NutrCyclAgroecosyst DOI10.1007/s10705-013-9554-0 ORIGINAL ARTICLE Stream water nutrient and organic carbon exports from tropical headwater catchments at a soil degradation gradient John W. Recha • Johannes Lehmann • M. Todd Walter • Alice Pell • Louis Verchot • Mark Johnson Received:23July2012/Accepted:15January2013 (cid:2)SpringerScience+BusinessMediaDordrecht2013 Abstract Carbonandnutrientlosseswerequantified watershed OC losses, even 50 years after the forest fromfoursmallheadwatercatchmentsinwesternKenya wasremoved.Flow-weightedstreamwaterconcentra- intheyear2008.Theyincludeaforestedcatchmentand tionsofDOCandcoarseparticulateOC,allNspecies, three catchments under maize continuously cultivated totalP,KandNasignificantly(P\0.05)increasedin for5,10and50 yearsfollowingforestconversion.The streams after forest conversion and long-term cultiva- CisotopiccompositionofdissolvedorganicC(DOC)in tion. Solute concentrations increased despite the fact stream discharge suggested that soil organic C (SOC) that soil contents decreased and total water flow derived from the original forest rather than OC from increasedindicatingmobilizationofCandN,PandK maize may have contributed to a large extent to from soil with progressing cultivation. In contrast, Ca andMgconcentrationsinstreamwaterdidnotsystem- aticallychangeafterdeforestationandcultivation,and may be controlled by geochemical weathering rather J.W.Recha(cid:2)J.Lehmann(&) thanbychangingwaterflowpathsortopsoilcontents. DepartmentofCropandSoilSciences,Cornell All OC and nutrient exports increased with longer University,Ithaca,NY,USA e-mail:[email protected] cultivation over decadal time scales (P\0.05) to the same or greater extent than through deforestation and M.T.Walter the first years of cultivation. Fluvial OC and total N DepartmentofBiologicalandEnvironmental losseswere2and21 %oftotalSOCandtotalNdecline, Engineering,CornellUniversity,Ithaca,NY,USA respectively,inthetop0.1 mover50 years.FluvialOC A.Pell losses therefore played a minor role, and SOC losses DepartmentofAnimalSciences,CornellUniversity, were mainly a result of microbial mineralization. Ithaca,NY,USA ResultingtotalNlossesbystreamdischarge,however, L.Verchot were large with 31 kg ha-1 year-1 after 50 years of CenterforInternationalForestryResearch,Bogor, continuous cropping in comparison to fertilization of Indonesia 40 kg N ha-1 year-1. Most (91 %) of the N losses occurred as NO -. In contrast, P losses by stream M.Johnson 3 InstituteforResources,EnvironmentandSustainability, discharge were negligible in comparison to plant Vancouver,Canada uptake.Waterlossesshouldbemanagedtoreducesoil fertility declines especially through large N export M.Johnson from agricultural headwater catchments. However, DepartmentofEarthandOceanSciences,University ofBritishColumbia,Vancouver,Canada stream concentrations of both P (0.01–0.15 mg L-1) 123 NutrCyclAgroecosyst and N (0.4–4.8 mg L-1) were moderate or low with gathered in both temperate (Nair and Graetz 2004; respecttopossibleconsequenceforhumanhealthandnot Schipper et al. 2007) and tropical regions (Pandey responsibleforeutrophicationobservedinLakeVictoria. etal.2010).However,studiesonnutrientlossesfrom entirecatchmentshavemostlybeenconfinedtoforests Keywords Carbon(cid:2)Cultivation(cid:2) rather than to different agricultural watersheds with Degradationgradient(cid:2)Forestconversion(cid:2) increasingperiodofcultivation.StudiesfromMalay- Headwatercatchment(cid:2)Nutrients siaandtheAmazonexaminedimpactsofdisturbances (Zulkifli 1990), pasture (Thomas et al. 2004) and forestclearing(Lesack1993;Malmer1996;Williams Introduction andMelack1997;Williamsetal.1997).Itisnotclear, to what extent continuous cultivation and accompa- The impact of land use change on nutrient and soil nyingsoildegradationovertimeaffectfluvialnutrient organic carbon (SOC) dynamics in the tropics is lossesinadditiontotheeffectsoflossofforestcover, particularlyseveresincesoildegradationismorerapid andhowthesefluviallossescomparetonutrientinput than in temperate zones (Spaans et al. 1989; Malmer andcropuptake. and Grip 1990; Grip et al. 2004; Hartemink et al. Not only the relationship between agricultural 2008). Continuous cultivation and land tillage cause nutrient input and fluvial export is unclear, but also rapid loss of SOC (Mann 1986; Davidson and what these exports mean for pollution of waterways. Ackerman 1993; Tilman et al. 2002; McLauchlan Lake Victoria, the largest tropical lake in the world, 2006;Solomonetal.2007)duetothedisruptionofthe is significantly contaminated with DOC and nutri- physical, biochemical, and chemical mechanisms of ents, notably P, posing a threat to people, the fish soilorganicmatter(SOM)stabilization.However,the industry and biodiversity (Verschuren et al. 2002). It reductioninSOCstocksmaynotbesolelyattributed is widely believed that the contamination originates to increased mineralization losses to CO , but could fromlandusechangeinducedbyagriculture(Scheren 2 includeasignificant,yetpoorlyquantifiedproportion et al. 2000; Hecky et al. 2010). This is a common of stream water losses (Davidson and Ackerman attributionandnon-pointsourcepollutionwithNand 1993). On a basin and headwater catchment scale of P by agriculture is indeed observed in many parts of forests, stream water OC losses appear to be a minor the world (Carpenter et al. 1998). Yet, fertilizer use fraction of total SOC losses compared to mineraliza- in tropical smallholder agriculture is typically much tionofCO (Richeyetal.2002;Johnsonetal.2006b). lower than in industrialized countries where detailed 2 However, the importance of fluvial OC losses during assessments are available. No data are available that thefirstyearsofconversionfromforeststocultivation, investigate the nutrient and OC loads of agricultural which arethe mostimportant intermsofSOClosses headwaters in Africa and the contribution that (Davidson and Ackerman 1993) have not been well deforestation by burning and progressive cultivation quantified. While mineralization losses typically would make. decrease with greater period of cultivation (Kimetu The dynamics of nutrient and C losses over the et al. 2009), the stream water discharge usually course of the year can provide insight into the increases (Williams et al. 1997; Germer et al. 2010; mechanism and pathways of nutrient and C in Recha et al. 2012), potentially increasing the impor- watersheds (Markewitz et al. 2004; Johnson et al. tanceoffluvialOClossesascultivationprogresses. 2006a).Typically,weexpectadilutionofnutrientand Soil organic matter decline leads to rapid nutrient Cconcentrationsduringtherainyseason(Likensand releaseofNandPaswellasreducedcationexchange Bormann 1995). However, under certain circum- capacity resulting in diminished nutrient retention of stances of highly weathered soils, concentrations the soil (Lal 2006). However, surprisingly little may increase during the wet season as a result of information is available on how soil degradation greatersurficialflowpathsthatmobilizesolutesfrom following land use conversion affects nutrient losses nutrient and C-rich topsoils (Markewitz et al. 2004). from watersheds. On a plot level, information on It is not clear whether forest conversion and soil accelerateddecreaseinsoilnutrientcontentsresulting cultivationmaychangetherelationshipbetweenwater from various types of agricultural practices has been and nutrient concentrations and how these dynamics 123 NutrCyclAgroecosyst are modified in tropical soils rich in weatherable The site has a mean elevation of 1,800 m above sea minerals (Davidson et al. 2004; Yusop et al. 2006) level, a maximum daily temperature of 26 (cid:3)C and a withhighexchangeablebasecationcontents(Kimetu minimumof11 (cid:3)C(Glenday2006). et al. 2008). The fluvial DOC losses may be a The Kakamega-Nandi forest in Western Kenya is significant fraction of C losses and of total solute the country’s only remaining tropical rain forest. losses in watersheds (Selva et al. 2007). In forest Massive deforestation has taken place to create land ecosystems,themajorityofthefluvialorganicClosses for settlement and farming. About 16 % of forest occurred as DOC rather than coarser organic C cover was lost between 1986 and 2001 (Awiti et al. fractions (Johnson et al. 2006b; Selva et al. 2007). 2008). The forest forms the easternmost relic of the DespitethelowerimportanceofparticulateorganicC, Guinean-Congolian rainforest belt, which once mostoftheorganicCwasmobilizedfromthetopsoil spanned from East to West Africa. The area around litterlayer(Johnsonetal.2006a,b).Whetherthesame thisforest isamongthemostdenselypopulatedrural istrueforagriculturalsoils,orwhethersubsoilorganic areasintheworld.Ithadapopulationdensityof778 CmaygaininimportanceforsoilsdepletedinSOCby persons per km2 in 2009 compared to 73 persons per long-termcultivationisnotclear. square kilometer for the entire country (Kenya Thepurposeofthisstudywastoquantifythefluvial National Bureau of Statistics 2010). Consequently nutrient and OC losses from headwater catchments theforestisunderhighanthropogenicpressure,which separating the effects of soil degradation and con- ismirroredbythedecreasingnaturalforestcoverand comitant losses of soil organic matter contents from intensive cultivation (deGraffenried and Shepherd thoseofforestclearingintheuppercatchmentsofthe 2009; Swallow et al. 2009). Past deforestation rates LakeVictoriabasin.Itishypothesizedthat(1)stream intheKakamega-Nandiforestindicatedadecreaseof waterOClossesincreaseinimportancewithlengthof forest area and an increase in the fragmentation of cultivation; (2) long-term cultivation is as or more natural,old-growthforest(LungandSchaab2006). important than deforestation and short-term cultiva- Soils in the Kapchorwa catchment are kaolinitic tion for increasing nutrient and OC losses; and (3) Acrisols (FAO-UNESCO-ISRIC 1988), which are headwaters with small holder agriculture are not classified as Ultisols in the US soil taxonomy (Soil important sources of N and P to rivers and lakes. SurveyStaff2003).Theparentmaterialofthesesoils The effects of continuous agricultural land use on is principally granite, with some inclusions of nutrientswereinvestigatedbycomparing threehead- Precambriangneisses,whichsupportsLuvisols(Wer- water catchments following forest conversion to ner et al. 2007) and other undifferentiated basement continuous maize production for either 5, 10, or system rocks at higher elevations (Jaetzold and 50 yearsinWesternKenyawithacatchmentthathas Schmidt 1983). Soils in the catchment have remainedforested. 45–49 % clay, 15–25 % silt, and 26–40 % sand (Kimetuetal.2008). The forest section of the Kapchorwa catchment is Methods part of the Kakamega-Nandi forest composed of tropical rainforest species. It is largely an indigenous Thestudysite forest with a 30-m closed canopy dominated by evergreen hardwood species. The most common The field measurements were done in Kapchorwa species are Funtumia africana, Ficus spp, Croton within Nandi County in Kenya, which is located in spp, and Celtis spp (Glenday 2006). Other species EasternAfrica.Thesiteislocated60 kmnortheastof include Aningeria altissima (A. Chev.), Milicia ex- Lake Victoria at longitude 35(cid:3)00000 E and latitude celsa (Welw.) C.C. Berg, Antiaris toxicaria (Lesch.) 0(cid:3)100000N(Rechaetal.2012).Theregionbelongstothe andChrysophyllumalbidum(G.Don),Oleacapensis sub-humidecologicalzonecharacterizedbyabimodal (L.) and Croton megalocarpus (Hutch.) (Kinyangi rainy season with a mean annual precipitation of 2008). The above and below ground net primary 2,000 mm(Awitietal.2008).The‘‘longrainseason’’ productionoftrees ina tropical forest is estimated at is from April to June (*1,200 mm) and the ‘‘short 15.2 Mg ha-1 year-1 (Hertel et al. 2009). The agri- rain season’’ from August to October (*800 mm). cultural catchments have maize as the sole crop, and 123 NutrCyclAgroecosyst have been under maize cultivation since conversion Stream stage was recorded using water capacitance from forest cover. The maize grain yields without probes (Odyssey Dataflow Systems Pty Ltd., New fertilizer input inthe 5-,10- and 50-year old agricul- Zealand) installed at the weir. The probes were turalcatchmentsare6.5,5.5and2.5 Mg ha-1 year-1, programmed to give a reading of the average stream respectively(Ngozeetal.2008). stage at 2 min. Data from these probes were down- loadedbiweekly.Theweirratingsweredeterminedat Hydrologicinstrumentationandfieldwork lowandintermediateflows.Theweirratingcorrelation coefficientswerer2 = 0.944,r2 = 0.915,r2 = 0.926, Inordertoidentifytheheadwatercatchments,thetime r2 = 0.931fortheforest,5-,10-and50-yearconver- whennativeforestswereconvertedtoagriculturewas sions, respectively. The stream hydrographs were determined from historical community settlement normalizedbycorrespondingcatchmentsizestoallow patterns over the last century (Bleher et al. 2006; comparisonbetweenresponsesforthe4catchments. Rechaetal.2012).Specificyearsofconversionwere Streamwatersamplingwasdonebiweeklyattheweir verified from data available from records of the outlet, at the beginning and mid of every month. The Kenyan government from the Department ofForests, watersampleswerefilteredthrough0.45-lmpore-size theMinistryofAgriculture,aswellasfrominterviews glass-fiber filter, into two separate 50-mL centrifuge with officials of local institutions and from county vials.Weaddedthymolintothefirst50-mLcentrifuge council records. Within each site, population settle- vial that was used for determination of calcium (Ca), mentpatternsweredistinctandnewlyacquiredfields magnesium (Mg), sodium (Na), potassium (K), total were excised from sections of the native forest for dissolvedphosphorus(TDP),nitratenitrogen(NO –N), 3 agriculture.Allfourheadwatercatchmentsarelocated ammonium nitrogen (NH ?-N) and total dissolved 4 within an area of 6 km2 (Fig. 1) and represent a soil nitrogen(TDN).Weadded10 %HClintothesecond degradation gradient that corresponds to years under 50-mLcentrifugevialthatwasusedfordeterminationof maize cultivation that has been used in several other dissolved organic carbon (DOC) and the DOC stable studies(Kimetuetal.2008,2009;Ngozeetal.2008; isotopicratioof13Cto12C.The0.45-lmpore-sizeglass- KimetuandLehmann2010;Rechaetal.2012).Such fiberfilterpaperwithsedimentwasairdriedandkeptfor chronosequencessubstitutetimeforspaceandhaveto the determination ofcoarseparticulateorganiccarbon be carefully selected to assure similar properties (CPOC), the CPOC isotopic ratio of 13C to 12C, and before the change (Huggett 1998). Also hydrological coarseparticulatenitrogen(CPON). differencesbetweencatchmentshavetobeconsidered Five soil samples from 0 to 0.1 m depth were (Elsenbeer,2001;Johnsonetal.2006b)andonlyclear collected randomly by stratifying for location within trendsacrosstheentiresetofthefourcatchmentsare slope and plateau of each catchment, after carefully interpretedhere. removing the litter layer (capturing only the mineral Hydrologic instrumentation was installed in 2006 horizon).TheywereanalyzedforSOM,SOC,totalN, (Rechaetal.2012)andcatchmentsmonitoredcontin- available P, Ca, K, Mg and Na. Similarly, four sites uously. The boundaries of each catchment were were randomly selected by stratifying for location determinedusingaGlobalPositioningSystem(GPS) within slope and plateau of each catchment and onthelandscapearoundeachspringuptotheplateau, additional soilsampled at depths of0.1–0.3, 0.3–0.9, using an average of five to eight independent assess- 0.9–1.5, 1.5–2.4 m. Bulk density was determined by ments. The GPS data were then used to generate a thesoilcoremethod;weusedfivesoilsamplesfroma Geographical Information System (GIS) output and depth of 0–0.1 m in each catchment (Campbell and map of the area. The sizes of the catchments were Henshall1991).Soilporosity(ø)wascomputedfrom 12.8 ha for the forest, 14.4 ha for the 5-year old bulk density (q ) and particle density (q ) using the b p conversion, 9.1 ha for the 10 year conversion and formula ø = 1 - (q /q ). The use of organic and b p 10.0 haforthe50-yearconversion.Therewereatotal inorganic fertilizers was assessed in all studied ofone,six,andelevenhouseholdslivinginthe5-,10-, agricultural watersheds by a full survey of all and 50-year conversion catchments, respectively. households using interviews. Farmer interviews in A standard V-notch weir was constructed at each the watersheds confirmed information provided catchment outlet for determining stream discharge. by the studies of Kinyangi (2008), Kimetu (2009), 123 NutrCyclAgroecosyst Fig.1 Mapofthestudy areaoftheKapchorwa headwaterchronosequence ofcatchmentsinwestern Kenya,anditsapproximate locationwithinKenya(inset showsKenyaandits districts).Theweirpositions areshownbydotsineach catchment.Lettersindicate landusehistoriesofthe catchments:forested catchment(a),maize cultivationfor5years(b), 10years(c)and50years (d)followingforest conversion and Moebius-Clune et al. (2011). The soils received Plasma Atomic Emission Spectroscopy (ICP-AES, nil, or only very little inorganic fertilizer since forest ARCOS, Germany).TDNwasanalyzedusing Shima- clearing,noanimalmanure,andwerealwayscropped dzu’sTotalNitrogenModule,TNM-1thatuseschemi- with maize. From the interviews, only di-ammonium luminescence. Dissolved organic nitrogen (DON) was phosphate (DAP, N:P:K of 18:46:0) fertilizer was computed using the formula DON = TDN-NO - ? 3 used at an application rate of6 and9 kg ha-1 within NH ?. DOC analysis was carried out on a Shimadzu 4 the 10 year and 50 year watersheds, respectively, for Total Organic Carbon-Visionary Series (TOC-V ) CSH theyear2008.Thefarmersinthe5 yearwatersheddid analyzerfollowingtheproceduredescribedbyQianand notuseanyfertilizer.TheNandPadditionwas1.02 Mopper(1996).NO –NandNH ?-Nweredetermined 3 4 and 2.76 kg ha-1, respectively, for the 10 year on a Seal AQ2-Automated Discreet Analyzer (Seal watershed. The addition for the 50 year watershed Analytical,England). was 1.62 and 4.14 kg ha-1 of N and P, respectively. Animal manure was notused in anyhousehold. Statisticalanalyses Laboratorymeasurements Statistical analyses were done using JMP Version 8 (SASInstituteInc,Cary,NC,USA)forsoilproperties SoiltotalCwasdeterminedbydrycombustionafterfine andstreamwatersolutesofthefourcatchments.Aone grinding soil using a Mixer Mill (MM301, Retsch, way ANOVA was used for the soil properties, and a Germany).SampleswereanalyzedfortotalCcontents repeated measure analysis controlling for date of withaEuropaANCA-GSLCNanalyzer(PDZEuropa sampling was performed for the stream water solutes. Ltd., Sandbach, UK). Soil organic matter (SOM) was Both analyses were followed with posthoc multiple determinedbylossonignition(Storer1984)andsoilpH comparisons using a Tukey correction for multiple (inwater)atthew/vratio1:2.5usingaglasselectrode comparisonswhenthecatchmenteffectwassignificant. (ThermoScientific,Beverly,MA,US).Filterscontain- ComparisonswereconsideredsignificantatP\0.05. ing CPOC were ground using a Mixer Mill (MM301, Retsch,Germany)andanalyzedfortotalC,theisotopic ratioof13Cto12C,andCPONusingacoupledEuropa Results 20–20continuousisotoperatiomassspectrometry(PDZ EuropaLtd.,Sandbach,UK).TheMehlich3extraction Soilproperties procedure (Mehlich 1984) was used for the available soil P, Ca, Mg, K and Na. The Ca, Mg, K, Na and P Within the top 0.1 m, the SOM and SOC concentra- concentrations were obtained by Inductively Coupled tions significantly decreased by 59 % and 75 %, 123 NutrCyclAgroecosyst Table1 Soilpropertiesinthetop0.1m Forest 5yearconversion 10yearconversion 50yearconversion Soilorganicmatter(%) 17.08a 13.67b 8.53c 7.00d Soilorganiccarbon(gkg-1) 108.3a 68.8b 36.4c 27.5c Soilorganiccarbon(tha-1) 86.6a 62.6b 37.5c 32.2c Soilnitrogen(gkg-1) 11.2a 6.9b 3.4c 2.3c Soilnitrogen(tha-1) 8.96a 6.28b 3.50c 2.69c Soilbulkdensity(gcm-3) 0.80c 0.91b 1.03ab 1.17a Totalporosity 0.70a 0.66b 0.61bc 0.56c Moistureat0.33bar(%v/v,fieldcapacity) 34.7a 34.1a 24.9b 23.0b pH(water,1:2.5) 7.39a 6.48b 6.23b 5.81c AvailableP(gkg-1) 0.011a 0.007a 0.003a 0.006a AvailableK(gkg-1) 0.69a 0.58ab 0.23b 0.44ab AvailableCa(gkg-1) 6.25a 4.98a 1.77b 2.91b AvailableMg(gkg-1) 0.73a 0.45b 0.20c 0.33bc AvailableNa(gkg-1) 0.026a 0.020a 0.010a 0.011a MeanswithinarowfollowedbythesamelettersarenotsignificantlydifferentfromeachotheratP\0.05(n=5) respectively,andthesoilNcontentsby79 %fromthe trendsatdepth.Similarly,thed13Cvaluesbecameless forest to the 50-year old agricultural catchment negative deeper in the profile for all the catchments (Table 1). The forest SOC and soil N were signifi- indicating a slight enrichment in the heavier isotope cantly (P\0.05) higher than the 5-year conversion. comparedtothetopsoil. Soilsinthe10-and50-yearconversiondidnotdiffer. TheSOM,SOCandNlosswasveryrapidinthefirst Streamwaterchemistry 10 yearsofconversiondecreasingby50,66and70 %, respectively. Available soil P, K and Na did not The d13C composition of CPOC (Table 2) did not change as a result of forest clearing and soil use change from the forest (-27.52 %) to the 10 year between the studied watersheds. In contrast, extract- conversion (-27.28 %). The 50-year old catchment able Ca and Mg significantly (P\0.05) decreased hadaslightlyhigherd13Cvaluecomparedtotheother overtime(Table 1).Similarly,pHvaluessignificantly catchments.ThestreamwaterDOCd13Ccomposition decreasedfromforest(7.39)toagriculture(5.81). Thesoilq increasedrapidlyby28.8 %inthefirst b Delta 13C (‰) 10 years of cultivation from 0.80 to 1.03 g cm-3 -28 -26 -24 -22 -20 -18 -16 -14 -12 (Table 1). Overall, the soil qb increased by 46.3 % 0.0 forest from 0.8 to 1.17 within 50 years of cultivation. The 5 year rapid increase in qb in the first 10 years after forest 0.5 1500 yyeeaarr conversion was followed by a 12 % drop in the total porosityfrom0.69to0.61inthesameperiod. m) 1.0 In the 0–0.1 m depth, the soil d13C values were h ( pt -26.07 ± 0.68 %, -25.37 ± 0.84 %, -21.21 ± e 1.5 D 0.87 % and -19.18 ± 0.44 % for the forest, 5-, 10- and 50-year conversion catchments, respectively 2.0 (Fig. 2). This indicates enrichment in the heavier isotope from the forest to the catchment under 2.5 cultivationfor50 years.Thed13Cvaluesatthetopsoil Fig.2 Longtermshiftintheisotopiccompositionofd13Cof (0–0.1 m) increased linearly over time from -26.07 thesoilsatvariousdepthsinthewatersheds.n=4.Barsare±1 to -19.18 % (r2 = 0.95; P\0.0001), with similar standarddeviationofthemeanforeachdepth 123 NutrCyclAgroecosyst Table2 Theisotopic Forest 5year 10year 50year compositionofd13Cofthe conversion conversion conversion CPOCandDOC,andthe flowweightednutrient Sedimentandfilteredstreamwater concentrationsofthe CPOCd13C(%) -27.52b -27.29b -27.28b -26.40a headwatercatchments DOCd13C(%) -28.59c -28.25c -26.79b -22.75a Streamflowconcentrations DOC(mgL-1) 1.31a 1.48a 1.47a 1.52a CPOC(mgL-1) 1.25a 1.43a 1.49a 1.65a CPON(mgL-1) 0.12a 0.14a 0.15a 0.16a TDN(mgL-1) 0.49b 0.67b 4.83a 4.64a NO -(mgL-1) 0.40b 0.58b 4.71a 4.52a 3 Meanswithinarow DON(mgL-1) 0.10a 0.09a 0.10a 0.12a followedbythesameletters TDP(mgL-1) 0.03b 0.01b 0.05b 0.15a arenotsignificantly K(mgL-1) 0.36c 0.80b 0.95a 1.18a differentfromeachotherat P\0.05(n=24for Ca(mgL-1) 7.23a 5.55b 5.48b 7.16a CPOCd13C,n=4for Mg(mgL-1) 3.01a 2.04c 1.61d 2.54b DOCd13C,andn=12for Na(mgL-1) 2.99c 3.70c 4.65b 5.61a allothers) followed the same trend as the sediment but with a ThetotalCexportsincreasedafterforestconversion significantly greater magnitude of enrichment in the andwithsubsequentcultivation(Table 3).TheDOCand heavierisotopefromtheforestandrecentconversion CPOC export doubled in the first 5 years following tothe10-yearoldconversion.Dischargeinthe50-year forestconversion.Theincreasecontinuedbyafurther43 conversionhadasignificantlyhighervaluethanthatof and 70 % from the 5- to the 50-year conversions, allotherheadwaters(-22.75 %).Theflow-weighted respectively. Amounts of CPON, TDN and NO - 3 averageconcentrationsofDOC,CPOCandCPONdid exported were also twice the first 5 years after forest not change with forest conversion and duration of conversion. Upon longer cultivation between 5 and cultivation (Table 2). TDN concentrations in the 10 years, CPON export increased 3.5 times, while the discharge from forest and 5-year catchments did not TDNandNO -exportincreasedeightfold.Between10 3 differ, but increased with longer cultivation. NO - and 50 years of continuous cultivation, the amount of 3 was the dominant form of dissolved N (91 %) in the CPONexportdoubledwhilethatofTDNandNO -did 3 Kapchorwawatershedfluvialecosystem,followedby notincreasefurther.ThemagnitudeofTDPexportwas DON (9 %), whereas NH ? was below our detection farlessthanthatofCandallothernutrients.Theamounts 4 limit.The10-and50-yearconversionshadninetimes ofTDPexporteddidnotchangeafterforestconversion higherTDNconcentrationscomparedtotheforestand (\0.03 kg ha-1), but increased with subsequent culti- most recent conversion. Stream TDP concentrations vationup to the0.98 kg ha-1. The cations(K, Ca Mg wereatleastoneorderofmagnitudelowercompared andNa)exhibitedasimilartrendwherebytheamounts to TDN, Ca, K, and Mg. TDP concentrations of the exported increased from the forest to the 5-, 10- and forest, 5- and 10-year conversions were threefold 50-yearconversions.Amongthemeasuredbasecations, lowerthanthatofthe50-yearconversion(P\0.05). KhadthelowestexportwhileCahadthehighest. StreamwaterKandNaconcentrationsbothincreased Therewasnotaclearseparationofthehydrograph upon forest conversion and subsequent cultivation. between the long rainy (April-June) and short rainy Overall,KconcentrationsincreasedthreefoldandNa (August-October) seasons in the Kapchorwa catch- concentrationstwofoldfollowing50 yearsofcultiva- ment in2008. The stream flow response toprecipita- tion after forest conversion. In contrast to all other tioneventsincreasedwithlongercultivation(Fig. 3). solutes, Ca and Mg concentrations did not show a Therewerenorelevantchangesintheconcentrations consistenttrendfromforesttocultivation. of DOC, TDN, TDP, Ca, and Mg across the whole 123 NutrCyclAgroecosyst Table3 Streamwater carbon and nutrient exports (kgha-1 fluvial C export is low compared to other studies. year-1)fromtheKapchorwaheadwatercatchments Mooreetal.(2010)observedafluvialtotalorganicC removal of 12 % by a Bornean blackwater river in a Forest 5year 10year 50year conversion conversion conversion peat-coveredarea ofIndonesia.CumulativefluvialN lossesoverthefirst50 years(usinglinearinterpolation DOC 3.87 6.85 8.46 9.79 of the sum of TDN and CPON) were 1.3 t ha-1 CPOC 3.94 6.60 8.57 10.63 compared to the total soil N losses of 6.3 t ha-1 CPON 0.33 0.64 0.87 1.03 (representing21 %)inthefirst0.1 m(calculatedfrom TDN 1.45 3.12 27.76 29.93 Table 1).SimilartofluvialClosses,watershedsatour NO - 1.18 2.68 27.09 29.16 3 siteshowedlowerexportthanatotherlocations. DON 0.27 0.44 0.67 0.77 Both fluvial CPOC and DOC were mainly mobi- TDP 0.09 0.03 0.29 0.98 lized from the topsoil, as d13C values in stream K 1.06 3.68 5.44 7.61 discharge were lower than in the topsoil and soil Ca 21.4 25.7 31.5 46.2 valuesincreasedwithdepthasalsoreportedbyKrull Mg 8.93 9.42 9.23 16.38 etal.(2002)atanearbysite.Similarconclusionswere Na 8.86 17.08 26.70 36.22 drawnbyFranketal.(2000)andRaymondandSaiers (2010) who found most of the DOC losses to be derived from the topsoil in catchments in central year.TheKandNaconcentrations,however,decreased Switzerland and eastern United States, respectively. slightly after the long rainy season in August and Alternatively, the low d13C values in our study may remainedlowfortheshortrainyseason.Overall,there alsoindicatethatOClossesoriginatedfromolderSOC was no relationship between the solute concentrations fractions characterized by small size and association anddischarge(Fig. 4). with minerals that showed lower d13C values in the same chronosequence (Kinyangi 2008). The isoto- Discussion piccompositionofthetopsoilcorrelatedstronglywith the DOC d13C (r2 = 0.84; y = 1.132x ? 7.1478) SourcesoffluvialOCandnutrientlosses as well as with the CPOC d13C (r2 = 0.83; y = 18.25x - 49.523),butwithalowslopeforCPOC. EventhoughannualfluvialOCexportsincreasedmore thanthree-fold(ourstudy)andCO evasionfromSOC Nutrientlosses,cropproductionandstream 2 mineralization decreased (Kimetu and Lehmann pollution 2010),annualDOCandCPOClossesbystreamwater of22 kg ha-1fromthehighlydegradedsoilswerestill TotalN(TDN ? CPON)lossesof31 kg ha-1year-1 low with only 2 % mineralization losses of the inthehighlydegradedsoilswererelevantcomparedto 1,200 kg ha-1 determined by direct measurements crop N uptake of maize at the same sites with from the soil surface using static vented chambers. 90 kg N ha-1 year-1 (Kimetu et al. 2008). There Similarly, cumulative fluvial organic C losses were wasnofertilizerapplicationintherecentconversion, 0.9 t ha-1 (2 %) over the first 50 years (using linear whileapplicationsinthe10-and50-yearconversions interpolation of the sum of DOC and CPOC) com- were about 40 kg N ha-1 (average from all farms in pared to SOC losses of 54 t ha-1 in the first 0.1 m thestudiedcatchments).Thetypicalapplicationrates (calculatedfromTable 1;similarSOClossesreported to maize by most farmers in the entire region are bySolomonetal.2007;KimetuandLehmann2010). 20 kg N ha-1 (Ngoze 2008), less than the stream It is possible that inorganic C and particularly waterlosses reportedhere.Evenmoredramatic were dissolved CO may significantly contribute to total the fluvial Ca and Mg losses of 46 and 16 kg ha-1 2 fluvial C losses from watersheds as shown in the year-1, respectively, which were greater than uptake Amazon(Johnsonetal.2008).However,evenwitha by a single maize crop of 20 and 10 kg ha-1, possibly larger inorganic than organic C loss, micro- respectively (Kimetu et al. 2008). In contrast, annual bialmineralizationfromthesoilstillexceeds90 %of fluvial P losses of 1 kg ha-1 were low compared to total SOC losses over the 50 years. The quantified plant uptake of 10 kg ha-1 (Kimetu et al. 2008) and 123 NutrCyclAgroecosyst 8 8 Forest 5 year 4 4 10 -1mg L) 6 3 -1day) -1mg L) 6 3-1day) Concentration ( 24 12 Discharge (mm Concentration ( 24 12 Discharge (mm 0 0 0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Day Day 8 8 10 year 50 year 4 4 -1) 6 1) -1) 6 1) L - L - mg 3 day mg 3 day Concentration ( 24 12 Disharge (mm Concentration ( 24 12 Discharge (mm 0 0 0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Day Day DOC TDP TDN Calcium Potassium Magnesium Sodium Discharge Fig.3 Bi-weeklydischargeandconcentrationsofDOC,Na,Ca,KandMgoftheKapchorwaheadwatercatchments 8 the watershed was not measured and stream water losses are only a lower boundary of actual losses of -1)L 6 soilP.Nonetheless,bothCaandMgaswellasPlosses g by stream water may not be as serious as N losses m s ( 4 because available P, Ca, and Mg in soil decreased n o muchlessthantotalN(Table 1;similarforthesame ati ntr 2 sites reported by Ngoze et al. 2008; Kinyangi 2008), e and crops were found to respond less to P than to N c n o additions (Ngoze et al. 2008). In addition, TDN C 0 concentrations and total losses by stream discharge increased the most by forest conversion and contin- 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 uous cultivation compared to any other nutrient Discharge (mm day-1 ha-1) studiedhere. DOC Calcium Magnesium Despite the significant N losses from stream TDN Potassium Sodium discharge for agricultural productivity, the observed TDP concentrations in the studied headwater catchments Fig.4 TherelationshipbetweendischargeandDOC,Naand are not an environmental concern for aquatic envi- nutrientconcentrationsinthe50yearwatershed ronments. Average annual N concentrations at or fertilizationofabout7.5 kg P ha-1(sumofinorganic below 3 mg L-1 and maximum concentrations of and organic fertilizers averaged for all farms) in the 6 mg L-1 during the dry season of December to oldestconversion.However,redistributionofPwithin March (Fig. 2) do not pose a risk for human health 123 NutrCyclAgroecosyst (Rubio-Arias et al. 2010) and are not expected to The increase in streamwater concentrations of cause eutrophication (Hill et al. 2011; King and TDN, NO - and DON with longer duration of 3 Balogh 2011). Similarly, P concentrations of cultivation can be attributed to three sources. (1) 0.01–0.15 mg P L-1 (Fig. 2) in the stream discharge MineralizationofSOMledtoaccumulationofmineral are moderate but not a cause for significant environ- N and especially nitrate (as seen from stream water mental concern (Auer et al. 1986; King and Balogh concentrations) which is weakly adsorbed to soil 2011;Yuetal.2011).Valuesupto6.7 mgPO -PL-1 mineralsandeasilyleached.(2)Thegreaternumberof 4 havebeenfoundintheRiverThamescatchmentinthe households on older conversions presumably led to UK(Youngetal.1999)andupto10 mgPO -PL-1in higher inputs of farm manure from the cattle kept by 4 the Spree River in Germany (Lewandowski and the farmers. The manure was shown to consist of Nutzmann 2010). Also DOC concentrations are sig- 22 g N kg-1(Kimetuetal.2008).Halfofthefarmers nificantlylower inthestudiedheadwatersthaninthe in the older conversions applied inorganic fertilizer lower Yala and Lake Victoria (Lalah and Wandiga equivalentto40 kg N ha-1.(3)Asstatedabove,only 2006).Therefore,thewelldocumentedcontamination farmers with fields that have been cultivated for ofthelowerYalaRiver(Aloo2003)andLakeVictoria 10 yearsorlongerapplymineralfertilizers.Fieldsthat withDOC,NandPwhichhasbeensuspectedtocause havebeenrecentlyclearedatthesamesitesarerarely the observed eutrophication (Scheren et al. 2000; fertilizedduetotheirhighNmineralizationratesand LalahandWandiga2006;Heckyetal.2010)doesnot theirlowresponsetomineral(Ngozeetal.2008)and seem to be caused by agriculture in the headwater organicNapplications(Kimetuetal.2008). catchments,butmustoccurfurtherdownstream. The increase in fluvial K and Na concentrations with longer cultivation may be explained by lower Carbonandnutrientlossesafterlandusechange SOC contents and the soil mineralogy being domi- nated by kaolinite, both resulting in low cation Our DOC concentrations and total losses were in the retention (Kimetu and Lehmann 2010). Krull et al. same range as reported by Lesack et al. (1984) from (2002) reported mineralogy dominated by quartz, theGambiaRiverintheWestAfricansavannahwitha kaolin and mica, with minor components associated DOC concentration of 1.98 mg L-1 and DOC export with feldspars and oxides in the adjacent Kakamega of 2.67 kg ha-1 year-1. Similar DOC concentrations part of the forest. They observed an abundance of but 2–10 times greater export were found by Cairns quartz, kaolinite, muscovite and microcline with a etal.(2009)intheforestsoftheOregonCascadesand slight increase in goethite with increasing soil depth. of southern Amazonia by Johnson et al. (2006b), yet The observed greater proportion of surface runoff of values are within the range reported for a variety of 4–10 %ofrainfallfromforestcomparedtolong-term watershedselsewhere(Lesacketal.1984;Selvaetal. cultivatedfields(Rechaetal.2012)mayalsohaveled 2007; Stutter et al. 2008; Waldron et al. 2009). The toagreatermobilizationofKandNanearthesurface. CPOC and CPON concentrations and export were in Extractable K concentrations did not decrease with the same order of magnitude as those of DOC and forest conversion and cultivation (Table 1; compare DON. This finding differs from that of Vidal-Abarca withKinyangi2008).SimilartoN,agreaternumberof etal.(2001)whofoundDOCtobethemostimportant households in the older conversion watersheds may fraction (98 %) of organic C flowing in a saline and haveaddedmorekitchenwastethatarerichinKand semi-arid stream in Spain. Similarly greater propor- Na such as ash from firewood cookstoves and farm tions of DOC were also reported by Johnson et al. yard manure. Kimetu et al. (2008) documented (2006a, b) from southern Amazonia. Therefore, par- contentsof23.2 g K kg-1inthemanure. ticulateOClossesaregreaterthaninotherstudies,but The lack of change or even slight decrease in Ca theobservedfluvialOClossesfromtheforestseemto andMgstreamwaterconcentrationsmaybearesultof berepresentativeofthosefoundinotherstudies.The adominanceofweatheringastheprimarysourceofCa examined change of solute losses as a response to and,toacertainextent,Mg.Inaddition,thesignificant landuse change may therefore be indicative of losses decrease in extractable Ca in the topsoil (Table 1) in inotherregions. combination with the rather decreasing stream water 123
Description: