Alterations of soil aggregates and intra-aggregate organic carbon fractions after soil conversion from paddy soils to upland soils:Distribution,mineralization and driving mechanism

Longfei KANG ,Jiamei WU ,Chunfeng ZHANG ,Baoguo ZHU and Guixin CHU

1College of Life Science,Shaoxing University,Shaoxing 312000(China)

2The Key Laboratory of Oasis Eco-agriculture, Xinjiang Production and Construction Group/Department of Resources and Environmental Sciences,Agronomy College,Shihezi University,Shihezi 832000(China)

3Hunan Institute of Agricultural Environment and Ecology,Changsha 410000(China)

4Jiamusi Branch,Heilongjiang Academy of Agricultural Sciences,Jiamusi 154007(China)

ABSTRACT Investigating the impacts of soil conversion on soil organic carbon(OC)content and its fractions within soil aggregates is essential for defining better strategies to improve soil structure and OC sequestration in terrestrial ecosystems.However,the consequences of soil conversion from paddy soil to upland soil for soil aggregates and intra-aggregate OC pools are poorly understood.Therefore,the objective of this study was to quantify the effects of soil conversion on soil aggregate and intra-aggregate OC pool distributions.Four typical rice-producing areas were chosen in North and South China,paired soil samples(upland soil converted from paddy soil more than ten years ago vs.adjacent paddy soil)were collected(0-20 cm)with three replicates in each area.A set of core parameters(OC preservation capacity,aggregate carbon(C)turnover,and biological activity index)were evaluated to assess the responses of intra-aggregate OC turnover to soil conversion.Results showed that soil conversion from paddy soil to upland soil significantly improved the formation of macro-aggregates and increased aggregate stability.It also notably decreased soil intra-aggregate OC pools,including easily oxidized OCa(EOCa),particulate OCa(POCa),and mineral-bound(MOCa)OC,and the sensitivity of aggregate-associated OC pools to soil conversion followed the order:EOCa(average reduction of 21.1%)＞MOCa(average reduction of 15.4%)＞POCa(average reduction of 14.8%).The potentially mineralizable C(C0)was significantly higher in upland soil than in paddy soil,but the corresponding decay constant(k)was lower in upland soil than in paddy soil.Random forest model and partial correlation analysis showed that EOCa and pH were the important nutrient and physicochemical factors impacting k of C mineralization in paddy soil,while MOCa and C-related enzyme(β-D-cellobiohydrolase)were identified as the key factors in upland soil.In conclusion,this study evidenced that soil conversion from paddy soil to upland soil increased the percentage of macro-aggregates and aggregate stability,while decreased soil aggregate-associated C stock and k of soil C mineralization on a scale of ten years.Our findings provided some new insights into the alterations of soil aggregates and potential C sequestration under soil conversion system in rice-producing areas.

Key Words:decay constant,easily oxidized organic C,macro-aggregate,meso-aggregate,micro-aggregate,mineral-bound organic C,particulate organic C,potentially mineralizable C

INTRODUCTION

Soil carbon(C)has been a major focus globally in this century.This is mainly because conservation and storage of soil organic C (OC) have been considered as critical components of soil quality for sustaining crop productivity in agricultural system (Jianget al.,2017) and soil OC sequestration play a crucial role in mitigating global climate change(Lehmann and Kleber,2015;Arachchiet al.,2016).It’s reported that soil was the largest terrestrial store of C that was the second largest to the ocean(Stockmannet al.,2013).The mineralization of soil OC was recognized as an important process that is associated with nutrient release,greenhouse gas emissions,and food production(Lianget al.,2014;Mustafaet al.,2020).Therefore,understanding the alterations of soil OC mineralization processes is vital for realizing better management of soil fertility,food security,and mitigation of climate change.

In agricultural ecosystems,C sequestration potential is influenced by many factors,such as edatope(Chabbiet al.,2009),climate condition(Wiesmeieret al.,2014),cropping system(Jagadamma and Lal,2010),and fertilization(Bhattacharyyaet al.,2012).Total content of OC in a given soil is dominated by the balance between C output and input(Xieet al.,2017).Several studies have shown that the increase of soil OC was closely associated with organic amendments,such as manure,straw,and litter(Mustafaet al.,2020)and the loss of soil OC was also related to edatope and land use change(Weiet al.,2021).Moreover,soil biochemical and physical properties also play a crucial role in soil OC sequestration.For instance,soil particles(e.g.,soil aggregates)with different sizes resulted in significant spatial heterogeneity of soil OC (Xuet al.,2021).Though the effects of soil physicochemical and biological properties on soil OC pools have been extensively investigated,the response mechanism of soil OC accumulation to the alteration of aggregates received little attention and knowledge on the distribution of intra-aggregate OC pools remains fragmentary.

It was reported that the distribution of different aggregate sizes was the key parameter in characterizing soil structure and it plays major roles in several processes including the storage of plant nutrients,the optimization of soil air regime,and the accumulation of soil OC(Dorjiet al.,2020).Previous studies illustrated that macro-aggregates contained more labile C fractions that from litter or manure.In contrast,micro-aggregates possessed more recalcitrant C substrates,which was prompted by the humification of microbial residues or organic matter(OM)(Sixet al.,2006;Totscheet al.,2018).Moreover,Six and Paustian (2014)illustrated that soil OC is a key factor in maintaining soil aggregate stability because it primarily involved in the processes of aggregate formation and destruction.According to hierarchical theory,Tisdall and Oades(1982)emphasized that:i) soil binding agents (plant root,litter,or hyphae)bind soil micro-aggregates into macro-aggregates,ii)fresh soil OC promotes the formation of macro-aggregates and particulate OC in macro-aggregates promote the production of micro-aggregates,and iii) the turnover of OC in soil aggregates(formation/disruption)occurs continually with time.Nevertheless,knowledge about the responses of soil aggregate-associated C to soil conversion from paddy soil to upland soil has not been well-documented.

In this century,soil conversion has become one of the predominant global changes,which was driven by the urgent and high demand for variety of foods,fibers,and other products (Maranguitet al.,2017).Paddy soil makes up the largest anthropogenic wetlands and cover approximately 0.7 to 0.9 billion ha on earth and is mostly formed under flooded lowland conditions (Kögel-Knabneret al.,2010;Chenet al.,2021).Paddy soil is widely distributed across temperate,subtropical,and tropical climate zones in China,it is highly affected by human activities,and it plays vital roles in maintaining biodiversity and regulating climate(Wuet al.,2021).Simultaneously,rice(Oryza sativaL.)is the major source of daily calorie intake for half of the global population(Carrijoet al.,2017).Hence,paddy soils play crucial roles in regulating ecosystem balance and providing food.However,large areas of paddy soils have been converted to upland soils with the acceleration of urbanization,economic interest,and market requirement in China during the past decades(Yuanet al.,2015).It is reported that paddy soil area has decreased from 33.3 million ha in 1981 to 26.5 million ha in 2003(Sunet al.,2011).This trend is even more pronounced in East and South China,where approximately 1.19 million ha paddy soil has been converted into upland cultivation(e.g.,orchard farms) from 1997 to 2011 in Zhejiang Province,China(Yang and Zhang,2014).Consequently,to ensure food security,it is urgent to understand the mechanisms of soil nutrient transformation and sequestration under land conversion condition,especially aggregate-associated C pools in soil.

Many studies have been developed to assess the kinetics of soil C mineralization from various C pools under anthropogenic disturbances(Leifeld and Fuhrer,2005;Cooperet al.,2011;Saviozziet al.,2014;Ghoshet al.,2019).However,only bulk soil was incubated in most studies,and the mineralization characteristics of various soil OC fractions was obtainedviaconceptual equations/models(Muelleret al.,2014).In this study,wet sieving procedure was used and incubation was carried out to determine soil aggregate fractions with different sizes.The parallel incubation of these aggregate fractions with different sizes will deepen our understanding of the relationships between soil aggregate fractions and OC pools.Hence,the specific objectives of this study were to:i)investigate the effects of soil conversion from paddy soil to upland soil on aggregate stability and OC distribution,ii)assess kinetics of C mineralization for bulk soil and soil aggregates with different sizes,and iii)explore the driving mechanisms of aggregate-associated OC distribution and mineralization under soil conversion system.

MATERIALS AND METHODS

Site description

Four typical rice-producing areas in China were selected(Fig.1),which distributed at Changsha City in Hunan Province(HN),Shaoxing City in Zhejiang Province(ZJ),Jiamusi City in Heilongjiang Province(HLJ),and Urumqi City in the Xinjiang Uygur Autonomous Region(XJ).The climate types belong to subtropical monsoon climate in HN and ZJ,temperate monsoon climate in HLJ,and temperate continental climate in XJ.The mean annual temperature are 18.1,17.5,3.5,and 8.1°C in HN,ZJ,HLJ,and XJ,respectively.The mean annual precipitation are 1 517,1 461,531,and 307 mm in HN,ZJ,HLJ,and XJ,respectively.Meteorological data during 2008-2020 were obtained for each area from the China Meteorological Data Sharing Service System(http://data.cma.cn).Soil textures are dominated by loam in HN and ZJ and silt loam in HLJ and XJ based on the World Reference Base(WRB)classification system issued by FAO.

Fig.1 Visualized photographs of the study sites at four typical rice-producing areas in China,including Changsha City in Hunan Province(HN),Shaoxing City in Zhejiang Province(ZJ),Jiamusi City in Heilongjiang Province(HLJ),and Urumqi City in the Xinjiang Uygur Autonomous Region(XJ).

A paire of locations(paddy and upland soils)was selected(0-20 cm)with three replicates in each area.The paddy and upland soils were adjacent to each other with similar soil parent materials and meteorological conditions.Paddy soil had been consecutively cultivated for rice monocropping system for more than 42 years,while the adjacent upland soil had been converted from paddy soil 9-12 years before(Table I,Fig.1).The dominant plant species were shrub forest(Fraseri dress),arbor forest(Malus halliana),grain crops(rotation of maize-soybean),and arbor forest(UlmuspumilaL.) at HN,ZJ,HLJ,and XJ sites in upland soils,respectively.Fertilization regimes and cultivation times were shown in Table SI(see Supplementary Material for Table SI).Soil pH values were 4.97,5.25,6.71,and 7.89 at HN,ZJ HLJ,and XJ sites in paddy soils,respectively,and 4.89,4.51,6.22,and 8.11 at HN,ZJ HLJ,and XJ sites in the corresponding upland soils,respectively.Other basic soil physicochemical properties were shown in Tables SII and SIII(see Supplementary Material for Tables SII and SIII).

Sample collection

Soil samples were collected from September to November 2019 after rice harvest.At each location,three randomized sampling plots (50 m× 40 m) were selected with 3 repetitions.Eight surface soil samples(“S”shaped sampling strategy)were sampled with an auger(M-0943,New Landmark Soil Equipment Co.Ltd.,China;diameter=5.0 cm,length=35 cm with graduated scale)after removing the biological crust,litter,and fine roots manually,and then thoroughly mixed to form a composite sample(24 soil samples in total).

TABLE IGeographical featuresa) and alteration models of land use at the study sites of four typical rice-producing areas in China

Soil samples were transported to laboratory promptly,and each composite soil sample was divided into two portions.One portion was air-dried at room temperature.All intrusions(e.g.,small stones and iron-manganese concretion)and plant tiny roots in soil samples were removed by a tweezer,then soil samples were sieved through meshes(5,2,or 0.15 mm) for determination of physicochemical properties and nutrient contents.Another portion(fresh soil sample)was sieved through a 2-mm mesh and immediately stored at 4°C for measurement of microbial biomass and enzymatic activities.

Aggregate fractionation scheme

Soil aggregate fractions were separated by wet sieving procedure(Elliott,1986),and three soil aggregate fractions were obtained based on aggregate diameter (D):i)D ＞0.25 mm,ii)0.053 ≤D≤0.25 mm,iii)D ＜0.053 mm.Briefly,air-dried soil sample(100.0 g)on a 0.25-mm sieve was watered in deionized water at 25°C to promote disintegration(Kemperet al.,1985).Aggregates were separated by moving the 0.25-mm sieve for 50 times within 6min.The material remaining on the sieve was rinsed into containers,while material passing through the 0.25-mm sieve mesh was further fractionated using the same process with 0.53-mm sieve.All fractions were dried in an oven at 50°C.These aggregate fractions were defined as large macro-aggregates(D ＞0.25 mm),meso-aggregates(0.053 ≤D≤0.25 mm),and micro-aggregates (D ＜0.053 mm) according to its particle size(Wanget al.,2014).

For each soil aggregate and bulk soil,soil samples were passed through a 0.15-mm sieve prior to OC and easily oxidized OC(EOC)determination,and another subsample(macro-aggregates and meso-aggregates)was used to determine soil intra-aggregate particulate OC (POC)and mineral-bound OC(MOC)using potassium dichromate external heating method(Shaw,1959)after sieving through a 0.053-mm sieve.Remaining sub-sample were kept to assess C-mineralization kinetics.Detailed analysis method for different C pools(OC,EOC,POC,and MOC)and basic physicochemical properties were shown in Texts S1-S4(See Supplementary Material for Texts S1-S4).To obtain more reliable results,the determination was repeated three times.

Incubation experiment and C mineralization measurement

Soil C mineralization in intra-aggregate fractions(macroaggregates,meso-aggregates,and micro-aggregates) and bulk soil samples were assessed according to Gregorichet al.(1989)method in the laboratory under controlled aerobic conditions.It’s worth noting that the sieving procedure and then oven drying(60°C)can result in microbiome disappears(e.g.,spore-forming microbes)for each soil aggregate(Creameret al.,2011).Therefore,soil inoculum solution was prepared from corresponding bulk soil.The detailed processes can be found in Mustafaet al.(2020).

All soil samples were pre-incubated for 12 d in the dark at 25°C before incubation(Mustafaet al.,2020).Then 15.0 g bulk soil sample and each aggregate were taken in 265-mL jars with vials containing 0.1 mol L-1sodium hydroxide solution to trap the respired CO2content.The respired CO2content was determined at 1,3,7,14,21,28,and 35 d after incubation.

Comprehensive analysis,calculation and assessment

Assessment of aggregate stability.Following the wet sieving procedure,mean weight diameter(MWD,mm),unstable aggregate index(UAI),and geometric mean diameter(GMD,mm)of soil aggregates were calculated according to Kemper and Rosenau(1986)and Sekaranet al.(2021):

whereXi(mm)is the mean diameter of fractioni,Wi(kg kg-1soil)is the weight percentage of fractioniin soil,Wt(kg)is the total weight of the sample used for the analysis,andW0.25(kg) is the weight of soil aggregates with size＜0.25 mm.

Turnover of C in aggregate fractions.For each aggregate fraction,the enrichment factor(EF)(Guggenbergeret al.,1994),OC preservation capacity(CPC)(Sekaranet al.,2021),and C turnover(ACT)characteristic(Sixet al.,2000)were calculated as follows:

whereCfandCt(g kg-1)refer to the content of different C forms(e.g.,OC,POC,or EOC)in the respective aggregate fraction and bulk soil,respectively,Ci(g kg-1)is the OC content in aggregate fractioni,POCc(g kg-1)is the content of coarse POC originated from soil macro-aggregate fraction(D ＞0.25 mm),and POCf(g kg-1)is the content of fine POC originated from meso-aggregate fraction(0.053 mm＜D ＜0.25 mm).

Sequestration of C.The stock of OC(OCstock,mg C ha-1) in bulk soil or aggregate fractions at 0-20 cm was calculated using the equation given by Bhattacharyyaet al.(2010).Additionally,the difference in OC sequestration(ΔOCstock)and annual rate of OC sequestration(OCr,mg C ha-1year-1)were calculated as follows(Lianget al.,2021):

whereC(g kg-1)is the soil C content,A(1 000 m2ha-1)is the area,dis the soil depth(0.20 m in the current study),andρ(g cm-3) is the soil bulk density.The OCstock1and OCstock2refer to OC stock in treatment soil (upland soil)and reference soil (paddy soil),respectively,andy(year)represents the years after paddy soil was converted to upland soil.

Carbon management index.The C management index(CMI)and its components were assessed based on the method of Blairet al.(1995),where paddy soil was considered as reference sample and CMI was calculated as follows:

where LC refers to the lability of C,CKMnO4andCtotal(g kg-1) are the KMnO4-oxidized and total C contents,respectively,LI refers to the lability index,and LCsampleand LCreferenceare the lability of C in sample soil(upland soil)and reference soil(paddy soil),respectively.Then C pool index(CPI)was expressed as the ratio of total C in sample soil(TCsample)and reference soil(TCreference).

Modeling and assessment of soil C from microbial respiration.The C mineralization from bulk soil and each soil aggregate fraction was analyzed through the total CO2-C respired content for 35 d.The potentially mineralizable C(C0,mg kg-1)and corresponding decay constant(k,d-1)were modeledviaa first-order decay model(Stanford and Smith,1972).

where Ctis the amount of C mineralized(mg kg-1)during time interval timet(d).

Statistical analysis

Data analysis was performed using SPSS 21.0 (SPSS Inc.,USA).The significant differences between different treatments were evaluated using one-way analysis of variance(ANOVA)(Tables SII-SIV,see Supplementary Material for Table SIV)and three-way ANOVA(Tables II and SV,see Supplementary Material for Tables SV)with Duncan’s test at the level ofP ＜0.05.Partial correlation analysis was performed using the R package“ppcor”,and the random forest(RF)model was visualized to estimate the importance of physicochemical factors and various C pools in explainingkof C mineralization by using the R package“randomFores”(R Core Team,2022).The Mantel test was carried out to compare the correlations of distance matrices between soil C mineralization(C0and correspondingk)and various factors(Euclidean distance)viathe R package“ggcor”.Partial least squares path modeling(PLS-PM)was constructed to explore the direct and indirect effects of soil conversion on various soil C pools by the“pls-pm”package in R.Other figures were visualized by GraphPad Prism 8.0(GraphPad Software Inc.,USA).

RESULTS

Aggregate size distribution and stability

Macro-aggregates (＞0.25 mm) were found to be the most abundant class in both paddy and upland soils,which ranged from 36.1% to 69.1% (Table SIV).This indicated that macro-aggregates were the dominant component of soil aggregate.Soil conversion from paddy soil to upland soil significantly promoted the formation of water-stable macro-aggregates across all sites.For example,the proportions of macro-aggregates at HN,ZJ,HLJ,and XJ sites were significantly increased by 46.9%,23.3%,27.9%,and 36.9%,respectively,in upland soil relative to paddy soil.While this conversion markedly decreased the proportions of meso-aggregates(0.053-0.25 μm)and micro-aggregates(＜0.053 mm)by 21.5%and 37.1%,respectively,across all sites.Consequently,significantly higher GMD and MWD and lower UAI were found in upland soil as compared to paddy soil(Table SIV).

Soil OC pools in bulk soil and soil aggregate fractions

Soil conversion significantly and synchronously decreased OC content in macro-aggregates (＞0.25 mm),meso-aggregates(0.053-0.25 mm),and micro-aggregates(＜0.053 mm).Additionally,EOC contents in macro-aggregates,meso-aggregates,and micro-aggregates were 21.4%,22.6%,and 19.5% lower in upland soil than in paddy soil,respectively(Table II).Similar trends were obtained for POC and MOC contents under this soil conversion system.Interestingly,soil conversion from paddy soil to upland soil significantly(P ＜0.05)altered CPC values of soil aggregates.The CPC values of soil micro-aggregates at HN,ZH,HLJ,and XJ sites were 56.0%,42.3%,56.4%,and 29.3%lower in upland soil than in paddy soil,respectively.While this conversion significantly reduced CPC values of soil meso-aggregates(33.0%on averagely)and micro-aggregates(46.0%on average)across all sites.

Bulk soil EOC(EOCt),POC,and MOC(MOCt)contents were significantly decreased by 22.5%,15.5%,and 16.7%in upland soil compared to paddy soil,respectively,across all sites(Table II).For instance,EOC contents at HN,ZJ,HLJ,and XJ sites were 25.4%,19.3%,13.8%,and 31.3%lower in upland soil than in paddy soil,respectively.Consequently,significantly(P ＜0.05)lower OC content was obtained in upland soil(19.5 g kg-1on average)relative to paddy soil(22.6g kg-1on average).

Carbon mineralization kinetics under soil conversion system

The cumulative amount of mineralized C was lower in upland soil,as compared to paddy soil in both bulk soil and different aggregate fractions (Figs.2a-h and S1,see Supplementary Material for Fig.S1).The determination coefficient (R) values of kinetic parameters of C mineralization potential C0andkshowed that they well fitted the first-order kinetic model (R2＞0.982).The C0values in bulk soil,macro-aggregates,meso-aggregates,and microaggregates were 12.8%,17.9%,23.4%,and 36.7%higher in upland soil compared to paddy soil,respectively.Whilekvalues in bulk soil,macro-aggregates,meso-aggregates,and micro-aggregates were 12.8,16.5,21.7,and 19.1 lower in upland soil than in paddy soil,respectively,according to the first-order decay model(Fig.2,Table SVI,see Supplementary Material for Table SVI).Moreover,in both paddy and upland soils,C0values were higher in bulk soil(801.4 mg CO2-C kg-1soil)and meso-aggregates(795.2 mg CO2-C kg-1soil)than in macro-aggregates(754.1 mg CO2-C kg-1soil) and micro-aggregates (737.2 mg CO2-C kg-1soil),whilekvalues were higher in bulk soil (0.063 d-1) and macro-aggregates(0.060 d-1)than in meso-aggregates and micro-aggregates(Fig.2).

Fig.2 Cumulative C mineralized(a-h)and potentially mineralizable C(C0,i)and corresponding decay constant(k,j),as well as their regression lines,in bulk soil and different soil aggregate fractions(SA)in upland soils converted from paddy soils(a-d)and paddy soils(e-h)from four typical rice-producing areas in China,including Changsha City in Hunan Province(a and e),Shaoxing City in Zhejiang Province(b and f),Jiamusi City in Heilongjiang Province(c and g),and Urumqi City in the Xinjiang Uygur Autonomous Region(d and h).In a-h,vertical bars represent standard errors of the means(n=3).In i and j,the shaded areas around the regression lines represent the 95%confidence intervals.

Driving factors of C mineralization under soil conversion system

Random forest model was used to predict the contributions of continuous variables,such as various C pools in aggregates and bulk soil,total nutrients(e.g.,total N(TN)and P(TP)),and soil biochemical variables(i.e.,pH,mineral N(Nm),available P(AP),cation exchange capacity(CEC),microbial biomass C (MBC),clay content,CaCO3,and enzyme activities),tok(Fig.3).In paddy soil,aggregateassociated EOC(EOCa),aggregate-associated POC(POCa),and EOCt were calculated as the key variables impactingkbased on soil C pool level,while soil TN,pH,and MBC were identified as the key variables impactingkbased on soil environmental factor level.Likewise,in upland soil,EOCa,POCa,and aggregate-associated MOC(MOCa)were dominant variables in predictingkin comparison with other predictors (P ＜0.05),while soilβ-D-cellobiohydrolase(CBH),clay content and TP were the dominant variables in predictingkin comparison with other predictors(P ＜0.05)(Fig.3a,b).

Fig.3 Random forest(RF)model(a and b)and partial correlation analysis(c and d)of the corresponding decay constant(k)for potentially mineralizable C using nutrient,physicochemical,and biological factors as predictors in paddy soils(a and c)and upland soils converted from paddy soils(b and d)from four typical rice-producing areas in China.OCa,EOCa,POCa,and MOCa are the organic C,easily oxidized organic C,particulate organic C,and mineral-bound organic C in aggregates,respectively,and OCt,EOCt,POCt,and MOCt are the corresponding C fractions in bulk soil,respectively.TN=total N;Nm=mineral N;TP=total P;AP=available P;CEC=cation exchange capacity;CBH=β-D-cellobiohydrolase;BG=β-1,4-glucosidase;MBC=microbial biomass C;AV=all variables;Var=variance;IncMSE=increase in mean squared error.

Moreover,partial correlation analysis illustrated that EOCa was the most important variable impactingkin paddy soil,while MOCa was the most important predictor in predictingkin upland soil(Fig.3).This further confirmed the results obtained from the RFmodel.In addition,Spearman’s and Mantel correlation analyses showed that C0was significantly correlated with CPC(r=0.79),GWD(r=0.73),and OCt(r=0.66)in paddy soil,while OCt(r=0.88),CPC(r=0.70),and MOCt(r=0.68)were closely associated with C0(P ＜0.001)in upland soil(Fig.4).Additionally,the soil aggregate properties of MWD and GWD were positively associated with different OC fractions both in paddy and upland soils(P ＜0.05).

Fig.4 Relationships and Mantel tests between environmental variables(biological factors,chemical properties,physical features,and nutrient status)and characteristic parameters of C mineralization(potentially mineralizable C(C0)and corresponding C mineralization decay constant(k))in paddy(a)and upland(b)soils.MWD=mean weight diameter;GMD=geometric mean diameter.See detailed description of other variables in Fig.3.

Direct and indirect effects of soil conversion on aggregateassociated C pools

The PLS-PM was conducted to explore the direct and indirect relationships between aggregate-associated C pools and soil physicochemical and biological variables.These variables well explained OC distributions in both bulk soil and different aggregate fractions (R2＞0.82) (Fig.5).In macro-aggregates,soil physical properties directly and positively affected nutrient status (path coefficient (pc)=0.45,P ＜0.001),while soil chemical properties directly and adversely affected soil nutrient status (pc=-0.92,P ＜0.001) and biological factors (pc=-0.88,P ＜0.001).As a result,nutrient status was directly and positively associated with EOCa,POCa,and MOCa(P ＜0.01),and biological factors were directly and negatively related with MOCa (pc=-0.55,P ＜0.001) (Fig.5a).Overall,soil conversion from paddy soil to upland soil largely altered the distributions of soil aggregate-associated C pools through altering soil physical properties,regulating soil nutrient status,and decreasing soil chemical and biological properties of different aggregate fractions(Fig.5b).

Fig.5 Partial least squares path modeling(PLS-PM)of the effects of soil conversion(a-d)and standardized total effects of nutrient,physicochemical,and biological factors(e-h)on different soil C pools in different aggregate fractions:macro-(a and e),meso-(b and f),micro-(c and g),and all(d and h)aggregates.Solid black and red lines represent positive and negative relationships,respectively,and the width of these lines denotes the strength of the causal relationship.Numbers associated with single-headed arrows are the partial regression coefficients of multiple regressions.Asterisks*,**,and***indicate P ＜0.05,P ＜0.01,and P ＜0.001,respectively.ns=not significant;PF=physical feature;CP=chemical property;NS=nutrient status;BF=biological factor.See detailed description of variables in Figs.3 and 4.

Comprehensive indexes of C turnover, sequestration and management

Soil conversion from paddy soil to upland soil significantly changed the EFvalues of CPC,EOCa,POCa,and MOCa(P ＜0.05).The macro-aggregates,meso-aggregates,and micro-aggregates exhibited different C EFvalues(Table SV).For instance,soil conversion from paddy soil to upland soil resulted in an increase of 7.6% in EOC EF(EFEOC)value of macro-aggregates,while it resulted in a decrease of 3.4% in EFEOCvalue of micro-aggregates (P ＜0.05)across all sites.Meanwhile,the EFvalues of various C fractions(CPC,EOCa,POCa,and MOCa)decreased with the increase of aggregates size both in paddy and upland soils.Additionally,significantly higher LI value was obtained in macro-aggregates(0.94 on average)than in meso-aggregates(0.88 on average)and micro-aggregates(0.89 on average).Similar trend was found for CMI.Soil ACT values at HN,ZJ,HLJ,and XJ sites were 2.9%,2.4%,22%,and 4.2%lower,respectively,in upland soil relative to paddy soil.Consequently,soil conversion from paddy soil to upland soil significantly reduced OCrat HN and HLJ sites,while it notably enhanced OCr at ZJ and XJ sites.The OCrvalue was lower in meso-aggregates(0.024 mg cm-2year-1)than in macro-aggregates(0.032 mg cm-2year-1)and microaggregates (0.036mg cm-2year-1) across all sampling sites.In bulk soil,similar trend was obtained for OCr(Table SV).Moreover,soil biological activity index(BAI)(Text S4)at HN,ZJ,HLJ,and XJ sites were 31.6,%18.5%,24.9%,and 40.0% lower,respectively,in upland soil relative to paddy soil.

DISCUSSIONS

Response mechanisms of soil aggregates and OC stock to soil conversion

Soil aggregate is a key component of soil structure and plays a major role in nutrient cycling,OM decomposition,and OC retention(Dorjiet al.,2020).Our findings showed that soil conversion from paddy soil to upland soil significantly increased the proportion of macro-aggregates(＞0.25 mm),but decreased the proportions of meso-aggregates(0.053-0.25 mm)and micro-aggregates(＜0.053 mm).Therefore,the notably higher GMD and MWD,but lower UAI occurred in upland soil compared to paddy soil(Table SIV).These results are mainly due to different land management measures.Ploughing is considered to be an agricultural activity closely related to soil structural stability (e.g.,aggregate stability),and frequent ploughing could largely destroy soil structure(Jianget al.,2011).In the present study,paddy soil was long-term flooded and deeply ploughed each year,while upland soil was planted by shrub,arbor,or legumes with no deep ploughing each year(Table SIV),which resulted in stronger disturbance of the surface layer(0-20 cm)in paddy soil than upland soil.Eventually,a higher fraction of macroaggregates was obtained in the upland soil.This result was evidenced by another study based on land use change from paddy soil to vegetable field,which showed that intensive ploughing can cause the break-up of macro-aggregates and the reduced stability of soil aggregates(Wanget al.,2014).

Another possible explanation was that the roots of plants in upland soil (Fraseri dress,Malus halliana,Ulmus pumilaL.,maize,and soybean)and paddy soil(Oryza sativaL.)are different.The roots of shrub and arbor belong to tap root,while rice root belongs to fibrous root and is thinner and grows slower than tap root(Montagnoliet al.,2021).Therefore,the destroyed degree of surface layer(0-20 cm)was more intensive in paddy soil than in upland soil.Similar to our findings,Wanget al.(2019)also reported that different roots(e.g.,tap and fibrous roots)strongly disturbed the distribution of soil aggregate fractions and soil structure stability under soil conversion condition.In conclusion,plant species,tillage measures,and management modes(e.g.,fertilization)are considered as the factors affecting soil aggregate distribution and stability.

Additionally,soil OC sequestration is considered as important components of soil fertility and quality,and it plays a crucial role in mitigating global climate change(Arachchiet al.,2016).Our study showed that soil OC was significantly declined after soil conversion from paddy soil to upland soil across all sites(Table II).This phenomenon mainly owed to that:i) C input is greater in paddy soil(plough layer) than in upland soil (Liuet al.,2019) and paddy soil formed more complex OC structures than upland soil,resulting in slower utilization of OC by microorganisms(Qiuet al.,2018);ii) the phenolic functional groups are largely enriched in paddy soil due to the condition of O2deficiency(Olket al.,2006);iii)soil minerals stabilize OC in paddy soil by decreasing the accessibility for microorganisms and enzymesviaisolation in pores as well as sorption and physical occlusion within aggregates,which decreased the turnover rate of OC(Baileyet al.,2019);and iv)rice root exudates and dissolved organic substances are bound with Fe minerals(iron plaques mechanism),which can largely reduce OC mineralization (Kuzyakov and Razavi,2019).In line with our findings,Aliet al.(2021) showed that significantly higher OC appeared in paddy soil compared with upland soil.All in all,soil conversion from paddy soil to upland soil not only significantly increased the proportion of macro-aggregates,but also notably reduced soil C stock.

Previous studies emphasized that soil aggregate stability was positively related to the preservation of OC and C sequestration (Chenget al.,2015;Eganet al.,2018).In our study,soil conversion from paddy soil to upland soil significantly improved MWD,while this soil conversion significantly decreased OC(Tables SIII and SIV).A possible explanation could be that the loss of OC driven by biochemical mechanisms offset the physical preservation of OC by aggregates under this soil conversion condition.In fact,the degradation and loss of OC(i.e.,OC stability)can be affected by physical occlusion within soil aggregates(physical mechanism),chemical absorption with minerals(chemical mechanism),and OC inherent molecular structure resistance to microbial degradation(biochemical mechanism)(Schmidtet al.,2011).Especially compared with paddy soil,the reduction of soil moisture could enhance the liberation of protected OC and its exposure to oxidizing conditions in upland soil,which largely accelerates the loss and degradation of OC(Weiet al.,2021).

Response mechanisms of aggregate-associated C pools to soil conversion

Based on hierarchy concept,soil aggregate-associated OC showed a significant effect on the transformation and turnover of OC,and it was regarded as the main agent for soil aggregation(Sixet al.,2004).Hence,OC content in various aggregate fractions plays a key role in assessing soil fertility and soil health (Heet al.,2021).Soil macro-aggregates were comprised of different sizes of soil micro-aggregates under the action of temporary binding agents(Cooperet al.,2020).Soil macro-aggregates played a significant role in protecting OC,resulting in the accumulation of OC in macroaggregates(Eganet al.,2018).In agreement with this,our findings showed that higher soil OC content was observed in soil macro-aggregates rather than in micro-aggregates at HN,ZJ and HLJ sites(Table II).Similarly,Mustafaet al.(2020)reported that soil macro-aggregate fraction possessed the highest OC content(3.7-8.6g kg-1)instead of microaggregates and meso-aggregates in a red soil.However,another theory showed that the specific surface area of soil micro-aggregates was larger relative to macro-aggregates,thus micro-aggregates can capture more OC(e.g.,litter residues and root exudates)(Sixet al.,2004).Considering this,opposite result was obtained at XJ site,where the content of OC was higher in micro-aggregates than that in macroaggregates(Table II).This result was probably attributed to the fact that soil micro-aggregates were derived from the combination of persistent binding agents and primary particles,and this structure was relatively stable(Xuet al.,2021).Therefore,OC associated with micro-aggregates was relative stable and difficult to be decomposed by microorganisms at XJ site.

Labile OC(LOC),as temporary binding agents,plays a vital role in the formation and stabilization of soil aggregates (Heet al.,2021).Revealing the distribution of soil LOC pools in different aggregate fractions will shed valuable insights into the dynamics of soil aggregates in agricultural ecosystems(Liet al.,2018).In the present study,irrespective of land use pattern (paddy and upland soils),EOC content was significantly higher in macro-aggregates compared to micro-aggregates(Table II).Our results were consistent with the results of Heet al.(2021),who found that EOC in macro-aggregates exhibited low sensitivity to microbial decomposition and environmental disturbances.Thus,macro-aggregates were regarded important OC pool for accumulation of EOC in agricultural soil.Additionally,the content of POC was higher in macro-aggregates than in meso-aggregates and micro-aggregates.The distribution of MOCa content in aggregates showed the same trend with that of POCa (Table II).Similarly,this phenomenon also was evidenced by Heet al.(2021) in a fir plantation soil mainly because soil POC can be rapidly decomposed by microorganisms and converted into MOC in soil(Muruganet al.,2019).

As discussed above,soil OC and LOC fractions (i.e.,EOC,POC,and MOC) were concentrated in the macroaggregates regardless of land use pattern.According to soil aggregate theory(Edwards and Bremner,1967),the formation of soil micro-aggregates was envisioned as a solid-phase reaction occurring among the electrically neutral clay,polyvalent metals,and OC.Soil macro-aggregates were further produced by clustering of these micro-aggregates(Xuet al.,2021).Therefore,soil OC composited into macro-aggregates would be physically protected OC,which made it inaccessible for decomposition by microorganisms.Consequently,the results obtained in this study suggest that the formation of soil macro-aggregates plays a crucial role in C sequestration under soil conversion condition.

Generally,the responses of various soil LOC fractions to different disturbances(e.g.,tillage pattern,fertilization,warming,and precipitation)were variable(Jiang and Xu,2006).In this study,soil conversion from paddy soil to upland soil significantly decreased soil aggregate-associated C pools(EOCa,MOCa,and POCa)across all sites,and the sensitivity of soil aggregate-associated C fractions to this conversion was:EOCa(21.1%)＞MOCa(15.4%)＞POCa(14.8%)(Table II).Similar results were observed by Songet al.(2012)and Luoet al.(2019),who found that soil EOC was more sensitive to land use change than other soil OC pools in agricultural soil.Moreover,soil OC pools showed the highest stability under soil conversion from paddy soil to upland soil in the meso-aggregate fraction in this study,followed by macro-aggregates(Table II).This phenomenon was mainly attributed to the fact that micro-aggregates possess large surface area and the exposure of OC substances provides an opportunity for microbial acquisition.In contrast,the structure of soil macro-aggregates is more conducive to the colonization of mycorrhizal fungi,thus more OC was accumulated in the macro-aggregates(Juarezet al.,2013;Qianet al.,2018).Furthermore,soil conversion from paddy soil to upland soil notably decreased soil ACT across all sites in the present study(Table SV).The possible explanations for this phenomenon were:i)abundant OC input(e.g.,root and litter residues)in paddy soil enriched different speciation of OC substances (e.g.,phenolic functional groups) (Olket al.,2006) (Table I),and ii) microbial biomass and enzyme activities(biological activity index)were significantly higher in paddy soil than those in upland soil (Tables II and SV).Additionally,fertilization was also an important factor influencing the distribution of various OC pools and alterations of aggregate-associated C turnover.

Response of aggregate-associated C mineralization to soil conversion

Variations in soil OC mineralization are considered to be closely associated with soil aggregate sizes and aggregateassociated C contents(Mustafaet al.,2020).In this study,the higher mineralization of OC was obtained in macroaggregates relative to micro-aggregates across all sites(Fig.2a-h),suggesting that macro-aggregates had more readily mineralizable OC than micro-aggregates.Our results agreed with the findings of Rabbiet al.(2014),who showed that OC associated with macro-aggregates could produce higher CO2-C than OC associated with micro-aggregates.However,Rabbiet al.(2015)and Caiet al.(2016)found the contradictory results.After analysis and comparison of many study results,the contradictory results can be attributed to inconsistencies in experiment methods(e.g.,wet or dry sieving procedures),cultivated conditions(25 or 35°C),and content and complexity of OC(Gregorichet al.,1989).Moreover,soil conversion from paddy soil to upland soil resulted in the decrease of CO2-C production across all aggregate fractions and bulk soil(Fig.2a-h).Similarly,Xieet al.(2017)also reported that the production of CO2-C was decreased after soil conversion.In the present study,the considerably lower CO2-C production was observed in upland soil,which might be due to the decline of microbial activity and extracellular enzyme activities after soil conversion from paddy soil to upland soil(Tables II and SV).

Additionally,C0andkvalues were estimated using the first-order decay equation to assess the sensitivity of soil C mineralization under soil conversion system.In this study,the changes in C0andkvalues showed the opposite trend(i.e.,kvalues decreased,while C0values increased)(Fig.2i,j,Table SVI),implying that each of them independently reflects the different kinetics of OC mineralization(Zhanget al.,2010).Moreover,soil conversion from paddy soil to upland soil significantly increased C0values in various aggregate fractions,while it notably decreasedk(Fig.2i,j).Therefore,soil conversion from paddy soil to upland soil largely declined OC stock and prompted the transformation of stable C pools to potential mineralized C pools.Ultimately,it promoted the rapid mineralization of OC in upland soil.

In agriculture ecosystems,OC mineralization is driven by various factors(Rakeshet al.,2021;Kanet al.,2022).In this study,we found that soil pH and EOCa were the important physicochemical and nutrient factors impacting thekvalues of OC in paddy soil,respectively.While MOCa and CBH were identified as the key nutrient and physicochemical factors in upland soil,respectively(Fig.3a-d),Xiaoet al.(2018)reported that pH was the vital factor influencing OC mineralization in a red soil.Additionally,Mantel test and Partial least squares path model further proved that the nutrient factor was the dominant variable in driving soil mineralization (i.e.,C0andk) under this soil conversion system(Fig.4).Hence,in the future,to achieve soil C sequestration and decrease OC loss,it is crucial to identify the dominant physicochemical(e.g.,soil texture,porosity,pH,CEC,and nutrient availability)and biological(e.g.,C/N/P-acquiring enzymes activities and microbial biomass,diversity,and community composition) factors under the condition of artificial disturbances (e.g.,soil conversion,fertilization,irrigation,and tillage measures)in agroecosystems.

CONCLUSIONS

In summary,our findings showed that soil conversion from paddy soil to upland soil significantly increased the percentage of macro-aggregates and enhanced soil aggregate stability,while decreasing soil intra-aggregate OC pools(EOCa,POCa,and MOCa).The sensitivity of intra-aggregate OC pools to soil conversion followed the order of EOCa＞MOCa＞POCa.The significantly higher C0values appeared in upland soil,while higher correspondingkvalues occurred in paddy soil.Random forest model and partial correlation analysis showed that EOCa and pH were the important nutrient and physicochemical factors,respectively,impactingkof C mineralization in paddy soil,while MOCa and CBH were identified as the key factors in upland soil.All in all,our study demonstrated that soil conversion from paddy to upland soil increased the percentage of macro-aggregates and aggregate stability,but decreased soil aggregate-associated C stock andkof C mineralization on a scale of 10 years.The outcomes of this study provided some valuable insights into the alterations of soil aggregates and potential C sequestration under soil conversion system in rice-producing areas.

Paddy soils make up the largest anthropogenic wetland on Earth.It plays crucial roles in ecosystem services and ecological functions(e.g.,food security,climate regulation,and biodiversity).Currently,the conversion from paddy soil to upland soil is seriously threatening food security,animal and microbial inhabitants,and the diversity of ecological landscape.Although this study provided some valuable insights into the alterations of soil aggregates and potential C sequestration under soil conversion system in rice-producing areas,there are limitations for this study because:i)information about above-ground crop yield/biomass should also be taken into account,and a larger scale should be considered and ii)the processes in soil with different properties(e.g.,acidic and alkaline soils) should be compared and environmental variables(e.g.,temperature and precipitation)should also be taken into consideration when assessing the effects of soil conversion on soil aggregate distributions and aggregate-associated C pools.

ACKNOWLEDGMENT

This work was jointly supported by the National Natural Science Foundation of China(No.41161047),and the Scientific Development and Technology Innovation Project of Xinjiang Production and Construction Group(XPCG)in China(No.2017BA041).

SUPPLEMENTARY MATERIAL

Supplementary material for this article can be found in the online version.