Forested soils in the tropics contain a large carbon pool that may respond to global environmental changes such as climate warming and land-use change. A better understanding of the distribution of tropical soil carbon (C) pools is necessary in order to manage soil C as well as to predict its potential responses to global change. The goals of this study were to quantify the relationships among soil C and environmental variables for 35 forest plots in a 140,000-ha landscape in northeastern Costa Rica, and to identify variables that can predict soil C storage at unsampled sites. The biophysical variables included indices of net primary productivity (forest floor mass, root biomass, and an index of productivity derived from satellite imagery), soil particle-size distribution and mineralogy, elevation, and slope. Soil carbon storage in these volcanic soils was relatively high, ranging from 51.1 to 138.6 Mg C ha-1 in the top 30 cm of mineral soil. The relationships among forest soil C and biotic and abiotic variables were different for low-elevation (< 120 m) and high-elevation (120-800 m) sites, and elevation explained much of the variability in soil C concentrations. Soil particle-size distribution and mineralogical variables are correlated in this landscape and co-vary in predictable ways along the elevation gradient. Thus, elevation represents a weathering gradient with younger, allophanic soils at higher elevations and older soils with gibbsite, goethite, and kaolinite as dominant clay minerals in the lowlands. We propose two mechanisms of C stabilization: soil C concentrations and contents are positively correlated to the amount of noncrystalline clays (e.g. allophane, imogolite, and ferrihydrite) in the high elevation soils, and positively correlated to aluminum in organo-metal complexes in the low elevation sites. The strong correlations among soil C concentrations, contents, and elevation (mediated through effects on soil mineralogy) indicate that it is possible to predict soil C in this landscape using variables that are easily mapped in a geographic information system.
Bibliographical noteFunding Information:
We would like to thank Phil Sollins, Dan Richter, Elissa Levine, and Jason McLachlan for helpful reviews of earlier drafts of this manuscript. We especially would like to thank I.C. Grieve and D. Van Dam, who provided excellent suggestions in their reviews of the manuscript. Jane Read provided the NDVI data, and David Clark generously lent us his GPS and community base station. Adolfo Downs, Rigo Gonzalez, and Rodolfo Vargas provided invaluable support in the field and lab. This work was made possible by funding from the following sources: NASA Graduate Student Researcher Program Fellowship, NSF Dissertation Improvement Grant (No. 9972569), Andrew W. Mellon Foundation Fellowship for Ecosystem Studies awarded through the Organization for Tropical Studies, National Security Education Program Fellowship, Center for International Studies Travel Grant from Duke University, and a Tinker Field Research Grants from the Duke University Latin American Studies Program.
- Costa Rica
- Soil carbon
- Soil mineralogy
- Tropical rain forests