Estimating soil ammonium adsorption using pedotransfer functions in an irrigation district of the North China Plain

2021-03-17 09:06:02ChunyingWANGDefengWUXiaominMAOJingmingHOULeiWANGandYupingHAN
Pedosphere 2021年1期

Chunying WANGDefeng WUXiaomin MAOJingming HOULei WANG and Yuping HAN∗

1State Key Laboratory of Eco-hydraulics in Northwest Arid Region,Xi’an University of Technology,Xi’an 710048(China)

2North China University of Water Resources and Electric Power,Zhengzhou 450045(China)

3College of Water Conservancy and Civil Engineering,China Agricultural University,Beijing 100083(China)

4Henan Key Laboratory of Water Resources Conservation and Intensive Utilization in the Yellow River Basin,Zhengzhou 450046(China)

(Received May 9,2019;revised August 27,2019)

ABSTRACT Extensive use of chemical fertilizers in agriculture can induce high concentration of ammonium nitrogen(-N)in soil.Desorption and leaching of -N has led to pollution of natural waters.The adsorption of -N in soil plays an important role in the fate of the -N.Understanding the adsorption characteristics of -N is necessary to ascertain and predict its fate in the soil-water environment,and pedotransfer functions(PTFs)could be a convenient method for quantification of the adsorption parameters.Ammonium nitrogen adsorption capacity,isotherms,and their influencing factors were investigated for various soils in an irrigation district of the North China Plain.Fourteen agricultural soils with three types of texture(silt,silty loam,and sandy loam)were collected from topsoil to perform batch experiments.Silt and silty loam soils had higher -N adsorption capacity than sandy loam soils.Clay and silt contents significantly affected the adsorption capacity of-N in the different soils.The adsorption isotherms of -N in the 14 soils fit well using the Freundlich,Langmuir,and Temkin models.The models’adsorption parameters were significantly related to soil properties including clay,silt,and organic carbon contents and Fe2+and Fe3+ion concentrations in the groundwater.The PTFs that relate soil and groundwater properties to soil-N adsorption isotherms were derived using multiple regressions where the coefficients were predicted using the Bayesian method.The PTFs of the three adsorption isotherm models were successfully verified and could be useful tools to help predict -N adsorption at a regional scale in irrigation districts.

Key Words:adsorption capacity,adsorption isotherm,ammonium nitrogen,groundwater quality,soil properties

INTRODUCTION

Large amounts of N fertilizers,e.g.,urea and ammoniumbased fertilizers,are used in agriculture to increase crop yield.Extensive use of fertilizers causes high concentrations of ammonium nitrogenin soil.Ammonium nitrogen can be converted to nitrate nitrogenthrough the nitrification processes;however,is easier to be lost from soil owing to its negative charges.Nitrification inhibitors are used widely for reducing nitratethus increasingconcentrations in soils(Vogeleret al.,2011).Ammonium nitrogen is adsorbed in soils;however,desorption and leaching of-N from the soils could be increased because of the high-N concentrations in soils.Besides,intensive rainfall or high application of irrigation water could also enhance the leaching of-N(Zhanget al.,2003).When the adsorption capacity is satisfied or equilibrium is reached,surplus-N can also leach out and enter the surface water or groundwater(Zhanget al.,2007).Previous studies worldwide have shown that considerable-N leaching from soils can occur and cause serious agricultural non-point source pollution(Giles,2005;Zhaoet al.,2015).The presence of high nitrogen concentrations in groundwater can result in a decrease in dissolved oxygen in natural waters,be toxic to aquatic organisms,and pose great threats to human health(Giles,2005;Yanet al.,2011;Sieczka and Koda,2016).In the fate of-N,adsorption can retard and delay the migration of nitrogen,and to some extent can inhibit nitrogen loss.The-N adsorption in soil plays an important role in maintaining N nutrient and decreasing the risk of-N leaching.

There are many factors affecting the adsorption ofN.Soil properties including clay,silt,organic matter,cation exchange capacity(CEC),iron(Fe)and Mn oxides,as well as irrigation water quality,including Fe and Mn ions can influence the adsorption of-N(Wang and Alva,2000;Fernandoet al.,2005;Yanet al.,2013).Sieczka and Koda(2016)reported that the adsorption amount ofN in soils with high clay and high organic matter content was higher than that in sand.Phillips and Sheehan(2005)found that the adsorption of-N was associated with CEC in six different soils.Sunet al.(2015)reported that the adsorption amount of-N in soil with higher CEC was larger.Alshameriet al.(2017)conducted batch adsorption tests of-N on six natural clay minerals and reported that the higher-N adsorption capacity could be mainly attributed to the CEC.However,Sasakiet al.(2002)drew different conclusions and reported that CEC had no significant relationship with the adsorption amount ofN in soil solution.In addition,many studies have shown that Fe oxides or Fe ions at a certain concentration have a significant effect on nutrient adsorption,e.g.,phosphate and-N(Hsu,1964;Borggaard,1983;Chenet al.,2013).Hsu(1964)found that the rapid fixation of phosphate in soils was because of the native surface reactive Fe oxides already present in soils.Borggaard(1983)found that the decreases in the soil phosphate adsorption capacity were attributed to the removal of Fe oxides.Chenet al.(2013)found that the concentration of Fe2+ions had significant effects on the adsorption ofN in soils.Their study shows that a low concentration of Fe2+ion(0.7 mg L−1)would increase the maximum adsorption of-N by soil.However,excessive addition of Fe2+ions(15.0 mg L−1)will compete with-N for adsorption sites,thereby reducing the amount of-N adsorbed by the soil.Besides,they reported that Fe2+ions in groundwater may cause desorption and leaching of-N during irrigation.Shuman(1988)reported that the combined action of organic matter and inorganic matter(Fe oxides)in soil colloids is the main factor determining the amount of soil adsorption.Sasakiet al.(2002)reported that the Fe2+ion promotes the desorption of-N,as the Fe2+ion occupies the adsorption point of the cation in the soil.Both the soil properties and groundwater quality parameters act on-N adsorption in irrigation districts.As mentioned above,the influence of these factors on the adsorption of-N is still inconclusive and even contrasting.Therefore,adsorption of-N in various soils in irrigation areas needs further research and verification.

Equilibrium adsorption based on theoretical or empirical models have been well applied to describe the adsorption behavior of-N.Experimental data are commonly used to fit adsorption isotherms to establish the adsorption isotherm models.The most common adsorption isotherm models include the Langmuir model(Langmuir,1916),the Freundlich model(Freundlich,1906),the Redlich-Peterson model(Redlich and Peterson,1959),and the Temkin model(Temkin and Pyzhev,1940).Many studies have reported that some theoretical or empirical models are well applied to describe the adsorption behavior of-N.Sieczka and Koda(2016)found that the Redlich-Peterson model fit well with-N adsorption data.The results of Sunet al.(2015)and Alshameriet al.(2017)showed that the Langmuir adsorption isotherm can explain the-N adsorption process well.However,the adsorption isotherms of various soils in large areas are not readily available,and the acquisition of specific parameters of the models is time consuming and expensive based on batch experiments(Vogeleret al.,2011).Understanding the effect factors on the parameters of the adsorption isotherm models can help elaborate the-N adsorption process of various soils.

The use of pedotransfer functions(PTFs),which relate easily measurable soil properties to parameters of-N adsorption isotherm models,can provide a reasonable estimate of model parameters when no direct data are available and can be useful and efficient tools to predict-N adsorption in soils(McBratneyet al.,2002).For adsorption studies,PTFs were developed to estimate the adsorption of heavy metals(Anderson and Christensen,1988),pesticides(De Gerónimoet al.,2018),and nutrients such as phosphate(Borggaardet al.,2004),sodium(Rashidi and Seilsepour,2008),and potassium(Wanget al.,2016).However,few studies developed PTFs that relate soil properties to parameters of-N adsorption isotherm models.Vogeleret al.(2011)successfully constructed PTFs relating parameters of the Freundlich model with CEC and clay content for predicting-N adsorption in soils with high clay content,low organic matter content,and high-N concentration.A simple function was also derived for-N adsorption(Liet al.,2006),relating-N adsorption to the CEC of soils.The parameters of the adsorption isotherm models were related to the adsorption capacity and/or the adsorption strength.However,there are many factors,including soil properties and irrigation water quality,influencing soil-N adsorption as mentioned earlier.The relationship between these factors and parameters of the adsorption isotherm models remains unclear and needs to be further studied to develop PTFs for predicting-N adsorption in soils.

The North China Plain is one of China’s most important agricultural regions.High levels of crop production depend on irrigation and fertilizer application in this plain and groundwater is the primary source of irrigation.This region suffers serious groundwater decline and-N and-N pollution problems(Zhanget al.,2010;Yang,et al.,2015).In this study,14 soil samples with three types of soil texture(silt,silty loam,and sandy loam)were collected,and batch experiments were carried out in a typical irrigation area with shallow groundwater buried depth in the North China Plain.The main objectives of this study were:i)to determine the adsorption capacity and adsorption isotherms of-N in the soils with varying soil properties and fit the adsorption isotherms with the Freundlich,Langmuir,and Temkin models;ii)to assess the relationship between the factors,i.e.,soil properties and groundwater quality,and theN adsorption capacity and the parameters of-N adsorption isotherm models;and iii)to derive PTFs that relate the parameters of-N adsorption isotherm models to selected soil properties and groundwater quality for predicting the-N adsorption in various soils,and to verify the PTFs using the experimental data.

MATERIALS AND METHODS

Study site

People’s Victory Canal Irrigation District(35°00′-35°30′N,113°31′-114°25′E)is located in the lower reaches of the Yellow River in the North China Plain(Fig.1),and the irrigation area is 1 500 km2.The study area was the most famous Yellow River irrigation district;however,it is subjected to intensive groundwater-based irrigation owing to the higher price of surface water irrigation and the shallow groundwater buried depth in recent years.This irrigation district belongs to the temperate continental monsoon climate.The average annual temperature is 14.5°C,the maximum temperature is 41°C,and the minimum temperature is−16°C.The average annual potential evaporation is 1 865 mm and the average annual rainfall is 580 mm.The rainfall is mainly concentrated in June to September.The main crops are winter wheat-summer corn(44%),winter wheat-summer peanut(27%),and winter wheat-summer rice(11%)in the irrigation district.Other land uses including rural residential areas,urban areas,and water body areas constitute the remaining 18% of area in the irrigation district.

Data collection and analysis

Fourteen sampling sites with varying soil properties were selected to collect soil and groundwater samples in the irrigation district.Locations of the sampling sites in the study area are shown in Fig.1.The sampling sites include silt soil sites of Fengzhuang(Si-FEZ)and Niangniangmiao(Si-NNM);silty loam soil sites of Fuzhuang(SiL-FUZ),Daxinzhuang(SiL-DXZ),Wangguanying(SiL-WGY),Nanweizhuang(SiL-NWZ),Qianlicun(SiL-QLC),Lidazhao(SiL-LDZ),Chenzhuang(SiL-CZ),Sunxingcun(SiL-SXC),and Guandimiao(SiL-GDM);and sandy loam soil sites of Tapu(SaL-TP),Xishizhuang(SaL-XSZ),and Zhangbanzao(SaL-ZBZ).The land use at each sampling site in the irrigation area is shown in Table I.Soil samples and groundwater samples were collected at each sampling site on September 25,2017 after the final groundwater irrigation for corn.The soil samples were collected at a depth of 0-5 cm from the soil surface.The groundwater samples were collected from the irrigation well near the soil sampling sites and stored in 500-mL plastic bottles.The pH and electric conductivity(EC)of groundwater collected from the irrigation well were measured in the field using a U-50 multi-parameter water quality meter(HORIBA,Ltd.,Japan).After that,the water and soil samples were stored at 4°C and transported to the laboratory for testing.

Fig.1 Location of People’s Victory Canal Irrigation District in the North China Plain and distribution of 14 soil sampling sites with three types of soil texture,silt(Si),silty loam(SiL),and sandy loam(SaL).FEZ=Fengzhuang;NNM=Niangniangmiao;FUZ=Fuzhuang;DXZ=Daxinzhuang;WGY=Wangguanying;NWZ=Nanweizhuang;QLC=Qianlicun;LDZ=Lidazhao;CZ=Chenzhuang;SXC=Sunxingcun;GDM=Guandimiao;TP=Tapu;XSZ=Xishizhuang;ZBZ=Zhangbanzao.

A Malvern laser particle size analyzer(Malvern Panalytical Ltd.,UK)was used to analyze the particle size of the sampled soil.According to the USDA soil texture classification(Bormann,2010),soil textures in the irrigation area included silt,silty loam,and sandy loam as shown in Table I.The soils in the irrigation area are mainly silty loam,followed by sandy loam and silt(Fig.1).The soil organic carbon(SOC)content and total organic carbon(TOC)concentration in groundwater were colored by potassium dichromate oxidative and measured by spectrophotometry(Zhengzhou Gee-nong Technology Co.,Ltd.,China).The Fe2+and Fe3+in soils were extracted with diethylenetriaminepentaacetic acid(DTPA).Concentrations of Fe2+and Fe3+ions in the extracted solution from soil and groundwater were colored by phenanthroline and measured by spectrophotometry.The Mn2+in the soil was extracted with ammonium acetate.The concentrations of Mn2+ion in the extracted solution from soil and groundwater were colored by potassium periodate and measured by spectrophotometry.The physical and chemical properties of the soil and groundwater at each sampling site in the irrigation district are shown in Tables I and II,respectively.Soil clay and SOC were measured in this study.The CEC of soil is usually significantly correlated with clay and SOC(Aminiet al.,2005;Lakeet al.,2009).Therefore,CEC was not measured,instead clay content and SOC were selected as the potential factors related to-N.

Equilibrium adsorption test of NH+4-N

The soil was naturally air-dried and sieved through a 2-mm mesh to conduct the batch experiments.For the preparation of soils for testing,5 g soil was weighed,placed in 100-mL conical flasks,and 50 mL solution added with different concentrations of-N(5,50,100,150,200,250,300,and 500 mg L−1).Three replicates were performed for each experiment.The conical flasks were sealed with plastic wrap,placed in an air-constant oscillator,and shaken for 2 h at room temperature(25°C).After shaking,15 mL of solution was placed in a centrifuge tube and centrifuged at 2 220×gfor 30 min.At the end of the centrifugation,the supernatant was removed quickly and filtered through 0.2-µm membrane filters.The-N concentration in filtered liquid was colored using Nessler’s reagent and measured using a spectrophotometer.Finally,the following formula was used to calculate the amount of soil adsorbed-N:

whereqeis the adsorption amount in equilibrium(mg kg−1),ρ0is the initial mass concentration of-N(mg L−1),ρeis the mass concentration of-N in solution in equilibrium(mg L−1),Vis the total volume of adsorption solution(L),andmis the weight of soil(kg).

Adsorption isotherm models

Three commonly used adsorption isotherm models,i.e.,Freundlich,Langmuir,and Temkin,were selected to fit the experimental data of-N equilibrium adsorption.TheFreundlich model was expressed as:

TABLE ISoil physical and chemical properties of the selected sites in the People’s Victory Canal Irrigation District of the North China Plain

TABLE IIGroundwater quality of the selected sites in the People’s Victory Canal Irrigation District of the North China Plain

whereCsolidis the-N amount adsorbed by soil in equilibrium(mg kg−1),Csolutionis the concentration of-N in the solution in equilibrium(mg L−1),kfis the parameter related to adsorption capacity and adsorption affinity,andnis the parameter related to the adsorption intensity.

The Langmuir model was expressed as:wherekmis the theoretical adsorption capacity of-N in the adsorption process(mg kg−1)andkais the adsorption equilibrium constant in the adsorption equation,reflecting the adsorption affinity(L mg−1).

The Temkin model was expressed as:

whereTais the parameter related to the heat of sorption and adsorption affinity andktis the parameter related to the heat of sorption(J mol−1).

The adsorption isotherm models were used to fit the measured data from the batch experiments and the model performance was evaluated using the coefficient of determination(R2)and the root mean square error(RMSE).

Derivation of PTFs

The PTFs that relate soil properties and/or groundwater quality to parameters of-N adsorption isotherm models were developed for predicting soil-N adsorption.The parameters of each isotherm model were obtained using the measured data.To determine potential factors(i.e.,clay,silt,sand,SOC,and Fe2+,Fe3+,and Mn2+contents of soil,and Fe2+,Fe3+,and Mn2+ion concentrations of groundwater)that relate to the parameters of-N adsorption isotherm models,the Pearson correlation analysis was performed.Multiple regression analysis was performed to develop the relationship between parameters of adsorption isotherm models and soil properties and groundwater quality,thereby deriving the PTFs to predict-N adsorption.Bayesian calibration was performed to derive the coefficients of multiple regression equations with the statistical package R(3.5.0)(http://www.rproject.org),particularly the R2jags package(https://cran.rproject.org/web/packages/R2jags/index.html).After performing the Pearson correlation analysis,if there was a positive correlation,the calibrated coefficient of the multiple regression equation was considered as positive during the Bayesian calibration,andvice versa.The performance of the PTFs of the adsorption isotherm models were evaluated using theR2and RMSE values.

RESULTS AND DISCUSSION

Adsorption of NH+4-N in different soils

Fig.2 Adsorbed -N amount over the initial -N concentration range for soils with different types of textures,silt(a),sandy loam(b),and silty loam(c and d),at the selected sites.Vertical bars are standard deviations of means(n=3).See Table I for the description of sites.

TABLE IIIAdsorption capacity of -N for soils with different types of texture at the selected sites

TABLE IIIAdsorption capacity of -N for soils with different types of texture at the selected sites

a)See Table I for the description of sites.

Soils Sitea) images/BZ_151_1254_1029_1325_1080.png-N adsorption capacity mg kg−1 Silt soils Si-FEZ 701 Si-NNM 1 000 Silty loam soils SiL-WGY 627 SiL-DXZ 679 SiL-FUZ 617 SiL-NWZ 591 SiL-QLC 732 SiL-CZ 653 SiL-SXC 691 SiL-GDM 875 SiL-LDZ 491 Sandy loam soils SaL-TP 557 SaL-XSZ 540 SaL-ZBZ 362

Fitting analysis of Freundlich,Langmuir,and Temkin models

Three adsorption isotherm models,i.e.,Freundlich,Langmuir,and Temkin,were selected to fit the experimental data.Table IV shows the parameters and performance of the three models for predicting the-N adsorption in the studied soils.The comparison between soil adsorbed-N amounts measured and predicted by the adsorption isotherm models is shown in Fig.3.All the three models were fitted well to the experimental data of most studied soils.The performance of the Freundlich model was the best compared to the Langmuir and Temkin models(Table IV,Fig.3).TheR2values of the Freundlich model for all soils were greater than 0.90,except for the soil of SiL-QLC(R2=0.87).The performance of the Temkin model was also goodwithR2values greater than 0.80 for all soils.However,it should be noted that at low concentration of initial-N solution(5 mg L−1),the prediction of the Temkin model underestimated the adsorbed-N amount.TheR2values of the Langmuir model for all the sites were above 0.80,except for the sampling sites of SiL-WGY(R2=0.68)and SiL-CZ(R2=0.57).Overall,theR2values of the three models were higher than 0.50 for all soils,indicating that the three models were suitable for predicting soil-N adsorption in this study.

TABLE IVFitting parameters and performancesa)of the -N adsorption isotherm models for soils at the selected sites

TABLE IVFitting parameters and performancesa)of the -N adsorption isotherm models for soils at the selected sites

a)kf=the parameter related to adsorption capacity and adsorption affinity;n=the parameter related to the adsorption intensity;km=the theoretical adsorption capacity of -N in the adsorption process;ka=the adsorption equilibrium constant in the adsorption equation,reflecting the adsorption affinity;Ta=the parameter related to the heat of sorption and adsorption affinity;kt=the parameter related to the heat of sorption;RMSE=root mean square error;R2=coefficient of determination.b)See Table I for the description of sites.

Siteb) Freundlich Langmuir Temkin kf 1/n R2 RMSE km ka R2 RMSE kt Ta R2 RMSE mg kg−1 L mg−1 J mol−1 Si-FEZ 22.99 0.68 0.97 99.4 909 0.017 0.97 61.9 155.6 −141.2 0.95 59.3 Si-NNM 22.39 0.74 0.98 91.4 1 250 0.013 0.97 210.0 198.3 −198.8 0.94 78.7 SiL-WGY 6.74 0.90 0.95 134.2 1 250 0.005 0.68 56.3 163.6 −247.9 0.96 50.9 SiL-DXZ 22.89 0.67 0.97 84.3 909 0.016 0.96 46.8 147.6 −135.2 0.93 66.7 SiL-FUZ 15.28 0.72 0.97 221.1 909 0.011 0.94 208.0 142.7 −169.1 0.92 169.3 SiL-NWZ 15.08 0.71 0.94 90.2 833 0.011 0.86 62.4 131.7 −151.6 0.90 70.8 SiL-QLC 16.56 0.72 0.87 118.9 1 000 0.015 0.98 88.8 164.5 −189.8 0.80 134.3 SiL-CZ 9.10 0.84 0.92 148.1 1 111 0.007 0.57 85.9 157.8 −217.9 0.88 87.8 SiL-SXC 17.31 0.73 0.97 200.1 1 000 0.012 0.93 49.5 161.1 −181.9 0.92 75.7 SiL-GDM 18.03 0.73 0.97 125.8 909 0.015 0.95 58.4 157.3 −166.7 0.93 71.2 SiL-LDZ 20.38 0.63 0.92 100.1 556 0.027 0.96 41.3 107.0 −84.2 0.95 39.1 SaL-TP 12.47 0.73 0.98 77.0 833 0.010 0.89 71.5 129.6 −171.6 0.85 81.4 SaL-XSZ 10.86 0.76 0.97 69.5 833 0.009 0.93 50.1 126.6 −170.5 0.93 54.6 SaL-ZBZ 5.98 0.79 0.93 165.5 455 0.012 0.80 252.0 90.2 −127.5 0.81 178.3 Mean 0.95 123.3 0.89 95.9 0.91 87.0

Fig.3 Comparison between soil adsorbed -N measured and predicted by the Freundlich(a),Langmuir(b),and Temkin(c)adsorption isotherm models and by the pedotransfer functions(PTFs)of these models.

For the parameters of the Freundlich model(Table IV),kfvalues of silt soils were 22-23,and thekfvalues of all silty loam and sandy loam soils were 5-23.Thekfvalues were the largest in silt soils,followed by silty loam soils and sandy loam soils.The 1/nvalues of silt soils were 0.68-0.74,while the 1/nvalues of silty loam and sandy loam soils were 0.63-0.90.For the parameters of the Langmuir model(Table IV),except for the soil of SiL-LDZ,thekmvalues of all silt and silty loam soils were 800-1 250 mg kg−1,and thekmvalues of all sandy loam soils were 450-850 mg kg−1.Thekmvalues of the silty loam soil of SiL-LDZ(556 mg kg−1)were smaller than those of the other silty loam soils.Thekavalues of all soils were 0.005-0.027 L mg−1.Among them,thekavalue of the soil of SiL-LDZ was the largest.The parameters of the Temkin model are shown in Table IV.Thektvalues ranged from 90 to 200 J mol−1.Thektvalues were 150-200,100-170,and 90-130 J mol−1for silt,silty loam,and sandy loam soils,respectively.The range ofTawas from−250 to−80.TheTavalues ranged from−200 to−140,from−250 to−80,and from−180 to−120 for silt,silt loam,and sandy loam soils,respectively.Results showed that the values of the adsorption isotherm models’ parameters might be related to the soil properties,such as,soil particle content.

Correlations of NH+4-N adsorption with soil properties and groundwater quality

Correlations between soil properties and actual adsorption capacity.Relationships between soil clay,silt,and sand contents and the actual adsorption capacity(i.e.,measured adsorption capacity)of-N are shown in Table V.Soil-N adsorption capacity was positively correlated with soil clay and silt contents but negatively correlated with soil sand content.As soil clay and silt contents increase,the-N adsorption capacity becomes larger,and the greater the sand content in soils,the smaller the-N adsorption capacity of soil.The Pearson correlation coefficient(r)between soil clay content and-N adsorption capacity was0.79,which was greater than that between soil silt content and the adsorption capacity(r=0.71).Although the clay content of soil was much lower than the silt content,the-N adsorption capacity of clay was much greater than that of the silt.The more the sand content in soils,the less the adsorption capacity,as indicated by their negative correlation(r=−0.72).For all the studied soils,the higher the soil clay and silt contents,the higher the-N adsorption capacity.The correlation between SOC content and-N adsorption capacity was analyzed as shown in Table V.The SOC content is also an important factor influencing-N adsorption.The SOC content was positively correlated with-N adsorption capacity for the studied soils withr=0.61(Table V).This indicated that the more the SOC content,the higher the-N adsorption capacity of soil.

TABLE VPearson correlation coefficients among adsorption capacity of -N,parameters of adsorption isotherm models(Freundlich,Langmuir,and Temkin),soil properties,and groundwater quality

TABLE VPearson correlation coefficients among adsorption capacity of -N,parameters of adsorption isotherm models(Freundlich,Langmuir,and Temkin),soil properties,and groundwater quality

*,**Significant at P<0.05 and P<0.01,repectively.a)SOC=soil organic carbon;TOC=total organic carbon;ADC=actual adsorption capacity;kf=the parameter related to adsorption capacity and adsorption affinity;n=the parameter related to the adsorption intensity;km=the theoretical adsorption capacity of -N in the adsorption process;ka=the adsorption equilibrium constant in the adsorption equation,reflecting the adsorption affinity;Ta=the parameter related to the heat of sorption and adsorption affinity;kt=the parameter related to the heat of sorption.b)Without data from the silty loam soil of Sunxingcun(SiL-SXC).c)Without data from the the silty loam soil of Lidazhao(SiL-LDZ).

Variablea) ADC Freundlich Langmuir Temkin kf 1/n km ka kt Ta Soil property Clay 0.79** 0.50,0.68**b) −0.16 0.44 0.055 0.60* −0.19 Silt 0.71** 0.70** −0.29 0.57* 0.30 0.72** −0.16 Sand −0.72** −0.71** 0.29 −0.58* −0.29 −0.73** 0.16 SOC 0.61* 0.083 0.35 0.70** −0.23 0.71** −0.60*Fe2+ 0.01 −0.32 0.42 0.23 −0.27 0.11 −0.36 Fe3+ 0.24 0.11 0.20 0.46 −0.21 0.41 −0.32 Mn2+ 0.33 0.10 0.06 0.19 0.11 0.25 0.15 Groundwater quality TOC 0.4 0.024 0.11 0.28 0.036 0.37 −0.32 Fe2+ 0.01 −0.52 0.75** 0.45 −0.42,−0.78**c) 0.19 −0.60*Fe3+ −0.32 −0.79** 0.88** 0.28 −0.66**,−0.83**c) −0.048 −0.60*Mn2+ −0.10 0.29 −0.39 −0.32 0.76**,0.26c) −0.23 0.48 ADC 1.00 0.56* −0.062 0.77** −0.051 0.92** −0.45 Adsorption parameter kf 1.00 −0.81** 0.093 0.69** 0.39 0.41 1/n 1.00 0.45 −0.83** 0.18 −0.80**km 1.00 −0.49 0.93** −0.86**ka 1.00 −0.23 0.80**kt 1.00 −0.67**Ta 1.00

Correlations between soil properties and parameters of adsorption isotherm models.Correlations between parameters of the three adsorption isotherm models(i.e.,Freundlich,Langmuir,and Temkin)and soil properties were analyzed(Table V).For the Freundlich model,correlation analysis was performed onkfand soil particle contents.Results showed thatkfwas weakly related to soil clay content.However,with the exclusion of the SiL-LDZ site,a significant positive correlation was found betweenkfand clay content.Thekfvalue was positively correlated(r=0.70)with soil silt content.Thekfwas negatively correlated(r=−0.71)with soil sand content.Therefore,the more the clay and/or silt in soils,the larger thekfvalue,and the greater the soil-N adsorption capacity.Correlation analysis was also performed amongkf,1/n,and the actual adsorption capacity(Table V).Thekfvalue was positively correlated with the actual adsorption capacity of-N,indicatingkfwas a parameter related with adsorption capacity,which was affected by soil particle contents.It demonstrated that there was a relationship betweenkfand soil particle contents.Thekfvalue was negatively correlated with 1/nwith anrof−0.81.This might indicate that parameterskfand 1/nin the Freundlich model were inherently related.However,results of correlation analysis revealed no correlation between 1/nand soil particle contents.It contradicted the results of Vogeleret al.(2011),who reporte that the parameter 1/ncould be predicted with soil clay content and CEC.Besides,correlation analysis showed that bothkfand 1/nin the Freundlich model were not significantly related to SOC content.

Correlations between parameters of the Langmuir model and soil properties are shown in Table V.Results showed thatkmis positively correlated with soil clay content,but the correlation coefficient was low(r=0.44).The parameterkmwas also positively correlated with soil silt content(r=0.57).The parameterkmwas negatively correlated with soil sand content(r=−0.58).In addition,kmwas positively correlated with SOC content(r=0.70).This indicated that the more the soil clay,silt,and SOC contents,the more favorable the-N adsorption.Correlation analysis was performed betweenkmand the actual adsorption capacity.The relationship betweenkmand the actual adsorption capacity is shown in Table V.Results showed that the theoretical adsorption capacity(km)was positively correlated with the actual adsorption capacity,with anrof 0.77.This indicated that the Langmuir model fitted well with the experimental data,andkmwas a parameter that reflects-N adsorption capacity of soil.Parameterkaof the Langmuir model was related to adsorption affinity.The larger thekavalue,the greater the adsorption affinity.There was no correlation betweenkaand soil particle contents.Correlations between parameterskaandkfand betweenkaand 1/nare shown in Table V.The parameterkawas positively correlated withkf(r=0.69)andkawas negatively correlated with 1/n.This suggested that the physical meaning ofkfandnin the Freundlich model was similar toka,both being related to the-N adsorption affinity of soil.The negative correlation betweenkaand 1/nshowed that parameternwas also related to the adsorption affinity.

Correlations between parameters of the Temkin model and soil properties were analyzed,as shown in Table V.Results revealed thatktwas positively correlated with clay content with anrof 0.60.The parameterktwas positively correlated(r=0.72)with soil silt content.The parameterktwas negatively correlated(r=−0.73)with soil sand content.Therefore,the more the clay and silt particles in soils,the larger thektand the greater the soil-N adsorption capacity.The parameterktwas positively correlated with SOC with anrof 0.71.This indicated that the larger the SOC content,the larger thektvalue.The correlation betweenktand the adsorption capacity was analyzed.Results showed thatktwas positively correlated(r=0.92)with the actual adsorption capacity.In addition,ktwas positively correlated(r=0.93)withkm.The higher the adsorption capacity of-N,the larger thektvalue.The parameterktwas positively related to the soil clay,silt,and SOC contents(Table V).These findings were consistent withktas a parameter related to the adsorption capacity of-N.In addition,this phenomenon was consistent with the results of the study of Boatman and Murray(1982),who reported that SOC was positively correlated with the adsorption of-N in soils.Correlation analysis betweenTaand SOC showed that there was a negative correlation between them with anrof−0.60.Because parameterTais related to the adsorption affinity,it indicates that the higher the SOC content,the lower the adsorption affinity.The parameterTawas positively correlated withkawith anrof 0.80.This was consistent with the physical meaning ofTaandkain the models being similar,both being related to the adsorption affinity of-N of soil.The larger theTa,the greater the adsorption affinity.ParameterTawas not related to the soil particle content.This is similar to the relationship betweenkaof the Langmuir model and the soil particle content.However,results indicated thatTawas related to the SOC content,whilekahad no significant correlation with SOC content(Table V).

Correlations between groundwater quality and parameters of adsorption isotherm models.To further explore the influencing factors of the parameters of the adsorption isotherm models,the correlation between model parameters and the groundwater quality was analyzed,as shown in Table V.For the Freundlich model,kfwas negatively correlated with Fe2+ion(r=−0.52)and Fe3+ion(r=−0.79)concentrations in groundwater.This indicated that the higher the concentrations of Fe2+and Fe3+ions in the groundwater,the smaller the-N adsorption capacity and affinity in soils.The correlation between 1/nand groundwater quality was also analyzed.The parameter 1/nwas positively correlated with Fe2+ion(r=0.75)and Fe3+ion(r=0.88)concentrations in the groundwater.This indicated that the greater the concentration of Fe ions in groundwater,the smaller thenvalue and the smaller the adsorption affinity.

For the Langmuir model,correlation analysis showed thatkawas negatively correlated with Fe3+ion(r=−0.66),and Fe2+ion(r=−0.42)concentrations in groundwater.The Fe2+and Fe3+ion concentrations were important factors toka.When the SiL-LDZ study site was excluded from the correlation analysis,the correlation betweenkaand Fe ion concentration became stronger,with anrof−0.78 and−0.83 betweenkaand Fe2+ion and betweenkaand Fe3+ion,respectively,as shown in Table V.Moreover,correlation analysis betweenkaand Mn2+ion concentration in groundwater showed thatkawas positively correlated with Mn2+ion concentration for all sites,with anrof 0.76.However,when the SiL-LDZ site was excluded,the correlation betweenkaand Mn2+ion concentration became weaker,with anrof 0.26 betweenkaand Mn2+.The concentration of Mn2+ion in groundwater was almost three times higher at the SiL-LDZ site than the other sites,which might affect-N adsorption.Overall,Fe2+and Fe3+ion concentrations in groundwater andkawere closely related for the studied soils.Althoughkawas positively correlated withkf,kfhad no clear correlation with Fe2+and Fe3+ion concentrations in groundwater.

For the Temkin model,as shown in Table V,Tawas negatively correlated with Fe2+and Fe3+ion concentrations in groundwater,with anrof−0.60 for both.This suggested thatTain the Temkin model was related with the Fe ion concentrations in groundwater.As mentioned before,kain the Langmuir model was also negatively correlated with Fe ions.It could be further concluded thatTahad the similar physical meaning aska,both being related to soil-N adsorption affinity.Overall,the concentrations of Fe ions were important factors related to the-N adsorption in soils.The larger the Fe ion concentrations in groundwater,the smaller the-N adsorption affinity of soil.

In this study,results indicated that Fe2+and Fe3+ion concentrations in groundwater were important factors related to the adsorption process of-N in soils.As 1/n,ka,andTaare parameters related to the adsorption affinity,andkfis a parameter related to adsorption capacity and affinity,there are negative relationships between Fe ion concentrations in groundwater with adsorption affinity and adsorption capacity.The negative relationships might be because the Fe ions in groundwater formed Fe oxides in soils after groundwater irrigation(Chenet al.,2013).Another possible reason is that the higher Fe concentration in groundwater could be due to the higher active amorphous Fe oxides content in soil.A previous study reported that active Fe oxides could release Fe ions to the soil solution and leach into groundwater during precipitation and irrigation events(Kumaret al.,1981).Iron oxides might have lower adsorption capacity and affinity compared with clay,silt,organic carbon,and Mn oxides in soils(Shuman,1988).Consequently,when Fe oxides were increased or higher in soils,soil-N adsorption affinity and capacity could be decreased or lower.At the SiL-LDZ site,the higher Mn2+ion concentration in groundwater showed a positive relationship with the-N adsorption affinity.The possible reason could be that Mn2+ion in groundwater can form Mn oxides in soils after irrigation.Another possible reason is that high Mn2+ion concentration in groundwater might indicate high concentration of active Mn oxides in soil.The Mn oxides in soil have much stronger adsorption affinity than Fe oxides,and can thus significantly enhance the adsorption affinity(Shuman,1988).According to the study of Shuman(1988),Mn oxides can promote the adsorption of cations in soils.Therefore,the adsorption affinity of-N in the soil of SiL-LDZ was high when Mn2+ion concentration in groundwater or Mn oxides in soils reached a high value.It should be noted that the available Fe2+and Fe3+extracted with DTPA and available Mn2+extracted with ammonium acetate from the studied soils were not significantly related to parameters of the adsorption isotherm models.Other extracting solutions should be used for selective dissolution of Fe and Mn oxides in soils and the effects of different forms of Fe and Mn oxides on soil-N adsorption should be evaluated in future studies(Lindsay and Norvell,1978;Chao and Zhou,1983).

Pedotransfer functions for NH+4-N adsorption

For the parameters of the Freundlich model,it can be seen from the analysis in the previous section thatkfwas positively correlated with clay and silt contents,andkfwas negatively correlated with Fe2+and Fe3+concentrations in groundwater(Table V).In order to obtain PTFs of the Freundlich model for estimating-N adsorption in soils,multiple regression analysis was performed to explain the change inkfusing the clay and silt contents of soil,and Fe2+and Fe3+ion concentrations of groundwater.The parameter 1/nwas positively correlated with the Fe2+and Fe3+ion concentrations of groundwater(Table V).Multiple regression analysis was performed to describe the change in 1/nusing the Fe2+and Fe3+ion concentrations of groundwater.Thus,the PTFs of the Freundlich model can be described as:

whereb0-b4are the fitting coefficients,clay and silt are the clay and silt contents of soil,respectively(%),and Fe2+and Fe3+are the Fe2+and Fe3+concentrations of groundwater,respectively(mg L−1).

For the parameters of the Langmuir model,as mentioned in the previous section,kmwas positively correlated with soil clay,silt,and SOC contents(Table V).In order to obtain PTFs of the Langmuir model for estimating soil-N adsorption,multiple regression analysis was performed to predict parameterkmusing the soil clay,silt,and SOC contents.The parameterkawas negatively correlated with Fe2+and Fe3+ion concentrations of groundwater(Table V).Multiple regression analysis was performed to predict the parameterkausing Fe2+and Fe3+ion concentrations of groundwater.The PTFs of the Langmuir model can be written as:

whereb5is the fitting coefficient.

For the parameters of the Temkin model,as mentioned in the previous section,the parameterktwas positively correlated with the soil clay,silt,and SOC contents(Table V).In order to obtain PTFs of the Temkin model for estimating the adsorption of-N in soils,multiple regression analysis was performed to predictktusing the soil clay,silt,and SOC contents.The parameterTawas negatively correlated with Fe2+and Fe3+ion concentrations of groundwater,and SOC content(Table V).Multiple regression analysis was performed to predictTausing Fe2+and Fe3+ion concentrations of groundwater and SOC content.The PTFs of the Temkin model can be written as:

Eleven soils,except SiL-DXZ,SiL-QLC,and SiL-XSZ,were used to calibrate the coefficients of the PTFs using the Bayesian method,and the values of the coefficients are shown in Table VI.The soils of SiL-DXZ,SiL-QLC,and SiL-XSZ were used to validate the performance of the PTFs.

Verification of PTFs using experimental data

Comparison of the parameters predicted with the PTFs and the parameters of adsorption isotherm models fitted with experimental data are shown in Fig.4.The parameters predicted with the PTFs showed similar trends as parameters of the adsorption isotherm models fitted with experimental data for all studied soils in the irrigation district.Some parameters predicted with the PTFs had obvious bias from the parameter values of the three models.At the SiL-LDZ site,the 1/nvalue of the Freundlich PTF was slightly lower than that of Freundlich model.In addition,kavalue of Langmuir PTFs was much higher than that of the Langmuir model.The results of the comparison indicated that PTF of the Freundlich model showed higherR2values and betterperformance in predicting the model parameters than PTFs of the other two models.

TABLE VIValues of the multiple regression coefficients(b0-b5)based on pedotransfer functions for parametersa)of adsorption isotherm models(Freundlich,Langmuir,and Temkin)

Fig.4 Comparison of parameters predicted from pedotransfer functions(PTFs)of the adsorption isotherm models and the models(Freundlich(a and b),Langmuir(c and d),and Temkin(e and f))fitted with experimental data at the selected sites.See Table I for the description of sites.kf=the parameter related to adsorption capacity and adsorption affinity;n=the parameter related to the adsorption intensity;km=the theoretical adsorption capacity of -N in the adsorption process;ka=the adsorption equilibrium constant in the adsorption equation,reflecting the adsorption affinity;Ta=the parameter related to the heat of sorption and adsorption affinity;kt=the parameter related to the heat of sorption.

The PTFs of the three adsorption isotherm models(Freundlich,Langmuir,and Temkin)were used to fit the-N adsorption data of the 14 studied soils.The performances of the PTFs of the three models for predicting soil adsorbed-N are shown in Fig.3 and Table VII.In addition,the performance of the PTFs of models was compared with that of the models fitted with experimental data(Tables IV and VII).The soil adsorbed-N predicted from the PTFs was close to the measured data(Fig.3).For the studied soils,theR2and RMSE values of the Freundlich PTF were close to that of the Freundlich model in predicting soil adsorbed-N,except for the SiL-LDZ,SaL-TP,and SaL-ZBZ sites(Tables IV and VII).Overall,theR2values of the Freundlich PTF were from 0.53-0.97,while theR2values of the Freundlich model were from 0.87-0.98.At the SiL-LDZ site,the RMSE values of PTF of the Freundlich model weremuch higher than that of the Freundlich model.It could be due to effect of Mn2+ion in irrigation groundwater or Mn oxides in the soil at the SiL-LDZ site.At the SaL-TP and SaL-ZBZ sites,PTF of the Freundlich model had lowerR2values(0.53 and 0.60)than the Freundlich model(0.98 and 0.93).It could be due to the high soil sand content at the SaL-TP and SaL-ZBZ sites,and the low SOC content.This is consistent with results reported by Vogeleret al.(2011)that PTF of the Freundlich model showed lower performance for soils with high sand content and low SOC content.TheR2and RMSE values of the Langmuir PTF were similar and even better than those of the Langmuir model.TheR2values of the Langmuir PTF were from 0.71 to 0.98,whileR2values of the Langmuir model were from 0.57-0.98.At the SiL-CZ site,the the Langmuir model showed a lowerR2value of 0.57.For the soil with high sand content at the SaL-TP site,PTF of the Langmuir model showed slightly lower performance in fitting measured data than the PTF of the Freundlich model,indicated by lowerR2and RMSE values.TheR2and RMSE values of the Temkin PTF were close to those of the Temkin model,except for the SaL-TP site.At the SaL-TP site,the Temkin PTF showed lowerR2and RMSE values(R2=0.55 and RMSE=85.6)than the Temkin model in fitting the measured data(R2=0.81 and RMSE=81.4).Overall,the PTFs of the three adsorption isotherm models(Langmuir,Freundlich,and Temkin)can be applied to predict-N adsorption of soil.However,it should be noted that at low concentration,both PTF of the Temkin model and the Temkin model could not predict the-N adsorption well.

TABLE VIIModel performance for the pedotransfer functions(PTFs)of the adsorption isotherm models(Freundlich Langmuir,and Temkin)

Previous studies by Vogeleret al.(2011)and Liet al.(2006)developed PTFs for predicting Freundlich model parameters for soil-N adsorption using clay and/or CEC.However,results of this study indicated that along with clay,silt,and SOC in soil,Fe ion concentrations in groundwater were also important factors related to parameters of the Freundlich,Langmuir,and Temkin models for soil-N adsorption in the studied soils of the irrigation district of the North China Plain.Thus,all the factors of clay,silt,and SOC in soil and Fe ion concentrations in groundwater were considered when developing PTFs of soil-N adsorption isotherm models.The PTFs relating both the soil properties and groundwater quality to-N adsorption isotherm models were successfully verified in this study.In future studies,the PTFs presented in this study should be applied and tested on other irrigation regions with similar soil physical and chemical properties,buried depth and quality of irrigation groundwater,and management operations.Besides,the relationship between Fe and Mn ion concentrations in irrigation groundwater and parameters of soil-N adsorption isotherm models indicated that Fe and Mn oxides in soils affected soil-N adsorption.In future studies,the forms of Fe and Mn oxides affecting-N adsorption and suitable extraction solutions should be clarified to develop PTFs related to soil physical and chemical properties with soil-N adsorption.

CONCLUSIONS

The PTFs developed in this study improved previous simplistic PTFs taking the factors of soil clay and CEC into account which were not suitable to predict adsorption affinity in this study.The PTFs demonstrated the relations of soil properties and groundwater quality with-N adsorption in soil at a regional scale in irrigation districts.The PTFs developed could be applied to other regions with similar soil types,groundwater condition,and management practices.However,other influencing factors of soil-N adsorption such as clay mineralogy,pH,or Mn2+ion and Mn oxides were not considered in this study.The estimated coefficients of the PTFs should be considered as site and/or soil-specific coefficients and could vary temporally and spatially.Nonetheless,the PTFs are useful in regional scale N cycle simulation models in which-N adsorption is important.It should be noted that the Temkin model and its PTF should not be applied for soils with low-N content.When Mn2+ion concentration of groundwater and Mn oxides of soil are high,the effects of Mn2+ion in irrigation groundwater and Mn oxides in soil on soil-N adsorption deserve further attention.

ACKNOWLEDGEMENTS

This work was supported by the Open Fund of State Key Laboratory of Eco-hydraulics in Northwest Arid Region of China(No.2019 KFKT-8),the National Natural Science Foundation of China(Nos.51609084,51679089,and 41807161),and the“948”Program of the Ministry of Water Resources of China(No.201328).The authors also sincerely thank Dr.Wei HU at the New Zealand Institute for Plant & Food Research Limited for his valuable suggestions on the manuscript.


登录APP查看全文