Effects of flow hydraulics on total nitrogen loss on steep slopes under simulated rainfall conditions

Soil erosion and loss of soil nutrients has become a global environmental problem of growing concern. Physically, nutrients can move from soil to water as they become dissolved in runoff water or may be moved in particulate form when bound to suspend sediment (Panuska et al. 2008). The loss of soil nutrients depends on soil conditions, rainfall intensity, flow hydraulic, slope, and vegetation cover (Fraser et al. 1999; Pan & Shangguan 2006). Understanding how nutrients are transported from hillslope to water and their dynamic can also provide an important basis for ecological restoration.

Soil nutrient from the sloping cropland to the stream is controlled by the interaction of nutrient source factors (soil, crop, and land use) with its transport factors (surface runoff, subsurface runoff (SSR), and sediment). Zöbisch et al. (1995) reported that total loss of nutrients was dependent on the total amount of runoff and soil loss. Researches indicate that both surface and SSR are important hydraulic pathways for nutrient transport (Jia et al. 2007). The generation of SSR depends on soil properties, water conductivity, slope gradient, and rainfall intensity (Naef et al. 2002). The hydrological processes and the associated erosional responses are known to be related to soil properties. Many reasons, such as shallow and loose soil layer, the excessive use of nitrogen, and numerous rainstorms, have caused serious soil erosion in the purple soil region, and is a major environmental problem in China (Fu et al. 2011; Peng et al. 2014). Average soil erosion rates in the purple soil region have reached 3,400 t km⁻² year⁻¹, and in some areas even reached 9,000 t km⁻² year⁻¹ (Wen et al. 2005). Considering purple soil showing the overlying soil and underlying rock which is known as ‘the Geotechnical dual structure’, SSR is a prominent runoff process in purple soil slope (Niu et al. 2010). Hubbard & Sheridan (1983) reported that from a small, upland, coastal plain watershed, 99% of total NO₃-N loss occurred in SSR. Moreover, a vast number of studies have investigated relationships that describe sediment detachment and nutrient transport. The general agreement is that suspended sediment serves as the delivery mechanism for nutrient pollutants (Shi et al. 2012; Molder et al. 2015). The amount of nutrient loss is directly proportional to the amount of soil loss (Kothyari et al. 2004). More information on slope hydrological processes and nutrient transformation of purple soil is needed to better understand the behavior and interaction of the different pathways from runoff and sediments that make up nutrients' transport processes.

Approximately 800 million people worldwide depend directly on sloping croplands with a slope gradient >10° for sustenance (Drees et al. 2003). Many classifications of slope steepness for soil take 10° as a starting point for ‘steep slopes’ (Shi et al. 2012; Peng et al. 2015).

The farmlands in the Three Gorges Reservoir Area are distributed almost entirely across steep sloped land. We thus focused on steep slopes (e.g., slope gradient higher than 10°), which are typical in the Wangjiaqiao watershed, in combination with different rainfall intensities and with purple soil as the test sediment. The objectives of this study were: (i) to measure and analyze the production mechanism of surface flow, subsurface flow, and sediment based on rainfall simulations; (ii) to investigate the characteristics of total nitrogen (TN) loss in different pathways from runoff and sediments under varied slope gradients and rainfall intensities; and (iii) to investigate the relationship between TN transport and hydraulic parameters in purple soil. We hypothesized that the soil erosion processes and the associated nitrogen loss on the purple soil slope can be described in terms of the combination of rainfall intensity and slope gradient.

Soil samples were collected from Wangjiaqiao watershed, Zigui Country in Hubei province, China to study soil and nutrient loss in the sloped cropland. The region has a subtropical monsoon climate with a mean annual precipitation of 1,016 mm that occurs mostly between May and September. Slope gradients within the watershed range from 2° to 58° with an average of 23°. Selected soils were derived from purple sandy shales, which are the main parent materials in the study watershed. Based on the US Soil Taxonomy, they were classified into entisols. In order to analyze the soil erosion and the distribution of nutrient on sloping cropland under simulated rainfall conditions, soil samples were taken from a 0 to 40 cm soil layer of sloping cropland. To remove impurities such as stones and grass, the air-dried soil samples were sieved through a 5 mm mesh. Selected soil physical and chemical properties, shown in Table 1, were determined by standard methods described in ISSAS (1978). The size distribution of soil aggregates was measured by wet sieving with three replicates. Fifty grams of the soil samples were placed in the top of a set of sieves with 5.0, 2.0, 1.0, 0.5, and 0.25 mm pore openings from top to bottom. The sieve-set was placed on the shock rack of a Yoder aggregates analyzer, immersed in distilled water. Each sample was sieved for a duration of 30 min at a frequency of 30 min⁻¹ and 3 cm amplitude of the movement.

The study was carried out in the soil erosion laboratory. The experimental equipment (Figure 1) included a rainfall simulator (side-sprinkle rainfall simulating set-up) and the runoff plot (the soil flume was 3 m in length, 1 m in width, and 0.5 m in depth). The runoff plot is a soil flume with a slope gradient adjustment device to allow the slope to be adjusted from 0 to 25°. Simulated rainfall was applied with lateral spraying nozzles parallel to the slope. The height of simulated rainfall was 16 m. The rainfall simulator can be set at any selected rainfall intensity ranging from 0.5 to 5.83 mm·min⁻¹ by adjusting the nozzles' aperture size and water pressure. The raindrop diameter distribution was 0.2 to 2 mm, and >85% of raindrop diameters were <1.0 mm. The raindrop kinetic energy distribution was 19.48 to 37.81 J/(m²·mm). Calibration of rainfall intensities was conducted before each experiment. A ‘V’-shape trough was inserted at the soil surface and via plastic pipes, into a surface runoff collection container. The soil profile at the downslope end of the flume was covered with an amino-plastic web and the gap between the aminoplastic web and the bottom of the flume was filled with gravel (Jia et al. 2007). A trench about 0.05 m deep and 1 m wide was constructed at the bottom of the flume. The outlet of the trench was connected to a SSR collection container.

A total of 15 treatment combinations were performed. According to the rainfall statistical characteristics in Wangjiaqiao watershed, five different rainfall intensities (0.6, 1.1, 1.61, 2.12, and 2.54 mm·min⁻¹) were selected for this study. These rainfall intensities were representative of the maximum 5-min precipitation (I₅), the maximum 10-min precipitation (I₁₀), the maximum 15-min precipitation (I₁₅), the maximum 30-min precipitation (I₃₀), and the maximum 60-min precipitation (I₆₀) in the study watershed (Wang et al. 2010); approximately 2.54, 2.12, 1.61, 1.1, and 0.6 mm·min⁻¹, respectively (the return period is 10-year), following a similar method as Shen et al. (2015), since high-intensity and short-duration rainstorms are primarily responsible for soil erosion in the Three Gorges Reservoir Area, China. Three slope gradients (10°, 15°, and 20°) were selected, representing typical slope gradients of the farmland of Wangjiaqiao watershed. Every simulated rainfall test was performed varying rainfall intensity but with a fixed rainfall volume of 50 mm. The duration of rainfall was controlled by the precipitation. The duration of rainfall was 83, 45, 31, 24, and 20 min for rainfall intensities of 0.6, 1.1, 1.61, 2.12, and 2.54 mm·min⁻¹, respectively, following a similar method as Ding & Zhang (2016). Four indices were used to research the TN transport process, including surface (SR) and SSR volume, the loss characteristics of TN, and the sediment yield (S).

The soil sample was packed in the flume layer with 10 cm increments with a final height of 40 cm, and its sand, silt, and clay contents were 46%, 23%, and 31%, respectively. The bulk density of packed soil in the flume was 1.35 g·cm^–3. Before each rain simulation, a known amount of rainfall was sprayed to minimize the differences in antecedent soil moisture. Both surface and SSR-collection pails at the flume outlet were changed periodically. At the beginning of runoff generation, samples were collected every 2 min. When the flow reached a stable state, the sampling interval was set to 5 min. After each rainfall, the amount of runoff in each pail was weighed on a balance. The sediment was deposited in the pails, and then a siphon was used to transfer the supernatant into polyethylene bottles. Soil samples were air dried and weighed, and all samples were preserved in a refrigerator at 4 °C; analysis was complete within 24 h of collection. Sediment yield was determined as the dry sediment mass, while sediment yield rate was defined as dividing sediment yield per unit area by the period of time. All treatments were conducted in three replicates. All runoff samples were analyzed for TN, using standard methods (APHA 1985). For the sediment samples that were yielded from the plot, sediment TN was determined by the Kjeldahl method.

Surface flow velocities (V_s) were measured by the dye method. Three measured sections were set up along the flume length direction at 1-m intervals. The mean surface flow velocities were the average of the measured values at each section. V_s was used to estimate the profile average runoff velocity (V) by the relation of ⁠, where n is a coefficient, laminar flow; it is 0.67 for laminar flow, 0.70 for transient flow, and 0.80 for turbulent flow (Abrahams et al. 1986).

One-way analysis of variance (ANOVA) was used to analyze the difference of runoff and sediment and the least significant difference (LSD) method at the 0.05 probability level was used to identify statistical differences among treatments. The relationships between flow hydraulics and TN losses were analyzed by a simple regression method. All of these analyses were conducted using SPSS 20.0 software.

The ANOVA showed that rainfall intensity and slope gradient had significant effects on runoff distribution and sediment yield. Mean surface runoff volume ranged from 38.6 to 49.1 mm (Figures 2 and 3). Surface runoff increased with increasing slope gradient for low (0.6 mm·min⁻¹) and high (2.12 and 2.54 mm·min⁻¹) rainfall intensities. For moderate rainfall intensities (1.1 and 1.6 mm·min⁻¹), surface runoff initially decreased to a minimum volume when the slope gradient was 15°, and then increased as the gradient increased further. Defersha & Melesse (2012) and Fang et al. (2015) found that the effects of slope and rainfall intensity on runoff and soil loss varied with soil type. In our experiment, surface runoff showed an decreasing–increasing trend as rainfall intensity increased. This may be associated with the physical properties of purple soil and effects of different rainfall intensity. Many studies indicate that soil surface sealing and rill development complicate runoff and erosion characteristics (Fang et al. 2015). Runoff discharge was increased when a seal was formed at the soil surface during rainfall (Assouline & Ben-Hur 2006). However, in this study where a seal layer is present at the soil surface, it seems that infiltration increases as rainfall intensity changed from low to moderate. Seal formation was less intense as rainfall intensity increased due to higher splash and sheet erosion, and increase of rill density and depth. Therefore, surface runoff decreased when the rainfall intensity changed from low to moderate. As rainfall intensity changed from moderate to high, infiltration excess runoff occurred when rainfall intensity exceeded the infiltration capacity. Consequently, surface runoff increased when the rainfall intensity changed from moderate to high. Moreover, average SSR volume ranged from 0.12 to 1.03 mm (Figures 2 and 3). Compared with the SSR, surface runoff for different slope gradients and simulated rainfall intensities was approximately 37.4 to 402.8 times the volumes for SSR. Furthermore, under the same slope gradient, SSR increased at first, then decreased as rainfall intensity increased. It is obvious that a long, low intensity rainfall can cause more SSR than a short high intensity rainfall (Jia et al. 2007).

According to Figures 2 and 3, the average sediment yield ranged from 1,008.13 to 3,447.77 g·m⁻². Both slope gradient and rainfall intensity had significant effects on the sediment yield. Sediment yield increased with increasing slope gradient. This pattern accorded with other studies (Defersha & Melesse 2012) in slope gradient and rainfall intensity increasing sediment yield. Ziadat & Taimeh (2013) reported that rainfall intensity was the most important factor affecting soil erosion and that erosion could occur at a relatively small intensity on wet soils as a result of subsequent rainfall events. In this study, sediment yield was more sensitive to the rainfall intensity than the slope gradient. Under conditions of low (0.6 mm·min⁻¹), moderate (1.1 and 1.61 mm·min⁻¹), and high (2.12 and 2.54 mm·min⁻¹) rainfall intensities, the sediment yield increased by 1.15, 1.19, 1.46, 1.58, and 1.89 times, respectively, as the slope gradients increased from 10° to 20°. On the other hand, for the slope gradients of 10°, 15°, and 20°, the sediment yield increased by 1.75, 2.48, and 2.91 times, respectively, as the rainfall intensities increased from 0.6 to 2.54 mm·min⁻¹.

The mass loss and concentrations of TN under different rainfall intensities and slope gradients are shown in Figure 4. In all three delivery mechanisms, the mass loss of TN (M-TN) was greatest for the sediment delivery mechanism, intermediate for the surface runoff pathway, and least for the SSR pathway. The M-TN_S values ranged from 4.13 to 10.97 kg·hm⁻², the M-TN_SR values from 0.87 to 3.56 kg·hm⁻², and the M-TN_SSR values from 0.04 to 0.55 kg·hm⁻². Our results are different from those of Luo et al. (2009), who argued that in purple soil nitrogen is easily leached out to the interface of rock–soil, thus the subsurface flow will carry more TN loss out of the soil at a slope gradient of 10°. The cause of these differences may be explained by the effective input water rate (greater rainfall intensities and steep slope gradients). There existed a critical slope of the purple soil slope, which was about 10°. Infiltration decreased greatly with slope when the slope was larger than the critical slope gradient (Qian et al. 2016). Moreover, an increase in slope gradient enhanced flow velocity and reduced the chance that runoff would be infiltrated into the soils (Fang et al. 2015). On steep slopes ranging from 10° to 20°, the influence of slope gradient on SSR volume was significant.

The range of M-TN values differed widely, and values did not show the same trend in all three delivery mechanisms. The mass of sediment-associated TN increases, and the mass of runoff-associated TN decreases, as the slope gradient increases. This may be due to the fact that nutrients such as nitrogen and phosphorus are basic components of the soil, which undergoes rainfall-driven erosion. Thus, the amount of nutrient loss is directly proportional to the amount of soil loss (Kothyari et al. 2004). In addition, this may be associated with the particle selectivity of erosion of deposited sediment. Zhang et al. (2011) found that nutrients were mainly contained within the clay fraction. Researchers have reported that interrill erosion results in selective removal of clay and silt-sized particles (Pan et al. 2016a). Gilley et al. (2012) reported that nutrient desorption, which was an important control for runoff-associated nutrient loss, was limited by the contact time between runoff and the soil. The transport of clay and silt-sized particles increases, and the contact time between runoff and the soil decreases, as the slope gradient increases.

The concentration of TN (C-TN) was another major factor that influenced TN loss. As presented in Figure 4, the C-TN_S, C-TN_SR, and C-TN_SSR values ranged from 0.3 to 0.43 mg·g⁻¹, 1.90 to 7.93 mg·L⁻¹, and 31.89 to 93.71 mg·L⁻¹, respectively. The average TN concentration in the SSR was over 12 times greater than that in the surface runoff. The results are in agreement with many studies pertaining to the impacts of SSR on nitrogen transport (Pionke et al. 2000; Jia et al. 2007).

The statistical model, Equation (9), may provide a better correlation coefficient (R² = 0.94) than Equation (8) (R² = 0.90) for estimating sediment-associated TN transport rate in purple soil.

The results of the current study indicate that TN transport was closely related to the rate of Manning roughness coefficient to average flow depth, Reynolds numbers, stream power, and flow velocity. Equations (7) and (9) were convenient and effective for predicting runoff-associated TN transport rate and sediment-associated TN transport rate, respectively. However, some studies found that particle size characteristics of sediments are important for controlling the off-site effects of sediment-bounded nutrients (Novara et al. 2013). Fine soil particles are readily transported by overland runoff (Issa et al. 2006). Palis et al. (1990) reported that fine soil particles are usually richer in soil-sorbed nutrients than are coarse sediments. Further studies should be conducted to investigate the effects of rainfall intensity on temporal change in eroded soil particles and sediment-bound nutrient loss.

This study investigated the loss characteristics of TN and the relationship between flow hydraulics and TN losses within a rainfall simulation system at various rainfall intensities and steep slope gradients. The results showed that surface runoff volume ranged from 38.6 to 49.1 mm, which was 37.4 to 402.8 times that for SSR. The average TN concentration in the SSR was over 12 times greater than that in the surface runoff. A high correlation between the rate of Manning roughness coefficient to average flow depth and the unit area runoff-associated TN transport rate (R² = 0.71) indicated that flow hydraulics are the major controlling factors of runoff-associated TN transport. Moreover, a significant positive linear relationship between sediment yield rate and MTN_S was obtained (R² = 0.98) indicating that sediment-associated TN transport was driven by soil loss. In this study, the main hydraulic indicators were selected to be used for sediment-associated TN loss modeling. Among the studied hydraulic parameters, Reynolds numbers (R² = 0.90), flow velocity (R² = 0.87), and stream power (R² = 0.73) showed a good linear relationship with the unit area sediment-associated TN transport rate, whereas Fr, f, and n/h were not good hydraulic indicators for the sediment-associated TN loss on purple soil erosion. The statistical model, Equation (9) (including stream power and flow velocity), provides a better correlation coefficient (R² = 0.94) than Equation (8) (R² = 0.90) for estimating sediment-associated TN transport rate in purple soil. The results can be used to improve the mechanisms of soil erosion and the effects of different pathways on nutrient loss.

Special thanks should be given to all of the crew members in Zigui County Water Resources Bureau. Funding for this research was provided by the National Natural Science Foundation of China (No. 41301297) and the Basic Research Fund for Central Public Research Institutes (No. CKSF2014026/TB).