Complimenting modern records of tropical cyclone activity with longer historical and paleoclimatological records would increase our understanding of natural tropical cyclone variability on decadal to centennial time scales. Tropical cyclones produce large amounts of precipitation with significantly lower δ18O values than normal precipitation, and hence may be geochemically identifiable as negative δ18O anomalies in marine carbonate δ18O records. This study investigates the usefulness of coral skeletal δ18O as a means of reconstructing past tropical cyclone events. Isotopic modeling of rainfall mixing with seawater shows that detecting an isotopic signal from a tropical cyclone in a coral requires a salinity of ~33 psu at the time of coral growth, but this threshold is dependent on the isotopic composition of both fresh and saline end-members. A comparison between coral δ18O and historical records of tropical cyclone activity, river discharge, and precipitation from multiple sites in Puerto Rico shows that tropical cyclones are not distinguishable in the coral record from normal rainfall using this approach at these sites.

Forests are often subject to multiple, compounded disturbances, representing both natural and human-induced processes. Predicting forest dynamics requires that we consider how these disturbances interact to affect species demography. Here we present results of an individual-based, spatially explicit forest simulator that we developed to analyze the compounded effects of hurricane disturbance and land use legacies on the dynamics of a subtropical forest. We used data from the 16-ha Luquillo Forest Dynamics Plot in Puerto Rico, together with a reconstruction of historical wind damage, to parameterize the simulator. We used the model to ask two questions. (1) What are the implications of variation in hurricane frequency and severity for the long-term dynamics of forest composition, diversity, and structure? Both storm severity and frequency had striking effects on forest dynamics, composition, and structure. The periodicity of disturbance also played an important role, with periods of high hurricane activity fostering the establishment of species that may become rare in the absence of severe storms and quiescent periods allowing these species to reach reproductive size. Species responses to hurricane disturbance could not be predicted from their life history attributes. However, species perceived to be primary forest species exhibited low temporal variation in abundance through the simulations. (2) How do hurricanes and legacies from human land use interact to determine community structure and composition? Our results suggest that, over time, regardless of the storm regime, land use legacies will become less apparent but will lead to a forest community that contains a mixture of secondary and primary forest species formerly dominant in areas of different land use. In the long term, hurricane disturbance generated two communities with slightly greater similarity than those not subject to storms. Thus, the inclusion of hurricane disturbance does not alter the prediction that land use legacies in this tropical forest will diminish over time. Our simulations also highlight the contingent effects of human legacies on subsequent community dynamics, including the response to hurricane disturbance, therefore supporting the notion that compounded disturbances can interact in ways that cannot be predicted by the study of single disturbances. The widespread importance of land use as a large-scale disturbance makes it imperative that it be addressed as a fundamental ecological process.

The forests and soils at Luquillo Experimental Forest (LEF), Puerto Rico, are frequently disturbed by hurricanes occurring at various frequencies and intensities. We have derived a forest version of the Century soil organic matter model to examine the impact of hurricanes on soil nutrient availability and pool sizes, and forest productivity in the tabonuco forest at Luquillo. The model adequately predicted aboveground plant production, soil carbon, and soil nitrogen levels for forest conditions existing before Hurricane Hugo. Simulations of Hurricane Hugo and of an historical sequence of hurricanes indicated a complex pattern of recovery, especially for the first 10 yr after the hurricanes. After repeated hurricanes, forest biomass was reduced, while forest productivity was enhanced. Soil organic matter, and phosphorus and nitrogen mineralization stabilized at higher levels for the LEF than for hurricane-free tabonuco forest, and organic soil phosphorus was substantially increased by hurricanes. Results from these simulations should be regarded as hypotheses. At present there is insufficient data to validate the results of hurricane model simulations.

The Luquillo Mountains of eastern Puerto Rico are used as a case study to evaluate possible single- or multiple-factor controls of productivity in montane forests. A review of published studies from the Luquillo Mountains revealed that canopy height, productivity, and species richness decline while stem density increases with elevation, as is typical of other montane forests. A mid-elevation floodplain palm stand with high levels of productivity provides a notable exception to this pattern. Previous basic and applied studies of productivity in the Luquillo Mountains have consistently considered the overall gradient in productivity to be important in understanding forest structure and function. Recent observational and experimental studies have determined that disturbance of all types is an important factor mediating productivity in both low- and high-elevation (cloud) forests. For example, low-elevation forest recovers more quickly from hurricane disturbance and is more responsive to nutrient additions than is cloud forest. All of the factors proposed for limiting productivity are supported in one way or another by research in the Luquillo Mountains. What is critically lacking is both an appreciation for the way that these factors interact and experiments appropriate to evaluate multiple controls acting simultaneously.