Phosphorus pools in Oxisols under shaded and unshaded coffee systemson farmers’ fields in Brazil

Irene M. Cardoso1,2,*, Bert H. Janssen2, Oene Oenema2 and Thomas W. Kuyper2

1Department of Soil Science, Federal University of Viçosa (UFV), CEP: 36571000 Viçosa, Minas Gerais,Brazil; 2Department of Environmental Sciences, Wageningen University, P.O. BOX 8005, 6700 ECWageningen, The Netherlands; *Author for correspondence

Received 30 May 2002; accepted in revised form 27 March 2003

Key words: Atlantic Coastal Rainforest, On-farm experiment, Sequential phosphorus extractions

Abstract

Phosphorus (P) is a primary limiting nutrient for crop production in weathered tropical soils. The deficiency ismainly caused by sorption of phosphate onto Al- and Fe- (hydr)oxides. We hypothesise that the distribution ofsoil P among various pools is influenced by land use. Our objective was to characterise the soil inorganic (Pi)and organic P (Po) pools and to compare the various pools at different depths in agroforestry (shaded) and mo-nocultural (unshaded) coffee cultivation systems. The study was carried out in the Atlantic Coastal Rainforestdomain, Brazil, with Oxisols as the dominant soil type. Soils were collected from four farmers’ coffee (Coffeaarabica L.) fields, two agroforestry and two monocultural systems. Three profiles were sampled per field, atdepths of 2-3, 10-15 and 40-60 cm. A simplified sequential P fractionation was carried out, using resin, 0.5 MNaHCO3, 0.1 M NaOH, 1 M HCl and concentrated HCl as extractants. Sum-P (resin, NaHCO3, NaOH, 1 M HCland concentrated HCl) ranged from 370 to 830 mg kg−1. Concentrated HCl extracted the largest portion (74%),followed by NaOH (22.5%). Labile (sum of resin, NaHCO3 and NaOH) P ranged from 13 to 40% of Sum-P. Themajor part (62%) of the labile fraction was Po. In the agroforestry fields, the amount of Po decreased less withdepth and the percentage of Po in labile pools was higher than in monocultural fields. This suggests that agro-forestry maintains larger fractions of P available to agricultural crops by influencing the dynamics of P throughthe conversion of part of the Pi into Po, thereby reducing P losses to the unavailable pools.

Introduction

Phosphorus is an important nutrient in relatively shortsupply in most natural ecosystems, and the primarylimiting nutrient for crop production in highly weath-ered tropical soils (Linquist et al. 1997). The defi-ciency is mainly caused by strong adsorption ofH2PO4

− to Al- and Fe- (hydr)oxides, which turns largeproportions of total P into a form that is unavailableto plants (Fontes and Weed 1996). Various standard-ised tests are being used to estimate the fraction of Pavailable to plants (Neufeldt et al. 2000). However, ithas been suggested that in highly weathered tropicalsoils, plants have access to P fractions, in particularto moderately labile organic P, which are not ac-

counted for in these standard P tests (Beck andSanchez 1994).

Estimation of the various P pools in the soil, in-cluding organic P, is usually done by P fractionation.One of the fractionation procedures often used is thatof Hedley (Tiessen and Moir 1993). The procedurecomprises the sequential extraction of inorganic P (Pi)and organic P (Po) with increasingly aggressive re-agents. This allows characterisation of the differentpools of Pi and Po that are supposed to be differen-tially available to plants (Beck and Sanchez 1994;Tiessen and Moir 1993; Novais and Smyth 1999).However, any chemical fractionation can at best ap-proximate biological functions and, therefore, differ-ent definitions of the ‘labile’ pools are found in theliterature. The term labile is defined here as the ex-

55Agroforestry Systems 58: 55–64, 2003.© 2003 Kluwer Academic Publishers. Printed in the Netherlands.

tracted pools considered available for the plants, i.e.,the extractable pools that plants can absorb (Novaisand Smyth 1999). Some authors only consider the Piand Po extracted by resin and NaHCO3 as labile, andthe Pi and Po extracted by NaOH and HCl as recal-citrant or unavailable P (Cross and Schlesinger 1995).Other authors argue that forms of P extractable withsodium hydroxide are involved in long-term transfor-mations of P in the soil, and contribute to labile P(Neufeldt et al. 2000; Magid et al. 1996). Fractionsthat are labile in the tropical soils are not necessarilylabile in temperate soils, and what is labile in the longterm may not be labile in the short term (Magid et al.1996). From a perspective of long-term P transforma-tion, which is relevant for perennial crops such as thecoffee (Coffea arabica L.) studied here, NaOH frac-tions can be considered as labile. In this study, wetherefore consider Resin-Pi and NaHCO3-Pi and −Poas labile P, and NaOH-Pi and −Po as moderately la-bile P, reflecting P transformations during one andmore seasons respectively (Tiessen and Moir 1993;Agbenin and Tiessen 1995; Lilienfein et al. 2000;Neufeldt et al. 2000). The 1 M HCl pool representsprimary Ca-associated P, and the concentrated HClpool represents the recalcitrant pool that does notreadily participate in P transformations (Agbenin andTiessen 1995).

Transformations and availability of soil P not onlydepend on soil characteristics, but also on interactionswith plants and on associations of plants with micro-organisms (Magid et al. 1996; Novais and Smyth1999). Microbial activity plays a major role in nutri-ent turnover in general and in P transformation andredistribution into different forms in particular (Stew-art and Tiessen 1987). Micro-organisms can solubi-lise, take up and immobilise P. The importance ofmycorrhiza in increasing nutrient uptake by plants,especially phosphorus, is well documented (Bolan1991). Changes in the quantity and nature of Po arethe net result of the transformations in a highly dy-namic P cycle (Stewart and Tiessen 1987). The addi-tion of organic and inorganic P fertilisers can lead tobuild-up of organic P in soils, as a consequence of in-creasing activities of soil biomass. Immobilisation ofP depends on the availability of Pi and of organicsubstrates to soil biota (Stewart and Tiessen 1987;Magid et al. 1996).

It has been suggested that intercropping with plantspecies that are efficient in P acquisition, due to spe-cial plant characteristics and/or special interactionswith micro-organisms, may increase the availability

and supply of P to the main crop (Ae et al. 1990;Cooper et al. 1996). One form of intercropping isagroforestry, which includes trees as a component ofthe agroecosystem. These trees may increase P sup-ply by retrieving nutrients from lower soil horizons(Young 1997), and increase P availability through ac-celerating P cycling by (1) improving the chemicaland physical quality of soils, and (2) by enhancingmicrobial activity (Cooper et al. 1996).

The hypothesis of the present study is that agro-forestry modifies the distribution of soil P among thevarious pools. Some trees, for instance, take up partof the inorganic P that is not available to agriculturalcrops, and transform it to organic P, which, after be-ing released and mineralised in the soil environmentcan be used by the crops (Szott and Melendez 2001).This process can take place not only close to the sur-face but also deeper in the soil. Thus, the objective ofthis paper is (i) to characterise the inorganic and or-ganic P pools of an Oxisol (USDA classification, Fer-ralsol in FAO classification) in two medium-aged andtwo old coffee fields with either agroforestry (shaded)or monocultural (unshaded) cultivation systems; and(ii) to compare the soil P pools at different depths inthese two coffee cultivation systems. These are farm-ers’ fields.

Materials and methods

Study sites

We studied the effect of different cropping practiceson soil P distribution in the Zona da Mata, Brazil, aspart of ongoing research on agroforestry systems. TheZona da Mata is situated in the Southeast of the stateof Minas Gerais, in the Atlantic Coastal Rainforestdomain. The area has steep slopes (20 to 45%) withaltitudes ranging from 200 to 1800 m, and a tropicalhighland climate (average temperature 18 °C, averageprecipitation 1500 mm, and one dry period of 2-4months). The parental rock is gneiss and the domi-nant soil types are deep and well drained Oxisols,acid and poor in available nutrients. In the past, for-est covered almost the entire region, but nowadaysonly little more than 7% of the original vegetation re-mains. The most important crops are pasture and cof-fee, often inter-cropped with maize and/or beans.Other crops are sugar cane, cassava and rice. Cardosoet al. (2001) give details on pedology, agriculture andsociology.

56

Soils were collected from two on-farm agrofor-estry (shaded) and two on-farm monocultural (un-shaded) systems planted with coffee (Coffea arabica).The trees of the agroforestry systems were plantedbetween coffee rows. The agroforestry fields were 15and 19 years of age, further referred to as AFMedium

(medium-aged) and AFOld (old) respectively. The mo-nocultural fields were 15-20 and 20-24 years of age,from now on called MCMedium (medium-aged) andMCOld (old) respectively. Because this was an on-farm study, an exact match of the age of the systemswas impossible. The distance among coffee trees was3.5 m × 1.5 m (between rows x within rows) or 4 m× 2.0 m. The AFMedium, AFOld and MCMedium hadaround 2000 coffee trees per hectare whereas theMCOld had around 1300 coffee trees per hectare. Be-sides coffee trees, each agroforestry field had around370 trees per hectare. The plant species observedaround the sampled spots in AFMedium and the AFOld

belonged to the genera Araucaria, Sparatosperma,Tabebuia, Schizolobium, Carica, Casuarina, Cecro-pia, Commelina, Tibouchina, Musa, Stenocalix, Pinusand Eriobotrya(Cardoso et al. 2001).

The farmer mainly harvested coffee berrieswhereas agroforestry trees were planted for soil con-servation and nutrient cycling. To allow more lightand air circulation within the agroforestry fields, thelow parts of the tree crowns (except coffee) werepruned in winter, leaving the leaves and branches inthe field. Moreover, some species such as Tabebuiaserratifolia and Schizolobium parahyba lose theirleaves in the winter, which also favours light and aircirculation in the system. Spontaneous vegetation wascut down once a year in the agroforestry fields andthree times a year in the monocultural fields. Otherresearch in the Zona da Mata showed that 67% of thesoil was covered by vegetation in agroforestry fieldsand 16% in monocultural fields (Cardoso et al. 2001).The more common spontaneous vegetation observedaround the sampled spots was from the generaBidens, Vernonia, Commelina, and Sida.

Samples were taken in a landscape that is repre-sentative of the region, with slopes at the collectionsites ranging from 21 to 36% and altitudes from 685to 720 m. All fields were located in the same water-shed and the agroforestry fields were adjacent to eachother. The monocultural fields were one and five kmapart from the agroforestry fields. The sites of AFMe-

dium, AFOld and MCMedium were fallow before theywere implanted with agroforestry or monocultural

coffee, the site of MCOld was pasture. No fertiliser ortillage was used before coffee being implanted.

Since the farmers only use manual tools, no morethan the first five centimetres of the soil were mixedby annual management (minimum tillage). Fertiliserswere applied in the form of compounds (N-P-K), atrates of 130 (AFMedium and AFOld), 300 (MCMedium)and 100 (MCOld) grams of 20-5-20 per coffee tree peryear. Fertiliser was always applied under the coffeecanopy, i.e., not between rows. Limestone was broad-cast at a rate of 0.6 ton per ha per year in AFMedium

and AFOld. In MCOld the rate was 100 g per tree, un-der the canopy, once every four years. In MCMedium,limestone was applied for the last time 4-5 years agobut the rate and method are not known. Chicken ma-nure was applied in AFMedium in the year of the soilsampling at a rate of around 3000 kg per ha. InMCOld, about 3600 kg of chicken manure per hectareper year were applied during 10-14 years, the lasttime it was applied being 10 years before we sampledthe soil. The manure was applied between coffeerows, where the sampling took place. Herbicides wereapplied in the monocultural fields.

Sampling and analysis

Samples were collected in June 1999 (dry season), inthree pits (considered here as replicates), approxi-mately 10 m apart, in between the coffee rows, atseveral depths. For this study we used samples col-lected at 2-3 cm, 10-15 cm and 40-60 cm. The treesin the agroforestry systems were also located betweencoffee rows. Approximately one litre of soil was col-lected, from each depth, air-dried and passed througha two-millimetre sieve. Care was taken that equal vol-umes were collected from each layer.

Using routine procedures, the soils were analysedfor texture (pipette method); pH (in water), organicmatter (Walkley-Black method); calcium and magne-sium (extractor KCl, 1 mol); and phosphorus and po-tassium (extractant Mehlich-1).

The sequential P fractionation followed the flowdiagram of Figure 1, based on Tiessen and Moir(1993), but without filtering and centrifugation at5000 rpm at room temperature, as proposed by Beckand Sanchez (1994) for tropical soils. Tiessen andMoir (1993) ultra-centrifuge the samples at zero de-grees. Each sample was done in duplicate and the re-sults were averaged. For Resin-Pi, two strips (9 ×62 mm) of BDH no. 55164 anion exchange mem-brane were shaken in 30 ml of water in centrifuge

57

bottles. The resin strips were taken from the centri-fuge bottles, soaked in 20 ml of 0.5 M HCl andshaken for 16 h. The bottles with soil and water werecentrifuged; the supernatant was discarded and 30 mlof 0.5 M NaHCO3 was added to the residues, shaken(16 h), centrifuged again and the supernatants werekept in polyethylene bottles in a refrigerator prior tomeasurement. The procedures for NaHCO3 were re-peated with 0.1 M NaOH and with 1 M HCl. Afterextraction with 1 M HCl, 10 ml of concentrated HClwas added to the residue and allowed to react for10 min at 80 °C, whereupon another 5 ml of concen-trated HCl was added and the mixture was hand-shaken every 15 min and left at room temperature for1 h. The samples were centrifuged and the super-natant was decanted in a 50-ml volumetric flask. Thesoil was washed with 10 ml deionised water, centri-fuged, and washed twice with deionised water. Deio-

nised water was added to reach a total volume of50 ml. Inorganic P in the extracts was determinedcolorimetrically with the molybdate-ascorbic acidmethodology (Tiessen and Moir 1993). Total P wasalso determined using the method of molybdate-ascorbic acid (Tiessen and Moir 1993), after digestionof the NaHCO3 and NaOH extracts. For each extract,the amount of organic P was calculated as the differ-ence between total and inorganic P. The soil residueafter extraction with concentrated HCl was not fur-ther digested in H2SO4+H2O2, because this extract isunlikely to contain anything but highly recalcitrant Pi(Tiessen and Moir 1993). We refer to the sum of allfractions as Sum-P instead of total P because we didnot include the last step of Tiessen and Moir (1993),i.e., digestion with H2SO4+H2O2. Phosphorus poolswere both expressed as fractions (part of Sum-P) andas concentrations (mg kg−1 soil).

Figure 1. Flow diagram of the P fractionation method as developed by Tiessen and Moir (1993) and modified by Neufeldt et al. (2000) andLilienfein et al. (2000).

58

Statistical analyses

Statistical analyses were performed using Statistica(StatSoft Inc.). Analysis of variance, followed by anLSD (least significant difference, p < 0.05) post-hoc(equivalent to a contrast) test, was used to test fordifferences among treatments, after sequential Bon-ferroni corrections (Rice 1988). The fractional data(mg: mg, Po: labile P in Table 2 and C: Po in Table 3)were arcsine (sqrt) transformed prior to statisticalanalyses to satisfy normality and homogeneity of var-iance assumptions. In these investigations we com-pared “on-farm” fields that differed in managementpractices (agroforestry and monocultural coffee sys-tems), age (medium-aged and old) at three depths.Using criteria suggested by Hairston (1989), we con-cluded that individual samples could be treated as in-dependent replicates for a proper ANOVA (testingagainst error, not testing against interaction). Hairston(1989) suggested that if the F-ratio of the mean squarebetween fields divided by the mean square within afield is smaller than 1, the use of individual samplesas replicates is justified. We therefore analysed soildata most likely to be unaffected by the various treat-ment factors. For this, we used soil texture (sand), pHand exchangeable cations in the 40-60 cm depth.

Results

The characteristics of the soils and post hoc compari-sons (LSD) are shown in Table 1, and can be inter-preted as follows, according to the general criteria ofSoil Fertility Committee of Minas Gerais State, Bra-zil (CFSEMG 1989). Soils were clayey and acidic;the amount of organic matter was largest in 2-3 cmand intermediate at deeper soil layers, except in theMCOld system, which had low amounts at 40-60 cm.The concentration of exchangeable cations (calcium,magnesium and potassium) in the agroforestry fieldswas highest at the surface and medium to low deeperin the soil. Mehlich 1-P is low at all depths in allfields.

Table 2 shows the average values and the post-hoccomparisons of Sum-P, the labile P, the sum of or-ganic P (Sum-Po) and the mass fractions of P in eachextract analysed per depth (2-3 cm, 10-15 cm and40-60 cm) in the systems (AFMedium, AFOld, MCMe-

dium, MCOld). For the same depths and systems Ta-ble 3 shows the total carbon:sum-Po (C:Po). OrganicP in concentrated HCl was not detected, probably be-cause it was very low, as demonstrated by Lilienfeinet al. (1996) for Brazilian Oxisols. Sum-P rangedfrom 370 to 830 mg kg−1. Concentrated HCl ex-

Table 1. Soil characteristics of two ages group of agroforestry (AF) and monocultural (MC) coffee fields at three depths, in the Zona da Mataof Minas Gerais, Brazil

soil characteristics depths (cm) AFMedium AFOld MCMedium MCOld

Clay (%) 2-3 44 b 43 b 39 b 56 a

20-30 65 a 49 b 45 b 65 a

40-60 66 a 45 b 47 b 64 a

Sand (%) 2-3 46 a 37 ab 41 a 34 bc

20-30 31 bc 34 ac 36 ab 25 c

40-60 30 b 34 ab 35 ab 25 bc

pH (H2O) 2-3 5.0 a 5.0 a 4.8 ab 4.6 b

10-15 5.0 a 5.0 a 4.8 a 4.7 ab

40-60 4.9 ab 4.8 ab 5.1 a 4.6 b

organic matter (g kg−1) 2-3 40.6 a 40.0 a 39.1 a 38.9 a

10-15 26.5 ab 22.6 ab 25.4 ab 21.7 ab

40-60 17.2 b 15.9 b 17.2 b 11.2 b

exchangeable cations: Ca2+ + Mg2+ + K+(mmolc kg−1) 2-3 55.8 a 68.5 a 20.9 b 23.0 b

10-15 20.7 b 25.7 b 12.2 bc 7.8 c

40-60 11.4 bc 7.5 c 7.9 c 5.5 c

Mehlich-1 P (mg kg−1) 2-3 3.1 b 4.5 ab 5.3 a 3.8 ab

10-15 1.2 cd 1.1 cd 2.2 c 1.5 c

40-60 0.7 d 0.6 d 0.3 d 3.0 bc

For a given soil characteristic, significant differences are indicated by different letters (P<0.05, LSD, with sequential Bonferroni corrections),either among fields for a given depth (rows) or among depths for a given field (columns); n = 3.

59

tracted the largest portion of Sum-P (74% on aver-age), followed by NaOH (22.5%). Resin, 1 M HCland NaHCO3 extracts contained on average 0.5, 0.7and 2.3% of Sum-P, respectively. The organic P frac-tions, NaHCO3-Po and NaOH-Po, were on average1.3% and 14% of Sum-P.

There were effects of system and age, and a sig-nificant system x age interaction on Sum-P (p <0.001), which can be ascribed to the high values inMCOld. The labile pool (Resin-Pi + NaHCO3-Pi, −Po+ NaOH-Pi, −Po) showed significant effects of sys-tem, age, depth and a system x depth interaction (p <

Table 2. Concentration of phosphorus (P) in the various extracts and ratio Sum Po: labile of two ages group of agroforestry (AF) and mo-nocultural (MC) coffee fields at three depths, in the Zona da Mata of Minas Gerais, Brazil

P fractions (mg P kg soil−1) Depths AFMedium AFOld MCMedium MCOld

resin-Pi1 2-3 2.6 b 6.0 a 5.7 a 6.4 a

10-15 0.7 b 1.7 b 2.0 b 2.7 b

40-60 0.7 b 1.5 b 0.0 b 2.5 b

NaHCO3-Pi 2-3 5.2 d 4.7 d 12.7 a 8.8 b

10-15 0.8 e 2.9 de 8.2 c 3.1 de

40-60 0.7 e 2.9 de 6.8 c 6.5 bc

NaHCO3-Po2 2-3 8.5 a 8.7 a 11.9 a 9.8 a

10-15 7.0 a 7.4 a 7.3 a 6.9 ab

40-60 4.8 ab 6.8 ab 2.2 b 2.0 b

NaOH-Pi 2-3 47.9 c 29.5 de 55.6 c 90.2 b

10-15 28.2 d 20.8 de 24.2 de 54.3 c

40-60 41.9 cd 24.3 e 13.7 e 116.8 a

NaOH-Po 2-3 95.0 ab 67.7 bc 68.8 b 129.5 a

10-15 88.5 ab 48.7 c 51.4 bcd 119.5 a

40-60 117.2 a 46.3 cd 24.9 d 75.3 c

HCl-Pi 1M 2-3 2.4 bc 6.0 b 11.8 a 1.6 c

10-15 1.8 cd 2.0 d 6.1 b 0.3 cd

40-60 1.7 bc 4.3 bcd 5.4 b 1.0 c

HClc-Pi3 2-3 283 c 440 bc 300 c 535 ab

10-15 254 c 414 bc 334 c 573 ab

40-60 246 c 446 b 318 bc 625 a

sum-P4 2-3 444 b 562 b 467 b 781 a

10-15 381 b 498 b 433 b 760 a

40-60 413 b 532 b 371 b 829 a

sum-Po5 2-3 104 ab 76 bc 87 b 139 a

10-15 96 ab 56 c 59 bc 126 a

40-60 122 a 53 cd 27 d 77 bc

labile6 2-3 159 cd 117 e 155 de 245 a

10-15 125 df 81 ef 93 f 187 b

40-60 165 c 82 eg 48 g 203 b

sum-Po: labile 2-3 0.65 ab 0.65 ab 0.52 b 0.57 b

10-15 0.76 a 0.69 ab 0.63 b 0.68 ab

40-60 0.74 a 0.66 ab 0.57 b 0.37 c

1. Pi = inorganic P;2. Po = organic P;3. HClc. = concentrated HCl;4. sum-P = labile P + 1M HCl-Pi + concentrated HCl-Pi;5. sum-Po = NaHCO3-Po + NaOH-Po;6. labile P = Resin-Pi + NHCO3-Pi and −Po + NaOH-Pi and −Po.For a given P fraction, significant differences are indicated by different letters (P<0.05, LSD, with sequential Bonferroni corrections), eitheramong fields for a given depth (rows) or among depths for a given field (columns); n = 3.

60

0.001). Labile P was higher in AFMedium and MCOld

than in the other fields (Table 2). In general, labile Pdecreased with increasing depth for the monoculturalbut not for the agroforestry fields (Table 2). Labile Pwas on average 26% of Sum-P (range 13-40%). Or-ganic P represented on average 62% of the labile P(range 38-76%, Table 2).

There were significant effects of depth and signif-icant interactions of system and depth on Sum-Po,NaHCO3-Po and NaOH-Po (p < 0.01). The concen-tration of Po did not change in depth in the agrofor-estry fields, but was higher at 2-3 and 10-15 cm thanat 40-60 cm of the monocultural fields (Table 2). Sys-tem, age and depth had significant effects (p < 0.001)on the ratio organic P: labile pools (Table 3).

There were significant effects in system, age, anddepth on Resin P with the amount of P closer to thesurface being higher than the amount at the otherdepths (Table 2). Within each depth (Table 2) theamount of P in the resin fractions did not differ amongthe fields except for 2-3 cm where AFMedium waslower than the others.

There were effects of system (agroforestry or mo-nocultural) and depth on NaHCO3-Pi and NaOH-Pi(p < 0.01). The monocultural fields contained higherconcentrations of NaHCO3-Pi than the agroforestryfields and higher concentrations of NaHCO3-Pi werefound at 2-3 cm than at other depths (Table 2). Theeffect of system on NaOH-Pi was mainly due to theMCOld field, which had a higher amount than theother fields (Table 2).

The amount of Phosphorus extracted by one molarHCl was affected by system, age and depth (p < 0.05),with higher amounts of Pi at 2-3 cm than at otherdepths (Table 2). Higher amounts of 1M HCl-Pi werefound at every depth in the MCMedium field than inother fields (Table 2). The values of concentrated HClare higher in MCOld than in the other fields (Table 2),

which caused an effect of system and age (p < 0.01).No effects of depth were found.

C:Po was on average 202, ranging from 82 to 367(Table 3). The post hoc comparisons (Table 3) did notshow any difference among systems, age or depth ex-cept for the lowest depth of AFMedium and MCMedium,which caused significant effects of age and depth onC:Po (p < 0.05). Po correlates significantly with car-bon contents for all fields (AFOld, r = 0.77, p < 0.01;MCMedium, r = 0.83, p < 0.005; MCOld, r = 0.74, p <0.025) except for AFMedium (r = 0.50). When lookingat the separate pools as Po-NaHCO3 and Po-NaOH,the AFMedium showed a significant correlation be-tween carbon with Po-NaHCO3 (r = 0.89, p < 0.005)but not correlation with Po-NaOH (r = 0.56).

Discussion

In general, the differences in soil characteristics (Ta-ble 1) are considered slight according to the criteriaused to interpret soil fertility in Minas Gerais (CF-SEMG 1989). The agroforestry systems had largeramounts of exchangeable cations than the monocul-tural systems, especially closer to the surface. Thishas to be attributed to the fact that limestone wasbroadcast in agroforestry systems and was appliedaround coffee trees in the monocultural fields. Theapplication rates of limestone were higher in the agro-forestry than in the monocultural systems, and erosionwas probably less, both contributing to a higher im-pact of the limestone. Despite the use of chemicalfertiliser, the Mehlich 1-P is low for every system inevery depth, which is a consequence of the high fixa-tion capacity of the Oxisols (Novais and Smyth1999).

All systems contained similar amounts of soil or-ganic matter if the same depths are compared. Theamount of organic P in the soil is roughly correlatedwith the amount of carbon, but inositol is known toobscure this correlation (Dalal 1977). Since NaOHprobably extracts compounds related to ortho-P mo-noesters, mainly inositol (Robinson et al. 1998), wedo not expect a clear correlation between C and Poextracted with NaOH. Nevertheless, we found a sig-nificant correlation in all fields except for the AFMe-

dium.Sum-P values were similar in all depths within

fields (Table 2) and may be due to similarity of soil-Padsorption properties with depth (characteristic ofOxisols). Sum-P was highest in MCOld

Table 3. Carbon(C): Sum organic Po (C:Po) of two ages group ofagroforestry (AF) and monocultural (MC) coffee fields at threedepths, in the Zona da Mata of Minas Gerais, Brazil

Depths (cm) AFMedium AFOld MCMedium MCOld

C: Po 2-3 228a 304ab 282a 162ab

10-15 161ab 234ab 251a 100b

40-60 82b 173b 367a 84b

Significant differences are indicated by different letters (P<0.05,LSD, with sequential Bonferroni corrections), either among fieldsfor a given depth (rows) or among depths for a given field (col-umns); n = 3.

61

(760-830 mg kg−1), probably as a result of large ap-plications of chicken manure in the past. Because allsystems were subject to minimum tillage, the evendistribution is likely to be the result of soil mixing bysoil fauna. Although Pi is immobile in Oxisols, theaddition of manure to soil triggers the movement ofP, including organic and inorganic forms (Novais andSmyth 1999) and mineralisation of transported or-ganic P occurs (James et al. 1996). High amounts(mg) of P in MCOld are mainly found in the recalci-trant fraction (concentrated HCl) followed byNaOH-Pi fractions, which can be due to inorganic Pin the chicken manure or due to mineralisation of thePo also present. When compared to the P fraction-ation using the method of Tiessen and Moir (1993)for Brazilian Oxisols, the values of the Sum-P in theyounger fields (AFMedium and MCMedium,370-470) arerather similar to the 255 to 471 mg kg−1 (excludingthe residue-P) found by Lilienfein et al. (2000), andthe 225 to 386 mg kg−1 of Neufeldt et al. (2000), buthigher than the amounts found by Tiessen and Moir(1993) in differently-textured sand Oxisols (110 to130 mg kg−1). The difference is due to difference insoil texture. Based on soil survey reports concerningthe first 20 cm of soil and assuming soil density as1.3 kg dm3, Resende (1997) calculated the average oftotal P (sulphuric acid method) in Brazilian soils tobe 153 mg kg−1, which equals 1800 kg of P2O5

(20 cm depth) per ha. This amount of P is not con-sidered low, but most of it is unavailable for crops,although part of it is used by other plants than agri-cultural crops, for instance Inga sp, Erythrina, Caja-nus etc. (Ae et al. 1990; Rheinheimer et al. 1999).Brazilian soils with these amounts of P can thereforebe considered as a P reserve (Resende 1997), cer-tainly if one takes into account that these soils(mainly Oxisols), are often more than 20 meters deepand can potentially be explored by deep tree roots, asis the case in agroforestry systems.

No effect of system was found on either of the or-ganic fractions (NaHCO3-Po or NaOH-Po) or theirsum (Sum-Po). Although higher inputs of organic ma-terial (leaves, roots etc.) are to be expected in theagroforestry fields, and this could increase the amountof organic P, the cycling rate is also expected to behigh which could increase the turnover. Therefore, alarge difference in P pools among fields was not ex-pected. However, the significant effect of depth andthe significant interaction between system and depthdo suggest that trees had an impact on organic P dis-tributions. The amount of organic P decreases less

with depth in the agroforestry systems than in monoc-ultural systems (Table 2). This can be an indicationof the impact of agroforestry systems in P transfor-mation deeper in the soil compared to the monocul-tural systems.

The ratio of labile P to Sum-P is low (on average26%). Within the labile pool, a larger fraction is or-ganic than inorganic. The agroforestry systems hadconsistently higher fraction of Po than the monocul-tural systems. The larger percentage of organic Pwithin the labile pool and the significant effect ofagroforestry systems on the absolute amount of or-ganic P at 40-60 cm (effect on the interaction system× depth, p < 0.01 and no difference among depths inthe agroforestry systems, Table 2) are likely due totree biomass additions. Using 31PNMR (PhosphorusNuclear Magnetic Resonance), we showed that in ourstudy area the agroforestry coffee systems the ratiosof organic P to total P and of diester to monoesterphosphates are higher than in monocultural coffeegrowing systems (Cardoso et al. 2003a). We alsoshowed that agroforestry fields had more roots andmore mycorrhizal fungal spores deeper in the soilthan monocultural fields (Cardoso et al. 2003b).Moreover, we observed more earthworms in the agro-forestry fields than in the monocultural fields duringthe soil sampling.

The largest P fraction was found in the concen-trated HCl fraction. This is in agreement with otherdata for tropical Oxisols (Novais and Smyth 1999;Cross and Schlesinger 1995; Tiessen and Moir 1993).The second largest fraction of P was extracted withNaOH. This is in agreement with the data of Tiessenand Moir (1993) for Brazilian Oxisols, but in dis-agreement with the results of Lilienfein et al. (2000)for a cerrado Oxisol in Minas Gerais. They found thelargest fraction in NaOH, followed by concentratedHCl. For this soil type, Lilienfein et al. (1996) foundlow resin-P, which is in agreement with the presentstudy, and no P in the 1 M HCl-P. The large amountof P in the recalcitrant fractions (HCl) emphasises theneed to develop sound management strategies to uti-lise native soil P more effectively, which can perhapsbe achieved by manipulation of plants, such as inagroforestry systems, associated with micro-organ-isms (Richardson 2001).

Fractionation of P, as done here, is a static tool tostudy P and is not sufficient to draw firm conclusionson organic P transformation. The rate and the impactsof changes on P cycling by agroforestry systems andefficiency of P use by the crops in the long-term need

62

to be further examined and understood for full evalu-ation of the importance of agroforestry in soil P utili-sation. Moreover, detailed studies are required for abetter understanding of the P transformation in thesoil through microbial activity.

Conclusions

In the agroforestry fields the amount of organic P de-creased less with depth and the percentage of organicP in labile pools was higher in agroforestry fields thanin monocultural fields. This suggests that agroforestryinfluences the dynamics of P through the conversionof part of the inorganic P into organic P. Because therate of cycling is higher for organic P than for inor-ganic P, agroforestry would maintain larger fractionsof P available to agricultural crops, thereby reducingP losses to the unavailable pools.

Acknowledgements

The authors would like to thank the CTA-ZM, farm-ers and farmer unions in the Zona da Mata, withoutwhom this work would not exist. We thank Arne Jan-ssen for commenting on previous versions of themanuscript and Rein van Eck for the assistance in thelaboratory. Valuable comments by anonymous refer-ees are gratefully acknowledged. Thanks are also dueto CAPES, the Brazilian sponsor of I.M.C.

References

Ae N., Arihara J., Okada K., Yoshihara J. and Johansen C. 1990.Phosphorus uptake by pigeon pea and its role in cropping sys-tems of the Indian Subcontinent. Science 248: 477–480.

Agbenin J.O. and Tiessen H. 1995. Phosphorus forms in particle-size fractions of a toposequence from Northeast Brazil. SoilSci. Soc. Am. J. 59: 1687–1693.

Beck M.A. and Sanchez P.A. 1994. Soil phosphorus fraction dy-namics during 18 years of cultivation on a Typic Paleudult. SoilSci. 34: 1424–1431.

Bolan N.S. 1991. A critical review of the role of mycorrhizal fungiin the uptake of phosphorus by plants. Plant Soil 134: 189–208.

Cardoso I.M., Guijt I., Franco F.S., Carvalho A.F. and Ferreira-Neto P.S. 2001. Continual Learning for Agroforestry SystemDesign: University, NGO and farmer partnership in MinasGerais, Brazil. Agric. Syst. 69: 235–257.

Cardoso I.M., van der Meer P., Janssen B.H., Oenema O. andKuyper T.W. 2003a. Analysis of phosphorus by 31PNMR inOxisols under agroforestry and conventional coffee systems inBrazil. Geoderma 112: 51–70.

Cardoso I.M., Boddington C.L., Janssen B.H., Oenema O. andKuyper T.W. 2003b. Distribution of mycorrhizal fungal sporesin soils under agroforestry and monocultural coffee systems inBrazil. Agrofor. Syst. (in press).

CFSEMG 1989. Recomendação para o uso de corretivos e fertili-zantes em Minas Gerais, 4a aproximação. Lavras, Brazil, 176pp.

Cooper P.J.M., Leakey R.R.B., Rao M.R. and Reynolds L. 1996.Agroforestry and the mitigation of land degradation in the hu-mid and sub-humid tropics of Africa. Exp. Agric. 32: 235–290.

Cross A.F. and Schlesinger W.H. 1995. A literature review andevaluation of the Hedley fractionation; applications to the bio-geochemical cycle of soil phosphorus in natural ecosystems.Geoderma 64: 197–214.

Dalal R.C. 1977. Soil organic phosphorus. Adv. Agron. 29: 83–113.Fontes M.P.F. and Weed S.B. 1996. Phosphate adsorption by clays

from Brazilian Oxisols: relationship with specific surface areaand mineralogy. Geoderma 72: 37–51.

James D.W., Kotuby-Amacher J., Anderson G.L. and Huber D.A.1996. Phosphorus mobility in calcareous soils under heavy ma-nuring. J. Environ. Qual. 25: 770–775.

Hairston N.G. 1989. Ecological experiments. Purpose, design andexecution. Cambridge University Press, Cambridge, 370 pp.

Lilienfein J., Freibauer A., Neufeldt H., Westerhof R., Ayarza M.A.,Silva et al. 1996. Influence of land-use on the distribution ofwater stable aggregates and P status of sandy and clayey cer-rado Oxisols, Brazil. In: Pereira R.C. and Nasser L.C.B. (eds),Biodiversity and sustainable production of food and fibers inthe tropical savannas. Proceedings of the 1st International Sym-posium on tropical savannas. EMBRAPA-CPAC, Brasilia, Bra-zil, pp. 323–328.

Lilienfein J., Wilcke W., Ayarza M.A., Vilela L., Lima S.C. andZech W. 2000. Chemical fractionation of phosphorus, sulphur,and molybdenum in Brazilian savannah Oxisols under differentland use. Geoderma 96: 31–46.

Linquist B.A., Singleton P.W., Cassman and K.G. 1997. Inorganicand organic phosphorus dynamics during a build-up and de-cline of available phosphorus in an Ultisol. Soil Sci. 162: 254–264.

Magid J., Tiessen H. and Condron L.M. 1996. Dynamics of organicphosphorus in soils under natural and agricultural ecosystems.In: Piccolo A. (ed.), Humic Substance in Terrestrial Ecosys-tems. Elsevier, Amsterdam, pp. 429–466.

Neufeldt H., Silva J.E., da Ayarza M.A. and Zech W. 2000. Land-use effects on phosphorus fractions in Cerrado Oxisols. Biol.Fertil. Soils 31: 30–37.

Novais R.F. and Smyth T. 1999. Fósforo em solo e planta emcondições tropicais. UFV/DPS, Viçosa, Brazil, 399 pp.

Resende M. 1997. O manejo do solo na agricultura sustentável. In:Almeida J. and Navarro Z. (eds), Reconstruindo a Agricultura.Porto Alegre, Brazil, pp. 253–288.

Rheinheimer D., Cassol P.C., Kaminski J. and Anghinoni I 1999.Fósforo orgânico do solo. In: Santos G.A. and Camargo F.A.O.(eds), Fundamento da matéria orgânica do solo: ecossistemastropicais e subtropicais. Gênesis, Porto Alegre, Brazil, pp. 139–157.

63

Rice W.R. 1988. Analysing tables of statistical tests. Evolution 43:223–225.

Richardson A.E.R. 2001. Prospects for using soil microorganismsto improve the acquisition of phosphorus by plants. Aust. J.Plant Physiol. 28: 897–906.

Robinson J.S., Johnston C.T. and Reddy K.R. 1998. Combinedchemical and 31P-NMR spectroscopic analysis of phosphorusin wetland organic soils. Soil Sci. 163: 705–712.

Stewart J.W.B. and Tiessen H. 1987. Dynamics of soil organicphosphorus. Biogeochemistry 4: 41–60.

Szott L.T. and Melendez G. 2001. Phosphorus availability underannual cropping, alley cropping, and multistrata agroforestrysystems. Agrofor. Syst. 53: 125–132.

Tiessen H. and Moir J.O. 1993. Characterisation of available P bysequential extraction. In: Carter M.R. (ed.), Soil sampling andmethods of analysis. Canadian Society of Soil Science. LewisPublishers, pp. 75–86.

Young A. 1997. Agroforestry for Soil Management. ICRAF andCAB International, Wallingford, 2nd ed., 320 pp.

64