Effects of Model Choice and Forest Structure on Inventory-Based Estimations of Puerto Rican Forest Biomass

-TO~~~ aboveground live tree biomass in Puerto Rican lower montane wet, subtropical wet, subtropical moist and subtropical dry forests was estimated using data from two forest inventories and published regression equations. Multiple potentially-applicable published biomass models existed for some forested life zones, and their estimates tended to diverge with increasing tree diameter at breast height. Inventoried forests showed structural characteristics typical of secondary tropical forests and stand successional trends of increasing stem density from initial reversion to young closed forest, followed by a decrease in stem density and gradual increase in basal area and biomass as the stands developed. Stems with D,, < 10 cm contributed 9.9-50.9% of total aboveground biomass. When present, 50-90+ cm DBH trees greatly increased aboveground biomass values on individual plots, but these effects subsided when averaged over the forested landscape. Inherent variability in large-tree form combined with equations that extrapolate beyond the range of sampled trees impedes accurate aboveground biomass estimation for larger trees. Application of equations developed in areas that most closely match potential study sites should improve overall estimation accuracy of all forest stands, however, locally developed biomass equations are lacking for subtropical dry and perhaps subtropical moist forest life zones in Puerto Rico. K ~ ~ ~ o ~ ~ ~ . B i o r n a s s , forest inventory, Puerto Rico, Rio Grande de Arecibo Following extensive deforestation (Wadsworth 1950), Puerto Rico's secondary forests developed into a unique combination of native Caribbean tree species and tree species introduced from tropical areas around the world (Lugo and Helmer 2004). Assessing the health and function of these new forest ecosystems requires a long-term inventory and monitoring effort. Estimating carbon fixed by and stored in tropical forest ecosystems provides insight into local and global biogeochemical cycles (Brown and Lugo 1982; Brown and Lugo 1990; Dixon et al. 1994; Grau et al. 2004; Phillips et al. 1998), so while past forest inventories in Puerto Rico focused on wood ms. received October 25, 2004; accepted May 22, 2005. volume and tree basal area (Birdsey and Weaver 1982; Franco et al. 1997), current inventories require accurate assessments of biomass and carbon. However, studies in other tropical forests have shown that uncertain biomass estimates complicate the quantification of forest carbon storage and fluxes (Brown et al. 1995; Houghton et al. 2000; Houghton et al. 2001; Nascimento and Laurance 2002). Regression equations used to estimate individual tree biomass have been cited as a source of significant error, particularly for larger trees (Overman et al. 1994; Clark and Clark 2000; Keller et al. 2001; Chave et al. 2004). Derivation of regression equations through destructive sampling allows researchers to model live tree aboveground biomass (AGB) as a function of one or more standing tree dimensions such as tree stem diameter at breast height (DDH at 1.37 m) and tree height (H,) (Parresol 1999). Largeoped from data that included sapling-sized trees. Crews located FIA plots with GPS units. A Geographic Information System (GIs) was used to assign Holdridge ecological life zones and parent material stratum from plot geographic coordinates. The FIA data set used in this study consists of 7 plots in the lower montane rain/wet forest with 315 trees, 72 plots in subtropical wet forest with 2,128 trees, 136 plots in subtropical moist forest with 3,640 trees, and 24 plots in the subtropical dry forest with 532 trees, for a total of 6,615 trees. An additional 38 plots categorized as recent reversions back to forest (< 10% forest tree canopy coverage) with 312 trees were separated from the main analyses because the goal was to first quantify AGB in more developed forests. Reversion plots were used in a separate analysis examining AGB changes over the course of stand development. The second data set comes from a forest inventory of the upper reaches of the Rio Grande de Arecibo watershed (RGA). After subtracting non-sampled areas such as water bodies, landslides, alluvial deposits, agriculture and urban/developed area, RGA has an area of 27,300 ha. It was inventoried in the summer 2003 by students of the University of Puerto Rico, Recinto Mayagiiez using a stratified random design. Details on the inventory design are available in Suarez-Rozo (2004). The number of plots was proportional to the area of each stratum and the number of days available to carry out field work. A total of 4,469 trees with DBH 2 10.0 cm were measured in 91, 0.1 ha square plots that were stratified by Holdridge ecological life zone and geological parent material. Total area for each forest type and parent material stratum was derived from 2001 Landsat imagery and GIs coverages (Suarez-Rozo 2004). The RGA inventory data used for this analysis gave 4 plots in lower montane wet forest with 234 trees, 47 plots in subtropical wet forest with 2,119 trees, and 8 plots in subtropical moist forest with 427 trees. Although standing dead trees were tallied in both inventories, only live trees were included in the AGB estimates. Shrubs and woody vines were not included in either inventory, so their biomass is not included in the estimates of AGB. Mangroves and other forest types of lesser extent that occur on Puerto Rico were not addressed by this study because the data sets used did not provide a sufficient number of samples.