Agroforestry Systems

, Volume 86, Issue 2, pp 159–174

The effects of management and plant diversity on carbon storage in coffee agroforestry systems in Costa Rica

Authors

    • Center for Sustainable Development StudiesSchool for Field studies
Article

DOI: 10.1007/s10457-012-9545-1

Cite this article as:
Häger, A. Agroforest Syst (2012) 86: 159. doi:10.1007/s10457-012-9545-1

Abstract

Agroforestry systems can mitigate greenhouse gas (GHG) emissions, conserve biodiversity and generate income. Whereas the provision of ecosystem services by agroforestry is well documented, the functional relationships between species composition, diversity and carbon (C)-storage remain uncertain. This study aimed to analyze the effects of management (conventional vs. organic), woody plant diversity and plant composition on aboveground and belowground C-storage in coffee agroforestry systems. It was expected that organic farms would store more C, and that an increase in plant diversity would enhance C-storage due to complementarity effects. Additionally, it was expected that steep slopes decrease C-storage as a result of topsoil erosion. Woody plants were identified on 1 ha plots within 14 coffee farms (7 conventional and 7 organic). C-stocks in trees, coffee plants and roots were estimated from allometric equations. C-stocks in litter and topsoil (0–25 cm) were estimated by sampling. On average, farms stored 93 ± 29 Mg C ha−1. Soil organic carbon accounted for 69 % of total C. Total C-stocks were 43 % higher on organic farms than on conventional farms (P < 0.05). Conventional and organic farms differed in vegetation structure, but not in species diversity. It was found that the combined effect of farm type, species richness, species composition and slope explained 83 % of the variation in total C-storage across all farms (P < 0.001). Coffee agroforestry in general and organic farms in particular may contribute to GHG mitigation and biodiversity conservation in a synergistic manner which has implications for the effective allocation of resources for conservation and climate change mitigation strategies in the agricultural sector.

Keywords

AgrobiodiversityFunctional diversityGreenhouse gas mitigationOrganic coffeePayment for environmental services

Introduction

Agriculture is directly responsible for 10–12 % of anthropogenic greenhouse gas (GHG) emissions and annual emissions are expected to increase further during the next decades (Smith et al. 2007). Cropland and pastures currently cover between 40 and 50 % of the terrestrial surface area (FAO 2007; Smith et al. 2007) and agriculture is a main driver for deforestation and habitat destruction in the tropics (Wright and Muller-Landau 2006). However, Smith et al. (2008) estimated that agriculture has a mitigation potential of 5.5–6.0 Gt year−1 CO2-equivalent by 2030, with as much as 89 % of this from soil organic carbon (SOC) storage. One strategy for combining GHG mitigation with economic benefits and food production in the tropics is agroforestry (Watson et al. 2000). Researchers have increasingly found evidence that agroforestry provides carbon sequestration and other ecosystem services, such as biodiversity protection and soil conservation (Jose 2009).

Coffee is one of the most important export commodities for many developing countries and represents a source of income for millions of producers, mostly smallholders (Donald 2004). Intensified coffee production in monocultures has expanded during the last decades, resulting in habitat destruction, biodiversity loss and soil degradation (Donald 2004). At the same time, it has also been established that traditional coffee agroforestry represents a viable strategy for sustainable agriculture in the tropics. For example, Dossa et al. (2008) found that shaded coffee has an important potential for GHG mitigation, while, e.g. Perfecto et al. (2003) and Philpott et al. (2008) emphasized the high importance of shade grown coffee for biodiversity conservation. Organic coffee farming has the potential to further enhance the environmental benefits from sustainable agriculture, because it eliminates agrochemicals, decreases fossil fuel dependency, controls erosion and accumulates soil organic matter (SOM) (Pimentel et al. 2005; Martinez-Torres 2008).

Uncertainties remain about the fundamental relationships between management and agroecosystem functioning. Few studies have explored the interactions between species composition, biological diversity, and carbon storage potential in tropical agroforestry systems (Kirby and Potvin 2007; Henry et al. 2009; Saha et al. 2009). These interactions are particularly intriguing, because they may indicate potential synergies between biodiversity conservation and GHG mitigation.

Evidence for the positive effects of species diversity and composition on ecosystem function (e.g. productivity) has been accumulated mostly for relatively simple experimental assemblages and temperate grasslands (e.g. Tilman et al. 1997, 2001; Spehn et al. 2005). Tscharntke et al. (2005) reviewed different mechanisms that can explain relationships between species diversity and the function of agroecosystems. For example, species complementarity may increase the efficiency of resource exploitation in diverse communities and species redundancy stabilizes ecosystem functioning in the face of environmental change (insurance hypothesis). Finally, higher species richness increases the chance that an assemblage contains disproportionately important species which contribute more to ecosystem functions, than the effect of overall diversity (sampling or selection effect). In the case of tropical agroforestry, positive relationships between plant species are well established. Vandermeer (1995) pointed out that intercropping can increase productivity by reducing interspecific competition (competitive production principle), or as a result of positive effects that some plant species have on the growing environment of other species (facilitation).

The objective of this study was to analyze the effect of farm management (conventional vs. organic), woody plant diversity and species composition on the carbon storage potential of 14 coffee agroforestry systems in the Rio Grande watershed in the Central Valley of Costa Rica. The study area has a steep, broken topography. It is known that slope affects erosion processes and soil carbon storage in agricultural systems (e.g. Smith et al. 2007; Martinez-Torres 2008). Consequently, the effect of slope on carbon storage was also taken into consideration.

It was expected that total carbon storage would be higher under organic management, because organic farmers rely on the integration of trees and SOM for the maintenance of soil fertility. It was further predicted that carbon storage would increase with woody plant diversity across farm types, due to complementary relationships and the presence of functional groups that potentially enhance the accumulation of biomass and SOM. Steep slopes were expected to limit carbon storage across farms types, due to increased topsoil erosion.

Methods

Study site and sampling design

Fourteen (seven conventional and seven certified organic) coffee farms were assessed within the Río Grande watershed near Atenas (9.98°N, 84.38°W), in the Central Valley of Costa Rica between November 2008 and April 2011. Farms sizes varied between 1.4 and 24.5 ha, and elevation ranged from 800 to 1,250 m a.s.l. According to the classification by the Unites States Department of Agriculture (USDA), the dominant soil order in this area is Alfisols. Twelve of the farms were located on Haplustalfs and the remaining two farms were located on small patches of Ultisols and Inceptisols (ITCR 2008). Precipitation at the study site ranges from 2,000 to 2,500 mm. All farms were located either in the premontane wet forest, or in the tropical moist forest (transition to premontane) Holdridge Life Zones (ITCR 2008).

As the number of suitable certified organic farms was limited, seven organic farms with a coffee crop area >1 ha were selected first within the Rio Grande Watershed. All farms had been managed organically for at least 7 years at the time of the study. The organic farms were then paired with seven conventional farms which were selected based on highest possible biophysical similarity (micro-watershed, elevation, precipitation, soil type). All 14 farms can be classified as agroforestry systems as shade trees were incorporated.

Data collection

A 1 ha plot (100 m × 100 m) was established in the approximate center of each farm and the respective GPS coordinates were recorded. In some cases the shape of the plot varied or had to be divided due to the topography and shape of the farm. Within the 1 ha plot all woody plant species with a diameter at breast height (DBH, measured 1.3 m above the ground) >5 cm were identified in the field. If a plant could not be readily identified, samples were taken to a botanist at the University of Costa Rica. When a plant could not be identified at all, it was treated as a morpho-species. Across all farms nine woody plants could neither be identified, nor be unequivocally distinguished as a morpho-species; these plants were omitted from the species richness estimates.

At each farm 4 subplots (20 × 25 m) were established within the 1 ha plot, constituting a total area of 0.2 ha per farm. For the estimation of aboveground carbon (AGC), the DBH and height for all woody plants with a DBH > 5 cm were measured within the subplots. Furthermore, the height and diameter (at 15 cm above the ground) of 25 coffee plants were recorded in the center of each subplot, totaling 100 plants per farm. Leaf litter, including branches less than 10 cm in diameter, was collected from an area of 30 cm × 30 cm in the corners of each subplot, totaling 16 samples per farm. Litter samples were dried in the laboratory of the Center for Sustainable Development Studies, in Atenas, at 80 °C for 48 h and weighed.

A total of 16 soil cores were taken, one in the corner of each subplot to a depth of 25 cm. The four samples of each subplot were pooled into a single sample, resulting in four samples per farm, which were sent to the University of Costa Rica for analysis of SOC content (see below). Additionally, one soil core sample of 510 cm3 was taken at the center of each subplot (four samples per farm) and dried at 105 °C for 48 h to determine bulk density.

The slope of the terrain was determined by overlaying the GPS coordinates for the 1 ha plots on geo-referenced maps (scale 1:10,000; IGNCR 2008), with 5 m contour lines in ArcGIS 9.2 (ESRI 2006). The slope varied considerably across farms, however, average slope was similar for conventional (27 ± 8°) and organic farms (24 ± 15°).

Structured interviews were conducted with all farmers between November 2008 and April 2011, as well as in February 2012 to collect information about farm history (previous land use, age of the plantations, year of conversion to organic practices) and farm management (fertilizer applications, herbicide use, soil conservation practices, coffee plant pruning and coffee yield).

Plant diversity and similarity

As the number of woody plants varied greatly (84–641 individuals per 1 ha plot), individual-based rarefaction techniques were applied to obtain comparable values of species richness across all farms. According to Gotelli and Colwell (2011) rarefaction allows for the comparison of communities, based on the number of individuals (n) from the smallest sample:
$$ {\text{E}}\left( {\text{s*|n*}} \right) = {\bar{{\text{s}}}} * , $$
where n* = a random subsample of n individuals (n = the size of the smallest sample) out of a larger sample with N individuals and S different species, s* = the number of species found in the subsample with n* individuals, and \( {\bar{{\text{s}}}} \)* = the mean number of species from repeated subsamples, which estimates E(s*|n*), the expected number of species found in a subsample with n* individuals out of a larger sample with a size of N. For this study rarefied species richness (\( {\bar{{\text{s}}}} \)* ± standard deviation) was calculated based on 1000 iterations of randomly drawn subsamples of n* = 84 for the 13 farms with N > 84 individuals per hectare.
Evenness was calculated using Simpson’s Measure (E1/D), following Krebs (1999):
$${\text{E}}_{{{\text{1}}/{\text{D}}}} = {\text{ }}\left( {1/\left( {\sum {\text{p}}_{{\text{i}}} ^{2} } \right)} \right)/{\text{s}} $$
where pi = the proportion of the species i of the total number of individuals per 1 ha plot and s = the total number of species per plot.
Species composition between all farms was compared by calculating pairwise Bray-Curtis similarity measures (1 − B) for all farms. The Bray Curtis Index takes into account the abundance for each species in the compared samples (Krebs 1999):
$${\text{B }} = {\text{ }}\sum {\text{ }}\left| {{\text{X}}_{{{\text{ij}}}} - {\text{ X}}_{{{\text{ik}}}} } \right|{\text{ }}/{\text{ }}\sum {\text{ }}\left( {{\text{X}}_{{{\text{ij}}}} + {\text{ X}}_{{{\text{ik}}}} } \right) $$
where B = Bray Curtis measure of dissimilarity and Xij, Xik = number of individuals of species i in the two different samples j and k.

Estimation of aboveground and belowground carbon stocks

Aboveground biomass (AGB) of shrubs and shade trees was estimated by applying the allometric equation for tropical moist forest, proposed by Chave et al. (2005):
$$ {\text{AGB}} = 0.0509*\rho *{\text{D}}^{2} *{\text{H}}, $$
where AGB = aboveground biomass (kg), ρ = wood density (g cm−3), D = diameter at breast height (cm) and H = tree height (m).

Specific wood densities were compiled from different sources (Brown 1997; Fearnside 1997; Cordero and Boshier 2003; Flores-Vindas and Obando-Vargas 2003; Penman et al. 2003; Chave et al. 2006; Orwa et al. 2009). If a species was included in different databases, then the lowest published wood density was applied to achieve conservative AGC estimates. A total of 76 different species was found on the subplots of all farms. Of those, 49 could be identified to species, 8 to genus and 6 to family; 13 morpho-species could not be identified at all. Wood densities were found for 40 species and 5 genera in the literature. The average wood density of these was 0.46 ± 0.16 g cm−3. This value was assigned for the remaining 31 species that could either not be identified or were not found in any database.

For the estimation of AGB stored in coffee plants an allometric equation from Segura et al. (2006) was applied:
$$ {\text{Log10}}\left( {\text{AGB}} \right) = - 1.113 + 1.578*{\text{Log10}}\left( {{\text{D}}_{ 1 5} } \right) + 0. 5 8 1*{\text{Log10}}\left( {\text{H}} \right) , $$
where AGB = aboveground biomass (kg), D15 = diameter in 15 cm above the ground (cm) and H = plant height (m).

Following Kirby and Potvin (2007), the factors 0.47 and 0.45 were used for the conversion of AGB into AGC in trees and organic litter, respectively. AGB of coffee plants was converted into AGC by applying a factor of 0.5, according to Medina-Fernandez et al. (2006).

Root biomass was estimated from total AGB in trees and coffee plants, using the allometric equation for tropical forests developed by Cairns et al. (1997):
$$ {\text{RB}} = { \exp }\left[ { - 1.0587 + 0.8863*{\text{Ln}}\left( {\text{AGB/1000}} \right)} \right] , $$
where RB = root biomass (Mg ha−1), AGB = total aboveground biomass from trees and coffee plants (kg ha−1). Carbon content in roots was assumed to be 50 %, according to Penman et al. (2003).

SOC content (%) of soil samples was determined by dry combustion (Elementar vario EL Cube) at the University of Costa Rica. Total estimated SOC (Mg ha−1) was projected by multiplying average SOC % per farm (n = 4) with the sample depth (25 cm), the area (1 ha) and the respective average bulk density per farm (n = 4).

Statistical analysis

The number of woody plants per hectare, species richness and evenness indices were compared between conventional and certified organic farms by applying non-parametric Mann–Whitney tests. The similarity structure of woody plant species composition across farms was explored by feeding the matrix of pairwise Bray-Curtis measures for all farms into a Principal Components Analysis (PCA). The first two principal components (PC), which explained 43.5 % of the variance in the data were plotted to determine if farms segregate into groups with distinctive species assemblages according to management system (conventional vs. organic). To compare the primary use of trees in different assemblages, information was compiled from Cordero and Boshier (2003), Flores-Vindas and Obando-Vargas (2003), and Orwa et al. (2009).

DBH, height and wood density were not normally distributed and could not be transformed. Consequently these variables were compared between conventional and organic farms using univariate, non-parametric Mann–Whitney tests. Mann–Whitney tests were further used to analyze differences in the individual carbon storage components between conventional and organic farms.

The univariate effects of species richness (rarefied number of species, \( {\bar{\text{s}}} \)*), species composition (the first component from the PCA on the Bray-Curtis similarity matrix, explaining 24.2 % of variance) and the slope of the terrain on total C-storage were analyzed by simple linear regression. The independent variables showed significant relationships among each other, as slope was correlated to both species richness (r = −0.63, P = 0.02) and species composition (r = 0.63, P = 0.02). To eliminate collinearity, the variables for species richness, species composition and slope were fed into a PCA. The resulting first PC accounted for 71.6 % of the variance in the data. This PC provided a consolidated variable, allowing to examine the combined effect of farm type (conventional vs. organic), species richness, species composition and slope on total C-storage across farms by using an ANCOVA without violating the assumption of no collinearity among covariates.

All statistical tests were performed in JMP 7.0 (SAS 2007). Rarefied species richness was estimated using EcoSim 7.0 (Gotelli and Entsminger 2001). The Bray Curtis indices were calculated with PAST 1.94b (Hammer et al. 2001).

Results

Farm history and management

Farm sizes ranged between 1.4 and 24.5 ha and on average the organic farms were smaller (3.1 ± 2.1 ha, average ± standard deviation) than the conventional farms (8.4 ± 9.3 ha). The year of conversion into coffee ranged from 1912 to 2005. Four of the farmers indicated that coffee had been grown for “a long time” on their farms, and three farmers did not know what the previous land had been (Table 1). The most common land use before conversion into coffee was pasture. Four of the conventional and one of the organic farms had been covered by cattle pastures in the past. Four farms (two conventional, two organic) had grown annual crops like maize, beans or tobacco, and two of the organic farms had been covered by a mix of land uses (coffee and plantains, coffee and forest), before switching entirely to coffee production (Table 1). The organic farms had converted from conventional management between 1998 and 2002 (Table 1). At the time of the study each farm had been managed organically for at least 7 years.
Table 1

Information on the history of the 14 coffee farms assessed in the Rio Grande watershed between November 2008 and April 2011

Farm

Area (ha)

Previous land use

Coffee production since

Organic management since

CON 1

5.6

No information

Before 1989a

CON 2

2.5

Pasture

2005

CON 3

4.0

Maize, beans

1982

CON 4

1.5

Pasture

1992

CON 5

2.1

Maize, beans

1996

CON 6

24.5

Pasture

1994

CON 7

18.9

Pasture

1992

ORG 1

2.4

Pasture

1985

2002

ORG 2

2.8

Coffee, forest

1952

2002

ORG 3

2.5

Maize, beans, tobacco

1962

1998

ORG 4

2.8

No information

Before 1989a

1999

ORG 5

1.8

No information

Before 1989a

2002

ORG 6

7.7

Coffee, maize, beans

1912

2002

ORG 7

1.4

Plantains, coffee

Before 1989a

2002

CON conventional, ORG certified organic farms

aHistorical land use of the year 1989 obtained from IGNCR (1991)

Average coffee plant density was 6,054 ± 1,563 plants per ha, most farmers pruned their plants regularly at a height of 0.3–0.5 m, and all farmers left the organic material from the pruning on their farm to decompose. The conventional farmers applied between 600 and 3,300 kg ha−1 of fertilizer per year, all of them used synthetic NPK fertilizers, and two farmers complemented synthetic products with organic material (Table 2). The organic farmers applied between 0 and 10,500 kg of organic soil fertilizers ha−1 annually, which included manure, compost, coffee pulp and molasses. Organic matter from green manure (herbs, grasses) was used on four organic farms as the only means of improving soil fertility (Table 2). Six conventional farmers used herbicides to eliminate weeds extensively on the entire plantation and one conventional farmer (CON 6) indicated to use herbicides only very localized, because the relatively dense shade canopy avoided the extensive growth of weeds on his farm. Four conventional and four organic farmers used living fences as erosion barriers. There was a higher density of living fence species (Dracaena fragrans and Yucca guatemalensis) on the organic farms (95 ± 117 plants ha−1) than on the conventional farms (48 ± 68 plants ha−1). Additionally, three of the organic farmers used ditches or holes to trap organic litter on steep slopes.
Table 2

Information on the management of the 14 coffee farms assessed in the Rio Grande watershed

Farm

Coffee plant density (n ha−1)

Coffee pruning height (m)

Pruning frequency (years)

Soil fertilizer

Fertilizer application (kg ha−1 a−1)

Average coffee yield (kg ha−1 a−1)

CON 1

3,880

0.3

4

NPK 18-5-15

1,400

7,300

CON 2

6,440

0.3

4

NPK 18-2-30, chicken manure, lime

2,900

12,300

CON 3

9,170

0.3

8–10

NPK-Mg 18-5-15-6.2

800

4,500

CON 4

6,830

0.5

4

NPK-Mg 18-5-15-6.2, NPK 20-3-20, Urea

1,800

9,100

CON 5

5,140

0.4

4

NPK 18-2-14

3,300

19,200

CON 6

6,770

0.3

5

NPK-Mg 18-5-15-6.2, Magnesamon

600

13,900

CON 7

6,460

0.8

5–6

NPK 15-3-31, coffee pulp

700

9,100

ORG 1

5,330

0.3

Irregular, selective

None (herbs and grasses as green manure)

0

3,800

ORG 2

7,110

0.3

6

Compost

800

1,400

ORG 3

7,310

0.5

10

Chicken and cow manure, coffee pulp

10,500

3,600

ORG 4

7,370

0.3

4–5

Fruit pulp, molasses, zinc

400

2,700

ORG 5

4,910

0.3

Irregular, selective

None (herbs and grasses as green manure)

0

3,800

ORG 6

3,500

0.3

3–4

None (herbs and grasses as green manure)

0

1,700

ORG 7

4,540

0.3

3

None (herbs and grasses as green manure)

0

1,800

Coffee yield refers to the average during the last 5 years. All data were obtained by structured interviews in February 2012 (CON conventional, ORG certified organic farms)

Coffee yield (cherries, approximate average during the last 5 years) on the conventional farms was 10,771 ± 4,833 kg ha−1. The organic farmers produced 2,686 ± 1,059 kg ha−1, roughly 25 % compared to conventional production (Table 2).

Aboveground and belowground carbon stocks

The average total C (AGC + BGC) stored across farm types was 92.6 ± 29.0 Mg C ha−1 (Table 3). The largest single carbon pool was SOC with 63.1 ± 21.4 Mg C ha−1, which represented an average of 69 ± 12 % of total C. On average, BGC (SOC and C stored in roots) was more than 2.5 times higher than AGC (trees, coffee plants and litter). Within the carbon-subplots of all farms a total of 1,337 woody plants were measured, representing 76 different species. Trees and non-coffee shrubs represented the second largest carbon pool after SOC (Table 3). There was a high variability in individual carbon pools across farms, especially for C stored in trees and coffee plants (Table 3).
Table 3

Estimated carbon stored (mean ± standard deviation) in different pools of aboveground (AGC) and belowground carbon (BGC) on 14 coffee farms in the Rio Grande watershed

Carbon pool

All farms (Mg ha−1)

Conventional farms (Mg ha−1)

Organic farms (Mg ha−1)

Trees

17.9 ± 12.1

12.6 ± 12.4 ns

23.2 ± 10.1 ns

Coffee plants

2.8 ± 2.0

3.4 ± 2.2 ns

2.3 ± 1.7 ns

Litter

4.1 ± 1.6

3.3 ± 0.9 ns

4.8 ± 1.8 ns

Total AGC

24.8 ± 12.0

19.3 ± 12.1 ns

30.2 ± 9.7 ns

Soil organic C

63.1 ± 21.4

53.1 ± 11.6 ns

73.0 ± 25.1 ns

Roots

4.8 ± 2.5

3.8 ± 2.6 ns

5.8 ± 2.1 ns

Total BGC

67.8 ± 22.1

56.9 ± 11.7 ns

78.8 ± 25.3 ns

Total C

92.6 ± 29.0

76.1 ± 18.4 a

109.1 ± 29.1 b

Different letters a, b indicate statistical difference between conventional and organic farms, Mann–Whitney test P < 0.05

ns not significant (P > 0.05)

On average the amount of total C stored on organic farms was 43 % higher compared to the conventional farms (Mann–Whitney-test, Z = 2.04, P = 0.04), although no statistically significant differences were found between individual carbon pools (P = 0.06 and P = 0.07 for AGC and BCG, respectively) (Table 3). Vegetation structure differed clearly between management systems. Whereas only 9 % of the trees measured across all conventional farms exceeded 10 m in height, 23 % of the trees measured on organic farms ranged between 10 and 30 m. Consequently, trees were significantly shorter under conventional management (6.2 ± 4.6 m, compared to 7.2 ± 4.4 m, Mann–Whitney test, Z = −3.99, P < 0.0001). Average wood density was significantly lower under conventional management as well, with 0.37 ± 0.12 g cm−3, compared to 0.41 ± 0.13 cm−3 under organic management (Mann–Whitney test, Z = −5.72, P < 0.0001).

Woody plant species diversity and similarity between farm systems

In total 5,091 woody plants (DBH ≥ 5 cm) were counted on the 1 ha plots of all 14 farms. These plants belonged to 108 different species. Of these, a total of 87 species were found on all 7 organic farms and 65 were found across all conventional farms (Table 4).
Table 4

Average number of woody plants (±standard deviation) with a DBH > 5 cm, rarefied number of species per ha and evenness (Simpson’s Measure, E1/D) for seven conventional and seven certified organic coffee farms in the Rio Grande watershed, Costa Rica

Farm type

Individuals (n ha−1)

Total species

(n)

Rarefied

species

(n ha−1)

Evenness

Conventional

249 ± 151 a

65

13.0 ± 5.0 ns

0.27 ± 0.13 ns

Organic

479 ± 131 b

87

13.9 ± 4.4 ns

0.22 ± 0.05 ns

Overall

364 ± 181

108

13.4 ± 4.6

0.24 ± 0.10

Different letters a, b indicate statistical difference between conventional and organic farms, Mann–Whitney test P < 0.05

ns not significant (P > 0.05)

The number of individuals per ha ranged between 84 and 641 and on average was almost twice as high on the organic farms than on the conventional farms (Mann–Whitney test Z = 2.43, P = 0.02) (Table 4). The total number of species per ha ranged between 5 and 37, and the rarefied number of species \( {\bar{\text{s}}} \)* varied between 5 and 21.6 ± 1.0 (average from 1,000 iterations ± standard deviation). The average number of rarefied species was not significantly higher on the organic farms, compared to the conventional farms (Mann–Whitney test Z = 0.64, P = 0.52) (Table 4). Evenness values were overall relatively low (average Simpson’s Measure E1/D = 0.24 ± 0.10), reflecting that the coffee farms were generally dominated by a few woody plant species. There were no significant differences in evenness between management systems (Mann–Whitney tests Z = −0.26, P = 0.78) (Table 4).

The first 5 PC that were extracted from a PCA on a Bray Curtis similarity matrix, presented eigenvalues >1 and together accounted for 77.5 % of the variance in the data across all farms (Table 5). The loadings of the eigenvectors show that the similarity indices for the farms CON 3, CON 5, ORG 6 and ORG 7 had the most important influence on the first PC. The indices for the farms CON 2, ORG 3 and ORG 4 had the highest leverage on the second PC. ORG 4 and CON 2 had the highest average Bray-Curtis similarity indices from pairwise comparisons with all other farms (0.30 and 0.28, respectively). The loadings of the eigenvectors for the farms CON 1, CON 7 and ORG 2 had the highest influence on the third PC (Table 5).
Table 5

Coefficients of the first five principal components, based on pair-wise Bray-Curtis similarity indices for woody plant species composition across 14 coffee farms in the Rio Grande watershed

Farm

PC 1

Eigenvectors

PC 2

Eigenvectors

PC 3

Eigenvectors

PC 4

Eigenvectors

PC 5

Eigenvectors

CON 1

0.173

0.156

0.435

−0.194

0.196

CON 2

0.238

0.494

−0.066

0.041

0.039

CON 3

−0.351

0.324

0.065

0.105

−0.064

CON 4

0.331

0.245

0.104

−0.099

−0.505

CON 5

0.386

0.091

0.330

−0.109

−0.011

CON 6

0.084

−0.140

0.300

−0.455

0.362

CON 7

0.044

−0.177

−0.456

−0.257

0.322

ORG 1

0.250

−0.177

0.151

0.535

0.108

ORG 2

0.100

−0.080

−0.417

−0.267

−0.388

ORG 3

−0.278

0.373

−0.108

0.164

0.295

ORG 4

0.139

0.518

−0.219

0.079

0.140

ORG 5

0.284

−0.244

−0.062

0.510

0.047

ORG 6

−0.365

−0.053

0.273

0.026

−0.432

ORG 7

−0.373

−0.018

0.221

0.061

0.097

Eigenvalue:

3.39

2.70

2.15

1.60

1.01

Cumulative percentage of variance:

24.2

43.5

58.8

70.3

77.5

CON conventional, ORG certified organic farms. Bold numbers indicate eigenvectors that contribute most to the differences between farms

By plotting the scores of the first two PC (43.5 % of variance), farms were separated in three distinctive groups, based on the similarity of woody plant species composition. These groups did not consistently correspond to farm type (conventional vs. organic) (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10457-012-9545-1/MediaObjects/10457_2012_9545_Fig1_HTML.gif
Fig. 1

Principal component analysis, based on pair-wise Bray-Curtis similarity indices for woody plant species composition across 14 coffee farms in the Rio Grande watershed, assessed between November 2008 and April 2011. The first two components explain 24.2 and 19.3 % of the variation in the data, respectively. Filled circles correspond to conventional coffee farms, white circles to certified organic farms. Identical numbers indicate paired farms; letters indicate the micro-watershed where each farm is located

Species composition and C-storage

The farms within the three groups segregated by the PCA, based on species composition also differed in the primary function of dominant tree species and total C-storage. The first group consisted of 3 organic (ORG 3, ORG 6, ORG 7) and 1 conventional farm (CON 3) (Fig. 1). This group was characterized by high total C-storage (126 ± 28 Mg ha−1) and contained the three farms with the highest BGC and total C-storage across all sampled farms (ORG 3, ORG 6 and ORG 7). Woody plant species diversity was high; the average number of rarefied species across these four farms was 15.4 ± 4.8. In terms of species composition the farms were dominated by living fence species, mainly D. fragrans (27 % relative abundance), timber species such as Cedrela odorata, Cordia alliodora and Diphysa americana (altogether 24 % relative abundance) and a broad variety of fruiting trees (22 %), such as Citrus spp., Mangifera indica, Spondias purpurea, Annona muricata and Byrsonima crassifolia (Fig. 2). Nitrogen fixing shade trees (legumes that tolerate frequent pruning such as Erythrina berteroana, Inga vera, Inga densiflora and Albizia adinocephala) comprised 19 % of the relative abundance across this group. The proportion of species that mostly regenerate spontaneously (“forest species” such as Cecropia spp., Tecoma stans and Anacardium excelsum) was 8 %. On these four farms woody plants were distributed relatively evenly across different functional groups (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs10457-012-9545-1/MediaObjects/10457_2012_9545_Fig2_HTML.gif
Fig. 2

Relative abundance of woody plants with different primary functions across three distinctive groups of coffee farms. The groups have been segregated by principal component analysis (Fig. 1), based on species composition (pairwise Bray Curtis similarity of the studied 14 farms)

The second group of farms included 4 conventional farms (CON 1, CON 2, CON 4, CON 5) and 1 organic farm (ORG 4) (Fig. 1). Average storage of total C was low (73 ± 13 Mg ha−1). The three farms with the lowest AGC-storage (CON 1, CON 4 and CON 5) and the two farms with the lowest total C-storage (CON 4 and CON 5) were part of this group. Woody plant species diversity was low as well. The average number of rarefied species was 11.2 ± 3.9. These farms were clearly dominated by nitrogen fixing shade trees (Erythrina poeppigiana, Erythrina fusca and Inga spp., altogether 59 % relative abundance). E. poeppigiana comprised 38 % of all trees across these five farms. Living fence species, such as D. fragrans and Y. guatemalensis made up 26 % of woody plants, followed by fruit trees (Citrus spp., S. purpurea and Psidium guajava, 8 % relative abundance). Five percent of all individuals could be classified primarily as timber trees and only 1 % were forest species (Fig. 2).

The third group of farms consisted of three organic farms (ORG 1, ORG 2, ORG 5) and two conventional farms (CON 6 and CON 7) (Fig. 1). This cluster was intermediate in terms of total C-storage (87 ± 18 Mg ha−1). These farms were also dominated by nitrogen fixing legumes (45 % relative abundance), mainly by Inga spp. (19 %) and E. fusca (16 %). Timber species (Juglans olanchana, C. odorata, D. americana, Tababuia rosea and others) made up 30 % of the individuals, followed by trees with multiple purposes (Bursera simarouba, 14 %), living fence species (Y. guatemalensis and D. fragrans, 11 %) and fruit trees (S. purpurea, Citrus spp., 10 %). Forest species made up only 3 % of the species assemblage across these five farms (Fig. 2).

Contribution of individual tree species and species composition to C-storage on farms

In terms of total AGC-storage across all farms, E. poeppigiana was the most important species, due to its high abundance, high absolute frequency (it occurred on 9 out of 14 farms) and to relatively large tree diameters (Table 6). Together with two other species, the genus Erythrina accounted for almost a quarter of the total individuals counted on the 1 ha plots across all farms, and for a third of the total AGC from woody plants. Tree heights of Erythrina spp. were usually maintained below 5 m by pruning. E. berteroana has a slightly higher wood density, and on average was maintained at a larger stature than the other species of this genus. C. odorata, a native timber tree, was the second most important species in terms of its contribution to total AGC-storage across farms. The genus Inga was represented by three different species which altogether accounted for 15 % of AGC from woody plants across all farms. I. densiflora and I. vera are characterized by relatively high wood densities. Other species accumulated high sums of AGC primarily due to high abundance (D. fragrans, or Citrus aurantium), high specific wood densities (P. guajava, D. americana), or relatively large stature (Guazuma ulmifolia, Ficus jimenezii). D. americana and G. ulmifolia also stored the highest amounts of AGC per tree, besides the one exceptionally large individual of F. jimenezii that was estimated to store more than 2.5 Mg of AGC on one of the conventional farms (Table 6).
Table 6

The 20 most important woody plant species, in terms of their contribution to estimated AGC stored on seven conventional and seven certified organic coffee farms in the Rio Grande watershed, Costa Rica, according to measurements taken between November 2008 and April 2011 (measurements from four 0.05 ha subplots per farm)

Species

Wood density (g cm−3)

DBH (cm)

Height (m)

Total abundance (n)

Frequency (n farms)

Total C (Mg)

Share of AGC woody plants (%)

Average C per tree (Mg)

C per species (Mg ha−1)

Erythrina poeppigiana

0.20

26 ± 14

7.5 ± 7.1

162

9

10.304

20.6

0.064 ± 0.196

5.7 ± 7.9

Cedrela odorata

0.33

17 ± 13

8.7 ± 4.8

99

8

5.678

11.3

0.057 ± 0.175

3.5 ± 4.1

Inga densiflora

0.58

21 ± 12

8.6 ± 4.0

48

5

4.172

8.3

0.087 ± 0.161

4.2 ± 3.8

Erythrina berteroana

0.25

28 ± 18

8.8 ± 5.7

41

2

3.934

7.9

0.096 ± 0.230

9.8 ± 2.9

Inga vera

0.56

21 ± 6

7.8 ± 2.8

58

3

2.984

6

0.052 ± 0.037

5.0 ± 6.4

Ficus jimenezii

0.32

135

18.9

1

1

2.636

5.3

2.636

13.2

Juglands olanchana

0.42

13 ± 4

11.4 ± 3.4

94

2

2.286

4.6

0.024 ± 0.032

5.7 ± 2.5

Erythrina fusca

0.22

18 ± 11

5.0 ± 3.5

107

5

2.270

4.5

0.021 ± 0.050

2.2 ± 3.1

Spondias purpurea

0.31

20 ± 12

7.0 ± 2.6

62

8

2.151

4.3

0.035 ± 0.044

1.3 ± 1.8

Cordia alliodora

0.33

21 ± 14

10.5 ± 4.9

25

6

1.939

3.9

0.078 ± 0.131

1.6 ± 1.3

Diphysa americana

0.60

20 ± 19

7.9 ± 4.0

15

5

1.868

3.7

0.125 ± 0.207

1.9 ± 3.0

Citrus aurantium

0.46

11 ± 6

4.7 ± 1.8

112

6

1.130

2.3

0.010 ± 0.015

0.9 ± 1.1

Guazuma ulmifolia

0.45

35 ± 14

12.3 ± 3.1

5

2

0.899

1.8

0.180 ± 0.108

2.2 ± 0.1

Gliricidia sepium

0.50

17 ± 11

8.8 ± 7.5

14

2

0.864

1.7

0.062 ± 0.096

2.2 ± 2.4

Albizia adinocephala

0.46

14 ± 8

6.8 ± 5.6

21

5

0.745

1.5

0.036 ± 0.070

0.7 ± 1.0

Tecoma stans

0.46

12 ± 10

5.8 ± 3.0

22

5

0.539

1.1

0.025 ± 0.084

0.5 ± 0.9

Dracaena fragrans

0.46

7 ± 2

4.7 ± 1.1

175

6

0.469

0.9

0.003 ± 0.003

0.4 ± 0.4

Inga sp.

0.49

10 ± 6

3.6 ± 1.5

75

3

0.444

0.9

0.006 ± 0.011

0.7 ± 0.5

Psidium guajava

0.63

13 ± 12

5.0 ± 3.8

8

3

0.428

0.9

0.054 ± 0.139

0.7 ± 1.1

Albizia saman

0.42

45

19.9

1

1

0.405

0.8

0.404

2.0

The effects of management, topography, plant diversity and species composition on total carbon stocks

Individually, all of the examined independent variables farm type (Mann–Whitney-test, Z = 2.04, P = 0.04, Table 3), estimated species richness from rarefaction analysis (linear regression R2 = 0.41, P = 0.01), species composition (the first component from the PCA on the Bray-Curtis similarity matrix) (linear regression R2 = 0.61, P = 0.001) and the slope of the terrain (linear regression R2 = 0.46, P = 0.007) had a significant effect on total C-storage (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs10457-012-9545-1/MediaObjects/10457_2012_9545_Fig3_HTML.gif
Fig. 3

Univariate effects of a the rarified number of woody plant species (\( {\bar{\text{s}}} \)* ± standard deviation) b species composition (first component from a PCA on Bray Curtis similarities across all farms) and c) the slope of the terrain on total carbon storage on seven conventional and seven certified organic coffee farms in the Rio Grande watershed, Costa Rica. Filled circles correspond to conventional coffee farms, white circles to certified organic farms

The first PC extracted from a PCA on the three continuous independent variables for species richness, species composition and slope explained 71.6 % of the variance in the data, with an eigenvalue of 2.15. The remaining two components showed eigenvalues <1. The combined effect of farm type and the consolidated PC based on species richness, species composition and slope on total C-storage was highly significant (whole model ANCOVA, adjusted R2 = 0.83, F2,11 = 33.08, P < 0.0001). The individual effects of farm type, and the PC for species richness, species composition and slope both were highly significant within the ANCOVA model (P = 0.004 and P < 0.0001, respectively).

Discussion

C-storage in organic and conventional coffee agroforestry systems

Average total C-storage estimates from this study (93 ± 29 Mg ha−1) fall within or below ranges reported for comparable carbon pools in tropical agroforestry systems. For example, Albrecht and Kandji (2003) indicated a carbon storage potential between 39 and 102 Mg ha−1 for agroforestry in the humid tropics of South America. Soto-Pinto et al. (2010) found that coffee farms in Mexico stored between 122 and 150 Mg C ha−1 in living biomass and soils (0–20 cm). Studies from Costa Rica estimated between 93 and 195 Mg ha−1 of total C-storage for shaded coffee farms (Avila et al. 2001; Mena-Mosquera 2008). These two studies included SOC to a depth of 25 and 30 cm, respectively.

With an average of 24.8 ± 12.0 Mg ha−1 the estimates for AGC fall into the lower range of values reported for shade grown coffee in Costa Rica. Avila et al. (2001), Mena-Mosquera (2008) and Salgado-Vasquez (2010) gave total AGC estimates between 6 and 70 Mg ha−1, whereas De Melo and Abarca-Monge (2008) considered a wider range (9–154 Mg ha−1) for AGC stored only in the trees on coffee farms. One reason for the relatively low AGC values obtained by this study could be that the lowest wood density found in the literature was applied in the allometric equation from Chave et al. (2005) for a given tree species.

The average SOC-storage found in this study (63.1 ± 21.4 Mg ha−1) fell within or below the range reported by similar studies. Avila et al. (2001) and Mena-Mosquera (2008) reported high SOC values for coffee farms in Costa Rica (81–161 Mg ha−1, representing between 63 % and over 90 % of total C). Soto-Pinto et al. (2010) estimated SOC-storage between 83 and 108 Mg C ha−1 for coffee agroforestry in Mexico (0–20 cm soil depth), which represented over 70 % of total C found above 1,000 m a.s.l.

As expected, total carbon storage was significantly higher on organic farms than on conventional farms, although the differences between individual carbon pools were only nearly statistically significant (Table 3). The effect of farm type also was highly significant within the combined ANCOVA model that included a consolidated covariate for species diversity, composition and slope. Results from other studies that compare C-storage in organic and conventional coffee farms appear to vary. For example, Salgado-Vasquez (2010) did not find a significant effect of management type on the AGC of coffee farms in Costa Rica and Nicaragua, whereas Payan et al. (2009) reported that SOC concentrations were higher on organic coffee farms in Costa Rica, probably due to the higher input of organic matter from shade trees.

In the present study there were several differences in management between farm types that could have contributed to the significant differences in total C-storage. Average tree density per hectare, tree height and wood density were significantly higher on the organic farms, which potentially enhanced AGC-storage. Accordingly, Muschler (2000) argued that there are numerous motivations for organic coffee farmers to incorporate trees. Besides erosion control, benefits include N-fixation, enhanced nutrient cycling, as well as natural pest control.

The organic farms were certified according to criteria from USDA National Organic Program (USDA 2000) and the national legislation (MAG 2001), which both prescribe specific soil conservation practices, such as erosion control. It has been shown that slope is strongly correlated with soil erosion in coffee agroforestry systems (e.g. Martinez-Torres 2008). As the present study found that slope had a significant effect on total C-storage across management systems, it appears that preventing erosion can contribute to maintaining SOC-storage. Organic farmers incorporated more erosion barriers, which was reflected by a higher average density of living fence species (D. fragrans, Y. guatemalensis) per hectare. Three organic farmers applied additional erosion control techniques, such as ditches to retain SOM on steep slopes. Soil erosion may be further decreased by the absence of herbicides. All conventional farmers applied herbicides, whereas most of the organic farmers indicated to use the herb layer as green manure (Table 2). The application of herbicides leads to soil exposure (Hartemink 2006), and thus may contribute to soil erosion and SOC loss. According to USDA (2000) and Payan et al. (2009) C is directly added to the soil through inputs of manure and organic amendments under organic management. In the present study three organic farmers and two conventional farmers applied organic fertilizers to their soils, however, with the exception of one farm (ORG 3), soil fertilizer inputs on organic farms were very low (Table 2).

There are some factors that might confound the true effect of farm type on current SOC-storage. Sanderman and Baldock (2010) emphasized that SOC-storage between farming systems cannot be properly compared without knowing their baseline SOC status. In this study the SOC levels at the time of the conversion to organic management were unknown, which leads to uncertainties about the effect of farm management on SOC. Conventional farms on average were converted into coffee more recently, and four out of seven farms had previously been covered by cattle pasture, which could potentially affect baseline SOC levels in comparison to the organic farms (Table 1). Furthermore, Jobbagy and Jackson (2000) pointed out that high clay contents stabilize SOC. There is no detailed information on soil texture for the sampled farms, although the paired sample design aimed to minimize physical differences between management systems. Finally, there is no information on the vertical distribution of SOC on the studied farms. Consequently the SOC measurements to a depth of 25 cm might substantially underestimate this C-pool. SOC represents potentially one of the largest C-pools in tropical agroforestry systems. For future studies it is recommendable to compare time series between management systems, to analyze SOC in different depths across the plant rooting zone and to control for physical variables such as texture.

Relationships between woody plant diversity and C-storage in coffee agroforestry systems

Plant species diversity and species composition did significantly affect total C-storage across all sampled farms. The available results on relationships between plant diversity and carbon storage in tropical agroforestry systems so far are mixed. Kirby and Potvin (2007) did not find a correlation between tree species diversity and C-storage in Panamanian home gardens. Similarly, Henry et al. (2009) did not find a relationship between perennial plant diversity and AGC-storage in home gardens and other agricultural systems in Kenya. On the other hand Saha et al. (2009) did report a positive relationship between plant diversity and SOC-storage in Indian home gardens.

According to Tilman et al. (1997), both species composition and species diversity determine the variation in resource use efficiency and resource requirements within a community. Diverse communities may enhance ecosystem functions through niche differentiation and facilitation (complementarity effects). Additionally, they are more likely to include highly efficient, competitive species that may increase overall productivity of the system (sampling effects) (Tscharntke et al. 2005; Kirby and Potvin 2007).

Fridley (2001) argued that species composition is not random in managed systems and that stochastic immigration and extinction events are limited. In the case of coffee agroforestry systems, we are not likely to observe primarily the effects of plant diversity on C-storage, but the effects of widely intentional plant assemblages. These assemblages exploit and enhance functional traits which are in accordance with the management goals of the farmer. Coffee farmers may take advantage of well-known complementarity mechanisms that potentially increase overall productivity and thus C-storage.

One of the best documented facilitating mechanisms in plant communities is nitrogen fixation by legumes (Vandermeer 1995; Tilman et al. 1997; Fridley 2001). Shade trees like Erythrina spp. and Inga spp. fix nitrogen and offer desirable functional traits, such as fast growth and tolerance to pruning. Leaf litter and mulch from pruning may improve physical soil properties and increase SOC (Muschler 2000; Cordero and Boshier 2003; Youkhana and Idol 2009). Resh et al. (2002) found that tropical forest plantations with N-fixing trees accumulated more SOC than plantations of non N-fixing trees. Species complementarity can further be achieved by combining plants of different stature and shade tolerance into multilayered agroforestry. For example, tall timber trees with relatively small crowns such as C. alliodora and Tabebuia rosea do not seem to have a pronounced negative effect on coffee yield, (Muschler 2000; Cordero and Boshier 2003; Haggar et al. 2011). Niche differentiation by rooting depths may be another important factor (Fridley 2001). Jobbagy and Jackson (2000) found that plant functional types (grass, shrubs, trees) control the vertical distribution of SOC in soils on a global scale, due to different rooting zones and shoot/root allocations. These findings could be relevant for maximizing SOC-storage in agroforestry systems, as well.

According to Tilman et al. (1997) the number of functional traits within an assemblage might have a stronger effect on ecosystem processes than the number of species. In this study, three distinctive clusters of farms were determined, based on species composition (Figs. 1, 2). The farms which were strongly dominated by pruning-tolerant, N-fixing shade trees (mostly Erythrina spp.) seemed to store less carbon than those that showed a more even distribution of trees with different primary functions (living fence, timber, N-fixing shade trees, fruit trees and forest species). It can be assumed that the diverse tree functions utilized by farmers imply a broader variety of structural and physiological traits, which in turn may enhance agroecosystem function and C-storage.

Conclusions

Coffee agroforestry systems provide considerable carbon storage potential and SOC seems to be of particularly high importance. Organic management may further enhance C-storage. Organic systems incorporated more trees per hectare that significantly differed in stature and average wood density from those in conventional farms. Furthermore, organic management relies on soil improvement by incorporating vegetation elements, the application of organic amendments, green manure and erosion barriers. Increasing slopes negatively affected total C-storage across management types, which emphasizes the importance of erosion control as a strategy for increasing C-storage in tropical agriculture.

Woody plant diversity and composition had a significant effect on C-storage across management systems. It is suggested that managed diversity on multifunctional coffee farms enhances complementarity mechanisms and adds a broader variety of functional plant groups, which increases C-storage, either by biomass accumulation or by facilitating processes such as nutrient cycling or increasing SOM.

The existence of potential synergies between agrobiodiversity and carbon storage has important implications for the effective allocation of resources for conservation and climate change mitigation strategies in the agricultural sector. Multifunctional, multilayered agroforestry systems that integrate a variety of native woody plant species and ensure soil conservation should be a priority for incentive mechanisms such as payment for environmental services programs.

Acknowledgments

I would like to thank CoopeAtenas and APROCAFE for their friendly cooperation throughout this study. I am very grateful for the advice on data analysis from Gerardo Avalos. John DeLeo provided technical support for locating and mapping the study plots. The following SFS students collaborated in data collection and laboratory work: Abby Conroy, Rose Mankiewicz, Matt Tannenbaum, Jesi Felton, Jeff Desmarais, Lizzie Fox, Jeremy Weyl, Shannon Zaret, Dan Silvia, Brian Waterman, Alyssa Inman, Carolyn Chu, Chris Wagner, Patrick Boleman, Brian Gaulzetti, Chamae Monroe, Eddie Miller, Kalyn Campbell, Tessa Sanchez, Angela Marshall, Abby Beissinger, Elena Neibaur, Jacqueline Ford, Jeremy Thweatt, Elizabeth Friedrich, Marta Behling, Sheila Jarnes, Thomas Beneke, Anjel Carbajal, Anna Farb, Eunice Ko, Jennifer Burns, Julia VanderWoude, Elizabeth Keeffe, Matthew Gibbs, Rachael Wright, Romina Clemente, Beatriz Luraschi, Alexandra Beskrowni, Sarah Cafran, Daniel Grover, Miriam Gunderson, Samantha John, Erin Johnson, Caitlin Kirk, Pin Pravalprukskul and Radost Stanimirova. I am especially grateful for the invaluable assistance in the field from Mark Bennet. Rafael Acuña helped with plant identification at the University of Costa Rica. I would further like to thank Gerardo Avalos, Kate Henderson and three anonymous reviewers for their comments on the manuscript. Ana Contessa helped with the farmer interviews. I gratefully acknowledge the key financial and logistical support provided by the School for Field Studies (SFS) Center for Sustainable Development Studies in Atenas, Costa Rica.

Copyright information

© Springer Science+Business Media B.V. 2012