During the mid-1900s, most of the island of Puerto Rico was deforested, but a shift in the economy from agriculture to small industry beginning in the 1950s resulted in the abandonment of agricultural lands and recovery of secondary forest. This unique history provides an excellent opportunity to study secondary forest succession and suggest strategies for tropical forest restoration. To determine the pattern of secondary succession, we describe the woody vegetation in 71 abandoned pastures and forest sites in four regions of Puerto Rico. The density, basal area, aboveground biomass, and species richness of the secondary forest sites were similar to those of the old growth forest sites (>80 yr) after approximately 40 years. The dominant species that colonized recently abandoned pastures occurred over a broad elevational range and are widespread in the neotropics. The species richness of Puerto Rican secondary forests recovered rapidly, but the species composition was quite different in comparison with old growth forest sites, suggesting that enrichment planting will be necessary to restore the original composition. Exotic species were some of the most abundant species in the secondary forest, but their long-term impact depended on life history characteristics of each species. These data demonstrate that one restoration strategy for tropical forest in abandoned pastures is simply to protect the areas from fire, and allow natural regeneration to produce secondary forest. This strategy will be most effective if remnant forest (i.e., seed sources) still exist in the landscape and soils have not been highly degraded. Patterns of forest recovery also suggest strategies for accelerating natural recovery by planting a suite of generalist species that are common in recently abandoned pastures in Puerto Rico and throughout much of the neotropics.

The internal transformations of nitrogen in terrestrial ecosystems exert strong controls over nitrogen availability to net primary productivity, nitrate leaching into groundwater, and emissions of nitrogen-based greenhouse gas. Here we report a reductive pathway for nitrogen cycling in upland tropical forest soils that decreases the amount of nitrate susceptible to leaching and denitrification, thus conserving nitrogen in the ecosystem. Using 15N tracers we measured rates of dissimilatory nitrate reduction to ammonium (DNRA) in upland humid tropical forest soils averaging ;0.6 mg·g21·d21. Rates of DNRA were three times greater than the combined N2O and N2 fluxes from nitrification and denitrification and accounted for 75% of the turnover of the nitrate pool. To determine the relative importance of ambient C, O2, and NO3 concentrations on rates of DNRA, we estimated rates of DNRA in laboratory assays using soils from three tropical forests (cloud forest, palm forest, and wet tropical forest) that differed in ambient C and O2 concentrations. Rates of DNRA measured in laboratory assays ranged from 0.5 to 9 mg·g21·d21 in soils from the three different forests and appeared to be primarily limited by the availability of NO3, as opposed to C or O2. Tests of sterile soils indicated that the dominant reductive pathway for both NO2 and NO3 was biotic and not abiotic. Because NH4 is the form of N generally favored for assimilation by plants and microbes, and NO3 is easily lost from the ecosystem, the rapid and direct transformation of NO3 to NH4 via DNRA has the potential to play an important role in ecosystem N conservation.

Rainfall interception (I) was measured in 20 m tall Puerto Rican tropical forest with 4 complex topography for a one-year period using totalizing throughfall (TF) and stemflow 5 (SF) gauges that were measured every 23 days. Measured values were then compared to 6 evaporation under saturated canopy conditions (E) determined with the Penman-Monteith 7 (P-M) equation, using (i) measured (eddy covariance) and (ii) calculated (as a function of 8 forest height and wind speed) values for the aerodynamic conductance to momentum flux 9 (ga,M). E was also derived using the energy balance equation and the sensible heat flux 10 measured by a sonic anemometer (Hs). I per sampling occasion was strongly correlated with rainfall (P): I = 0.21P + 0.60 (mm), r2 11 = 0.82, n = 121. Values for canopy storage 12 capacity (S = 0.37 mm) and the average relative evaporation rate (E/R = 0.20) were 13 derived from data for single events (n = 51). Application of the Gash analytical 14 interception model to 70 multiple-storm sampling events using the above values for S and 15 E/R gave excellent agreement with measured I. For E/R = 0.20 and an average rainfall intensity (R) of 3.16 mm h-1, the TF-based E was 0.63 mm h-116 , about four times the value derived with the P-M equation using a conventionally calculated ga,M (0.16 mm h-117 ). 18 Estimating ga,M using wind data from a nearby but more exposed site yielded a value of E (0.40 mm h-119 ) that was much closer to the observed rate, whereas E derived using the energy balance equation and Hs was very low (0.13 mm h-120 ), presumably because Hs was 21 underestimated due to the use of too short a flux-averaging period (5-min). The best 22 agreement with the observed E was obtained when using the measured ga,M in the P-M equation (0.58 mm h-123 ). The present results show that in areas with complex topography, 1 strongly underestimated when calculated using 2 equations that were derived originally for use in flat terrain; hence, direct measurement of ga,M using eddy covariance is recommended. The currently measured ga,M (0.31 m s-13 ) 4 was at least several times, and up to one order of magnitude higher than values reported for forests in areas with flat or gentle topography (0.03–0.08 m s-15 , at wind speeds of about 1 m s-16 ). The importance of ga,M at the study site suggests a negative, downward, 7 sensible heat flux sustains the observed high evaporation rates during rainfall. More work 8 is needed to better quantify Hs during rainfall in tropical forests with complex 9 topography.

Little is known about ecosystem-level responses to multiple, climatic disturbance events. In the subtropical forests of Puerto Rico, the major natural disturbances are hurricanes and droughts. We tested the ecosystem-level effects of these disturbances in sites with different land use histories. From 1989 to 1992, data were collected to determine the effects of Hurricane Hugo and two droughts on litterfall inputs, fine-root biomass, and decomposition rates in three topographic locations (stream, riparian, upslope) within two watersheds. From 1994 to 1998, we added a third watershed and an experiment in which coarse-wood levels were manipulated to simulate hurricane inputs. Data were collected on tree and palm growth rates, litterfall inputs, fine-root biomass, and decomposition rates. From 1994 to 1998, four hurricanes and three droughts were recorded. Measured parameters had unique responses and recovery rates to hurricanes and droughts. Litterfall inputs returned to long-term mean rates within one month following droughts and small-to-moderate hurricanes but required five years to recover after an intense hurricane. In contrast, fine-root biomass recovered seven months after an intense hurricane but failed to recover after five years following a severe drought. Despite the dramatic effects of these weather events on some ecosystem parameters, we found that aboveground measures of tree and palm growth were more affected by preexisting site conditions (e.g., nitrogen availability due to past land use activities) than hurricanes or droughts. The addition of coarse woody debris increased tree and palm growth, fine-root biomass, and litter production; however, in the case of tree and palm growth, this effect was least measurable in the sites with the highest productivity. We found that decomposition rates were more controlled by litter quality than weather conditions. In conclusion, we found that certain ecosystem structures (e.g., canopy structure and fine-root biomass) generally recovered more slowly from disturbance events than certain ecosystem processes (e.g., plant growth rates, decomposition rates). We also found that past land use activities and disturbance legacies were important in determining the responses and recovery rates of the ecosystem to disturbance.