Chinese Geographical Science 2006 16(4) 342–346 DOI 10.1007/s11769-006-0342-5 www.springerlink.com

Characteristics of Channeling Flow in Cultivated Horizon of Saline Rice Soil LUO Jinming1, DENG Wei2, ZHANG Xiaoping1, YANG Fan1, LI Xiujun1 (1. Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130012, China; 2. Institute of Mountain Hazards and Environment, Chinese Academy of Sciences, Chengdu 610041, China) Abstract: By applying bromide ion as tracer, the channeling flow has been quantitatively described in saline rice soil and alkaline soil of Da,an City, Jilin Province of China. Breakthrough curves of bromide ion in the saline rice soils after 1-year cultivation and 5-year cultivation and alkaline soil have been attained. Results show that the rice cultivation practice can improve the alkaline soil structure, however, it can accelerate the development of channeling flow pathway. Therefore, the channeling flow pathway has been developed widely in saline rice soil, but rarely in the alkaline soil. Three models of convection-dispersion equation (CDE), transfer functional model (TFM) and Back-Progation Network (BP Network) were used to simulate the transportation process of bromide ion. The peaks of probability density function of saline rice soil are higher with left skewed feature compared with that of the alkaline soil. It shows that the TFM and CDE can simulate the transportation process of the bromide ion in saline rice soil after 5-year cultivation, however, some deviation exists when it was used to simulate transportation process of bromide ion in saline rice soil after 1-year cultivation and alkaline soil; BP network can effectively simulate transportation process of bromide ion in both saline rice soil and alkaline soil. Keywords: channeling flow; saline rice soil; alkaline soil; transfer function model; convection-dispersion equation; Back-Progation Network

1 Introduction The preferential flow tends to result in the loss of irrigation water and nutrients. The formation of the preferential flow was caused by various factors, such as soil structure, tillage, worm cavity and root channeling rupture (Pot et al., 2005). Besides, frozen rupture can not be ignored in the northeastern China where the lowest temperature reaches –40℃ (Wang et al., 1993). There are several types of preferential flow, such as bypass flow, channeling flow and macropores flow. Channeling flow was studied in this paper. The channeling flow in alkaline soil is very slight because of the strong blocky structure of the alkaline horizon in surface (Wang et al., 1993). However, in the Songnen Plain, Northeast China, the local people diverted water from the Nenjiang River to cultivate rice, which leaches the salinity down to the subsurface of soil. Such practice made the soil permeability be improved a lot, however, the channeling flow was developed due

to rice cultivation. In Da’an City, located in the Songnen Plain, the irrigation level was as high as 15,000–19,500 t/ha for rice cultivation (Wang et al., 1996), but it was only 8000t/ha in Australia. On the other hand, the Da’an City belongs to semiarid area, agricultural water resources are quite deficient. Therefore, it is necessary to thoroughly study the channeling flow of saline rice soil in Da’an City, Jilin Province. At present, there are many studies on various models to character the preferential flow in non-saline soil and saline soil. Simunek et al. (1998; 2000) developed Hydrus-1D, 2D and UNSATCHEM models to simulate transportation of solute in porous material, which belong to determinate models; White et al. (1984) presented a stochastic model—transfer function model (TFM) to simulate the solute transportation in porous material; Ye (1990), and Ren et al. (1999; 2000; 2001) used the TFM to describe the preferential flow in non-saline soil; Luo and Deng (2000), Li and Dong (2002) and Wang et al. (2004) applied the artificial neural network to simulating the movement of

Received date: 2006-06-20; accepted date: 2006-09-28 Foundation item: Under the auspices of the Key Innovation Project of Chinese Academy of Sciences (No. KZCX1-SW-19-02) Biography: LUO Jinming (1977–), male, a native of Chengdu of Sichuan Province, Ph.D. candidate, specialized in hydrological ecology and soil water transportation. E-mail: [email protected]

Characteristics of Channeling Flow in Cultivated Horizon of Saline Rice Soil

343

chloride ion and evaluating the phreatic groundwater quality. But few dissertations were proposed to simulate the transportation of solute in alkaline soil worldwide. The objective of this study is to character the channeling flow in saline rice soil and simulate the channeling flow by convection-dispersion equation (CDE), transfer function model (TFM) and Back-Progation Network (BP Network) models.

2 Material and Methods 2.1 Study site The experimental spot is located in Da’an City (45°36′42″N, 124°03′01″E) of Jilin Province. The irrigation ditches have been built to divert water of the Nenjiang River to improve the physical character of alkaline soil since 1993. Nowadays, the total area of paddy field in the study area has exceeded 1000ha. The undisturbed soil cores were sampled from three plots: two saline rice soil plots after 1-year and 5-year cultivation with an area of 600m2 and an alkaline soil plot of 540m2 (Fig. 1).

Fig. 1 Experimental design

The soil properties have been described in detail ① . Four main horizons were identified in typical alkaline soil (Fig. 2). Alkalized horizon (0–5cm), which is abbreviated as Aa. Next layer is alkaline horizon (5–15cm), abbreviated as A+Bth. The third soil horizon is basic horizon (15–100cm), named as Bz layer. The deepest horizon is mother material layer (100–120cm), shorted as Cg, which was not presented in Fig. 2. A represents arable horizon and P, the subarable horizon into which alkali was leached from Aa or A+Bth. 2.2 Displacement experiment The bromide ion tracer displacement was conducted

Fig. 2 Soil structure of three plot profiles

according to Lee et al. (2001). Totally 45 undisturbed soil cores with a volume of 173cm3 (10cm in length, 5.7cm in diameter) were taken. The wall of PVC (polyvinyl chloride) plastic pipe was smeared by molten paraffin before being pressed into soil to avoid soil structure smearing and prevent the bromide ion flowing directly along the wall. A wire screen was attached to bottom. The upper surface was surrounded by a plastic ring tightly with the length of 3cm to ensure a certain water head. A medical dropper with a pinhead was hanged above the soil core to work as artificial raindrop instrument, the inlet velocity was controlled by medical dropper, and 0.086mol/L of potassium bromide solution continued to flow into the cores until the effluent concentration of bromide ion reached 95% of the influent of bromide ion concentration. After that, the cores were leached by distilled water until the effluent concentration of bromide ion was less than 5% of the influent concentration of the bromide ion. Effluent from the core was measured in each 10 minutes in first 4 hours, 30 minutes after 4 hours, and 1 hour when the effluent concentration reached 50% of influent concentration. The concentration of bromide ion was measured by an ion-selective electrode (made in Shanghai, China). 2.3 Models Nowadays, there are three types of models to simulate the solute transportation in both unsaturated and saturated conditions. The traditional one is convection-dispersion equation (CDE). Under steady-state condition, the transportation of the non-absorption solute can be described as follows

① Zhang Xiaoping, 1997. The research report about the soil resources of Da’an residual riverway. Changchun: Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, 1–24.

LUO Jinming, DENG Wei, ZHANG Xiaoping et al.

344

(Richards, 1931): ∂ Qout / ∂ t=D* ∂ Qout/ ∂ z2– (1) v* ∂ Qout / ∂ z where Qout is the mean concentration of the effluent during the observed time t; D is the dispersion coefficient of the soil core; z is the depth of the soil core; and v is the mean pore-flow-velocity. The other model is transfer function model (TFM) (White et al., 1984; Jury et al., 1986), which is a stochastic model. TFM regards the transport process of solute as a stochastic process, ignoring specific physical-chemical and biological mechanism of the transportation of the solute in soil profile. The concentration of the effluent at a certain moment can be expressed as the integral of input solute and the probability density function: Qout(t)=

∫

t 0

Qin (t' ) g [(t - t'/t )]dt'

(2)

where Qout(t) is the mean concentration of effluent and Qin(t ' ) is the mean concentration of influent during the period of t' ; g[(t–t' /t' ] is the transfer function. The third model is Back-Progation Network (BP Net work) (Rumelhart and Bernd, 1986), which has been applied to various natural sciences including the solute transportation simulation. BP Network has the advantages of simplicity, excellent performance and non-linear reflection capacity. The structure of the BP network is composed of input layer, disposal layer and output layer. The input vector of the network is composed of input layers. Target vector of the network is the ideal value of the output from the network. Input vector, transferred from the pre-neural unit, is to be disposed. The result of the input vector from pre-neural unit is named as output vector B=(b1, b2,… bn). The responding value Ct of every unit can be obtained once the output value from every disposal unit Lt is calculated: p

Lt =

∑V j =1

jt bi

− γt

j=1, 2, …., p

sen to fit the curve. The fitted equations are listed as follows: P(t)5=0.0086+0.0083/(1+exp((t–0.392)/0.671)) R2=0.99 P(t)1=0.0244+0.1458/(1+exp((t–0.408)/0.966)) R2=0.98 P(t)A=0.0128+6.4E9/(1+exp((t–1.45E6)/0.435)) R2=0.97 where P(t)5 represents the probability density distribution of 5-year-cultivation plot; P(t)1 is the probability density distribution of 1-year-cultivation plot; P(t)A is the probability density distribution of alkaline soil; t is the time (min).

(3)

Ct = f(Lt) t=1, 2, …, q (4) where Vjt represents connect weight factor of the jth unit at time t, bi is output cell of output vector B, γt is threshold.

3 Results and Discussion 3.1 Characteristics of channeling flow The probability density distribution of three plots were illustrated in Fig. 3. Boltzmann growth model was cho-

Fig. 3 Probability density distribution (P) with time (t) of three plots

It is evident that there were significant differences in different soil types, although the samples were from adjacent three plots. The distinguished P(t) distribution suggests that the solutes transport in soil column is different. The slopes of three curves are steep, however the gradient of P(t)5 is steeper than P(t)1. Another feature is that the probability density function of alkaline soil appears to be a long tail. The feature of P(t)5 indicates that there is abrupt solute distribution probability process in a short time. And large portion of solute enters the column to reach the balance with the effluent. The features of P(t)5 make it clear that a longer time is needed for the effluent to reach balance with influent. The balance-need time of 1-year-rice cultivation is between the P(t)5 and P(t)A, but more similar to P(t)5. The probability density functions are illustrated in Fig. 4. It is clear that the g(t|0) of rice soils are different to that of alkaline soil. The peaks are far higher with positively left skewed feature, which was used to indicate the heterogeneous property (Ren et al., 2000). The normal distribution (symmetry) in alkaline soil means relative homogeneity (no significant channeling flow) of the alkaline soil. The comparative high peak of probability density function

Characteristics of Channeling Flow in Cultivated Horizon of Saline Rice Soil

indicates that the velocity of solute to transport through the columns is more rapid (Pot et al., 2005).

Fig. 4 Probability density function (g(t|0)) along time t of the three soil plots

Another feature in Fig. 4 is that the time of the solute to transport through the column with maximum PDF (Tm) is higher than T0.5. Such difference can be used to indicate the preferential flow (Ren et al., 2000). The values of T0.5 and Tm were obtained by interpolation method with Matlab (version of 6.5). Ren et al. (2000) suggested (T0.5–Tm)/Tm to indicate the preferential flow. White et al. (1984) and Jury et al. (1986) proposed the concept of transport volume (θst) to denote the soil water content in the process of solute transportation. White et al. (1984) utilized mobile volume (θm) to describe the soil water content state during the process of the solute infiltration. The parameters of channeling flow based on Equation (2) are listed in Table 1. Table 1 Probability density function parameters and estimated values of transport volume (θst) and mobile volume (θm) Plot

Tm (min)

T0.5

(T0.5–Tm)/Tm

θst

θm

(min)

(%)

(%)

(%)

5-year rice soil

33.14

51.42

55.17

88.93

100.00

1-year rice soil

81.14

94.42

16.37

60.77

93.26

121.22

133.22

9.89

19.60

78.71

Alkaline soil

From Table 1, it can be seen that (T0.5–Tm)/Tm reaches 55.17% in rice soil after 5-year cultivation, which indicates the existence of transparent preferential flow. The ratio was sharply descended to 16.37% in the rice soil after 1-year cultivation. And the ratio was only 9.89% in alkaline soil, denoting that there is scare preferential flow in alkaline soil. The transport volume has the similar properties, but the gap between the 5-year and 1-year rice soils presented a greater disparity. Ren et al. (2000), White et al. (1984) and Jury et al. (1986) reached similar results. The reason for such difference is that due to cultivation practice, the channeling flow pathway has been widely

345

developed in 5-year rice soil, therefore, most of soil water can participate the transportation of bromide ion, while only very small part of soil water can do in alkaline soil. From Table 1, it can be seen that 88.93% of soil water participates the bromide ion transportation, but only 19.60% does in alkaline soil. Besides, 100% of soil water is mobile in 5-year rice soil. Nevertheless, 93.26% and 78.71% of soil water belongs to mobile water in 1-year rice soil and alkaline soil, respectively (Table 1). 3.2 Simulation by three models The simulation of breakthrough curves can indicate the process of the solute transportation, and can express the difference among three types of soils. In Fig. 5, the simulation lines of 5-year rice soil by TFM, CDE and BP Network present similar effluent distribution compared with the observed values. However, there were some deviations between the simulation result and observed data in 1-year rice soil and alkaline soil, especially that of the alkaline soil. The reason for the simulation discrepancy in alkaline soil and 1-year rice soil results from the blocky physical structure property, the CDE and TFM can not effectively simulate the solute transportation in alkaline horizon. The BP network shows perfect simulation results in both rice soils and the alkaline soil. Therefore, the TFM and CDE are comparative feasible model to describe the solute transportation in rice soil (non-alkaline or slight saline soil), but not effective to describe solute transportation of the alkali soil. The BP Network is a satisfying model to simulate the solute transportation in both alkaline soil and non-alkaline soil.

4 Conclusions The character of channeling flow in 5-year rice soil, 1-year rice soil and alkaline soil was discribed by applying bromide ion as tracer. The breakthrough process of bromide ion has been observed. It indicates that the blocky soil has been transformed into fertile culitvated soil in the plot after 5-year cultivition, howver the channeling flow pathway developed at the same time, which induces heterogeneous flow and water and nutrient loss . The probability density function of rice soil is obviously demonstrated a positively left skewed feature, and in a relative short time, there are a great deal of solution to transport through the soil core. The probability density function of the alkaline plot presents a normal distribution (symmetry) with long tailing feature. The mobile water ratio indicates that channeling flow in alkaline soil is scare.

346

LUO Jinming, DENG Wei, ZHANG Xiaoping et al.

Fig. 5 Changes of relative concentration (C/C0) of bromide ion with time (t) simulated by TFM, CDE and BP Network

Convection-dispersion equation (CDE), transfer functional model (TFM) and Back-Progation Network (BP Network) were applied to simulating the solute transportation in three types of soils. The results suggest that the TFM and CDE can simulate transportation process of bromide ion in non-alkaline soil, but there are some deviations in alkaline soil. Such deviations mean that the TFM and CDE are not effective to simulate the transportation in alkaline plot. The BP Network is proved to be a satisfying model to simulate transportation of bromide ion in three types of soils. References Jury W A, Aposito G, White R E, 1986. A transfer function model of solute transport through soil. Water Resources Research, 22: 243–247. Lee J, Horton R, Noborio K et al., 2001. Characterizing of preferential flow in undisturbed, structured soil columns using a vertical TDR probe. Journal of Contaminant Hydrology, 51: 131–144. Li Xingwang, Dong Manling, 2002. RBF Network method of evaluating water quality. Bulletin of Soil and Water Conservation, 22(3): 51–54. (in Chinese) Luo Xianxiang, Deng Wei, 2000. Sensitivity analysis and forecast on dynamics of soil salinization in West Plain of Songnen. Journal of Water and Soil Conservation, 14(3): 36–40. (in Chinese) Pot V, Simunek J, Benoit P, 2005. Impact of rainfall intensity on the transport of two herbicides in undisturbed grassed filter strip soil cores. Journal of Contaminant Hydrology, 81: 63–88. Ren Li, Li Baoguo, Ye Shuping et al., 1999. Transfer function approach of chloride travel in saturated soil under steady flow condition. Advances in Water Science, 2: 107–112. (in Chinese) Ren Li, Qin Yaodong,Wang Ji, 2001. Stomachstic modeling of chloride travel in non-homogeneous saturated soil under condi

tion of preferential transportation. Acta Pedologica Sinica, 38(1): 104–113. (in Chinese) Ren Li, Wang Ji, Qin Yaodong, 2000. Transfer function model on Cl transport in heterogeneous saturated soils under steady flow. Advances in Water Science, 4: 392–400. (in Chinese) Richards L A, 1931. Capillary conduction of liquids through porous media. Physics, 1: 318–333. Rumelhart J, Bernd L, 1986. The theory and application of Back-Progation Network model. Mathematics Science, 21: 336–348. Simunek J M, Van Genuchten, Sejna M, 1998. The HYDRUS-1D software package for simulating the movement of water, heat, and multiple solutes in variably saturated media. Riverside: U.S. Salinity Lab., 1–270. Simunek J, Sejna M, Van Genuchten, 2000. The HYDRUS-2D software package for simulating the two-dimensional movement of water, heat, and multiple solutes in variably-saturated media. Riverside: U.S. Salinity Lab., 1–251. Wang Debin, Wu Chunsheng, Li Xiujun et al., 1996. Economized irrigation techniques of well-irrigated paddy in saline alkali waterlogging lowland soil of Songnen Plain. In: Qiu Shanwen et al. (eds.). Agricultural Overall Development in Alkalized Land and Sandy Land in Songnen Plain. Beijing: Science Press. 71–75. (in Chinese) Wang Jinman, Yang Peiling, Ren Shumei et al., 2004. Simulation and prediction of soil solute transport on the basis of artificial neural work. Journal of Shenyang Agricultural University, 35(5–6): 486–488. (in Chinese) Wang Zhunqing, Zhu Shouquan, Yu Renpei, 1993. The Saline Soil in China. Beijing: Science Press. 15–23. (in Chinese) White R E, Thomas G W, Smith M S, 1984. Modeling water flow through undisturbed soil cores using a transfer function model derived from 3HOH and Cl transport. Journal of Soil Science, 35: 159–168. Ye Zitong, 1990. The utilization of salt transfer function model in studying salt-water movements in soil layers under infiltration. Journal of Hydraulic Engineering, 2: 1–9. (in Chinese)