combination of ANOVA (or Kruskal-Wallis) and univariate regression trees we determined that the amount of carbon in the top 80 cm of soil was influenced primarily by forest type (c > p > t) probably driven by differences in litter and/or root C:N ratios. Topographic position was significantly correlated with C amount (v > s, r), with the higher C amounts in the valleys probably driven by low O2 levels. Bedrock type was significantly correlated with C amount in c and p stands, but not in the tabonuco type. N was strongly correlated with C as expected. Exchangeable Ca was different across forest types (t > c, p) and bedrock type (qd > vc). Mg and K were differed by forest type, but not by bedrock type (t > c, p) or any other variables.

The next phases of this project are (1) to determine levels of these nutrients below the root zone (80-140 cm) and the factors controlling their distribution; and (2) establish field experiments to test the results of the regression trees which indicate that the C:N ratio of litter and/or root inputs is the most important variable influencing C distribution. The latter represents a first step in exploring the usefulness of regression trees as a way of sorting out the relative importance of each of the state factors (climate, topography, organisms, parent material and time) in the classic paradigm relating environmental variables to soil properties.

Soil C differs markedly across forest types (c> p> t, p s, r, p