Soil phosphorus composition , loss risk and contribution to the aquatic environment in a typical agricultural area

River eutrophication risk increased significantly in agricultural areas. In this paper, spatial variability of soil phosphorus (P) and loss risk in the Jialu River Basin, China, were analyzed using a geostatistical approach. The correlation between soil and river sediment P was analyzed to identify the main aquatic P source. The results showed that inorganic phosphorus (IP) was the main form of soil TP (82.13%), but the ratio of apatite phosphorus (AP) and non-apatite phosphorus (NAIP) varied between different soil types. AP was the primary form of IP in fluvo-aquic cinnamon soil, while NAIP dominated in meadow aeolian sandy soil. Calculated soil total dissolvable P (TDP, 94–622 mg/kg) exceeded the environmental threshold. High TDP (>400 mg/kg) in mixed soil and sandy soil indicated a high P loss risk. The spatial variability of soil P was moderate to weak, indicating a low heterogeneity. In sediment, IP and AP showed a significant correlation with total organic carbon (p< 0.05), indicating a P source of soil erosion. Sediment AP had a significant positive correlation with soil AP (p< 0.05), confirming soil as the main source of sediment P. Furthermore, an accumulation of sediment P along the Jialu River and its consistency with water TP was revealed.


GRAPHICAL ABSTRACT INTRODUCTION
Phosphorus (P) is the primary limiting nutrient in most surface water, and a P concentration of over 20-35 μg/L will cause an algal bloom in a freshwater system (Ahmad et al. ). Non-point sources P in agricultural runoff contribute to a great portion of freshwater inputs, which accelerate eutrophication and arouse global environmental concern (Sharpley et al. ). It was believed that excess fertilization and manure production caused the accumulation of surplus P in soil, some of which is transported to aquatic ecosystems (Carpenter et al. ). Soil erosion from agricultural lands, which delivers large amounts of particulate P to surface water, has become the dominant source of P accumulation in the aquatic ecosystem (around 75%; Vilmin et al. ).
Therefore, the identification of soil P levels, risk of soil P loss and the impact of soil P migration on water quality are required for making sustainable P management to ensure both environmental safety and crop production (Zhou et al. ; El-Nahhal et al. a, b). ; NRCS ), which can be used for soil P loss risk evaluation in an agricultural area. In this case, a comprehensive soil P investigation at the basin scale was essential to provide sufficient basic information.
High-resolution mapping of soil P content is necessary to identify critical source areas where a large risk of loss coincides in agricultural landscapes. However, because of the high heterogeneity of soil type and property, it is difficult to precisely interpret soil P's content and distribution.
Spatial variability analysis based on geostatistics has been widely applied on soil property survey, fertilization evaluation and farmland nutrient management (Ahmad et  River Commission of the Ministry of Water Resources, P. R. China ), resulting in low velocity and relatively retained hydrodynamic state, which aggregated the risk of eutrophication. Therefore, it is necessary and urgent to thoroughly investigate the soil P loss risk and its contribution to aquatic P accumulation in this representative area.
The objectives of this research were to evaluate the soil P content and loss potential in the Jialu River Basin, as well as its contribution to aquatic P. To achieve these goals, the soil P composition and spatial variability in the Jialu River Basin were depicted using geostatistics. At the same time, the correlation between soil P and aquatic P was analyzed to interpret the P transport pathway and accumulation pattern at the basin scale. This research could provide a reference for the adjustment of agricultural management practices such as fertilization recommendations and non-point source pollution control in similar agriculture intensive areas.

Study area
Jialu River, one of the largest tributaries of the Huaihe River, with a total length of 255.8 km, and a catchment area of 5,896 km 2 . The geographical range of the Jialu River was   Table S1). One kg of surface soil (0-20 cm depth) was collected using a ring knife. The bulk density of the sample at each site was measured immediately. Soil samples were removed from stones, weeds and other debris, stored in plastic bags and sent back to the laboratory. All samples were airdried, sieved through 100 mesh for later use (Fu et al. ).
Ten sediment samples (0-10 cm depth) along the Jialu River and its tributaries were collected, air-dried, passed through 100 mesh sieve and stored in plastic bags for later use.

Statistical and geostatistical analysis
The summary statistics of soil P, including minimum, maximum, mean, standard deviation and coefficient of variation were calculated using SPSS 21. Additionally, a geostatistical method was applied on spatial variability analysis of soil P by the geostatistical tool in ArcGIS 9.2 for windows (Guoan & Xin ).
Variogram is the essential parameter in geostatistical analysis, representing the spatial correlation of regional variables, which is calculated by the following equation where h is the distance between sampling sites (lag), N (h) is is the measured value of a variable with a distance of h. The best-fitting semivariogram model with minimum root-meansquare error (RMSE) is selected for each soil P type.
where Z* (x i ) is the measured value of sample i, N is the number of the measured value and λ i is the unknown weight for the measured value of sample i.
Based on the fitted semivariogram model, spatial parameters such as nugget (C 0 ), sill (C þ C 0 ), range and

).
Cross-validation was conducted for each semivariogram model. The comparison between measured and estimated value, mean absolute error (MAE) are used to evaluate the accuracy of prediction.

Soil P and aquatic P composition
In the adopted speciation method in this study, the soil P  (Liu & Zhang ).
In Jialu River, sediment TP (n ¼ 10) ranged from 0.271 to 0.867 mg/g (Figure 3(a)).      types with an average of over 400 mg/kg, indicating a high P loss risk (Table 2). Therefore, stricter soil nutrient management strategies are recommended in the Jialu River Basin to control the loss of soil P.

Soil P spatial variability
The contrast between the semivariograms of different soil P types was carried out to analyze the spatial correlation and variability of soil P ( Figure 5 and Table 3). The crossvalidation results are shown in Table 4.
In this study, the nugget coefficient of IP was larger than 0.75, indicating a weak spatial correlation. The nugget coefficient of TP, AP and NAIP was between 0.5 and 0.75, indicating a moderate spatial correlation. The nugget coefficient of OP was between 0.25 and 0.5, indicating a significant spatial correlation (  properties (Ramzan & Wani ). In this research, a mixed soil sample from 0 to 20 cm, which was in the optimal range, was collected for P measurement. In conclusion, the spatial prediction of soil properties using the geostatistical approach is an alternative for the ordinary difference method, which will help in site-specific farming in the study area. In future research, a better sampling strategy  It was worth noticing that sediment AP had a significant positive correlation with soil AP (r ¼ 0.842, p < 0.05, n ¼ 7), while sediment OP was positively related to soil AP (r ¼ 0.0.841, p < 0.05, n ¼ 7) and negatively related to soil NAIP (r ¼ À0.812, p < 0.05, n ¼ 7) (   Based on the TP accumulation pattern analysis, TP in upstream and sub-streams sediments were lower than that of the downstream and the mainstream (Figure 6), revealing an accumulation behavior of sediments P along the Jialu River. At the same time, a good consistency of TP in the sediment and water can be seen, which was also subject to soil TP content, and further proved the extensive P input from soil to river water. Ning   **Significant correlation at 0.01 level (two-tailed). area ratio of 7.7%). In contrast, proportions of AP and NAIP in most of the soil types such as mixed soil were roughly equal (the area ratio of 34.6%). Calculated soil TDP varied from 94 to 622 mg/kg (393 mg/kg on average), which all exceeded the environmental threshold of 30 mg/kg. TDP in mixed soil and sandy soil (the area ratio of 49.8%) was the highest (average >400 mg/kg), indicating a high P loss risk. The spatial variability of soil P was moderate or weak, indicating a low heterogeneity in the Jialu River Basin.
Sediment IP, as well as AP (70.4-95.6% of IP), both showed a significant correlation with sediment TOC (r ¼ 0.795/0.838, p < 0.05, n ¼ 7). At the same time, sediment AP had a significant positive correlation with soil AP (r ¼ 0.842, p < 0.05, n ¼ 7), which confirmed that soil AP as the main source of sediment P was associated with organic matter from weathered soil. Furthermore, an accumulation behavior of sediment P along the Jialu River and a consistent trend with water TP along the flow path was revealed.