The availability of nitrogen (N) and phosphorus (P) controls the flow of carbon (C) among plants, soils, and the atmosphere, thereby shaping terrestrial ecosystem responses to global change. Soil C, N, and P cycles are linked by drivers operating at multiple spatial and temporal scales: landscape-level variation in macroclimate, seasonality, and soil geochemistry; stand-scale heterogeneity in forest composition and structure; and microbial community dynamics at the soil pore scale. Yet in many biomes, we do not know at which scales most of the biogeochemical variation emerges, nor which processes drive cross-scale feedbacks. Here, we examined the drivers and spatial/temporal scales of variation in soil biogeochemistry across four tropical dry forests spanning steep gradients of climate, soil parent material, and plant community structure. To do so, we quantified soil C, N, and P pools, extracellular enzyme activities, and microbial community structure across wet and dry seasons in sixteen plots located in Colombia, Costa Rica, Mexico, and Puerto Rico.

Soil biogeochemistry exhibited marked heterogeneity across the sixteen plots, with total organic C, N, and P pools varying four-fold, and inorganic nutrient pools by an order of magnitude. Most soil characteristics changed more across space (i.e., among sites and plots) than over time (between dry and wet season samplings). We observed stoichiometric decoupling among C, N, and P cycles, which may reflect their divergent biogeochemical drivers. Organic C and N pool sizes were positively correlated with the relative abundance of ectomycorrhizal trees and legumes. By contrast, the distribution of soil P pools was driven by soil geochemistry, with larger inorganic P pools in soils with P-rich parent material.

Most earth system models assume that soils within a texture class operate similarly, and ignore sub-grid cell variation in soil properties. Here we reveal that soil nutrient pools and fluxes exhibit as much variation among four Neotropical dry forests as is observed across terrestrial ecosystems at the global scale. Thus, the biogeochemical patterns we observed across the Neotropical dry forest biome challenge representation of soil processes in ecosystem models.

We collected replicated soil samples (0-10 cm of mineral soil) from multiple plots in four tropical dry forests in Costa Rica, Mexico, Puerto Rico, and Colombia in the wet and dry season.  All samples were analyzed for the following variables in the same laboratory: particle-size distribution, pH, nitrogen mineralization, microbial biomass carbon, microbial biomass nitrogen, microbial biomass phosphorus, soil mineralogy via selective dissolution analyses, soil phosphorus fractionations via a modified Hedley procedure, extracellular enzyme activitities, phospholipid fatty acid profiles and other variables.  This dataset reports data for all individual soil samples, i.e. 5 cores per plot, 4 plots per site, for four sites total, in both the wet and dry season.

The relevant metadata describing all columns in the dataset appear in a tab in this spreadsheet.