The ecological importance of trees lasts much longer than their life spans. Standing dead trees (snags) and fallen trunks and branches are an important component of above-ground carbon stocks and nutrient reserves, provide habitat for wildlife, and interact with disturbance regimes (e.g. by serving as fuel for fires) (Clark et al. Reference CLARK, CLARK, BROWN, OBERBAUER and VELDKAMP2002, Harmon et al. Reference HARMON, FRANKLIN, SWANSON, SOLLINS, GREGORY, LATTIN, ANDERSON, CLINE, AUMEN, SEDELL, LIENKAEMPER, CROMACK and CUMMINS1986, Pyle et al. Reference PYLE, SANTONI, NASCIMENTO, HUTYRA, VIEIRA, CURRAN, HAREN, SALESKA, CHOW, CARMAGO, LAURANCE and WOFSY2008). Despite these diverse functions, woody debris stocks remain poorly quantified in tropical forests in general (Brown Reference BROWN1997), and in tropical dry forests in particular (Harmon et al. Reference HARMON, WHIGHAM, SEXTON and OLMSTED1995). More empirical studies of the patterns of woody debris and processes that control its dynamics are needed to understand its role in global biogeochemical cycles and for ecosystem simulation models, many of which do not represent coarse woody debris (CWD) as a separate pool (Cornwell et al. Reference CORNWELL, CORNELISSEN, ALLISON, BAUHUS, EGGLETON, PRESTON, SCARFF, WEEDON, WIRTH and ZANNE2009).

In this study we quantified the stocks of standing and fallen CWD in secondary tropical dry forests in Costa Rica regenerating following pasture or agricultural land use. Secondary forests are an increasingly abundant land-cover type in the tropics. It is important to evaluate the extent to which secondary forests recover species composition and ecological functions that characterize mature forests. We studied forest patches in two conservation areas, Área de Conservación Guanacaste and Parque Nacional Palo Verde. These areas differ in soils, tree species composition and management history (Powers et al. Reference POWERS, BECKNELL, IRVING and PÈREZ-AVILES2009). Our specific goals were to evaluate how CWD stocks and the relative contribution of CWD to total live and dead above-ground biomass (AGM) stocks vary as a function of forest type and stand age. We predicted that total CWD stocks would increase with stand age, but that the relative proportion of CWD to AGM would decrease with stand age, as we expected biomass gain through successional time to exceed flow of necromass into CWD pools through tree mortality in these young stands.

Sector Santa Rosa of the Área de Conservación Guanacaste (10°50′N, 85°40′W) and Parque Nacional Palo Verde (10°21′N 85°21′W), have mean annual precipitations of 1575 and 1720 mm respectively, and 5–6-mo dry seasons with rainfall < 10 cm mo⁻¹ (Powers et al. Reference POWERS, BECKNELL, IRVING and PÈREZ-AVILES2009). Both areas have patches of forest in different successional ages, and the most common previous land use was cattle grazing, which has been occurring on this landscape for centuries (Janzen Reference JANZEN1988, Jimenéz et al. Reference JIMENÉZ, GONZÁLEZ-JIMENÉZ and MATEO-VEGA2001). We quantified CWD in eighteen 20 × 50-m plots located from ~100 m to kilometres apart. As species composition varies with soil properties (Powers et al. Reference POWERS, BECKNELL, IRVING and PÈREZ-AVILES2009), we stratified our sampling by forest cover type. Six plots were located in forests dominated by the evergreen oak species, Quercus oleoides, at Santa Rosa (referred to as OAK), and six were located in typical species-rich tropical dry forest at both Santa Rosa (referred to as SR) and Palo Verde (PV) (oaks are not present at PV). We grouped stands into young (5–18 y) and old (21–60 y) age classes. The older stands resemble mature forests in this region in terms of structure and biomass, but not composition (Powers et al. Reference POWERS, BECKNELL, IRVING and PÈREZ-AVILES2009).

We measured fallen and standing CWD for stems > 10 cm diameter using formulas that incorporate estimates of stem volume and decay-class-specific estimates of wood density. The volume of fallen coarse woody debris was quantified using the line intercept method (following Harmon & Sexton Reference HARMON and SEXTON1996). Two line intercepts of 50 m each were oriented parallel to each other 5 m inside each plot boundary (i.e. 10 m apart), creating a total intercept length of 100 m per 0.1-ha plot. Although our transect length is small, we suggest that this level of sampling is appropriate for the established plot areas of 50 × 20 m. Log diameter and hollow diameter were measured at the point of intersection with the transect line, correcting for any slope in the log. Neither log length nor orientation was measured, which may introduce an unknown amount of error into our estimates of fallen woody debris. However, plot orientation was haphazard and all plots were on level terrain, therefore there is no a priori reason to expect that the distribution of logs would be non-random. Non-circular debris was converted into a round diameter using an equation published by Harmon & Sexton (Reference HARMON and SEXTON1996). Suspended coarse woody debris was not recorded unless some portion of the debris touched the forest floor. The volume of standing CWD was calculated using Newton's formula (Harmon & Sexton Reference HARMON and SEXTON1996) and a fixed-area plot sampling of 20 × 50 m. Basal and breast-height diameters and hollow diameters were measured in the field with large calipers and an increment borer to estimate solid wood radial length. Upper diameters were calculated using a conversion factor of Chambers et al. (Reference CHAMBERS, HIGUCHI, SCHIMEL, FERREIRA and MELACK2000). Hollow volume, which ranged from 0% to 64%, was calculated using Newton's formula and subtracted from both standing and fallen volume estimates.

We estimated woody debris density for samples of each species (or unknown species) and five decay classes, following a system developed in tropical dry forests by Eaton & Lawrence (Reference EATON and LAWRENCE2006). In this method, the first sample of each species and decay state encountered in each plot served as the representative sample for the rest of the plot. We collected samples using a wood borer or machete, taking care to gather representative proportions of bark, sapwood and heartwood in each sample. Woody debris density was determined by the water-displacement method (Chave Reference CHAVE2005). In total we measured 53 samples, which ranged in density from 0.12 to 0.94 g cm⁻³ with an average density of 0.44 g cm⁻³. For comparison, wood density of 87 species of live trees in this forest varied from 0.18 to 0.90 (Powers & Tiffin Reference POWERS and TIFFIN2010), with a community weighted mean value of 0.67 g cm⁻³ (Powers, unpubl. data). In addition, we compared CWD stocks to unpublished data for live biomass of stems >10 cm dbh in these plots that were estimated from allometric equations and species-specific wood density values. In this paper, we abbreviate live plus dead mass as AGM (i.e. necromass plus biomass), and distinguish it from live above-ground biomass (AGBM).

We focused our statistical analyses on three main variables: fallen CWD, standing CWD and the per cent contribution of total CWD to AGM. Two-way analyses of variance (ANOVA) were performed to partition the variation in these aspects of CWD into components due to stand age (young, old) and forest type (OAK, SR, PV). In addition, we used multiple linear regression to investigate the relationship between total CWD stocks to stand age and total live above-ground biomass stocks (excluding one plot that was an obvious outlier). All values of zero were set to 0.01 Mg ha⁻¹ for data analyses, fallen and standing CWD values were log-transformed to normalize residuals for ANOVA, the per cent contribution of CWD to AGM was converted to a proportion and transformed using arcsine square root, but no transformation was necessary for the regression. All statistics were performed in Splus8.0 (Insightful Corp., Palo Alto, CA, USA).

Total coarse woody debris stocks varied from 0 to 32.0 Mg ha⁻¹ (Figures 1, 2). In general, fallen CWD stocks were larger than standing stocks (Figure 1a, b). One young oak plot stood out as having an exceptionally high number of standing dead trees, and this plot clearly affected the means and variation in standing CWD stocks and per cent contribution to AGM when summarized by forest type and stand age (Figure 1b, c). Consistent with our expectations, fallen CWD stocks were higher in older stands (ANOVA, F_1,12 = 12.5, P = 0.004; Figure 1a), but did not vary significantly with forest type (F_2,12 = 0.22, P = 0.80). Standing CWD did not vary significantly as a function of either forest type or stand age, even when the outlier oak plot was excluded (results not shown). On average, CWD accounted for low percentages of total AGM, ranging from < 0.1% to 28%, with a mean value of 3.8%. Although there was a trend for CWD to comprise a larger percentage of AGM in older stands (Figure 1c), this effect was not significant (F_1,12 = 3.36, P = 0.09). When analysed as continuous variables with multiple regression (excluding the one oak stand with high standing CWD), total CWD increased with both stand age and above-ground live biomass stocks (Figure 2); together, these two variables explained ~75% of the variation in total CWD (F_2,14 = 21.5, P < 0.0001; R² = 0.75). Although AGBM also increased with stand age, this term was significant in the multiple-regression model, indicating that it explains a proportion of the variation in CWD beyond stand age.

Figure 1. Coarse woody debris stocks (means of debris ≥ 10 cm diameter, plus SE for N = 3 plots per stand age/forest type combination) in Costa Rican tropical dry forest as a function of stand age and forest type (OAK refers to Quercus oleoides forests, SR is tropical dry forest in Santa Rosa and PV is tropical dry forest in Palo Verde), fallen CWD mass (Mg ha⁻¹) (a), standing CWD (Mg ha⁻¹) (b), and total CWD stocks as a percentage of total above-ground live and dead mass (AGM) (c).

Figure 2. Coarse woody debris stocks (standing plus fallen; Mg ha⁻¹) in secondary tropical dry forest in Costa Rica as a function of above-ground live biomass (a) and stand age (b). In panel (a) open symbols refer to old stands, filled symbols refer to young stands, oak stands are circles, PV stands are triangles and SR stands are squares. In panel (b) open circles are data reported in this study with one outlier plot identified as a filled circle. Other symbols represent CWD ≥ 10 cm diameter in secondary tropical dry forests worldwide (defined as those with a distinct dry season and MAP < 2000 mm), reported in the following studies: diamonds = Bartholomew et al. (Reference BARTHOLOMEW, MEYER and LAUDELOUT1953), triangles = Eaton & Lawrence (Reference EATON and LAWRENCE2006), and square = Singh (Reference SINGH1975).

Of the factors that affect CWD stocks, forest type and disturbance history are likely to have large effects (Woldendorp et al. Reference WOLDENDORP, KEENAN, BARRY and SPENCER2004). For example, Chao et al. (Reference CHAO, PHILLIPS, BAKER, PEACOCK, LOPEZ-GONZALEZ, VÁSQUEZ MARTÍNEZ, MONTEAGUDO and TORRES-LEZAMA2009) detected a nearly two-fold gradient in necromass stocks increasing from north-western to north-eastern mature forests in the Amazon, and attributed this large variation to forest dynamics, with regions that experience higher mortality inputs on a mass basis also having higher CWD. At regional or local spatial scales, disturbances such as hurricanes and previous land use are likely to determine CWD stocks and dynamics (Currie & Nadelhoffer Reference CURRIE and NADELHOFFER2002, Eaton & Lawrence Reference EATON and LAWRENCE2006, Harmon et al. Reference HARMON, WHIGHAM, SEXTON and OLMSTED1995).

Several interesting conclusions emerge from our measurements of coarse woody debris in secondary dry forests in Costa Rica. First, in our plots total CWD ranged from 0–32 Mg ha⁻¹, and the average value was 6.45 Mg ha⁻¹. Overall, we found no strong effect of forest type, although we admit that our power to detect forest type effects is limited by the small number of plots. However, one young oak plot stood out as having unusually large amounts of standing dead trees, reflecting a large mortality event localized to this stand in the past. Oak trees in this region are occasionally defoliated by a caterpillar (Orgyia sp.: Lymantriidae; Jeff Klemens, pers. comm.). Indeed, the trees in several of the oak plots including this stand were attacked in the year prior to our CWD measurements (Powers, pers. obs.). It is possible that the elevated mortality rates in this stand result from insect attack, although we do not have data to test this hypothesis. Nevertheless, this observation suggests a possible role of species-specific biotic interactions in determining the distributions of CWD through effects on tree mortality.

Second, although we cannot rule out whether other unmeasured factors affected CWD stocks in the stands we studied, stand age and live above-ground biomass explain a large amount of variation in CWD among plots, which indicates that plot history is largely responsible for heterogeneity in the studied sites. Plots in our study landscape may have experienced up to centuries of grazing prior to abandonment (Jiménez et al. Reference JIMENÉZ, GONZÁLEZ-JIMENÉZ and MATEO-VEGA2001), during which CWD was subject to the rapid decomposition of a tropical environment. Although CWD can persist in a temperate Douglas fir/hemlock forest 250 y after disturbance (Agee & Huff Reference AGEE and HUFF1987), it has been estimated that the average dead tree in the Amazon will lose 95% of its mass in only 18 y (Chambers et al. Reference CHAMBERS, HIGUCHI, SCHIMEL, FERREIRA and MELACK2000). Further accelerating decomposition, tropical pastures are highly combustible, making CWD vulnerable to loss through accidental and intentional fires (Uhl & Kauffman Reference UHL and KAUFFMAN1990) in addition to biological decomposition.

We speculate that the previous land use of long-term pasture has largely eliminated CWD from pastures in this region. As a result, CWD stocks in the regenerating forests we studied are relatively low when compared with CWD inventories from other secondary tropical dry forests (Figure 2b), many of which have regenerated following shifting cultivation, a form of land use that may leave some woody debris on the ground (Eaton & Lawrence Reference EATON and LAWRENCE2006). With no initial stocks of CWD after prolonged disturbance, it is likely that CWD accumulates only after regenerating forests have grown trees large enough to contribute to a dead-matter pool >10 cm diameter. In other secondary tropical dry forests in Mexico, India and the Democratic Republic of Congo, CWD as a percentage of AGM accounts for 3.2% to 23.1%, compared with an average value of 3.8% at our study sites (Bartholomew et al. Reference BARTHOLOMEW, MEYER and LAUDELOUT1953, Eaton & Lawrence Reference EATON and LAWRENCE2006, Singh Reference SINGH1975). To more fully understand ecosystem carbon budgets in regenerating secondary forests, future studies should focus on how the intensity, duration and type of previous land use affects CWD stocks, as well as accounting for mechanisms that explain the heterogeneity in CWD stocks and tree mortality across the landscape.