Phosphorus risk in an intensive agricultural area in a mid-high latitude region of China

Catena 127 (2015) 46–55

Contents lists available at ScienceDirect

Catena journal homepage: www.elsevier.com/locate/catena

Phosphorus risk in an intensive agricultural area in a mid-high latitude region of China Hezhen Lou a, Shengtian Yang a, Changsen Zhao a,⁎, Qiuwen Zhou a,d, Juan Bai c, Fanghua Hao b, Linna Wu a,e a State Key Laboratory of Remote Sensing Science, School of Geography, Beijing Normal University, Beijing Key Laboratory for Remote Sensing of Environment and Digital Cities, Beijing 100875, China b State Key Joint Laboratory of Environmental Simulation and Pollution Control, School of Environment, Beijing 100875, China c College of Urban and Environment Sciences, Shanxi Normal University, Linfen 041000, China d School of Geography and Environment Science, Guizhou Normal University, Guiyang 550001, China e College of Resource and Environment Engineering, Guizhou University, Guizhou, Guiyang 550025, China

a r t i c l e

i n f o

Article history: Received 28 March 2014 Received in revised form 21 October 2014 Accepted 4 December 2014 Available online xxxx Keywords: Phosphorus risk Remote sensing Spatial analysis Sanjiang Plain

a b s t r a c t The content and status of phosphorus (P) in agricultural soil have become critical environmental concerns because of the risk of excessive P reaching adjacent water bodies through leaching or runoff. The Sanjiang Plain, an important commercial grain base, is the largest swampy low plain in a mid-high-latitude region in China. The increasing risk of P leaching threatens the area's ecology and agricultural security, particularly because there have been more than 60 years of build-up of P in the soil as a result of intensive agricultural development. A remote sensing-driven model was used in this study to simulate the total phosphorus in the Sanjiang Plain soil from 2000 to 2010. A grid-random sampling method was employed to test the validity of the model in a typical in situ experimental area. Meanwhile, spatial analysis methods were used to analyze model outputs. The results show a significantly increasing trend in the area with higher concentrations of P in the Sanjiang Plain, phosphorus accumulation in the soil and an increasing area at risk of phosphorus leaching; among all land use types, cultivated land and water bodies were observed to have a higher phosphorus risk. We conclude that the growth in farmland converted from forest or wetland and the use of excessive fertilization to maintain high crop productivity are the two principal factors resulting in the growing concentration of soil P. The risk of higher phosphorus in the study area will pose a serious threat to the quality of both surface and ground water, which has important ramifications for future agricultural management and non-point source control in this agricultural area of the mid-high latitude region. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Phosphorus geochemistry is of agronomic and environmental importance. P, as one of the primary nutritional elements, is essential for terrestrial and aquatic plants and animals (Yang et al., 2012). Due to its low availability, P is also the most limiting factor for crop yield (Cheng et al., 2011; Qiao et al., 2013). In many regions worldwide, to match increasing food demand from population increases, mined rock P or anthropogenic P additions to farmlands have been necessary to maintain grain yields (Duan et al., 2011; Townsend and Porder, 2012). Longterm P fertilization causes significant accumulation of soil P, which represents a growing threat to the aquatic environment (Bolland et al., 1996; Zhuang et al., 2007). Runoff is the primary way that P migrates from soil to water bodies, but vertical P leaching draws more attention because it is another important vehicle for P loss from soil with high P or high ⁎ Corresponding author at: School of Geography, Beijing Normal University, #19, XinJieKouWai St., Beijing 100875, China. Tel.: +86 10 58805586. E-mail address: [email protected] (C. Zhao).

http://dx.doi.org/10.1016/j.catena.2014.12.013 0341-8162/© 2014 Elsevier B.V. All rights reserved.

organic matter (Djodjic et al., 2004; Sharpley et al., 1994). The build-up of P, and the potential for P leaching, represent a critical risk to water quality and eco-environmental safety. The mid-high latitude region in China is the core crop-production area in Northeast Asia. A very smooth slope grade, advantageous water condition and a natural supply of soil organic matter provide the region with great potential for reclamation (Richardson and Ho, 2003). In response to an increased demand for cropland, this region experienced more than 60 years of intensive agricultural activity; widespread land reclamation occurred during this period and its large area of natural wetland was diminished (Ouyang et al., 2013; Song et al., 2008). Sanjiang Plain is the largest swampy low plain in China, and it also represents the most important base for grain commodities in the country (Shi et al., 2007; Xu et al., 2012). According to local land use policy, virgin land, such as wetland and forest, have been reclaimed as paddy land and dry land, and increasing fertilization has been used to attain higher, stable crop production (Cao et al., 2006; Song et al., 2008). Research on the concentration of soil P has been primarily in situ experiments or studies conducted on a small spatial scale, and soil chemistry

H. Lou et al. / Catena 127 (2015) 46–55

tests have been the predominant method for obtaining P concentration data. In larger areas, the combination of chemistry tests and a simple empirical model are the preferred way to obtain the P data (Guo and Ma, 2011; Liu and Song, 2010; Wang et al., 2010; Wang et al., 2012a, 2012b). These classical methods can help us better understand the concentration of soil P at a small scale, and precise results can be obtained from sampling points. However, at a large spatial scale (e.g., basin scale) and for long time series the chemistry method is not effective; although some statistical models are used, these simple models cannot express the cyclical and dynamic mechanism of soil P in nature. Many complex mechanism models that can simulate and describe the migration and conversion of nutrients or heavy metals have been developed in recent decades, including the SWAT model, the PSYCHIC model, the EPIC model, the AnnAGNPS model, the ANSWERS model and the CREAMS model (Beasley et al., 1980; Knisel, 1980; Neitsch et al., 2005; Paul et al., 2008; Storm et al., 1988; Williams, 1990; Yuan et al., 2005). Among these models, the SWAT model has been used and validated in many countries and regions (Gassman et al., 2007). The statistical model coupled with a P-cycle mechanism model, or with a P-cycle model coupled with hydrological and erosion models, are widely used to simulate P migration and conversion (Hao et al., 2006; Zheng et al., 2009). Models such as the CREAMS model and the EPIC model can only simulate nutrient content at the catchment scale, but large-scale simulation of P-content and P-cycle using these models is not possible. One of the most efficient techniques for resolving the large-scale problem is the use of remote-sensing techniques. The simulation of P-concentration at large temporal and spatial scales necessitates a remote sensing driven model that can effectively describe the physical mechanism of the P cycle. The Ecohydrological Assessment Tool (EcoHAT) (Dong et al., 2013; Liu et al., 2009; Wang et al., 2009, 2010; Wang et al., 2011; Yang et al., 2011) is an ecohydrological simulation system that can comprehensively simulate regional ecohydrological processes based on the physical and chemical mechanisms. EcoHaT integrates hydrological cycle processes, nutrient cycle processes and the plant growth process, as well as taking into account the nutrient cycle and vegetation growth in the hydrological cycle. EcoHAT couples with remote sensing models to invert land surface parameters, providing a spatial data source for ecohydrological process simulation. Driven by remote sensing, EcoHAT takes full advantage of remote sensing in resolving the problem of simulating P-concentration in soil on a large scale and with a time series. At the same time, the EcoHAT system embeds GIS spatial analysis tools, which can analyze the simulation results effectively (Liu et al., 2009). The objectives of this study are as follows: 1) to study the variation in the area and the position of P risk in the soil, thereby obtaining the spatial and temporal variation pattern in the Sanjiang Plain; and 2) to obtain the concentration of P in the soil from 2000 to 2010 (11a) in the Sanjiang Plain, which has experienced more than 60 years of intensive agricultural development, using a remote sensing driven P model in EcoHAT as an alternative to soil chemistry methods. The study area, the input data for the model, and the framework of the P model in the EcoHAT system are introduced in detail in Section 2. In Section 3, the simulation results are analyzed, and the geographical variation in the concentration of P at different levels, as well as the potential for P leaching, is also described. In Section 4, the reliability and uncertainty of results, the factors driving P accumulation and the future risks associated with phosphorus in the Sanjiang Plain are discussed. 2. Methodology 2.1. Study area The Sanjiang Plain (43°49′55″–48°27′40″N, 129°11′20″–135°05′26″ E) (Fig. 1) is an area of low relief within the Heilongjiang Province of China, with a total area of 108,900 km2, occupying 22.6% of the total area of the Heilongjiang Province. Three rivers, the Amur River, the

47

Songhua River, and the Ussuri River, constitute its major waterway system and are responsible for the alluvial deposits in this area (Song et al., 2008; Yang et al., 2012). The climate is temperate and humid and midhumid continental, with an annual average temperature of 1.9 °C. January is the coldest month, with an average temperature of −20 °C, July is the hottest month, with an average temperature of 22 °C, and annual precipitation ranges from 500 to 600 mm. This region experiences seasonal freeze, and from November to April the soil undergoes frequent freeze–thaw cycles. The main soil types in the area are meadow swamp soil, humus swamp soil and peat swamp soil. Natural soil fertility is very good in the Sanjiang Plain; maize, wheat, soybean and other plants have the advantage of excellent growing-conditions. The plain is thus one of largest commodity crop bases in China. Crop production there plays a critical role in China's crop safety, even influencing Northeast Asia's crop safety. The large area of wetland in the Sanjiang Plain is another of its notable features. The total area of wetland was 1,100,000 ha in 2003 (Richardson and Ho, 2003), making it the most notable wetland in all of China. The Bawujiu Farm (Fig. 1), located in the northeast portion of the Sanjiang Plain in Heilongjiang Province, provided the case study site for this research project. The climate condition in the Bawujiu Farm is similar to the Sanjiang Plain. Its soil types include alluvium, loamy meadow grassland, clayey meadow grassland, lessive meadow grassland, gley meadow grassland, swamp soil, peat swamp soil, dark brow forest soil, albic soil, meadow lessive and clay lessive. These soil types are very common and typical in the Sanjiang Plain. Maize, rice and soybean are the primary crops in the Bawujiu Farm, similar to the Sanjiang Plain as a whole. The Bawujiu Farm is itself a record of the mass agricultural development in China. The current farm site was covered in extensive wetland and forest before its establishment in 1956; since then it has been affected by widespread land reclamation (Xu et al., 2012). In addition, the high level of crop production, the yearly increase in fertilization usage, the excess reclamation, and the local land use policy that seeks to convert more wetland and forest to farmland led us to choose this farm as our typical experimental site. 2.2. Data and experiment 2.2.1. Data The input data (Table 1) for the model include multisource remote sensing data and production, meteorological data, in situ experiment data, and survey and historical data. Multisource remote sensing data and meteorological data come from public free data platforms (e.g., USGS: United States Geological Survey, http://www.usgs.gov/ pubprod/); other data resources are from in situ experiments and statistical data. All the original input data were processed by IDL (Interactive Data Language) programs; IDL has the ability to process batched mass remote sensing images. Some of the data were resolved by the software ENVI5.0 or ArcGIS10.0. All the input data were imaged, then set according to a uniform projection and coordinate system. The spatial resolution of the input data is also a key step because only if the input data have the same spatial resolution can the P simulation model run smoothly, but different input data may have different spatial resolutions. For example, Landsat TM (Thematic Mapper) data have 30 m spatial resolution, while the MODIS (Moderate-Resolution Imaging Spectral Radiometer) data we used have 1000 m spatial resolution. At the same time, we tried to use GLDAS (Global Land Data Assimilation System) data to improve our spatial–temporal (three-hourly) resolution of meteorological data, but the GLDAS data have a coarse spatial resolution of 25 km. To obtain the uniform 1000 m spatial resolution we want to input into the model, we resampled and used some downscaling methods; the validation of the downscaling result was satisfactory. 2.2.2. Experiment for model validation To validate the performance of the P model in EcoHAT (EcoHAT-P), we sampled at the Bawujiu Farm at the end of April, 2011 (Fig. 1) by

48

H. Lou et al. / Catena 127 (2015) 46–55

Fig. 1. Location of the Sanjiang Plain and the Bawujiu Farm in Northeast China.

using the grid-random method. The whole region was separated into many small grids, the distance between any two of which was 1000 m; then coordinates of the sampling-grid center were randomly determined based on a random function. The trial proves that the Table 1 Important input data to model and method of data processing.a Data type Lands-RS data

Mete-RS data

Survey data

Data name

Data source

O-SP

Data processing method

In-SP

DEM Landuse

ASTER TM/ETM

30 m 30 m

Resample Visual-Interpre/Resample

1 km 1 km

LST LAI APAR Snowcover Ta U2

MODIS MODIS MODIS MODIS GLDAS GLDAS

1 km 1 km 1 km 1 km 25 km 25 km

Coord-Transformation/Clip Coord-Transformation/Clip Coord-Transformation/Clip Coord-Transformation/Clip IDL processing IDL processing

1 km 1 km 1 km 1 km 1 km 1 km

Rn RH RLA_W Tfall

GLDAS GLDAS Situ-survey Situ-survey

25 km 25 km – –

IDL processing IDL processing Data imaging Data imaging

1 km 1 km 1 km 1 km

fertminP Fert

Situ-survey Situ-survey

– –

Data imaging Data imaging

1 km 1 km

a Annotation: O-SP: original spatial resolution; In-SP: input spatial resolution; Lands-RS data: land surface remote sensing data; Mete-RS data: meteorology RS data; DEM: digital elevation model; LST: land surface temperature; LAI: leaf area index; APAR: absorbed photosynthetically active radiation; Ta: air temperature; U2: wind speed at 2 m height; Rn: net radiation; RH: relative humidity; RLA_W: biomass in unit leaf area; Tfall: threshold temperature to leaf wither and fall; fertminP: mineral phosphorus percentage in fertilization; Fert: fertilization use; ASTER: advanced spaceborne thermal emission and reflection radiometer; TM/ETM: thematic mapper/enhanced thematic mapper; MODIS: moderateresolution imaging spectral radiometer; GLDAS: global land data assimilation system; Visual-Interpre/Resample: visual interpretation/resample; Coord-Transformation/Clip: coordinate transformation/clip; – : point survey data.

grid-random method can produce a regional value more feasible than traditional sampling methods. One hundred and forty one soil samples were obtained as follows: we used a 5 cm ring knife to obtain the surface soil, while at the same time the geography coordinate, topography, and land use type for the sample point were recorded. When sampling, the soils that were on the farmland ridge, beside the road or farmhouse, as well as soils that clearly had been disturbed, were abandoned. First, five sub-samples were collected in a range of 100 cm × 100 cm and then were mixed together to obtain one final sample, which was then placed into a plastic sealing bag. All of the soil samples were air dried at room temperature (25 °C) to remove plant, organic residue and other visible extraneous material. The soil was pulverized with a stick and then ground in an agate mortar. A 2 mm nylon sieve was used to obtain the soil sample that could be used for the chemical experiment. Total soil P underwent mixed acid (HF, HNO3, HCLO4) digestion (Kara et al., 1997), and the inductively coupled plasma atomic emission spectrometry method (ICP-OES, IRIS Intrepid II XSP, ThermoElectron, USA) was used to obtain the total concentration of P in the soil samples. 2.3. Methods 2.3.1. P model in EcoHAT (EcoHAT-P) The P model (Fig. 2) in EcoHAT (EcoHAT-P) couples the plant and soil P cycle model, combining two processes, P cycle and P conversion in the vegetation and the soils. It includes seven sub-models (Fig. 2), which describe the two processes. EcoHAT-P is a physically and chemically based model that clearly describes the P-cycle process in plants and soil. EcoHAT-P is thus capable of simulating the generation and variation of soil P more precisely. Input data for the model include multisource remote sensing data and production, meteorological data, in situ experiment data and survey and historical data. The entire EcoHAT-P system is driven by remote sensing, parameters in the

H. Lou et al. / Catena 127 (2015) 46–55

49

Fig. 2. The basic construction and components of the soil P model. Annotation: Mine and Decom: mineralization and decomposition; Veg: vegetation.

model are substituted by remote sensing images and are spatially distributed in a pixel matrix. IDL was selected to make the model programming and combine all the sub-models.

integrated to describe the soil P-cycle process. All the arithmetical equations are enumerated in Table 2. 3. Results

2.3.2. Phosphorus related modules in EcoHAT The EcoHAT-P includes two major processes, the plant P cycle and the soil P cycle. The NPP (Net Primary Productivity) simulation model, a vegetation production distribution model, a vegetation nutrient absorption model, and a vegetation litter model are coupled to describe the vegetation P-cycle process; a fertilization model, a mineralization and decomposition model, and an inorganic P absorption model are

This study found the total concentration of P in the Sanjiang Plain from 2000 to 2010 (11a) by using EcoHAT-P. A grid-random sampling method was carried out in a typical in situ experiment at Bawujiu Farm to test the reliability of the simulation results, with the sampling dates and the range between neighboring points consistent with the simulation results. The spatial and temporal variation of total soil P

Table 2 Functions and parameters in the P model.a. No.

Model name

Equation

Reference

1

NPP simulation

NPP(x, t) = APAR(x, t) × ε(x, t)

2

Production distribution

CASA model (Zhu et al., 2007) ForNBM model (Zhu et al., 2003)

3

Nutrient absorption

4


 d FB ε LA dRB dNPP dF B − ¼ RLA W ¼ K ra dt dt dt dt dt dW B dNPP d F B dRB − ¼ − dt dt dt dt Xuptale = min(Xavail, Xdem)

Vegetation litter Llit

5

Fertilization

6 7

Mineralization & decomposition Inorganic P absorption

8 < a f h Fb; Fb; ¼ : afs F b ;

T air N T fall T air b T fall T air N T fall



afl FL; T air N T fall FL; T air b T fall ¼ a f wW B Rlit ¼ a f rRB

Llit ¼

W lit P solution;fert ¼ f ert ;minP  fert orgP frsh;fert ¼ 0:5  f ert orgP  fert orgP hum;fert ¼ 0:5  f ert orgP  fert Pmin a = 1.4 ⋅ βmin ⋅ (γtmp, γsw)1/2 ⋅ orgPact 8 pai > > P solution;ly − ; minP act;ly  < 1−pai 
P sol=sta;ly ¼ pai > > : 0:1  P solution;ly − ; minP act;ly  1−pai

ForNBM model (Paul and Tõnu, 1997; Zhu et al., 2003) ForNBM model (Zhu et al., 2003) SWAT model (Neitsch et al., 2005) Jones et al. (1984) SWAT model (Neitsch et al., 2005)

a Annotation: APAR(x,t): plant absorbed photosynthetically active radiation from pixel x in day t (MJ · m−2); ε(x,t): the real efficiency of radiation utilization (g C · MJ − 1 ); NPP: Net Primary Productivity (g C · m−2); FB: NPP the leaf obtained (g C · m−2); RLA_W: biomass in 1 m2 leaf (g · m−2); εLA: leaf area increment per month (m−2); RB: the NPP root obtained (g C · m−2); Kra: return coefficient of leaf nutrient, constant; WB: the NPP limb obtained (g C · m−2); Xavail: the plant available amount of element X (X = P) (g · m−2); Xdem: the plant demand amount of element X (X = P) (g · m−2); Llit: leaf wither amount (g C · m−2); afh: deciduous species wither threshold; afs: evergreen species wither threshold; afr: root wither threshold; afw: limb wither threshold; afl: herbage wither threshold; Tair: air temperature (°C); Tfall: the threshold temperature of Deciduous species (°C); Rlit: the root wither amount (g C · m−2); Wlit: the limb wither amount (g C · m−2); Llit: herbage wither amount (g C · m−2); FL: the NPP herbage obtained (g C · m−2); Tfall: the threshold temperature of herbage species (°C); Psolution, fert: the dissolved soil P that comes from fertilization (kg P · ha); fertminP: the percentage of mineral P in fertilization (%); orgPfrsh,fert: the fresh organic P in soil that comes from fertilization (kg P · ha); fertorgP: the percentage of organic P in fertilization (%); orgPhum,fert: the humus organic soil P that comes from fertilization (kg P · ha); fert: the amount of fertilization (kg P · ha); Pmina: the phosphorus mineralized from the humus active organic P pool (kg P · ha); orgPact: the amount of phosphorus in the active organic pool (kg P · ha); βmin: the rate coefficient for mineralization of the humus active organic nutrients; γtmp: the nutrient cycling temperature factor; γsw: the nutrient cycling water factor; Psol/act,ly: amount of phosphorus transferred between the active and stable mineral pools (kg P · ha); Psolution,ly: the amount of phosphorus in solution (kg P · ha); minPact,ly: the amount of phosphorus in the active mineral pool (kg P · ha); pai: the availability index.

50

H. Lou et al. / Catena 127 (2015) 46–55

was also obtained through the spatial analysis of GIS technology. At the same time, we obtained the variation in the potential P leaching area. To extract the high P risk area, “spatial multiplication and summation” in software ArcGIS10.0 were employed, and we then analyzed the high P risk area. 3.1. Validation of the EcoHAT-P We obtained 141 sampling points, plotted their geography coordinates, and overlaid the plotting results on the simulation image to obtain the simulation results. The observed average value of total P is 0.83 g/kg, and the simulated average value is 0.88 g/kg (Table 3). Other standards, the maximum value, the standard deviation, and the standard error of mean were also used to show the distribution characteristics in the two data groups and the degree of similarity between them. There is a scaling problem between the observation value and simulation value, however. The observation value can only represent one specific point in an area, but the pixel (1 × 1 km in this study) is the smallest distribution unit of the simulation result. To address this problem, we surveyed the soil types in the Bawujiu Farm and then zoned the site according to the soil types. There are 11 soil types at Bawujiu Farm: alluvium, loamy meadow grassland, clayey meadow grassland, lessive meadow grassland, gley meadow grassland, swamp soil, peat swamp soil, dark brow forest soil, albic soil, meadow lessive and clay lessive are included. We found points in the same soil type zone, then obtained the average observation and simulation values and used the averaged data to represent the value of this area. The test result (Fig. 3) shows that R2 of the observed and the simulated data is 0.649, demonstrating that the P model is reliable. 3.2. Increasing trend of higher phosphorus content The critical values for classifying the total concentration of P are the observed average value of 0.83 g/kg and the potential P-leaching threshold of 1.25 g/kg. The simulation results were reclassified into seven grades—0–0.21, 0.21–0.42, 0.42–0.62, 0.62–0.83, 0.83–1.04, 1.04–1.25, and 1.25–1.45 g/kg—and all of the processes were completed using ArcGIS10.0 software (Fig. 4). Long-term fertilization will result in an Olsen-P accumulation in the soil (Huang et al., 2011), and the Olsen-P leaches more easily into soil when the content exceeds a threshold of 60 mg/kg (Heckrath et al., 1995; Shan et al., 2004). The percentage of Olsen-P in total P in the Sanjiang Plain is 4.77% (Qu et al., 2008), so when the total soil P exceeds 1.25 g/kg, the soil is at high risk of P leaching. According to these references, the spatial and temporal distribution maps of total concentration of P in the Sanjiang Plain from 2000 to 2010 (Fig. 4) were obtained. We can see from the maps that the areas with total concentration of P content higher than the observed average value of 0.83 g/kg became more widespread from 2000 to 2010; the areas with 0.83–1.04 (orange), and 1.04–1.25 (red) g/kg had grown significantly; the areas with 0.42–0.62 (green), and 0.62–0.83 (yellow) g/kg had obviously decreased, and 0.62–0.83 (legend yellow) g/kg level decreased continuously. From this evidence, we can conclude that there are significantly more areas with higher P content in 2010 than in 2000.

Fig. 3. Observed and simulated results for the 11 soil types.

concentrations of P that approximated the observed average value clearly increased, and areas with higher concentrations of P extended at a rapid pace. The areas with P content in the 0.62–0.83 g/kg range decreased rapidly, ranging from 69,358 km2 in 2000 to 24,452 km2 in 2010, a decline of 4082 km2 per year. The areas with concentrations of P in the 0.83–1.04 g/kg range increased quickly, ranging from 15,963 km2 in 2000 to 38,274 km2 in 2010, or a 2028 km2 growth per year. The areas with concentrations of P at the 0.42–0.62 g/kg and 1.04–1.25 g/kg levels increased quickly, ranging from 17,216 km2 and 527 km2 in 2000 to 28,506 km2 and 8534 km2 in 2010. The area with a concentration of P at a high level of 1.25–1.45 g/kg increased in 2003, ranging from 79 km2 in 2004 to 1131 km2 in 2010. Based on the data, the soil concentration of P is increasing, and the accumulated soil P is increasing each year. 3.4. Increasing potential phosphorus-leaching risk The percentage variation between the two groups can be computed by adding the areas with P levels 0–0.21, 0.21–0.42, 0.42–0.62, and 0.62–0.83 g/kg, or those with lower than the average observed value, and adding the areas at P levels 0.83–1.04, 1.04–1.25, and 1.25– 1.45 g/kg, or those with higher than the average observed value (0.83 g/kg) (Fig. 6). When the percentage is lower than the average value, the P level has a decreasing trend, ranging from 84.8% in 2000 to 56.0% in 2010. Conversely, when the percentage is higher than the average value, the P level has an increasing trend, ranging from 15.1% in 2000 to 44.1% in 2010. Overall, the concentration of soil P grew from 2000 to 2010, the accumulation of soil P was accompanied by a time lapse. In our study, the soil P had higher leaching potential when the total P exceeded the threshold value (1.25 g/kg). These areas range from 79 km2 in 2004 to 1131 km2 in 2010. Unfortunately, the areas with leaching potential display rapidly increasing speed and amplitude; this means that more water bodies are at risk of non-point source pollution and that water conservation in this region will be more challenging.

3.3. Phosphorus accumulated in the Soil

3.5. Higher phosphorus risk in cultivated land and water bodies

The difference in P levels can reflect the P-concentration variation in the soil (Fig. 5). From 2000 to 2010 in the Sanjiang Plain, the areas with

“Spatial multiplication and summation” were adopted to estimate total concentration of soil P data by a spatial analysis module in

Table 3 Statistical analysis of the measured value and the model value. Type

Samples

Mean value (g/kg)

Max value (g/kg)

Standard deviation

Standard-error of mean

Observed value Simulated value

141 141

0.8308 0.8820

1.3970 1.1758

0.1783 0.1919

0.0150 0.0161

H. Lou et al. / Catena 127 (2015) 46–55

51

Fig. 4. Spatial and temporal distribution maps of total P distribution in the Sanjiang Plain from 2000 to 2010 (11a).

ArcGIS10.0 software. This method can easily extract areas with a high concentration of P. First, we divided the 11a time series (2000–2010) into three intervals, 2000–2003, 2003–2006, and 2007–2010. Each four-year group was subjected to a spatial multiplicative analysis, using the 1 g/kg as the classification node because the spatial multiplicative algorithm made the areas that have high P content become higher, and those with low P content become lower. This allowed for a simplified determination of the high P-concentration area, which was then analyzed for temporal and spatial variation. Second, the 11a simulation results were subject to spatial multiplication, again using 1 g/kg as the classification node to extract the area with high P concentration at a longer time scale (11a) and to compare the result to the short time

series (four-year interval). Finally, a spatial summation algorithm was applied to the 11a-simulation results from 2000 to 2010, and an equal interval classification was selected, to obtain the P-accumulation effect (Fig. 7). The spatial multiplication results show that the area of high P concentration grew rapidly, particularly from the first time range (2000–2003) to the second (2003–2006), but from the second to the third time range (2007–2010), the growth is hard to find. Compared to the 11a-spatial multiplication result, the long time series result was similar to the short time range (2007–2010), which might be because the soil accumulated more P from 2004 to 2010. Regarding the spatial variation, the high P-concentration area is distributed sporadically

52

H. Lou et al. / Catena 127 (2015) 46–55

Fig. 5. Areas of different concentrations of P levels in the Sanjiang Plain from 2000 to 2010 (11a).

from 2000–2003 and centralized in the western and southern regions. Over time, a large area of high soil P concentration appeared in the northern, middle and eastern parts of this region; what was a previously sporadic area became a vast area of high P concentration. Analysis of the 11a-spatial summation result shows that the high P-concentration areas in the north, northeast, and southeast of the Sanjiang Plain, e.g., in Hulin and Mishan counties, developed earlier. These counties have more cultivated land and their land use index is very high (Song et al., 2008). The areas at high risk of P leaching are in the mid-west and southeast of the Sanjiang Plain, a region with more rivers (Songhua, Naoli, and Songacha) and a complex water network. “Spatial multiplication and summation” were used to extract the area with a high concentration of P. Our analysis of the results found that the area with a high concentration of P is increasing over time and that the cultivated land and water bodies in this area are at the highest risk. 4. Discussion 4.1. The reliability and uncertainty of the P model The model's average simulation value is 0.88 g/kg and the average observed value is 0.83 g/kg, the two values are close. This study is the first to use a physically based model to ascertain the concentration of soil P in the Sanjiang Plain. Classical soil chemistry methods have been favored in previous studies; these methods can be precise in evaluating

point scale and the focus on the concentration of soil P, and changes in typical wetland and soil P concentration in response to land use change, are two major aspects of previous studies. Zhao et al. (2006) sampled in the wetlands and cultivated lands at the Naoli River basin and the Bielahong River basin in the Sanjiang Plain, detecting a total concentration of P (0–20 cm) of 0.896 g/kg; Qin et al. (2006) sampled from five typical land use types (soybean field, rice field, abandoned cultivation, artificial, and natural wetland) in the Bielahong River basin and the Nongjiang River basin, finding an average total concentration of soil P (0–20 cm) of 0.704 g/kg; and Qin et al. (2007) detected a total P concentration of 1.113–0.895 g/kg in Calamagrostis angustifolia, a typical wetland plant in the Sanjiang Plain. Although our study differs from other conventional methods with respect to the sampling location, the sampling time, and the test process, the results from the two methods are similar. The reliability of our use of the EcoHAT-P to simulate P concentration is thus confirmed. The uncertainty of the model is also an important factor in the differences between the observed and simulated values, however, and contributes to the differences between our study and previous studies. A scaling problem (Kumar et al., 2013; Samaniego et al., 2010) is the primary reason for the uncertainty of the model. There are three scaling issues in our study: scaling problems that derive from the input data to the model, the model's simulation scale and the scaling mismatch in the validation process. First, mass parameters have to be created to build a physically based P model, which creates an enormous database to ensure that the P model will run smoothly. The data in the database have different spatial or temporal scales, however, and the up-scaling or down-scaling process causes some damage to the original data. Second, the model's simulation scale, and whether it can describe the P cycle and transformation precisely, may also influence the final simulation result. Third, when we test the validity of the model, we can only obtain the “true” value in point scale, but the simulation result is distributed by pixel matrix (e.g. 1 × 1 km in this study), which creates a scale mismatching problem between the simulated and observed process. 4.2. Factors driving the growing concentration of soil P The overall concentration of soil P in the Sanjiang Plain is increasing and the areas at risk of P-leaching increased from 2004 to 2010. Regions with a history of cultivation and with large areas of farmland have experienced the most accumulation of soil P. From 2000 to 2005, cultivated land increased by 327,954 hm2, forest land was reduced by 162,048 hm2 , 633 hm2 of grassland disappeared, and natural wetland was reduced by 163,350 hm2 (Song et al., 2008). Diminished forest land, grassland, and natural wetlands that were converted to farmland, and permanent forest land and natural wetland that were converted to

Fig. 6. Percentage variation and potential P-leaching area.

H. Lou et al. / Catena 127 (2015) 46–55

53

Fig. 7. Areas with high phosphorus risk. (For interpretation of the references to color in the text related to this figure, the reader is referred to the web version of this article.)

cultivated land led to an increase in available phosphorus in the soil of 107.7% and 103.2%, respectively (Ouyang et al., 2013; Song et al., 2008). Inorganic P is the primary existing form of P in the soil; the total P will increase with more available forms of phosphorus in the soil. Since 1980, the amount of fertilization per unit of farmland increased by 2.9% per year in Heilongjiang Province; when inorganic fertilizer and organic manure are thrown into farmland, the risk to Olsen-P accumulation and leaching increases rapidly (Cao et al., 2006; Sharpley and Sisak, 1997). Land use change by which more farmland is converted from permanent forest or wetland and more fertilization is used to maintain a high level of crop production are the two most important factors that cause the phosphorus content in soil to increase. 4.3. Higher phosphorus risk in the future Overland flow is often considered the major pathway of soil P loss (Lehmann et al., 2005; Pote et al., 1999), but in many regions subsurface flow is also a significant pathway (Wang et al., 2012a, 2012b). Longterm, repeated application of fertilizer or livestock manure to soils has

led to increased losses of P in subsurface flow, particularly in organic soils (Majuire and Sims, 2002; Van Es et al., 2004). By 2006, the average soil Olsen-P had increased to 20.7 mg/kg, almost three times higher than was the case in China in 1980 (Bai et al., 2013). The increase in the total concentration of P is linked to an associated increase of Olsen-P in soil. From 2000 to 2010 in the Sanjiang Plain, the areas that had concentrations of P higher than the average observed value ranged from 15,963 km2 in 2000 to 38,274 km2 in 2010, which represents 68.73% of the total cultivated land in the Sanjiang Plain, and 2028 km2 growth per year. The risk for P leaching is thought to exist when the total concentration of P is higher than 1.25 g/kg. From 2000 to 2010 in the Sanjiang Plain, the areas that may be at risk of P leaching increased from 79 km2 in 2004 to 1131 km2 in 2010, a growth of 150.29 km2 per year. The growth in areas of P accumulation and potential P leaching are two dangerous trends that affect the safety of the ecological environment, and the crop production safety of the region. Once the phosphorus in soil migrates to adjacent lakes and rivers, a serious threat to water quality will also be posed. Total phosphorus increases follow Olsen-P growth, and more phosphorus may enter the groundwater

54

H. Lou et al. / Catena 127 (2015) 46–55

because of P leaching, contributing to the degradation of water quality and harming the aquatic environment. These developments have alarming repercussions for crop safety in the region. Higher phosphorus risk calls for more precise agricultural management measures and more science-based fertilization regimes. At a larger regional scale, detailed spatial–temporal maps of soil P concentration will be indispensable and will be particularly crucial to the science of soil fertilization. In summary, Sections 4.1–4.3 indicate the reliability and uncertainty of the model, the driving factors that contribute to the increasing concentration of soil P and the phosphorus risk that the Sanjiang Plain may face in the future. The Bawujiu Farm plays an important role in validating the P-model. Although the Farm has the climate condition, the soil types and the cultivated crops that are most similar with the Sanjiang Plain as-a-whole, and can very well represent the situation of the latter, a scale problem is unavoidable. Thus, the samples from the Bawujiu Farm have some uncertainties in calibrating parameters of the P-model when this model is applied to the whole Sanjiang Plain. To reduce the uncertainties introduced by the scaling problem, many more samples from other sub-study areas similar to the Bawujiu Farm should be collected in the future. 5. Conclusions We applied the remote sensing driven P simulation model in the EcoHAT system and obtained the spatial and temporal variation results for the total concentration of P in the Sanjiang Plain from 2000 to 2010 (11a). GIS spatial analysis technology was used to detect the spatial and temporal variation law of the concentration of soil P. An in situ experiment was completed on the Bawujiu Farm to validate the phosphorus simulation model, with the results showing that the model is reliable. Areas with higher concentrations of P increased remarkably from 2000 to 2010 in the Sanjiang Plain; phosphorus has accumulated in the soil and the area is at risk of dramatic increases in phosphorus leaching; the cultivated land and water bodies are at the highest risk of phosphorus-leaching. The growing concentration of soil P is driven primarily by humans; without intervention, the region may face a yet higher phosphorus risk in the future. Acknowledgments We acknowledge all reviewers and editors for their valuable advice. The authors would like to thank the management of the Bawujiu Farm. The project benefitted from the financial support provided by the National Natural Science Foundation of China (Grant Nos. 41271414, 40930740), the Supporting Programme of the “Twelfth Five-Year Plan” for Science & Technology Research of China (2012BAB02B00) and the Fundamental Research Funds for the Central Universities (HKY-JBYW-2013-22). References Bai, Z.H., Li, H.G., Yang, X.Y., Zhou, B.K., Shi, X.J., Wang, B.R., Li, D.C., Shen, J.B., Chen, Q., Qin, W., Oenema, O., Zhang, F.S., 2013. The critical soil P levels for crop yield, soil fertility and environmental safety in different soil types. Plant Soil 372, 27–37. Beasley, D.B., Huggins, L.F., Monke, E.J., 1980. ANSWERS: a model for watershed planning. Trans. ASAE Am. Soc. Agric. Eng. 23 (4), 938–944. Bolland, M.D.A., Yeates, J.S., Clarke, M.F., 1996. Effect of fertilizer type, sampling depth, and years on Colwell soil test phosphorus for phosphorus leaching soils. Fertil. Res. 44, 177–188. Cao, N., Qu, D., Chen, X.P., Zhang, F.S., Fan, M.S., 2006. Analysis of the contribution to nonpoint pollution made by balanced fertilizer in Northeast China. J. Northwest Sci.Technol. Univ. Agric. For. 34 (7), 127–133 (in Chinese). Cheng, L., Bucciarelli, B., Shen, J., Allan, D., Vance, C.P., 2011. White lupin cluster root acclimation to phosphorus deficiency and root hair development involve unique glycerophosphodiester phosphodiesterases. Plant Physiol. 156 (3), 1025–1032. Djodjic, F., Borling, K., Bergstrom, L., 2004. Phosphorus leaching in relation to soil type and phosphorus content. J. Environ. Qual. 33, 678–684.

Dong, G.T., Yang, S.T., Bai, J., Wang, Z.W., Zhang, Y.C., 2013. Open innovation in the Sanjiang Plain: a new paradigm for developing agriculture in China. J. Food Agric. Environ. 11 (3&4), 1108–1113. Duan, Y., Xu, M., He, X., Li, S., Sun, X., 2011. Long-term pig manure application reduces the requirement of chemical phosphorus and potassium in two rice-wheat sites in subtropical China. Soil Use Manag. 27 (4), 427–436. Gassman, P.W., Reyes, M.R., Green, C.H., Green, C.H., Arnold, J.G., 2007. The Soil and Water Assessment Tool: historical development, application, and future research directions. Am. Soc. Agric. Biol. Eng. 50 (4), 1211–1250 (ISSN 0001-2351). Guo, L., Ma, K.M., 2011. Seasonal dynamics of nitrogen and phosphorus in water and sediment of a multi-level ditch system in Sanjiang Plain, Northeast China. Chin. Geogr. Sci. 60, 799–807. Hao, F.H., Yang, S.T., Cheng, H.G., Bu, Q.S., Zheng, L.F., 2006. A method for estimation of non-point source pollution load in large-scale basins of China. Acta Sci. Circumst. 26 (3), 375–383 (in Chinese). Heckrath, G., Brookes, P.C., Poulton, P.R., Goulding, K.W.T., 1995. Phosphorus leaching from soil containing different phosphorus concentrations in the Broadbalk experiment. J. Environ. Qual. 24, 904–910. Huang, S.M., Guo, D.D., Zhang, S.Q., 2011. Effects of long-term application of organic fertilizer and superphosphate on accumulation and leaching of Olsen-P in Fluvo-aquic soil. Chin. J. Appl. Ecol. 22 (1), 93–98 (in Chinese). Jones, C.A., Cole, C.V., Sharpley, A.N., Williams, J.R., 1984. A simplified soil and plant phosphorus model: I. Documentation. Soil Sci. Soc. Am. J. 48 (4), 800–805. Kara, D., Ozsavasci, C., Alkan, 1997. Investigation of suitable digestion methods for the determination of total phosphorus in soils. Talanta 81 (1), 75–90. Knisel, W.G., 1980. CREAMS: a field scale model for chemicals, runoff, and erosion from agricultural management systems. United States Dept. of Agriculture Conservation Research Report. Kumar, R., Livneh, B., Samaniego, L., 2013. Toward computationally efficient large-scale hydrologic predictions with a multiscale regionalization scheme. Water Resour. Res. 49, 5700–5714. Lehmann, J., Lan, Z.D., Hyland, C., Sato, S., Solomon, D., Ketterings, Q.M., 2005. Long-term dynamics of phosphorus forms and retention in manure-amended soils. Environ. Sci. Technol. 39, 6627–6680. Liu, D.Y., Song, C.H., 2010. Effects of inorganic nitrogen and phosphorus enrichment on the emission of N2O from a freshwater marsh soil in Northeast China. Environ. Earth Sci. 21 (4), 437–455. Liu, C.M., Yang, S.T., Wen, Z.Q., Wang, X.L., Wang, Y.J., Li, Q., Sheng, H.R., 2009. Development of ecohydrological assessment tool and its application. Sci. China Ser. E Technol. Sci. 39 (6), 1112–1121. Majuire, R.O., Sims, J.T., 2002. Soil testing to predict phosphorus leaching. J. Environ. Qual. 31, 1601–1609. Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., 2005. Soil and Water Assessment Tool: Theoretical Documentation Version 2005, Texas, USA, pp. 201–259. Ouyang, W., Xu, Y.M., Hao, F.H., Wang, X.L., Chen, S.Y., Lin, C.Y., 2013. Effect of long-term agricultural cultivation and land use conversion on soil contents in the Sanjiang Plain. Catena 104, 243–250. Paul, A.A., Tõnu, O., 1997. A forest soil vegetation atmosphere model (ForSVA), I: concepts. Ecol. Model. 95 (2), 211–224. Paul, S.D., Withers, P.J.A., Lord, E.I., Betson, M.J., Stromqvist, J., 2008. PSYCHIC—a process-based model of phosphorus and sediment mobilization and delivery within agricultural catchments. Part 1: model description and parameterization. J. Hydrol. 350 (3), 290–302. Pote, D.H., Daniel, T.C., Nichols, D.J., Sharpley, A.N., Moore, P.A., Miller, D.M., Edwards, D.R., 1999. Relationship between phosphorus levels in three ultisols and phosphorus concentrations in runoff. J. Environ. Qual. 28, 170–175. Qiao, Y.X., Pei, B.D., Da, S.S., Cheng, L.S., Ling, Y.Q., Ostermann, A., Yao, H., Xian, Y.L., 2013. Effects of rainfall and manure application on phosphorus leaching in field lysimeters during fallow season. J. Soils Sediments 13, 1527–1537. Qin, S.J., Liu, J.S., Wang, G.P., Zhou, W.M., 2006. Phosphorus fractions under different land uses in Sanjiang Plain. Environ. Sci. 28 (12), 2777–2782 (in Chinese). Qin, S.J., Liu, J.S., Wang, G.P., Wang, J.D., 2007. Seasonal changes of soil phosphorus fraction under Calamagrostis angustifolia wetland in Sanjiang Plain. Acta Ecol. Sin. 27 (9), 3844–3851. Qu, J.F., Li, J.M., Xu, M.G., Dai, J.J., 2008. Total-P and Olsen-P dynamics of long-term experiment without fertilization. Plant Nutr. Fertil. Sci. 14 (1), 90–98 (in Chinese). Richardson, C.J., Ho, M., 2003. The wetland of China an over view: part I. Introduction and Sanjiang Plain. Wetl. Wire 6 (2), 4–6. Samaniego, L., Kumar, R., Attinger, S., 2010. Multiscale parameter regionalization of a gridbased hydrologic model at the mesoscale. Water Resour. Res. 46, W05523. Shan, H.Y., Yang, L.Z., Wang, J.G., 2004. Soil phosphorus loss to water: its pathways, environmental impact, and countermeasures. Soils 36 (6), 602–608 (in Chinese). Sharpley, A.N., Sisak, I., 1997. Differential availability of manure and inorganic source of phosphorus in soil. Soil Sci. Soc. Am. J. 61, 1503–1508. Sharpley, A.N., Chapra, S.C., Wedpohl, R., Sims, J.T., Daniel, T.C., Reddy, K.R., 1994. Managing agricultural phosphorus for protection of surface water: issues and options. J. Environ. Qual. 23, 437–451. Shi, F.C., Li, R.L., Wang, S.Q., Kaichiro, S., 2007. Profile distribution and accumulation characteristics of organic carbon and total nitrogen in typical marshes in Sanjiang Plain. Chin. J. Appl. Ecol. 18 (7), 1425–1431 (in Chinese). Song, K.S., Liu, D.W., Wang, Z.M., Zhang, B., Jin, C., Li, F., Liu, H.J., 2008. Land use change in Sanjiang Plain and its driving forces analysis since 1954. Acta Geograph. Sin. 1 (63), 93–104 (in Chinese). Storm, D.E., Dillaha III, T.A., Mostaghimi, S., 1988. Modeling phosphorus transport in surface runoff. Trans. ASAE Am. Soc. Agric. Eng. 31 (1), 117–127. Townsend, A.R., Porder, S., 2012. Agricultural legacies, food production and its environmental consequences. Proc. Natl. Acad. Sci. U. S. A. 109 (16), 5917–5918.

H. Lou et al. / Catena 127 (2015) 46–55 Van Es, H.M., Schindelbeck, R.R., Jokela, W.E., 2004. Effect of manure application timing, crop, and soil type on phosphorus leaching. J. Environ. Qual. 33, 1070–1080. Wang, X.L., Yang, S.T., Mannaerts, C.M., Gao, Y.F., Guo, J.X., 2009. Spatially explicit estimation of soil denitrification rates and land use effects in the riparian buffer zone of the large Guanting reservoir. Geoderma 150, 240–250. Wang, X.L., Mannaerts, C.M., Yang, S.T., Gao, Y.F., Zheng, D.H., 2010. Evaluation of soil nitrogen emissions from riparian zones coupling simple process-oriented models with remote sensing data. Sci. Total Environ. 408, 3310–3318. Wang, X.L., Wang, Q., Yang, S.T., Zheng, D.H., Wu, C.Q., Mannaerts, C.M., 2011. Evaluating nitrogen removal by vegetation uptake using satellite image time series in riparian catchments. Sci. Total Environ. 409, 2567–2576. Wang, Y., Liu, J.S., Wang, J.D., Sun, C.Y., 2012a. Effects of wetland reclamation on soil nutrients losses and reserves in Sanjiang Plain, Northeast China. J. Integr. Agric. 11 (3), 512–520. Wang, Y.T., Zhang, T.Q., O'Halloran, I.P., Tan, C.S., Hu, Q.C., Reid, D.K., 2012b. Soil test as risk indicators for leaching of dissolved phosphorus from agricultural soils in Ontario. Soil Water Manag. Conserv. 76, 220–229. Williams, J.R., 1990. The Erosion-Productivity Impact Calculator (EPIC) model: a case history. Philos. Trans. Biol. Sci. 329 (1255), 421–428. Xu, Y.M., Liu, Y., Ouyang, W., Hao, F.H., Ding, Z.L., Wang, D.L., 2012. Impact of long-term agricultural development on wetland landscape in Sanjiang Plain. Procedia Environ. Sci. 13, 922–932.

55

Yang, S.T., Dong, G.T., Zheng, D.H., Xiao, H.L., Gao, Y.F., Lang, Y., 2011. Coupling Xinanjiang model and SWAT to simulate agricultural non-point source pollution in Songtao watershed of Hainan, China. Ecol. Model. 222, 3701–3717. Yang, W.J., Cheng, H.G., Hao, F.H., Ouyang, W., Liu, S.Q., Lin, C.Y., 2012. The influence of land-use change on the forms of phosphorus in soil profiles from the Sanjiang Plain of China. Geoderma 189–190, 207–214. Yuan, Y., Bingner, R.L., Theurer, F.D., Rebich, R.A., Moore, P.A., 2005. Phosphorus component in AnnAGNPS. Trans. ASAE Am. Soc. Agric. Eng. 48 (6), 2145–2154. Zhao, H.Y., Wang, G.P., Liu, J.S., Zhang, G.Z., 2006. Phosphorus sorption/desorption characteristics of wetland soils in Sanjiang Plain. Ecol. Environ. 15 (5), 930–935 (in Chinese). Zheng, D.H., Wang, L., Zeng, H.J., Yang, S.T., Zhang, C.P., Liu, Y.S., 2009. A simulation of nonpoint source (NPS) pollution loads in Songtao Reservoir Catchment. Acta Sci. Circumst. 29 (6), 1311–1320 (in Chinese). Zhu, Z.X., Arp, P.A., Meng, F.R., Bourque, C.P.A., Foster, N.W., 2003. A forest nutrient cycling and biomass model (ForNBM) based on year-round, monthly weather conditions, part I: assumption, structure and processing. Ecol. Model. 169, 347–360. Zhu, W.Q., Pan, Y.Z., Zhang, J.S., 2007. Estimation of net primary productivity of Chinese terrestrial vegetation based on remote sensing. J. Plant Ecol. 31 (3), 413–424 (in Chinese). Zhuang, Y.H., Wu, Y.Q., Li, Y., 2007. Effect of combined application of organic and inorganic fertilizers on phosphorus leaching in vegetable soils. Soils 39, 905–909 (in Chinese).