Introduction

Tropical wetlands provide a range of ecosystem services, including ground water recharge, conservation of biodiversity, the removal of excess nutrients from surface waters, and the sequestration and storage of atmospheric carbon (e.g., Silvius and Giesen 1996; Jaenicke et al. 2008). However, they are under increasing threat from changes in land use and climate (Chimner and Ewel 2004; Day et al. 2008; Wosten et al. 2008; Keddy et al. 2009). The inaccessible nature of tropical peatlands has meant that few detailed investigations have been carried out on their formation, ecology, and function (but see Phillips et al. 1997; Page et al. 1999). This limits our ability to predict the consequences of increased direct and indirect anthropogenic pressures upon them (Phillips 1998; Limpens et al. 2008; Langner and Siegert 2009).

The humid tropics contain a substantial proportion (~13%) of global peatlands, mostly associated with coastal systems (Maltby and Immirzi 1993; Rieley et al. 1997; Hoekman 2007; Jaenicke et al. 2008). Tropical peatlands include self emergent ombrotrophic systems (Andriesse 1988) that are hydrologically isolated due to doming (Belyea and Baird 2006). In such peatlands, the processes of vegetation succession, peat accumulation, and transport of dissolved material from the wetland interior (Winston 1994; Phillips et al. 1997), lead to the formation of distinct concentric vegetation communities. The peats formed in these systems tend to be deep (>20 m in some places; Anderson 1983; Page et al. 1999) and, because they are formed from the roots and woody debris produced by swamp forests, to be relatively heterogeneous and low density (Shimada et al. 2001; Chimner and Ewel 2004).

In contrast to Southeast Asian wetlands, where most studies of ombrotrophic tropical peatlands have occurred to date, information on Central American wetlands is derived in large part from studies of calcareous minerotrophic wetlands in Belize and Mexico, in which productivity is limited strongly by phosphorus availability (e.g., Johnson and Rejmankova 2005). However, many wetlands in the region are not calcareous and the nutrient status of these systems remains poorly understood. Links between vegetation diversity and peat properties have been found in wetlands in both Southeast Asia (Page et al. 1999) and Central and South America, including Belize, Mexico, and Venezuela (Johnson and Rejmankova 2005; Vegas-Vilarrubia and Laseras 2008; Moreno-Casasola et al. 2009), which suggests that edaphic properties, in conjunction with hydrological processes, can drive diversity in tropical wetlands (Ferreira and Stohlgren 1999; Lopez and Kursar 2007; Rodriguez-Gonzalez et al. 2008). This is likely to have important implications for the carbon cycle, because peat properties (Bachoon and Jones 1992; Bridgham and Richardson 1992; Keller et al. 2006) and vegetation type (Chimner and Ewel 2004) exert strong controls on carbon storage and greenhouse gas fluxes from tropical and subtropical wetlands.

Here we report detailed information on tree diversity and biogeochemical properties, including nutrient concentrations, carbon gas fluxes, and microbial properties, along a strong nutrient gradient in a large peatland on the Caribbean coast of the Republic of Panama. This domed peatland is characterized by clear shifts in vegetation composition with distance from the peatland interior (Phillips et al. 1997) that correspond to marked changes in soil nutrients (Troxler 2007). Our aim was to strengthen the understanding of this ecosystem by quantifying changes in soil nutrients and forest community composition, and assessing their influence on below-ground processes linked to the carbon cycle.

Methods

Site description

The San San Pond Sak peatland is a 164 km2 mosaic of freshwater and marine-influenced wetlands in Bocas del Toro Province, Republic of Panama. Recognized internationally as a site of special scientific interest (Ramsar site # 611), it includes the significant 80 km2 Changuinola peat deposit, an ombrotrophic domed peatland to the southeast of the Rio Changuinola (Fig. 1). The oldest deposits within the Changuinola peat system are estimated to have been formed 4000–5000 years ago and vary from ~2 m at the wetland exterior to >8 m deep in the central areas of the peatland (Phillips et al. 1997).

Fig. 1
figure 1

Map of the study area showing the location of the San San Pond Sak peatland in Panama. The right panel shows the study transect. The locations of the five research sites (Site 1–5) are marked both in the overview and the inset. The map is based upon a panchromatic Landsat 7 Enhanced Thematic Mapper (ETM) + scene

Full size image

Seven distinct phasic plant communities have been identified in the peatland (Phillips et al. 1997). These form roughly concentric rings that, from the periphery, have been designated as: (i) Rhizophora mangle mangrove swamp, (ii) mixed species back-mangrove swamp, (iii) Raphia taedigera palm swamp, (iv) mixed-species forest swamp, (v) Campnosperma panamensis forest swamp, (vi) sawgrass/stunted forest swamp, (vii) Myrica–Cyrilla bog-plain.

The nearby town of Bocas del Toro, Isla Colon, approximately 10 km from the peatland, has a mean annual temperature of 27°C and a mean annual precipitation of 3209 mm (calculated from data for the years 1992–2001; Estadística Panameña 2001). Rainfall is relatively continuous throughout the year with no pronounced dry season, although there are two distinctive periods of lower rainfall (February–April and September–October). Samples in this study were collected during February 2007.

Study transect

We established five study sites with distinct differences in vegetation community along a 2.7 km transect into the central area of the peatland (Phillips et al. 1997; Troxler 2007). All sites were freshwater. The vegetation zones were defined as follows (letters in parentheses indicate the corresponding vegetation zone indentified by Phillips et al. 1997 as defined above):

  • Site 1. Raphia taedigera swamp (zone iii)

  • Site 2. Mixed-species forest swamp (zone iv)

  • Site 3. Campnosperma panamensis forest swamp (zone v)

  • Site 4. Stunted C. panamensis forest with a sawgrass (Cyperus sp.) understorey (zone vi)

  • Site 5. Sawgrass (Cyperus sp.) swamp with occasional stunted C. panamensis (this site was on the edge of the central bog-plain) (zone vi-vii).

At each site we established permanent plots to census vegetation and for soil and gas flux sampling (described below).

Vegetation census

To quantify forest structure and species diversity, we established a permanent 20 m × 50 m (0.1 ha) census plot at each study site, in which all stems ≥10 cm diameter at breast height were measured, mapped, and tagged. The clumping growth form of R. taedigera means that basal area for this species is likely to be a slight overestimation. Voucher samples of all species were collected and their identities verified at the National Herbarium at the University of Panama (http://www.up.ac.pa:8080/Herbario/inicio.php) following nomenclature by Tropics.org (Missouri Botanical Garden, www.tropics.org). The Shannon diversity index was calculated for each plot using EstimateS (Colwell 2005).

Gas flux sampling and analysis

Fluxes of CO2 and CH4 from the soil surface were measured in five 2 m × 2 m plots at each site along the transect in close proximity to the forest census plots (i.e., within the same vegetation type and <300 m away from the associated census plot). Plots for gas sampling were at least 20 m apart and each was divided into two subplots, with gas flux measurements made in each subplot. Values from the pair of subplots were averaged prior to statistical analysis. Gas flux samples were collected during 3 days in February 2007 with stable clear weather conditions and low wind speed. The soil temperature averaged 25°C at the time of sampling and did not vary significantly among sites or over time. Samples were collected between 10 am and 2 pm using the closed chamber technique (Denmead 2008). Headspace areas were 810 cm2 and ca. 10 cm high, with a total volume of 7.8 dm3. To minimize soil and root disturbance the headspace was sealed at the ground surface by a flexible plastic ‘skirt’ weighed down with a metal chain. The air in the head space was mixed prior to collecting a 25 ml air sample through a suba seal using a hypodermic needle and syringe. Samples were collected after 0, 2, 10, and 20 min, and injected into evacuated 12 ml exetainers for transport to the laboratory. Samples were analyzed by gas chromatography (GC 2014, Shimadzu, Milton Keynes, UK) with a 1 ml sampling loop, a molecular sieve column, and detection of CO2 by thermal conductivity and CH4 by flame ionization.

Potential soil respiration

To determine potential soil respiration, blocks of surface peat were collected from each plot in parallel with the peat samples collected for the nutrient analysis (described below). At the time of sampling the water table was either at the soils surface or just below (maximum 10 cm depth). The blocks were trimmed to 5 cm × 5 cm surface area and 10 cm depth and then allowed to drain to field capacity for 24 h at 35°C to simulate the maximum air temperature recorded under the tree canopy in the peatland. The flux of CO2 was then determined by placing the peat samples in air tight container with suba seals inserted in the lids as sampling ports. The volume inside the container excluding the peat was 0.46 dm3. Following mixing of the air in the head space a 25 ml sample was collected at the start of the incubation and then again after 10 min. The collection of two samples over time eliminated problems with under-pressure in the relatively small head space. The assumption of linearity during the 10 min sampling period was supported by the field data. Samples were stored and analyzed for CO2 concentration as described above.

Soil sampling

Surface soil samples were collected for biogeochemical characterization (i.e., total elements, extractable nutrients, microbial nutrients, and hydrolytic enzyme activities) by cutting blocks of peat (10 × 10 × 10 cm) adjacent to the plots used for gas sampling. Soils were transported to the laboratory (~4 h) and stored at 4°C prior to analysis. Bulk density was determined on a minimum of three separate intact cores per site (7.5 cm diameter, 0–10 cm depth). The majority of the roots were removed by hand prior to analysis, although fine roots inevitably remained in some samples. Moisture content was determined by drying subsamples of soil at 105°C for 24 h. Soil pH was determined using a glass electrode in a 1:1 ratio of fresh soil to deionized water.

Extractable nutrients

Dissolved organic carbon and nitrogen fractions were extracted by shaking 40 g of fresh soil in 75 mL of 0.5 M K2SO4 for 1 h. Extracts were centrifuged (8000g, 15 min) and dissolved organic carbon determined after a five-fold dilution by automated combustion and gas chromatography on a TOC-VCSH analyzer (Shimadzu, Columbia, MD). Ammonium and nitrate were determined by automated colorimetry using a Lachat Quickchem 8500 flow injection analyzer (Hach Ltd, Loveland, CO). Ammonium was determined by automated colorimetry at 660 nm following reaction with phenolate, while nitrate was determined at 520 nm following cadmium-catalyzed reduction to nitrite and reaction with sulfanilamide at pH 8.5. Total nitrogen was determined in the extracts by alkaline persulfate oxidation (Cabrera and Beare 1993) overnight at 80°C in sealed glass tubes, with detection as nitrate by automated colorimetry as described above. Dissolved organic nitrogen was calculated as the difference between total nitrogen and the sum of ammonium and nitrate.

Readily-exchangeable phosphate was determined by extraction with anion exchange membranes (AEM) using a method based on that described by Myers et al. (1999). Soil (20 g fresh weight) was shaken for 24 h with 80 ml deionized water and five anion-exchange resin strips (1 × 4 cm; manufactured by BDH Prolabo and distributed by VWR International, Lutterworth, Leicestershire, UK). The strips were rinsed in deionized water and the phosphate recovered by shaking for 1 h in 50 ml of 0.25 M H2SO4. Phosphate was determined in the acid solution at 880 nm following online neutralization and automated molybdate colorimetry using a flow injection analyzer (Lachat Quickchem 8500, Hach Ltd, Loveland, CO).

Total element concentrations

Subsamples of each soil were air-dried on shallow trays at ambient laboratory temperature (10 days at approximately 22°C), ground in a ball mill, and then stored in sealed vials at ambient laboratory temperature and humidity prior to determination of total element concentrations. Total carbon and nitrogen were determined by combustion and gas chromatography using a Flash NC1112 Soil Analyzer (CE Elantech, Lakewood, NJ). Total calcium, phosphorus, and potassium were determined by digestion of 200 mg soil in 2 ml of concentrated HNO3 under pressure in sealed PTFE vials at 180°C (PDS-6 Pressure Digestion System, Loftfields Analytical Solutions, Neu-Eichenberg, Germany) followed by analysis of diluted samples by inductively-coupled plasma optical-emission spectrometry (ICP–OES) using an Optima 2100 (Perkin-Elmer Inc., Shelton, CT). All total element concentrations were corrected for residual moisture in the air-dried sample.

Soil microbial biomass nutrients

Carbon and nitrogen contained within soil microbial biomass was estimated by CHCl3 fumigation and 0.5 M K2SO4 extraction using a correction factor of 2.64 to account for unrecovered biomass carbon (Vance et al. 1987) and 1.85 to account for unrecovered biomass nitrogen (Brookes et al. 1985). Microbial phosphorus was determined by extraction with anion-exchange membranes; soils were extracted as described above for available phosphate, but with the addition of 0.5 ml of hexanol (Myers et al. 1999). Fumigation-released phosphorus was calculated as the difference between phosphate determined in fumigation and unfumigated samples. We did not correct values for unrecovered biomass, since the correction factors used for CHCl3 fumigation (e.g., Brookes et al. 1982) gave microbial phosphorus values that exceeded the total soil phosphorus. We therefore assume that phosphorus released by hexanol fumigation provides a reasonable approximation of the microbial phosphorus concentration in these soils.

Hydrolytic enzyme assays

We measured the activity of seven enzymes involved in the release of carbon, nitrogen, and phosphorus from organic compounds. Assays were conducted using methylumbelliferone-linked fluorogenic substrates (Marx et al. 2001). Enzymes (and substrates) were:

  1. (i)

    Cellobiohydrolase (4-methylumbelliferyl β-d-cellobiopyranoside)

  2. (ii)

    β-Xylanase (4-methylumbelliferyl β-d-xylopyranoside)

  3. (iii)

    α-Glucosidase (4-methylumbelliferyl α-d-glucopyranoside)

  4. (iv)

    β-Glucosidase (4-methylumbelliferyl β-d-glucopyranoside)

  5. (v)

    N-acetyl-β-glucosaminidase (4-methylumbelliferyl N-acetyl-β-d-glucosaminide)

  6. (vi)

    Phosphomonoesterase (4-methylumbelliferyl phosphate)

  7. (vii)

    Phosphodiesterase (bis-(4-methylumbelliferyl) phosphate)

Samples were stored for 1 week at 4°C prior to analysis. This was not expected to significantly influence activity based on previous studies of tropical mineral soils, which indicated that activities were relatively stable for up to 2 weeks of refrigerated storage (Turner and Romero 2010). On the day of the assay, 2 g soil (dry weight basis) was added to 200 ml of 1 mM NaN3 and stirred for 10 min. Aliquots (50 μl) of soil suspension were dispensed into a 96-well microplate containing 100 μl of 200 μM substrate (100 μM final concentration in the assay mixture) and 50 μl of sodium acetate–acetic acid buffer adjusted to pH 4.0 (the approximate mean soil pH). Microplates were incubated at 30°C for either 30 min (β-glucosidase, N-acetyl-β-glucosaminidase, phosphomonoesterase, phosphodiesterase) or 4 h (other enzymes) dependent on activity. Following incubation, 50 μl of 0.5 M NaOH was added to terminate the reaction and fluorescence was determined immediately on a Fluostar Optima spectrofluometer (BMG Labtech, Offenburg, Germany).

Data analysis

Significant differences between sites were tested using analysis of variance (ANOVA) after variables that did not fulfill the normality assumption (as determined by visual inspection of quantile–quantile plots and Shapiro-Wilks test to assess the normality of the residuals) were log, square, or square root transformed. Significant differences between means were analyzed using least significant difference (LSD) at the 5% confidence level. Relationships between dependent and independent variables were tested using backwards elimination stepwise regression, with only significant regressions reported. All statistical analysis was performed in Genstat version 10.1 (Lawes Agricultural Trust, Rothamsted Experimental Station). Due to the high spatial variability in the wetland we set P = 0.1 as the limit for reporting the statistics from the data analysis.

Concentrations of total elements and microbial nutrients are reported both on a mass and an area basis. This was done to account for the variation in bulk density among sites (data shown below) and hence the amounts of nutrients per surface area. This is essential when considering the availability of nutrients to plants and when comparing among soils of markedly different bulk density. The extractable nitrogen and phosphorus were expressed on an area basis, since nutrient availability per surface area is most relevant from a plant perspective.

Results

Vegetation survey

The five study sites contained 16 tree species, including two palms. Tree species diversity (Shannon diversity index) ranged from 1.13 at site 1 (R. taedigera) to 1.90 at site 4 (stunted C. panamensis). Site 1 contained the lowest estimated stem density (106 stems ha−1), but the highest estimated basal area (103 m2 ha−1) (Table 1). The highest estimated stem density was at site 4 (529 stems ha−1) and the lowest basal area was recorded at site 5 (sawgrass; 5 m2 ha−1) (Table 1). Only C. panamensis was present at all five study sites and made the greatest contribution to the basal area at sites 2–5 (40–88%; Table 2). The basal area at site 1 was dominated completely by R. taedigera (>99% of the basal area).

Table 1 Estimated number of stems, species diversity (Shannon index) and basal area for trees ≥10 cm diameter at breast height at each of five sites across the sampling transect in the San San Pond Sak peatland, Panama
Full size table
Table 2 Contribution to the total basal area (%) of trees species for individuals ≥10 cm diameter at breast height at five sites across the sampling transect in the San San Pond Sak peatland, Panama, with the species contributing the greatest proportion of the basal area indicated in bold
Full size table

There was a negative relationship between the total basal area and tree species diversity (R 2 = 87.9, F 3 = 30.0, P < 0.05), with the highest basal area at sites 1 and 3, both of which were dominated by a single species (R. taedigera and C. panamensis, respectively; Table 2).

Peat chemical and physical properties

Water content and bulk density varied in parallel across the peatland, with the greatest water content and lowest bulk density in the wetland interior (Table 3). Although significant, the changes in pH of the surface peat were not substantial, ranging from 3.7 to 4.0 (Table 3).

Table 3 Chemical and physical properties of surface peat (0–10 cm) at five sites across the sampling transect in the San San Pond Sak peatland, Panama
Full size table

Total element concentrations expressed on a mass basis differed significantly among sites for phosphorus (a consistent decline towards the interior) and calcium (highest concentrations at site 3) (Table 3). However, all elements differed significantly among sites when concentrations were expressed on a area basis. The most marked changes were for potassium (four-fold) and phosphorus (seven-fold) between site 1 (R. taedigera swamp) and site 5 (sawgrass) (Table 3). Specifically, total phosphorus declined from 7.31 ± 0.31 to 1.10 ± 0.11 g m−2. The carbon and nitrogen concentrations expressed on a area basis declined towards the wetland interior by approximately two to three-fold (Table 3), although the greatest calcium concentrations was measured at site 3 (C. panamensis).

There were marked changes in C:P and N:P ratios along the study transect (Fig. 2). Values for C:P increased from 518 ± 38 at site 1 (R. taedigera) to 1313 ± 107 at site 5 (sawgrass), while N:P increased from 25 ± 3 to 67 ± 10 at the same sites. However, the C:N ratio did not vary significantly among sites (18.6–21.1; Fig. 2).

Fig. 2
figure 2

Total element ratios for carbon to nitrogen, carbon to phosphorus, and nitrogen to phosphorus in surface peat (0–10 cm) at five sites across the sampling transect in the San San Pond Sak peatland, Panama. The ratios are derived from data presented in Table 2. Values are the mean ± standard error of five replicate samples, with the least significant differences (LSD; 5%) shown

Full size image

The patterns seen in total nitrogen and phosphorus were reflected in significant site differences in extractable (plant-available) nutrients (Fig. 3), with a general decline in the concentrations of inorganic nitrogen and readily-exchangeable phosphate towards the interior of the peatland (F 4 = 3.5, P < 0.05 and F 4 = 28.2, P < 0.001, respectively). The trend in inorganic nitrogen was driven by a significant decrease in nitrate (lower concentrations towards the peatland interior), whereas ammonium concentrations did not vary significantly across the transect (data not shown). There was a noticeable low concentration of readily-exchangeable phosphate at site 2, despite a relatively high total phosphorus concentration (Fig. 3). There was a strong positive correlation between total phosphorus and readily-exchangeable phosphate (R 2 = 0.48, F 1,18 = 23.1, P < 0.001), although total nitrogen was not correlated with inorganic nitrogen (P > 0.1). There was also a significant positive regression describing the co-variation in exchangeable phosphate and nitrate (R 2 = 0.47, F 1,3 = 22.5, P < 0.001).

Fig. 3
figure 3

Variation in a extractable inorganic nitrogen (sum of ammonium and nitrate) and readily-exchangeable phosphate, and b dissolved organic carbon and dissolved organic nitrogen, in surface peat (0–10 cm) at five sites across the sampling transect in the San San Pond Sak peatland, Panama. Values are the mean ± standard error of five replicate samples, with the least significant differences (LSD; 5%) shown

Full size image

Organic carbon and organic nitrogen extractable in K2SO4 varied significantly among sites (F 4 = 3.9, P < 0.05), although the pattern along the transect differed from that of inorganic nitrate and exchangeable phosphate (Fig. 3). The extractable organic carbon and organic nitrogen concentrations were lowest at sites 3 and 4 (dominated by C. panamensis) and highest at sites 2 and 5 (mixed forest and sawgrass, respectively).

Microbial nutrients

Microbial carbon, nitrogen, and phosphorus concentrations varied significantly among sites, being highest in the peatland margin and decreasing towards the peatland interior (Table 4). The differences were approximately two-fold for microbial carbon and five-fold for microbial nitrogen and phosphorus. There were significant differences among sites in microbial C:N and C:P ratios (F 4 = 10.2, P < 0.001 and F 4 = 10.2, P < 0.001, respectively), but not in the microbial N:P ratio, despite the marked changes in the soil N:P ratios (Fig. 4). Microbial C:N ratios were relatively constant between sites 1 and 4, but increased markedly at site 5 (sawgrass).

Table 4 Concentrations of microbial carbon, nitrogen, and phosphorus in surface peat (0–10 cm) at five sites across the sampling transect in the San San Pond Sak peatland, Panama
Full size table
Fig. 4
figure 4

Microbial carbon to nitrogen, carbon to phosphorus, and nitrogen to phosphorus ratios in surface peat (0–10 cm) at five sites across the sampling transect in the San San Pond Sak peatland, Panama. Values are the mean ± standard error of five replicate samples, with the least significant differences (LSD; 5%) shown

Full size image

The microbial biomass contained between 32 and 47% of the total soil phosphorus, with the proportion increasing towards the peatland interior (Table 4). In contrast, microbial nitrogen accounted for only 1.8–3.8% of the total soil nitrogen, with values decreasing towards the peatland interior (Table 4). Thus, the microbial biomass contained a larger proportion of the phosphorus but a smaller proportion of the nitrogen at nutrient-poor sites compared to nutrient-rich sites. Microbial phosphorus was strongly correlated with total soil phosphorus (R 2 = 0.87, F 1,23 = 160.1, P < 0.001), but microbial nitrogen was not correlated with total soil nitrogen (P > 0.1). There was a strong relationship between microbial nitrogen and total soil phosphorus (R 2 = 0.70, F 1,23 = 57.6, P < 0.001).

Hydrolytic enzyme activities

All of the measured enzymes increased significantly from site 1 (R. taedigera) to site 5 (sawgrass) (Fig. 5). The two phosphatase enzymes showed a similar pattern across the transect (site effect, F 4 = 75.1, P < 0.001, and F 4 = 34.2, P < 0.001 for phosphomonoesterase and phosphodiesterase, respectively). For example, phosphomonoesterase increased from 58.7 ± 10.7 nmol MU g−1 soil min−1 at site 1 to 202.3 ± 22.2 nmol MU g−1 soil min−1 at site 5 (Fig. 5). However, there was a noticeable increase in phosphatase activity at site 2 (mixed forest), which corresponded with the extremely low concentration of readily-exchangeable phosphate at this site (Figs. 3, 5). The activity of N-acetyl β-glucosaminidase and the four enzymes involved in the carbon cycle increased markedly at site 5 (site effect F 4 = 15.6, P < 0.001). For example, N-acetyl β-glucosaminidase increased from 3.55 ± 1.86 nmol MU g−1 soil min−1 at site 1 to 34.3 ± 16.3 nmol MU g−1 soil min−1 at site 5 (Fig. 5). For the carbon enzymes, the activity of β-glucosidase (5.88–39.9 nmol MU g−1 soil min−1) showed a strong increase in the interior sites (F 4 = 15.2, P < 0.001) and was much greater than the activities of cellobiohydrolase, β-xylanase, and α-glucosidase (Fig. 5), although the activity of these enzymes were also highest at site 5 (site effect F 4 = 5.6, P < 0.01; F 4 = 21.8, P < 0.001; F 4 = 13.6, P < 0.001 for cellobiohydrolase, β-xylanase, and α-glucosidase, respectively)

Fig. 5
figure 5

Hydrolytic enzyme activities in surface peat (0–10 cm) expressed on the basis of microbial carbon: a phosphatases, b β-glucosidase and N-acetyl β-glucosaminidase, and c β-xylanase, α-glucosidase, and cellobiohydrolase, at five sites across the sampling transect in the San San Pond Sak peatland, Panama. Values are the mean ± standard error of five replicate samples, with the least significant differences (LSD; 5%) shown

Full size image

Carbon gas fluxes

Potential soil respiration differed significantly between sites (F 4 = 5.2, P < 0.01), being greatest at site 1 (R. taedigera) and least at site 4 (stunted C. panamensis) (Fig. 6a). However, this was not reflected directly in the surface CO2 fluxes measured in the field (Fig. 6b), for which there were no significant differences among sites (P > 0.1), and as for potential respiration the greatest CO2 efflux measured was from site 3 (C. panamensis) rather than from site 1 (R. taedigera). There was an order of magnitude greater soil respiration in response to drainage and elevated temperature compared to field measurements (Fig. 6a, b). Methane fluxes measured in the field differed significantly among sites (F 4 = 28.33, P < 0.001), with high CH4 efflux at site 1 (R. taedigera) and site 5 (sawgrass), but low or negative fluxes at the other three sites (Fig. 6c).

Fig. 6
figure 6

Carbon fluxes from surface peat (0–10 cm) at five sites across the sampling transect in the San San Pond Sak peatland, Panama: a potential CO2 flux from incubated samples, b CO2 fluxes in the field, c CH4 fluxes in the field. Values are the mean ± standard error of five replicate samples, with the least significant differences (LSD; 5%) shown

Full size image

Effect of nutrient availability on the tree community and microbial properties

Tree community

There was a strong (R 2 = 0.81) negative relationship between readily-exchangeable phosphate and tree species diversity described by the Shannon diversity index (F 1,3 = 17.8, P < 0.05), with lowest diversity in plots with the highest exchangeable phosphate. Similarly, nitrate accounted for a substantial proportion of the variance in species diversity (R 2 = 0.71; F 1,3 = 10.8, P < 0.05), but did not explain any additional variation when phosphate and nitrate were fitted together. In addition, there was a strong positive relationship (R 2 = 0.87, F 1,3 = 26.81, P < 0.05 (log-transformed basal area)) between readily-exchangeable phosphate and tree basal area. Again, nitrate was correlated to basal area (F 1,3 = 8.84, P = 0.059 (log-trans basal area)), but explained less variation (R 2 = 0.63) than readily-exchangeable phosphate and did not explain additional variation compared to exchangeable phosphate alone when the two variables were fitted together. Tree species diversity and tree basal area did not explain variation in either CO2 or CH4 fluxes (P > 0.1).

Microbial properties

There was a strong positive linear relationship between potential soil respiration and microbial carbon (R 2 = 0.58, F 1,18 = 27.7, P < 0.001), although CO2 fluxes where less strongly related to microbial carbon in the field: 25.2% of the variation in field-measured CO2 flux was attributed to moisture content and microbial carbon (F 1,18 = 5.04, P < 0.05). There was a strong positive linear relationship between the microbial C:P ratio and CH4 production in the field, with the microbial C:P ratio explaining 68% of the variation in the CH4 fluxes (F 1,23 = 51.3, P < 0.001). The microbial nutrient status (i.e. C:P and C:N ratio’s) was a strong predictor of the activity of all enzymes (Table 5). In addition, the availability of dissolved carbon, nitrogen and phosphous also explained some of the variation in the activity of all enzymes (Table 5). Total soil C:P ratio’s absorbed variation in β-Glucosidase and N-acetyl β-glucosaminidase.

Table 5 Results from stepwise regression analysis showing the significance of the overall model describing a range of soil microbial enzyme activities (data shown in Fig. 5), the amount for variance attributed to it, and the significance of the fitted variables
Full size table

Discussion

Tree community

Tree diversity in the San San Pond Sak peatland was low compared to surrounding lowland forests on mineral soils, which typically contain >100 species ha−1 and are characterized by extremely high β-diversity (e.g., Pyke et al. 2001). Low diversity tree communities are typical of Central American wetlands (Keogh et al. 1999; Ellison 2004; Ramberg et al. 2006; Lopez and Kursar 2007). For example, the low diversity in San San Pond Sak is similar to that in forested wetlands in Mexico (Shannon index 1.95; Moreno-Casasola et al. 2009) and lowland forest swamp in Costa Rica (Shannon index 1.99–2.56; Webb and Peralta 1998).

The basal area in the R. taedigera swamp (site 1) was high (103 m2 ha−1, but note that this was a slight overestimation due to the multi-stemmed growth form) compared to previous reports of tropical forest swamps, including 80 m2 ha−1 in French Guiana (Koponen et al. 2004) and 63–58 m2 ha−1 in Kalimantan, Indonesia (Page et al. 1999). In contrast, basal areas at the interior sites (5 and 10 m2 ha−1, although the lowest value was from the sawgrass site lacking a continuous forest canopy) are low in comparison to the lower range reported from other sites, including 22 m2 ha−1 in seasonally flooded evergreen forest in Brazil (Haase 1999) and 29 m2 ha−1 in fresh water swamps in Micronesia (Allen et al. 2005). The differences in basal area among sites with contrasting nutrient availability presented here were generally greater than changes in tree basal reported previously on contrasting mineral soils types (Allen et al. 2005) or under different flooding regimes (Koponen et al. 2004).

Within the peatland, species composition and basal area changed markedly over a relatively short distance and in parallel with considerable changes in nutrient status. This confirms previous reports (Phillips et al. 1997; Troxler 2007) and suggests that the nutrient gradient from the edge to the interior of the peatland exerts a strong control on the tree community. Low nutrients may contribute to the dwarfing of the trees and the sparse canopy, although this requires experimental confirmation. In contrast, the high basal area in the R. taedigera swamp and the relatively high concentrations of available nutrients there suggest that available phosphate and inorganic nitrogen concentrations are sufficient to maintain a relatively high productivity at that site. Strong edaphic controls of wetland vegetation communities were reported previously for the lower Orinoco delta plain of Venezuela, where species composition changed >80% among swamp forests, mangroves, and meadows, in response to a 10 km acidity gradient (Vegas-Vilarrubia and Laseras 2008).

We used a series of measurements to assess nutrient status within the San San Pond Sak peatland, including the concentrations of total and extractable nutrients, microbial biomass nutrients, and hydrolytic enzyme activities. These all indicated a marked decline in nutrient availability per unit area from the peatland margin to the interior sawgrass swamp, which we attribute to the doming of the peatland (i.e., the increased distance from the underlying mineral substrate) and gradual loss of nutrients dissolved in the soil solution as it move through the low density peat material as ground water and surface flow from the central areas to the edges of the wetland (Ingram 1982).

Several nutrients showed clear declines towards the center of the peatland, although most evidence points to phosphorus as being the key limiting nutrient. For example, the relatively high soil N:P ratios and the high proportion of the total phosphorus in microbial biomass across the entire transect indicate relatively low availability of phosphorus compared to nitrogen. The sawgrass swamp in particular exhibits clear evidence of extremely low phosphorus availability, as indicated both by the high soil C:P and N:P ratios, negligible exchangeable phosphate, high microbial C:P ratios, and the marked increase in phosphatase enzyme activity compared to other sites. In this context it is important to note the contrasting patterns in soil nutrients when expressing these on a per mass vs a per area basis: It is clear that some nutrients does not change on a per mass basis along the transects but the amounts of nutrients potentially available to the plants (i.e. per area) change markedly. Readily-exchangeable phosphate was the strongest predictor of basal area and species diversity of the forest. Furthermore, leaf tissue N:P ratio’s >16 suggests that the vegetation in the peatland is P limited (Troxler 2007) as a consequence of the low P availability. In contrast, nitrogen availability appeared to be the most important driver of the forest structure in Indonesian peatlands (Widyatmoko and Burgman 2006). Further work is therefore needed on nutrient turnover through litter inputs and decomposition to fully understand how nutrient availability impacts ecosystem function in this peatland.

Of particular interest in terms of below-ground nutrient status was that a considerable proportion of the total soil phosphorus was contained within the microbial biomass. This occurred at all sites and suggests that the microbial community exerts a strong control on phosphorus availability in the peatland, with intense competition between plants and microbes for available phosphorus. In contrast, a large proportion of the soil nitrogen is stabilized in organic matter and therefore is not available to microbes. High rates of leaching of phosphorus from litter material and subsequent assimilation by the microbial community have been demonstrated previously (Qiu et al. 2002; Battle and Golladay 2007). The microbial biomass declined markedly towards the peatland interior, yet contained a relatively constant and low microbial N:P ratio at all sites. This suggests a constraint on the stoichiometry of these nutrients in microbial biomass in the peatland, as reported recently in a meta-analysis of soil microbial nutrients in mineral soils (Cleveland and Liptzin 2007). A caveat is that the microbial nutrient concentrations almost certainly included a contribution from fine roots that were not removed prior to fumigation, leading to an overestimation of microbial nutrients.

The activities of all hydrolytic enzymes investigated increased towards the centre of the peatland in parallel with the decline in nutrient concentrations, while the microbial biomass concentration per surface area declined. Enzymes involved in the cycling of phosphorus and nitrogen respond strongly to changes in the availability of these nutrients, both in mineral soils (Olander and Vitousek 2000; Allison et al. 2007) and wetlands (e.g., Bragazza et al. 2006; Rejmánková and Macek 2008; Penton and Newman 2008), indicating a shift in resources allocation to increased enzyme activity to maintain metabolic rates. Similar rates of phosphatase activity were also reported from Malaysian peat swamps (Jackson et al. 2009).

There were also marked changes in activity from the edge to the interior of the peatland for enzymes involved in the degradation of long-chain plant-derived polymeric molecules (i.e., cellobiohydrase, N-acetyl-β-glucosaminidase, β-glucosidase). As these enzymes are induced by the presence of substrate, this perhaps indicates a change in the carbon chemistry of the peat, although the increase in N-acetyl-β-glucosaminidase activity may indicate either an increased microbial investment in nitrogen acquisition (Olander and Vitousek 2000; Bragazza et al. 2006) or increased fungal activity (Miller et al. 1998) in the interior of the peatland. β-Xylanase is linked to more slowly decomposing material, but differences in its activity among sites were relatively small. Interestingly, compared to finding in temperate peatlands (Bragazza et al. 2006), the activity of phosphatase, β-glucosidase and N-acetyl-β-glucosaminidase were several orders of magnitude lower in our study system.

The marked changes in nutrient status of below-ground organisms and species composition of the forest canopy are likely to influence the carbon cycle (Raich et al. 2007; Troxler 2007). Indeed, potential soil respiration determined on drained samples in the current study showed a clear link to nutrient status. This was not reflected in the pattern of CO2 flux in the field, however, suggesting that variables other than nutrient status, such as root respiration (Chimner 2004; Metcalfe et al. 2007) or water table fluctuations (Jauhiainen et al. 2005), regulate the net CO2 efflux from the peat surface. The CO2 flux rates were within the range of those measured from lowland forested swamps on deep peat in central Kalimantan, Indonesia (ca. 850 mg CO2 m−2 h−1; Jauhiainen et al. 2005), Sarawak, Malaysia (366–1954 mg CO2 m−2 h−1; Melling et al. 2005a, b), and Micronesia (32–381 mg CO2 m−2 h−1; Chimner 2004). Methane fluxes measured here were several orders of magnitude lower than CO2 fluxes and varied markedly among sites. The highest CH4 fluxes from site 1 (R. taedigera) and site 5 (sawgrass swamp) are comparable to fluxes reported in the literature from Kalimantan(e.g., 0.35–2 mg CH4 m−2 h1; Furukawa et al. 2005; Hadi et al. 2005; Jauhiainen et al. 2005), while the lowest CH4 fluxes at the three intermediate sites in the current study (C. panamensis and mixed forest) are comparable to those reported for Sarawak wetlands (maximum rates of 11.2 μg CO2 m−2 h−1; Melling et al. 2005a, b). Our results suggest that CH4 fluxes have high spatial variability in peatland systems and vary in response to differences in the water table (Inubushi et al. 2005; Hadi et al. 2005; Jauhiainen et al. 2005) and substrate availability (Bachoon and Jones 1992) to a greater extent than CO2 fluxes. Seasonal variability in gas fluxes has been shown in tropical peatlands (e.g. Hadi et al. 2005) and it is likely that both the CH4 and CO2 fluxes in the San San Pond Sak wetland will fluctuate between the drier and wetter periods of the year. Further detailed studies at a range of temporal scales are therefore required to assess the temporal stability of the spatial differences in gas fluxes in the peatland shown here as well as the mechanisms driving them.

As warming and drainage during the laboratory incubation increased CO2 production from surface peat by an order of magnitude compared to field fluxes, this demonstrates that the carbon storage potential of the system may be sensitive to climate warming and/or a lowering of the water table (Brady 1997; Inubushi et al. 2005; Hadi et al. 2005; Jauhiainen et al. 2005). However, this may be regulated to some extent by the apparent strong nutrient limitation of below-ground processes in the peatland interior, as found in subtropical wetlands (Bachoon and Jones 1992; Battle and Golladay 2007; Wright et al. 2009). Further experimental studies are needed to assess whether microbial activity in this ombrotrophic peatland is limited by nutrients or labile carbon or indeed the environmental conditions. In summary, our results revealed marked changes in nutrient status across a relatively short distance in this tropical peatland, which appear to exert a strong control on both above and below ground organisms. In particular, phosphorus limitation appears to be extreme in the peatland interior. Nutrient status therefore likely influences the carbon cycle in two distinct ways––by regulating carbon inputs through its influence on plant community composition and litter quality, and by influencing decomposition rates by below-ground organisms.