CARBON BALANCE IN ORGANIC CONILON COFFEE INTERCROPPED WITH TREE SPECIES AND BANANA

Over the last decade, conilon coff ee (Coff ea canephora) in consortium with wood trees has been established to improve environmental conditions. Little is known about how individual wood trees and banana aff ect soil quality when intercropped with conilon coff ee. The objective of the present study was to evaluate the impacts of intercropping organic conilon coff ee with diff erent wood tree species and banana on C balance. Five cultivation systems including conilon coff ee monoculture and intercropped with Inga edulis, Gliricidia sepium, BRS Japira banana (Musa sp.), or Bactris gasipaes were studied in a randomized complete block design, with four replicates at the south of Espírito Santo State, Brazil. A primary forest fragment adjacent to the experiment was also evaluated for comparison with the consortium. Samples of topsoil (0 to 10 cm) were collected in 2016 to evaluate the total organic C and total N. Soil temperature and moisture at 0 to 5 cm depth and the CO 2 emission were measured monthly over one year. The species planted with the conilon coff ee promoted a 5.52% decrease in the soil temperature and a 17% increase in the soil moisture content. They also promoted an increase in annual C balance, especially intercropped with Gliricidia and Inga (4.70 and 3.56 Mg ha, respectively), with a substantial increase in the soil total organic C and total N in both systems.


INTRODUCTION
Brazil is the world's largest coff ee producer and exporter and the second in consumption (Kist et al., 2017). The agro-industrial coff ee chain is one of the most important sectors of the Brazilian economy due to its signifi cant participation in the export basket and in the generation of employment, which represents in the medium and long term, one of the main strategic products for the country (Lopes et al., 2019). The country's coff ee production is based on two main species: arabica (Coff ea arabica) and conilon/robusta (C. canephora). Conilon is mainly produced and studied in the Espírito Santo state. This state is the second largest coff ee producer in the country, responsible for up to 78% of the conilon production. Conilon coff ee is the main source of income in 80% of rural properties in warm lands of Espírito Santo, with 283 thousand hectares in 63 municipalities, with 78 thousand producing families (Ferrão et al., 2019). According to CEDAGRO (2012), the conilon coff ee plantations areas are highly degraded compared to the others agricultural areas in the state. The degradation of the conilon coff ee areas is related to a series of factors, such as inadequate soil management practices, low vegetation cover, steep slopes, soil exposure to heave rain and solar radiation, which can promote losses in the stock of organic carbon and nutrients, reducing the soils productive capacity.
Tropical soil degradation can be monitored through its C content (Silva and Mendonça, 2007;Ghosh et al., 2012). Organic matter (OM) is considered a key component of any terrestrial ecosystem and is probably the most widely recognized indicator of the soil biological, physical and chemical properties, and is strongly associated with sustainable agroecosystems (Chen et al., 2009). Increasing or maintaining soil organic C is critical to optimize soil functions and crop production, especially in tropical regions (Ghosh et al., 2012).
Carbon sink in the soil is associated to geographic location, environmental factors, and management practices, among others (Jose, 2009). Soil management can infl uence organic matter decomposition, contributing to its degradation and can increase the atmospheric CO 2 concentration (Silva and Mendonça, 2007). Carbon losses may be mainly reduced by minimizing soil disturbance, with increasing soil cover, adopting agroecological or organic systems (Silva and Mendonça, 2007). The agroecological and organic systems have focus on soil health and productivity, associated to environmental preservation, agrobiodiversity, biological cycles and human quality of life (Souza, 2015). It is estimated that 89% of the greenhouse gas mitigation potential of agriculture depends on C sequestration (Smith et al., 2008). In addition, increasing the soil organic C is an important strategy for attenuating climate change eff ect regarding to CO 2 emissions to the atmosphere from agricultural land (Thomazini et al., 2015).
The use of agroforestry systems or intercropped trees with conilon coff ee seems to be an alternative to reduce soil degradation due to higher levels of organic C in the soil of these systems when compared to coff ee monocultures (Paudel et al., 2011). An abundant soil plant cover controls soil CO 2 emissions of the soil, because it regulates the microclimate and soil physical, biological and chemical conditions of the soil (Gomes et al., 2016). The agroforestry systems can also contribute to attenuate the climatic stresses (drought, high temperature, and insolation). These stresses lead to the decline of the conilon coff ee productivity mainly in the state of Espírito Santo (CONAB, 2017).
Agroforestry systems allow the diversifi cation of the allocation of C production into plant biomass and increase the input of organic matter residues to the soil (Ehrenbergerová et al., 2015). If implemented globally, agroforestry systems could remove signifi cant amounts of C (1.1-2.2 Pg C in fi fty years) from the atmosphere (Abbas et al., 2017). Furthermore, when intercropping with leguminous trees, the residues can also contribute to increase N pool (nutrient with high demand for coff ee crop) into the soil due to the biological fi xation (Paulino et al., 2009). Agroforestry systems increase the effi ciency of the system resources use (Eichhorn et al., 2006), providing goods (food, wood products and forage) and services (soil conservation, improvement of water and air quality, biodiversity and scenic beauty) (Chen et al., 2020).
Coff ee cultivation in agroforestry systems is common in Central and South America but is underrepresented in Brazil (Edenhofer et al., 2014), where coff ee plantations are predominantly a monocultures. Arabica coff ee has been studied in agroforestry systems, exhibiting benefi ts to edaphic and microclimatic conditions (Gomes et al., 2016).
In spite of the advances in research on tropical soils cultivated with arabica coff ee in agroforestry systems, little information is available for organic conilon coff ee. Knowledge is also lacking on the role of the individual plants intercropped with coff ee on C balance. The aim of this study was to evaluate the eff ect of conilon coff ee plantations, mixed with tree species and banana, on C balance.

Experimental design
The experiment was set in a randomized complete block design with four replicates. Treatments involved fi ve organic cropping systems of conilon coff ee: one in monoculture and four others mixed with diff erent tree species and banana: Inga edulis, Gliricidia sepium, BRS Japira banana (Musa sp.) and Bactris gasipaes. A primary fragment of native Atlantic Forest adjacent to the experiment was also evaluated as a reference with the conilon coff ee treatments.
Each experimental plot consisted of fi ve rows with 25 coff ee plants each (375 m 2 ) ( Figure 2). Measurements of soil temperature, moisture and CO 2 emission and soil samples were taken between two coff ee plants, 10 cm from the central row in each plot. Four 200-m 2 areas were set in the native forest fragment, which was under selective logging over the last 20 years, to make the measurements and take soil samples.

Planting and culture methods
Coff ee seedlings (variety 'Emcaper 8151 Robusta Tropical') were planted in a 3.0 x 1.0 m spacing, ie, 3,333 plants ha -1 . Other tree species and banana were established within coff ee rows in a 3.0 x 6.0 m spacing (1/6 of the planting pits received one of the intercropped species). Planting densities was 2.777 conilon plants ha -1 intercropped coff ee and 556 plants ha -1 for the intercropped species Irrigation was supplied only in the fi rst 4 months. At the beginning, fertilization consisted of 300 g reactive natural phosphate, 200 g limestone, and 2 kg chicken manure per pit, as recommended by Prezotti  Annually, the horizontal branches of the coff ee tree were cut and eliminated, keeping four vertical branches per plant as recommended by Ferrão et al. (2008), which reached 70% of the production. In 2016, the amount of waste generated by these practices was determined by randomly collecting the biomass of fi ve plants per experimental unit. After determining the fresh mass, a sample of 300 grams of each experimental unit was oven dried with forced air ventilation at 65°C until reaching constant mass. Subsequently, the production of dry mass per area was estimated.

Intercropped trees and banana
The Inga and Gliricidia were intercropped with conilon coff ee in order to provide non-market goods and services (soil cover, microclimate improvement, shading, C and N input to the soil, among others). The Musa and Bactris, besides providing some non-market goods and services (soil cover, shading, C supply, among others), are alternative crops, providing food or goods (banana and heart-of-palm).
The Gliricidia and Inga trees were fi rst pruned for shape in March 2015 to raise the crown above 4 m height. Pruning was conducted in the months of March and August every year. Musa was kept with four pseudo-stems per mat (the parent tree and three 'suckers'). Bactris was shaped without thinning the off shoots. Musa and Bactris stalks were fi rst cut for fruit and heart-of-palm, respectively, in 2014 and every subsequent 2 to 3 months, depending on to have fruit and palm. Residues from the prunings and cuts of the intercropped trees were deposited on the soil fl oor along the planting line.
In 2016, the amount of residues from pruning and cutting of the intercropped plants was determined by randomly collecting the biomass of four plants per experimental unit. After determining the fresh mass of each plant, a sample of 300 grams of each experimental unit was oven dried with forced air ventilation at 65 ° C until reaching constant mass.
2.5. CO 2 emissions, Q10, soil temperature and moisture content The CO 2 emissions, soil temperature and moisture content were measured monthly between March 2016 and February 2017. The CO 2 emissions were measured with a LI-8100 portable analyzer (Li-Cor, USA) coupled to a 10 cm survey chamber (LI-8100-102) placed in PVC collars at a 5 cm depth, 40 minutes before measurements were taken. Three measurements were taken in each experimental plot. For each CO 2 measurement, soil temperature and moisture values were also taken at the 5 cm depth with a soil moisture sensor (Decagon Devices, USA).
To compare sensitivity to soil temperatures across the diff erent systems, the proportional variation in the soil CO 2 emissions was calculated when the soil temperature increased by 10°C (Q10), based on the ratio between the soil temperature at the 5-cm depth and the amount of soil CO 2 emissions. An exponential regression was applied to obtain the relation between the CO 2 emissions and soil temperature (Eq. (1)): ECO2 = α × e (β1 × T) (1), where ECO 2 is the quantity of soil CO 2 emissions (µmol m -2 s -1 ), T is the soil temperature (°C), α is the intercepted CO 2 emissions at T=0, and β1 is the regression coeffi cient, obtained from the natural logarithm of the CO 2 emission amount and the soil temperature at a 5 cm depth. The Q10 values were thus obtained according to Eq. (2) (Gomes et al., 2016): Q10 = e 10 × β1 (2).

Annual net carbon balance
Annual net C balance was estimated as the diff erence between the means of the CO 2 emission and C input during the experimental period. Annual emissions were calculated based on the average measurements. Reported C input refers to the C from the coff ee and intercropped tree plant residues and to the C from the chicken manure and organic compost applied as fertilizer. Residues from spontaneous plants were disregarded. The equivalence between C and CO 2 was based on element molecular weights, in which one mole of CO 2 contains 12,011 g of C.
Coff ee and intercropped plant biomass, along with chicken manure and organic compost used as fertilizers, were collected (300 grams of each material) and dried in a continuous air circulation oven (65 °C) until a constant mass was reached. The total C of these materials was determined by loss on ignition at 430 °C for 24 h in a muffl e furnace (Kiehl, 1985) and showed the following C contents: 470 g kg -1 for Inga and Gliricidia plant residue, 480 g kg -1 for Musa and Bactris residue, 450 g kg -1 for coff ee plant residue and organic compost, and 140 g kg -1 for chicken manure.

Soil organic matter
Samples of top soil (0-10 cm), below the litter, were collected in March, June, September and December 2016. In each experimental plot, three soil samples were collected, forming a composite soil sample to determine total organic carbon (TOC) and total nitrogen (TN).

Data analysis
The annual average of the data were submitted to analysis of variance (ANOVA) through an F test, and the mean values were compared by Skott-Knott (p<0.10), 10% level of signifi cance due to high soil variability. Statistical analyses were performed using the software SISVAR (Ferreira, 2014).

Biomass production and carbon input in the soil
The pruning performed in the Gliricidia contributed to the greatest input of dry mass (12, 62 Mg ha -1 ) and C (5.93 Mg ha -1 ) in the soil, followed by Inga (9.59 and 4.51 Mg ha -1 , respectively). Musa and Bactris at the time of harvest promoted similar inputs of dry mass and C in the soil (mean of 2.40 and 1.16 Mg ha -1 , respectively).
In 2016, the amount of residues from the coff ee tree (branches and leaves) averaged 3.30 Mg ha -1 of dry mass in all cropping systems.

Soil temperature and moisture content, and CO 2 emission
The treatment CM presented the highest soil temperature (mean of 30.07oC). The lowest annual mean of soil moisture also occurred in the CM (20%), which did not diff er from coff ee intercropped with Inga (CI), Musa (CB) and forest (mean of 21.5%). The highest values occurred in coff ee intercropped with Gliricidia (CG) (mean of 25%) and Bactris (CP) (mean of 26%) ( Figure 4B).
The highest annual mean soil CO 2 emissions occurred in the CI, CG and CB, which did not diff er among them (mean of 3.10 μmol m -2 s -1 ), while the others systems had lower annual averages (mean of 2.58 μmol m -2 s -1 )( Figure 4C). The highest Q10 value was found in the forest (2.49), followed by CP (2.0), and the lowest value in the CI (1.16). The other systems presented close values of Q10 (mean of 1.51) ( Figure 4D).

Annual net carbon balance and total soil carbon and nitrogen content
In the estimation of annual net C balance, it was not considered the root systems and coff ee production ( Figure 5). The high C input from pruning by the diff erent intercropped plants contributed to the positive annual C balance (Figure 5a). The diff erences between the input and the losses were 3.56, 4.70, 0.57 and 0.84 Mg ha -1 for CI, CG, CB and CP, respectively. The annual net balance was negative in CM (-0.19 Mg ha -1 ), despite the contributions of 3.30 Mg ha -1 dry mass from coff ee tree residues and 2.50 and 5.67 Mg ha -1 dry mass from the chicken manure and organic compost, respectively. The soil under forest presented the highest TOC and TN values, followed by CG and CI. The lowest TN values were obtained in CM and CB ( Figure 5B, Figure 5C).

Soil temperature, moisture content, and CO 2 emission
The higher inputs of dry mass and C in the soil from the pruning residues of Gliricidia and Inga provide soil cover, microclimate, shading, C and N in the soil, wood, among others. Intercropping plants in the conilon coff ee culture resulted, on average, in a 5.52% decrease in the soil temperature and a 17% increase in the soil moisture content compared to CM, due to the reduction of the amount of solar radiation reaching the soil (Gomes et al., 2016). Under climate change scenarios (Stocker et al., 2013), the microclimate regulation provided by the canopies may become crucial for the production of Figure 4 -Annual average of soil temperature (a), moisture (b), CO 2 emissions (c) and Q10 (d) in conilon coff ee organic plots intercropped with tree species. Mean values followed by the same letter did not diff er by Scott-Knott test (p<0.10). Figura 4 -Média anual da temperatura do solo (a), umidade (b), emissões de CO 2 (c) e Q10 (d) em parcelas café conilon orgânico consorciado com espécies arbóreas. Os valores médios seguidos pela mesma letra não diferiram pelo teste de Scott-Knott (p <0,10).
Figure 5 -Net C balance between inputs (organic fertilization and intercropped species residues) and losses (CO 2 emissions) of carbon (a), and average annual total organic C (b) and N (c) contents of soil in organic conilon coff ee culture intercropped with diff erent tree species. Mean values followed by the same letter did not diff er by Scott-Knott test (p<0.10). Figura 5 -Balanço líquido de C entre aportes (fertilização orgânica e resíduos de espécies consorciadas) e perdas (emissões de CO 2 ) de carbono (a) e o teor médio anual total de C (b) e N (c) do solo na cultura do café conilon orgânico consorciado com diferentes espécies de árvores. Os valores médios seguidos pela mesma letra não diferiram pelo teste de Scott-Knott (p <0,10).
The soil moisture conditions in agroecosystems are also an important issue. The canopy helps to maintain high soil moisture (Gomes et al., 2016;Liu et al., 2013) as the soil cover favours reduction of soil evaporation. Coff ee production is extremely sensitive to soil water availability, as the adequate soil water supply is necessary for the development of coff ee fruits (Cannell, 1983). The water availability in the region concentrates in January and February (rainy season), however, short dry periods may lead malformation of coff ee beans. Thus, intercropping tree plants and banana may provide better conditions to retain (Liu et al., 2013) and improve the coff ee fruit size (Vaast et al., 2006). Bigger size coff ee fruits has positive economic impact once the production achieves high value in the market (Mehta and Chavas, 2008).
Annual CO 2 emissions are great in CI, CG and CB, in response to the greater input of organic matter and C from pruning of Inga and Gliricidia, and from Musa harvesting. Higher soil moisture content in CP might decrease soil porosity and consequently the CO 2 diff usivity (Daly et al., 2008) and CO 2 respiration (Hanson et al., 2000). Higher soil temperatures in CM may have blocked the microorganism respiration (Jiang et al., 2015). The lowest annual CO 2 emission from the forest system, as compared to agroecosystems, may be due to the low soil disturbance (Gomes et al., 2016).
Diff erent species in consortium with coff ee plantation may improve the environmental condition, at least at the soil surface, due to improving biological activity (Katayama et al., 2009). The trees and banana plants used for intercropping with the conilon plantation contributes to the organic matter pool, nutrient cycle and stimulates soil respiration and, therefore, the CO 2 emissions.
The Q10 has been used for the analyses of soil CO 2 emissions and to determine soil temperature sensitivity (Gomes et al., 2016;Thomazini et al., 2015), as the Q10 refl ects the ecosystem climatic variation (Raich and Schlesinger, 1992). The highest Q10 values were obtained in the native forest, and the lowest for the CI system. These data indicate lower sensitivity of CI system to global warming compared to the other systems. May be the low decomposition rate of Inga residue, increasing the amount of residue on the soil and the period of soil cover (Duarte et al., 2013), may turn the soil to be more resilient to climate changes.
Despite the lower input of dry mass and C in the soil by the Musa and Bactris cutting residues, both species are alternative crops and were intercropped with conilon coff ee with the main objective of providing food or merchandise (banana and palm heart).

Annual net carbon balance and total soil carbon and nitrogen content
The annual net C balance calculated represents 13.05, 17.23, 2.09 and 3.08 Mg ha -1 CO 2 equivalents for CI, CG, CB and CP, respectively. The coff ee monoculture presented a negative CO 2 , equivalent (-0.70 Mg ha -1 ), despite the input of dry mass from coff ee tree residues and organic manure. Jia et al. (2012) reported that C inputs in organic crop production systems consisted mainly of organic manure and crop residues, and they, usually, represent 23-73% and 11-16%, respectively, of the increase in total soil C. Carbon losses due to CO 2 emissions were higher in the CI, CG and CB systems. However, soil C inputs through plant from the pruning and cutting residues of intercropped tree species and banana contribute to a positive annual balance, especially in Inga and Gliricidia. In CI and CG, the crop residues contributed to about 50% of the C input.
The higher annual average values for the TOC and TN contents in CG and CI are associated with the larger input of organic residues from tree pruning (Abbas et al., 2017). Furthermore, Inga and Gliricidia are leguminous trees (Fabaceae), that accumulate signifi cant amount of N due to the biological fi xation (Romero-Alvarado et al., 2002), and build up soil organic matter (Silva and Mendonça, 2007). According to Paulino et al. (2009), Gliricidia may show up to 80% of the N due to the biological fi xation. Nitrogen is one of the most important nutrients for several plants for both the vegetative and reproductive growth (Cantarella et al., 2007). Thus, intercropping conilon coff ee with leguminous trees is an alternative to increase soil N content, especially in organic production, decreasing N fertilizers.

CONCLUSION
Organic cultivation of conilon coff ee intercropped with trees and banana, especially with Gliricidia and Inga, may increase C sequestration, increasing the levels of soil total organic C and N, and improve C balance. It results in lower soil temperature and increased soil moisture.
The use of tree species, especially leguminous trees, intercropped with conilon coff ee in organic or conventional production systems should be encouraged to improve the sustainability of the agricultural systems in the domain of Atlantic Forest biome.

ACKOWLEDGMENT
To CAPES (Coordination for Improvement of Higher Education), FAPES (Research Support Foundation of Espírito Santo) and CNPq (National Council for Scientifi c and Technological Development) for their fi nancial support and granting scholarships and research.