- Email: [email protected]
Monitoring and modelling draining and resident soil water nitrate concentrations to estimate leaching losses
Agricultural Water Management 97 (2010) 1779–1786
Contents lists available at ScienceDirect
Agricultural Water Management journal homepage: www.elsevier.com/locate/agwat
Monitoring and modelling draining and resident soil water nitrate concentrations to estimate leaching losses M. van der Laan a,∗ , R.J. Stirzaker a,b , J.G. Annandale a , K.L. Bristow a,b , C.C. du Preez c a b c
Department of Plant Production and Soil Science, University of Pretoria, Pretoria 0002, South Africa CSIRO Sustainable Agricultural National Research Flagship and Cooperative Research Centre for Irrigation Futures, Australia Department of Soil, Crop and Climate Sciences, University of the Free State, Bloemfontein 9300, South Africa
a r t i c l e
i n f o
Article history: Received 8 March 2010 Accepted 21 June 2010 Available online 16 July 2010 Keywords: Nitrate leaching Monitoring Modelling Draining and resident soil water SWB-Sci
a b s t r a c t Quantifying nitrogen (N) losses below the root zone is highly challenging due to uncertainties associated with estimating drainage fluxes and solute concentrations in the leachate. Active and passive soil water samplers provide solute concentrations but give limited information on water fluxes. Mechanistic models are used to estimate leaching, but require calibration with measured data to ensure their reliability. Data from a drainage lysimeter trial under irrigation in which soil profile nitrate (NO3 − ) concentrations were monitored using wetting front detectors (passive sampler) and ceramic suction cups (active sampler) were compared to NO3 − concentrations in draining and resident soil water as simulated by the research version of the Soil Water Balance model (SWB-Sci). SWB-Sci is a daily time-step, cascading soil water and solute balance model that provides draining NO3 − concentrations by accounting for incomplete solute mixing. As hypothesized, suction cup concentrations aligned closely with resident soil water concentrations, while wetting front detector concentrations aligned closely with draining soil water NO3 − concentrations. These results demonstrate the power of combining monitoring and modelling to estimate NO3 − leaching losses. Access to measured draining and resident NO3 − concentrations, especially when complemented with modelled fluxes, can contribute greatly to achieving improved production and environmental objectives. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Minimizing nitrogen (N) leaching losses from cropping systems requires a good understanding of the key physical, chemical and biological processes impacting on solute transformations and movement in soils. Predicting the movement of solutes through soil is far more challenging than predicting the soil water status (Flühler et al., 1996), and additional uncertainties due to the heterogeneous nature of soils (Addiscott, 1996) makes the quantification of N leaching losses even more difficult. Although physical monitoring provides direct estimates of solute concentrations in soil water, uncertainties regarding the pore volume being sampled and drainage fluxes make estimation of actual leaching losses subject to potentially large errors. Mechanistic modelling can be used to obtain concentrations as well as fluxes, but such models often require extensive calibration using measured data, and uncertainty remains regarding how well the key processes are represented in the model (Keating et al., 2001).
∗ Corresponding author. Permanent address: South African Sugarcane Research Institute, Mount Edgecombe 4300, South Africa. Tel.: +27 31 508 7449. E-mail address: [email protected] (M. van der Laan). 0378-3774/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.agwat.2010.06.012
A range of devices has been developed over the years to sample soil water solutions, and are classified as either active or passive samplers, depending on whether action needs to be taken by the operator to obtain a sample (Litaor, 1988; Paramasivam et al., 1997). Active samplers, such as ceramic suction cups (SC), are commonly used worldwide. The wetting front detector (WFD) is a funnel shaped passive sampler which is buried in the soil and is able to alert a user when a wetting front (−2 to −3 kPa matric potential) has passed a specific depth in the soil by means of a visual indicator (Fig. 1) (see www.fullstop.com.au; Stirzaker, 2003, 2008). Following an irrigation/rainfall event, the funnel shape results in unsaturated flow lines converging towards a small cavity in its base where free water forms and can be sampled for chemical analysis. WFDs have been used successfully to improve understanding of the leaching of salts and NO3 − in a system to which high rates of municipal sludge were applied (Tesfamariam et al., 2009). Solute concentrations of soil water samples collected by active and passive samplers under similar conditions can differ markedly, and understanding the processes leading to these differences remains challenging (Haines et al., 1982). Passive samplers collect samples under relatively wet conditions, where solute concentrations are indicative of those in the water that is draining from one soil layer to the next. Active samplers provide solute concentra-
1780
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786 Table 1 Properties for the drainage lysimeter soil. Soil property
Value
pH (H2 O) Bulk density (kg m−3 ) Base saturation (%) ECe (dS m−1 ) CEC (cmol(c+) kg−1 ) C (%) Sand (%) Silt (%) Clay (%) Bray I P (mg kg−1 )
4.73 1120 44.52 1.40 4.418 1.11 72.3 9.66 18 11
Approaches to and implications of using monitoring and modelling together to estimate NO3 − leaching are discussed.
Fig. 1. (a) Schematic of the wetting front detector (WFD) passive sampler and (b) example of how WFD’s can be strategically placed in the root zone to indicate when a wetting front has reached a specific depth enabling collection of a sample of the draining water.
tions indicative of those in the resident soil water, defined as all the soil water in a layer at a specific time. When sampling resident water the sample collected consists of the soil water held at suction less than the suction applied to the device (Magid and Christensen, 1993). Advantages and disadvantages in the in-field deployment of active and passive samplers have been extensively reviewed in the literature (Silkworth and Grigal, 1981; Barbee and Brown, 1986). Water infiltrating through a soil profile is associated with a spectrum of pore-water velocities (Turner, 1958; Coats and Smith, 1964; Clothier et al., 1995; Ilsemann et al., 2002). Non-uniform solute movement has been observed as a result of faster flow through larger pores and slower flow in smaller pores (White, 1985). To account for this, models incorporate incomplete solute mixing algorithms to improve the description of solute movement in soil (Tillman et al., 1991; Corwin et al., 1991). The mobile phase of the resident water undergoes miscible displacement by incoming irrigation or rainfall water, while the immobile phase of the resident water is largely bypassed (Corwin et al., 1991). SWB-Sci is a mechanistic, generic crop model which has undergone extensive testing regarding its ability to simulate crop growth and the soil water balance (Jovanovic and Annandale, 1999, 2000; Jovanovic et al., 1999, 2000; Annandale et al., 2000; Tesfamariam, 2004). Recently, N and P modelling subroutines have been incorporated into the model and tested using several datasets from Zea mays L. (maize) and Triticum aestivum L. (winter wheat) trials (Van der Laan, 2009). Soil water is simulated using a multi-layered cascading approach and crop growth is simulated by calculating a daily dry matter increment which is either radiation or water limited. Incomplete solute mixing is based on the approach developed by Corwin et al. (1991) in which a mobility coefficient was used to improve simulations of chloride movement in a soil column when compared to complete piston-type displacement. This approach is discussed in more detail later in the paper. Crop N model testing exercises often compare measured and simulated values for aboveground crop N and inorganic soil N levels (Addiscott and Whitmore, 1987; De Willigen, 1991; Yang et al., 2000), but to the best of our knowledge, the approach developed by Corwin et al. (1991) or any similar approach has not been tested against measured NO3 − concentrations from active and passive samplers. The hypothesis tested in this paper is that simulated resident soil water NO3 − concentrations align with concentrations measured with SCs, while simulated draining soil water NO3 − concentrations align with concentrations measured with WFDs.
2. Materials and methods 2.1. Drainage lysimeter trial A drainage lysimeter with a volume of 6.1 m3 , a surface area of 4.7 m2 and a depth of 1.3 m was used to represent a typical rootzone to study leaching losses at the local scale. The lysimeter was packed with sandy clay loam (18% clay) in mid-2006 and allowed to settle naturally for 17 months. The lysimeter is located at the University of Pretoria Experimental Farm (25◦ 44 S 28◦ 15 E, 1370 m above sea level). A gravel layer was placed at the conical base of the lysimeter to facilitate drainage. The following instrumentation was installed into the lysimeter: suction cups (SCs) (Sentek, Australia) at 15, 30, 45, 60, 80 and 100 cm depths; wetting front detectors (WFDs) at 15, 30, 45 and 60 cm depths; and ECH2 O-TE sensors (Decagon, Pullman, Washington) at 15, 30, 45, 60 and 80 cm depths (hereafter referred to as capacitance sensors). Data characterizing the initial soil properties were obtained by averaging results from samples collected at 0–15, 15–30, 30–45, 45–60, 60–80 and 80–100 cm depths (Table 1). The vegetable test crop swiss chard (Beta vulgaris ssp. cicla) was used for this trial due to its ease of cultivation, relatively deep root system (∼80 cm) and because multiple harvests of the outer leaves can be made without having to re-sow the crop. Seedlings were transplanted into the lysimeter on 10 June 2008 (mid-winter) at an effective spacing of 20 cm × 30 cm. Harvesting was done by removing all leaves except the middle three from each plant. A representative 1 m2 plot was harvested and dry mass determined by drying in an oven at 60 ◦ C for 4–5 days. Leaf samples were analyzed for N content at each harvest, except for the final harvest when samples were lost, so an average N percentage for the three previous analyses was used for the final harvest value. Suction was applied to the SCs using a 60 ml syringe immediately following irrigation/rainfall. According to the manufacturers, pulling the piston of the syringe back 2–3 times creates a suction of 60–70 kPa. If available, soil water samples were collected from both the WFDs and SCs the day following irrigation or rainfall. Drainage from the lysimeter was captured in large drums from which the quantity could be measured and a water sample taken for analysis. For each sample, NO3 − was analyzed using an RQEasy Nitrate Reflectometer (Merck, Darmstadt, Germany). Irrigation was applied with the primary objective of minimizing both plant water stress and N leaching. Following planting, small amounts of irrigation were applied at regular intervals. Thereafter, irrigation was applied to allow the WFD placed at 15 cm to respond, and as daily crop water demand increased, water was increased to allow the WFD placed at 30 cm to respond. Applications were made at weekly intervals, or more often if judged necessary to minimize plant stress. Nitrogen fertilizer (as calcium ammonium nitrate) was applied when average NO3 − concentration from WFD samples fell
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
1781
Table 2 Nitrogen (N) fertilization over the growth season. Days after planting
N applied (kg ha−1 )
7 108 132 148 175
10 10 10 30 30
below 100 mg l−1 . Timing and application rate for N fertilization is presented in Table 2. The soil was limed and all other nutrients were applied as deemed necessary following a comprehensive soil analysis and were assumed to be non-limiting. 2.2. Modelling incomplete solute mixing Fig. 2. Rainfall and irrigation for the growth season.
Incomplete solute mixing is based on the approach developed by Corwin et al. (1991). This approach utilizes a mobility coefficient () which represents the fraction of the resident soil water that is subject to piston-type displacement, with the fraction 1 − ␥ therefore representing the fraction of the resident water that is bypassed. According to this approach, the draining soil water NO3 − concentration exiting a specific soil layer is calculated as follows: Cout =
Vbi Cbi − Vfc Cin + Vin Cin + (1.0 − )Vbi Cbi Vout
maximum drainage rate sets an upper limit on the drainage that can take place in 1 day. Simulated NO3 − concentrations in the water draining from each layer, and NO3 − concentrations in the soil water of a layer following the drainage event were calculated on a daily time-step and compared to measured concentrations from WFDs and SCs, respectively.
(1)
when Vin > Vfc − (1.0 − )Vbi
3. Results
Cout = Cbi
3.1. Rainfall and irrigation
when
Total water input over the growing season to the lysimeter included 495 mm of irrigation and 251 mm of rain. Most of the rain occurred 130 days after planting (DAP) (Fig. 2). Depending on antecedent water content, irrigation applications of 14–22 mm were required for the WFD at 15 cm to respond, while irrigation applications of 20–36 mm were required for the WFD at 30 cm to respond.
Vfc − Vbi < Vin ≤ Vfc − (1.0 − )Vbi where Vin : volume of water entering (m3 ), Vfc : volume of water at field capacity (m3 ), Vbi : volume of water before irrigation/rainfall (m3 ), Vout : volume of water exiting (m3 ), Cbi : solute concentration before irrigation (kg m−3 ), Cin : solute concentration of entering water (kg m−3 ), Cout : solute concentration of exiting water (kg m−3 ). Corwin et al. (1991) observed that a mobility coefficient of 0.5 provided best overall fit between measured and predicted values for the loam soil used in their trial. The average ratio of all WFD to SC NO3 − concentration measurements across depths over the growth season yielded a value of 0.3. This value was therefore used as the mobility coefficient for the sandy clay loam used in this trial. Crop growth parameters for swiss chard were obtained from a trial conducted in close proximity to the lysimeter trial in the summer of 1996/1997 (Jovanovic and Annandale, 2000). Further calibration for N modelling, involving the estimation of crop N uptake factors was done using data from a preliminary trial conducted during the previous season (2007). Soil analysis results were used as inputs for the model, including organic carbon %, texture, soil pH(H2 O) and cation exchange capacity. Initial soil NO3 − levels were based on a visual estimate using the first four SC NO3 − measurements to account for uncertainty associated with beginning of season measurements. Guided by the inorganic N analyses done for the soil, NH4 + levels were initialized at 1/6 of NO3 − values. Finally, calibration was carried out to match simulated cumulative drainage with end of season measurements through adjustment of the drainage factor (0–1) and drainage rate (mm d−1 ), with the aim of ultimately assessing the ability of the model to simulate dynamic changes in draining and resident soil water NO3 − concentrations. The calibration yielded a drainage factor of 0.95 and a maximum drainage rate of 20 mm d−1 . For a layer, water in excess of FC can potentially drain to the next layer, and the drainage factor determines what fraction of that water will drain each day. The
3.2. Soil volumetric water content and response of WFDs Measured versus simulated profile volumetric water content (VWC) data to a depth of 90 cm is presented in Fig. 3. Although there was good agreement between measured and simulated data early in the season when plant water extraction was lower, the model under-estimated profile water content later in the season. Trends for measured and simulated values were, however, very similar. Measured and simulated VWC (), and WFD response at depths of 15, 30, 45 and 60 cm is presented in Fig. 4. While there were peri-
Fig. 3. Measured and simulated profile water content over the growing season (measurements are based on data from the capacitance sensors at five depths).
1782
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
Fig. 5. (a) Cumulative crop dry matter removed and (b) cumulative crop nitrogen removed over the growth season.
also coincided with times that high water contents were simulated (Fig. 4). 3.3. Cumulative aboveground dry matter production and nitrogen uptake Total aboveground dry matter (TDM) production ranged from 1600 to 2900 kg ha−1 per harvest and cumulative dry matter removed was slightly under-estimated by the model (Fig. 5a). Aboveground N mass ranged from 51 to 70 kg N ha−1 per harvest (Fig. 5b), and cumulative crop N removed was initially overestimated by the model but in good agreement for the final two harvests. 3.4. Drainage and leaching Over the growth period, 45 mm of cumulative drainage was measured from the lysimeter, with the first appearance of deep drainage occurring 150 DAP (Fig. 6). Despite calibrating the model to obtain a comparable final volume, the measured and simulated values differed notably through the growth season. Simulated
Fig. 4. Measured and simulated volumetric water content (VWC), and wetting front detector (WFD) response at depths of 15, 30, 45 and 60 cm (WFD responses are plotted at the VWC associated with a matric potential of −3 kPa).
ods of difference between measured and simulated VWC – which could be partly due to soil heterogeneity and sensor accuracy – the model was judged to perform reasonably well overall in simulating soil layer VWC. The WFDs were clearly observed to respond when spikes in VWC were measured by the automatic sensors, and
Fig. 6. Measured and simulated cumulative drainage (mm) over the growth season.
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
Fig. 7. (a) Measured and simulated cumulative NO3 -N leached and (b) drainage water NO3 − concentrations.
drainage commenced later but then occurred more rapidly in comparison to measured drainage. This may partly be attributed to the nature of drainage from a lysimeter, in which a positive pressure at the lower boundary is required to create free water for drainage (Flury et al., 1999). A total of 86 kg NO3 -N was measured to have leached from the soil profile (Fig. 7a). Measured NO3 − concentrations in the drainage water increased rapidly from 330 mg l−1 at 151 DAP to 1008 mg l−1 on 168 DAP and thereafter remained relatively constant at around 1000 mg l−1 (Fig. 7b). Simulated drainage concentrations ranged from 862 to 750 mg l−1 . Similar to drainage, NO3 -N leaching was initially under-estimated, then over-estimated through the midseason period, with the final end of season simulated cumulative NO3 -N leached (78 kg NO3 -N) in reasonable agreement with the measured value (Fig. 7a). 3.5. Soil water nitrate concentrations High soil solution NO3 − concentrations were observed at all depths at planting despite no fertilization having taken place since the previous season (Fig. 8). These high NO3 − concentrations are therefore attributed to mineralization occurring over a 4-month fallow period during which very little drainage took place. After planting, the removal of N from the system by an actively growing crop is observable in the measured data. In almost all cases, measured NO3 − concentrations from WFDs were less than those measured from SCs. This is consistent with lower solute concentrations found in the draining soil water phase due to incomplete mixing with the immobile soil water phase as has been observed in other experiments (Van Hoorn, 1981; Corwin et al., 1991). The addition of 10 kg N ha−1 7 DAP is observable by an associated increase in NO3 − concentration detected by the SCs placed at 15, 30 and 45 cm (Fig. 8). The effect of a second addition of 10 kg N ha−1 108 DAP is only observable in the SC and WFD at 15 cm. A third
1783
addition of 10 kg N ha−1 132 DAP does not result in a clear increase in SC NO3 − concentration at any of the depths. After an addition of 30 kg N ha−1 148 DAP, a sharp increase in NO3 − concentration followed by an immediate sharp decrease was observed in the SCs placed at 15 and 30 cm. An increase in NO3 − concentration for the WFD placed at 15 cm was also observed. The final application of 30 kg N ha−1 175 DAP did not cause clearly observable increases in NO3 − concentration in either the SCs or WFDs. As additions of fertilizer N were more clearly reflected at the beginning of the season when the crop did not yet have a fully developed root system, this NO3 -N ‘disappearance’ is therefore mostly attributed to rapid crop uptake. The onset of the rainy season 130 DAP clearly moved NO3 − down the soil profile, as observed from both the SC and WFD data. SCs placed at 45 and 60 cm showed an increase in NO3 − concentration after the onset of rain, with measurements demonstrating a pulse of NO3 − moving down the profile. A large increase in NO3 − concentration in the WFD placed at 60 cm 185 DAP is also consistent with the movement of a NO3 − pulse down the profile. Comparisons between measured and simulated data (Fig. 8) provide clear evidence that the SC concentrations reflect the concentrations in the resident soil water, while the WFD concentrations reflect those in the draining soil water. For both sets of comparisons, measured and simulated values showed similar trends to a depth of 60 cm, although simulated values did not fluctuate as much as the measured values. For the SCs at 80 and 100 cm, a sharp decline in NO3 − concentration can be observed after the onset of the rainy season (Fig. 9). This is after an initial increase in NO3 − concentration prior to 150 DAP. These data indicate that N is also moving past the 80–100 cm depth, as is confirmed by the leachate data collected at the base of the lysimeter. The initial increase in measured SC NO3 − concentrations at 80 and 100 cm was under-estimated by SWB-Sci. The rapid decrease in NO3 − concentration after 150 DAP was also under-estimated by the model, especially at the 100 cm depth.
4. Discussion Nitrate (NO3 − ) concentrations sampled from SCs were almost always higher than those sampled from WFDs for this particular soil and experiment. As measurements and comparisons were always made following a wetting event, this is a result of incomplete mixing between the resident and draining water phases. Good comparisons were observed between measured NO3 − concentrations from the SCs and simulated resident soil water concentrations, and measured concentrations from the WFDs and simulated soil water NO3 − concentrations draining from one layer to the next. This indicates that these samplers clearly sample different components of the soil water, and that the use of the Corwin et al. (1991) approach incorporated in a straightforward cascading soil water balance model with a daily time-step was effective in modelling the influence of incomplete mixing on solute transport. Furthermore, using WFD:SC NO3 − concentration ratios served as an effective estimate of the mobility coefficient. A major implication of this is that measuring and modelling can be used together to improve estimates of NO3 -N leaching losses. In using such an approach, a mechanistic crop N model (SWB-Sci in this case) is used together with data from WFDs and SCs to parameterise/calibrate and test the model for a specific site. Under field conditions, drainage data are rarely available to calibrate a model, as was done in this study. From the results presented in this paper, however, SC and corresponding WFD NO3 − concentration data can be used to estimate specific drainage parameters for the soil, and if
1784
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
Fig. 8. Measured nitrate (NO3 − ) concentrations from suction cups compared to simulated resident soil water concentrations (left) and measured NO3 − concentrations from wetting front detectors compared to simulated draining soil water concentrations (right) at depths of 15, 30, 45 and 60 cm.
SC and WFD NO3 − concentrations are well estimated by the model over time, then it can be assumed that drainage will also be accurately estimated. In addition, measured N concentrations together with water flux estimates obtained from a simple crop soil water balance model can also be used to estimate leaching. Although limitations in this dataset for model testing purposes are recognized, the ability of the model to estimate soil water, crop growth, N uptake and N leaching was judged to be adequate. Mechanistic crop N models aim to simulate key processes accurately.
Significant improvements have been made to the SWB-Sci model in order to achieve this. There is still some uncertainty, however, on how to best represent and test certain processes. For example a sharp decline in soil NO3 − concentrations was measured in the lower soil layers (80 and 100 cm) but not simulated by the model. Using data from instruments such as SCs and WFDs is proving useful in this regard. We recommend further work testing and applying the approaches proposed in this paper for a wide range of cropping systems a soil types to enhance their robustness in sup-
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
1785
doing so. Ultimately, it is envisaged that increased efforts to measure and model draining and resident NO3 − concentrations will lead to more efficient and less environmentally detrimental crop management practices. Acknowledgements We gratefully acknowledge the support of the Water Research Commission (Project 1516 ‘Development of an integrated modelling approach to prediction of agricultural non-point source (NPS) pollution from field to catchment scales for selected agricultural NPS pollutants’ and Project 1574 ‘Adapting the Wetting Front Detector to the needs of small-scale furrow irrigators and providing a basis for the interpretation of salt and nutrient measurements from the water sample’) and National Research Foundation of South Africa, and CSIRO and the Cooperative Research Centre for Irrigation Futures from Australia. References
Fig. 9. Measured nitrate (NO3 − ) concentrations from suction cups compared to simulated resident soil water concentrations at depths of 80 and 100 cm.
porting improved understanding of NO3 − dynamics and losses in agro-ecosystems. In addition to using SCs and WFDs to estimate leaching, basing adaptive management fertilization strategies on measured concentrations shows excellent potential. More work however still needs to be done before crop and soil specific guidelines based on measurements made from SCs and/or WFDs can be put into practice. Using predicted end of season total crop transpiration and N uptake to calculate the NO3 − concentration required in the soil water taken up by the roots to meet this total crop N demand could be a good way to start. Such an approach would help reduce over-fertilization, thereby maintaining crop yields while reducing N concentrations in the deep drainage leaving the root zone. Further research on such an approach is therefore recommended. Finally, access to the concentration data, especially resident water NO3 − concentration data, allows focus on improving N management from a crop production perspective, while access to the draining water NO3 − concentrations and fluxes allows focus on improving N management from an environmental perspective. 5. Conclusions Nitrogen leaching from agriculture can cost farmers dearly and pose a serious threat to receiving water bodies, but simple and effective ways of estimating these leaching losses are lacking. A diversity of approaches, ranging widely in levels of complexity, has been proposed to model solute concentrations in soil water. SWBSci, incorporating the functional solute bypass approach proposed by Corwin et al. (1991), was found to adequately simulate draining and resident soil water NO3 − concentrations that reflect the concentrations measured with WFDs and SCs, respectively. This work reinforces the value of combining monitoring and modelling to estimate solute leaching and proposes a pragmatic approach for
Addiscott, T.M., 1996. Measuring and modelling nitrogen leaching: parallel problems. Plant Soil 181, 1–6. Addiscott, T.M., Whitmore, A.P., 1987. Computer simulation of changes in soil mineral nitrogen and crop nitrogen during autumn, winter and spring. J. Agric. Sci. 109, 141–157. Annandale, J.G., Campbell, G.S., Olivier, F.C., Jovanovic, N.Z., 2000. Predicting crop water uptake under full and deficit irrigation. An example using pea (Pisum sativum cv. Puget). Irrig. Sci. 19, 65–72. Barbee, G.C., Brown, K.W., 1986. Comparison between suction and free-drainage soil solution samplers. Soil Sci. 141, 149–154. Clothier, B.E., Heng, L., Magesan, G.N., Vogeler, I., 1995. The measured mobile-water content of an unsaturated soil as a function of hydraulic regime. Aust. J. Soil Res. 33, 397–414. Coats, K.H., Smith, B.D., 1964. Dead end pore volume and dispersion in porous media. Soc. Pet. Eng. J. 4, 74–84. Corwin, D.L., Waggoner, B.L., Rhoades, J.D., 1991. A function model of solute transport that accounts for bypass flow. J. Environ. Qual. 20, 647–658. De Willigen, P., 1991. Nitrogen turnover in the soil-crop system; comparison of fourteen simulation models. Fert. Res. 27, 141–150. Flühler, H., Durner, W., Flury, M., 1996. Lateral solute mixing processes—a key for understanding field-scale transport of water and solutes. Geoderma 70, 165–183. Flury, M., Yates, M.V., Jury, W.A., 1999. Numerical analysis of the effect of the lower boundary on solute transport in lysimeters. Soil Sci. Soc. Am. J. 63, 1493–1499. Haines, B.L., Waide, J.B., Todd, R.L., 1982. Soil solution nutrient concentrations sampled with tension and zero-tension lysimeters: report of discrepancies. Soil Sci. Soc. Am. J. 46, 658–661. Ilsemann, J., van der Ploeg, R.R., Horton, R., Bachmann, J., 2002. Laboratory method for determining immobile soil water content and mass exchange coefficient. J. Plant Nutr. Soil Sci. 165, 332–338. Jovanovic, N.Z., Annandale, J.G., 1999. An FAO crop factor modification to SWB makes inclusion of crops with limited data possible: examples for vegetable crops. Water SA 25, 181–190. Jovanovic, N.Z., Annandale, J.G., Mhlauli, N.C., 1999. Field water balance and SWB parameter determination of six winter vegetable species. Water SA 25, 191–196. Jovanovic, N.Z., Annandale, J.G., 2000. Crop growth model parameters of 19 summer vegetable cultivars for use in mechanistic irrigation scheduling models. Water SA 26, 67–76. Jovanovic, N.Z., Annandale, J.G., Nel, A.A., 2000. Calibration and validation of the SWB model for sunflower (Helianthus annuus L.). S. Afr. J. Plant Soil 17, 117–123. Keating, B.A., Verburg, K., Smith, C.J., Probert, M.E., Gaydon, D., 2001. Assessing leakiness in Australia’s dryland farming systems. In: Proceedings of the International Congress on Modelling and Simulation, Australian National University, December, 2001, pp. 1811–1816. Litaor, M.I., 1988. Review of soil solution samplers. Water Resour. Res. 24, 727–733. Magid, J., Christensen, N., 1993. Soil solution samples with and without tension in arable and heathland soils. Soil Sci. Soc. Am. J. 57, 1463–1469. Paramasivam, S., Alva, A.K., Fares, A., 1997. Vadose zone soil solution sampling techniques to investigate pollutant transport in soils. Trends Soil Sci. 2, 115–136. Silkworth, D.R., Grigal, D.F., 1981. Field comparison of soil solution samplers. Soil Sci. Soc. Am. J. 45, 440–444. Stirzaker, R.J., 2003. When to turn the water off: scheduling micro-irrigation with a wetting front detector. Irrig. Sci. 22, 177–185. Stirzaker, R.J., 2008. Factors affecting sensitivity of wetting front detectors. Acta Hort. 792, 647–653. Tesfamariam, E.H., 2004. Modelling the soil water balance of Canola (Brassica napus L.) (Hyola 60). MSc(Agric.) Dissertation. University of Pretoria. Tesfamariam, E.H., Annandale, J.G., Steyn, J.M., Stirzaker, R.J., 2009. Exporting large volumes of municipal sludge through turfgrass sod production. J. Environ. Qual. 38, 1320–1328.
1786
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
Tillman, R.W., Scotter, D.R., Clothier, B.E., White, R.E., 1991. Solute movement during intermittent water flow in a field soil and some implications for irrigation and fertilizer management. Agric. Water Manage. 20, 119–133. Turner, G.A., 1958. The flow-structure in packed beds. Chem. Eng. Sci. 7, 156–166. Van der Laan, M., 2009. Development, testing and application of a crop nitrogen and phosphorus model to investigate leaching losses at the local scale. PhD Dissertation. University of Pretoria, Pretoria.
Van Hoorn, J.W., 1981. Salt movement, leaching efficiency, and leaching requirement. Agric. Water Manage. 4, 409–428. White, R.E., 1985. The influence of macropores on the transport of dissolved and suspended matter through soil. Adv. Soil. Sci. 3, 95–120. Yang, J., Greenwood, D.J., Rowell, D.L., Wadsworth, G.A., Burns, I.G., 2000. Statistical methods for evaluating a nitrogen simulation model, N ABLE. Agric. Sys. 64, 37–53.
Contents lists available at ScienceDirect
Agricultural Water Management journal homepage: www.elsevier.com/locate/agwat
Monitoring and modelling draining and resident soil water nitrate concentrations to estimate leaching losses M. van der Laan a,∗ , R.J. Stirzaker a,b , J.G. Annandale a , K.L. Bristow a,b , C.C. du Preez c a b c
Department of Plant Production and Soil Science, University of Pretoria, Pretoria 0002, South Africa CSIRO Sustainable Agricultural National Research Flagship and Cooperative Research Centre for Irrigation Futures, Australia Department of Soil, Crop and Climate Sciences, University of the Free State, Bloemfontein 9300, South Africa
a r t i c l e
i n f o
Article history: Received 8 March 2010 Accepted 21 June 2010 Available online 16 July 2010 Keywords: Nitrate leaching Monitoring Modelling Draining and resident soil water SWB-Sci
a b s t r a c t Quantifying nitrogen (N) losses below the root zone is highly challenging due to uncertainties associated with estimating drainage fluxes and solute concentrations in the leachate. Active and passive soil water samplers provide solute concentrations but give limited information on water fluxes. Mechanistic models are used to estimate leaching, but require calibration with measured data to ensure their reliability. Data from a drainage lysimeter trial under irrigation in which soil profile nitrate (NO3 − ) concentrations were monitored using wetting front detectors (passive sampler) and ceramic suction cups (active sampler) were compared to NO3 − concentrations in draining and resident soil water as simulated by the research version of the Soil Water Balance model (SWB-Sci). SWB-Sci is a daily time-step, cascading soil water and solute balance model that provides draining NO3 − concentrations by accounting for incomplete solute mixing. As hypothesized, suction cup concentrations aligned closely with resident soil water concentrations, while wetting front detector concentrations aligned closely with draining soil water NO3 − concentrations. These results demonstrate the power of combining monitoring and modelling to estimate NO3 − leaching losses. Access to measured draining and resident NO3 − concentrations, especially when complemented with modelled fluxes, can contribute greatly to achieving improved production and environmental objectives. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Minimizing nitrogen (N) leaching losses from cropping systems requires a good understanding of the key physical, chemical and biological processes impacting on solute transformations and movement in soils. Predicting the movement of solutes through soil is far more challenging than predicting the soil water status (Flühler et al., 1996), and additional uncertainties due to the heterogeneous nature of soils (Addiscott, 1996) makes the quantification of N leaching losses even more difficult. Although physical monitoring provides direct estimates of solute concentrations in soil water, uncertainties regarding the pore volume being sampled and drainage fluxes make estimation of actual leaching losses subject to potentially large errors. Mechanistic modelling can be used to obtain concentrations as well as fluxes, but such models often require extensive calibration using measured data, and uncertainty remains regarding how well the key processes are represented in the model (Keating et al., 2001).
∗ Corresponding author. Permanent address: South African Sugarcane Research Institute, Mount Edgecombe 4300, South Africa. Tel.: +27 31 508 7449. E-mail address: [email protected] (M. van der Laan). 0378-3774/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.agwat.2010.06.012
A range of devices has been developed over the years to sample soil water solutions, and are classified as either active or passive samplers, depending on whether action needs to be taken by the operator to obtain a sample (Litaor, 1988; Paramasivam et al., 1997). Active samplers, such as ceramic suction cups (SC), are commonly used worldwide. The wetting front detector (WFD) is a funnel shaped passive sampler which is buried in the soil and is able to alert a user when a wetting front (−2 to −3 kPa matric potential) has passed a specific depth in the soil by means of a visual indicator (Fig. 1) (see www.fullstop.com.au; Stirzaker, 2003, 2008). Following an irrigation/rainfall event, the funnel shape results in unsaturated flow lines converging towards a small cavity in its base where free water forms and can be sampled for chemical analysis. WFDs have been used successfully to improve understanding of the leaching of salts and NO3 − in a system to which high rates of municipal sludge were applied (Tesfamariam et al., 2009). Solute concentrations of soil water samples collected by active and passive samplers under similar conditions can differ markedly, and understanding the processes leading to these differences remains challenging (Haines et al., 1982). Passive samplers collect samples under relatively wet conditions, where solute concentrations are indicative of those in the water that is draining from one soil layer to the next. Active samplers provide solute concentra-
1780
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786 Table 1 Properties for the drainage lysimeter soil. Soil property
Value
pH (H2 O) Bulk density (kg m−3 ) Base saturation (%) ECe (dS m−1 ) CEC (cmol(c+) kg−1 ) C (%) Sand (%) Silt (%) Clay (%) Bray I P (mg kg−1 )
4.73 1120 44.52 1.40 4.418 1.11 72.3 9.66 18 11
Approaches to and implications of using monitoring and modelling together to estimate NO3 − leaching are discussed.
Fig. 1. (a) Schematic of the wetting front detector (WFD) passive sampler and (b) example of how WFD’s can be strategically placed in the root zone to indicate when a wetting front has reached a specific depth enabling collection of a sample of the draining water.
tions indicative of those in the resident soil water, defined as all the soil water in a layer at a specific time. When sampling resident water the sample collected consists of the soil water held at suction less than the suction applied to the device (Magid and Christensen, 1993). Advantages and disadvantages in the in-field deployment of active and passive samplers have been extensively reviewed in the literature (Silkworth and Grigal, 1981; Barbee and Brown, 1986). Water infiltrating through a soil profile is associated with a spectrum of pore-water velocities (Turner, 1958; Coats and Smith, 1964; Clothier et al., 1995; Ilsemann et al., 2002). Non-uniform solute movement has been observed as a result of faster flow through larger pores and slower flow in smaller pores (White, 1985). To account for this, models incorporate incomplete solute mixing algorithms to improve the description of solute movement in soil (Tillman et al., 1991; Corwin et al., 1991). The mobile phase of the resident water undergoes miscible displacement by incoming irrigation or rainfall water, while the immobile phase of the resident water is largely bypassed (Corwin et al., 1991). SWB-Sci is a mechanistic, generic crop model which has undergone extensive testing regarding its ability to simulate crop growth and the soil water balance (Jovanovic and Annandale, 1999, 2000; Jovanovic et al., 1999, 2000; Annandale et al., 2000; Tesfamariam, 2004). Recently, N and P modelling subroutines have been incorporated into the model and tested using several datasets from Zea mays L. (maize) and Triticum aestivum L. (winter wheat) trials (Van der Laan, 2009). Soil water is simulated using a multi-layered cascading approach and crop growth is simulated by calculating a daily dry matter increment which is either radiation or water limited. Incomplete solute mixing is based on the approach developed by Corwin et al. (1991) in which a mobility coefficient was used to improve simulations of chloride movement in a soil column when compared to complete piston-type displacement. This approach is discussed in more detail later in the paper. Crop N model testing exercises often compare measured and simulated values for aboveground crop N and inorganic soil N levels (Addiscott and Whitmore, 1987; De Willigen, 1991; Yang et al., 2000), but to the best of our knowledge, the approach developed by Corwin et al. (1991) or any similar approach has not been tested against measured NO3 − concentrations from active and passive samplers. The hypothesis tested in this paper is that simulated resident soil water NO3 − concentrations align with concentrations measured with SCs, while simulated draining soil water NO3 − concentrations align with concentrations measured with WFDs.
2. Materials and methods 2.1. Drainage lysimeter trial A drainage lysimeter with a volume of 6.1 m3 , a surface area of 4.7 m2 and a depth of 1.3 m was used to represent a typical rootzone to study leaching losses at the local scale. The lysimeter was packed with sandy clay loam (18% clay) in mid-2006 and allowed to settle naturally for 17 months. The lysimeter is located at the University of Pretoria Experimental Farm (25◦ 44 S 28◦ 15 E, 1370 m above sea level). A gravel layer was placed at the conical base of the lysimeter to facilitate drainage. The following instrumentation was installed into the lysimeter: suction cups (SCs) (Sentek, Australia) at 15, 30, 45, 60, 80 and 100 cm depths; wetting front detectors (WFDs) at 15, 30, 45 and 60 cm depths; and ECH2 O-TE sensors (Decagon, Pullman, Washington) at 15, 30, 45, 60 and 80 cm depths (hereafter referred to as capacitance sensors). Data characterizing the initial soil properties were obtained by averaging results from samples collected at 0–15, 15–30, 30–45, 45–60, 60–80 and 80–100 cm depths (Table 1). The vegetable test crop swiss chard (Beta vulgaris ssp. cicla) was used for this trial due to its ease of cultivation, relatively deep root system (∼80 cm) and because multiple harvests of the outer leaves can be made without having to re-sow the crop. Seedlings were transplanted into the lysimeter on 10 June 2008 (mid-winter) at an effective spacing of 20 cm × 30 cm. Harvesting was done by removing all leaves except the middle three from each plant. A representative 1 m2 plot was harvested and dry mass determined by drying in an oven at 60 ◦ C for 4–5 days. Leaf samples were analyzed for N content at each harvest, except for the final harvest when samples were lost, so an average N percentage for the three previous analyses was used for the final harvest value. Suction was applied to the SCs using a 60 ml syringe immediately following irrigation/rainfall. According to the manufacturers, pulling the piston of the syringe back 2–3 times creates a suction of 60–70 kPa. If available, soil water samples were collected from both the WFDs and SCs the day following irrigation or rainfall. Drainage from the lysimeter was captured in large drums from which the quantity could be measured and a water sample taken for analysis. For each sample, NO3 − was analyzed using an RQEasy Nitrate Reflectometer (Merck, Darmstadt, Germany). Irrigation was applied with the primary objective of minimizing both plant water stress and N leaching. Following planting, small amounts of irrigation were applied at regular intervals. Thereafter, irrigation was applied to allow the WFD placed at 15 cm to respond, and as daily crop water demand increased, water was increased to allow the WFD placed at 30 cm to respond. Applications were made at weekly intervals, or more often if judged necessary to minimize plant stress. Nitrogen fertilizer (as calcium ammonium nitrate) was applied when average NO3 − concentration from WFD samples fell
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
1781
Table 2 Nitrogen (N) fertilization over the growth season. Days after planting
N applied (kg ha−1 )
7 108 132 148 175
10 10 10 30 30
below 100 mg l−1 . Timing and application rate for N fertilization is presented in Table 2. The soil was limed and all other nutrients were applied as deemed necessary following a comprehensive soil analysis and were assumed to be non-limiting. 2.2. Modelling incomplete solute mixing Fig. 2. Rainfall and irrigation for the growth season.
Incomplete solute mixing is based on the approach developed by Corwin et al. (1991). This approach utilizes a mobility coefficient () which represents the fraction of the resident soil water that is subject to piston-type displacement, with the fraction 1 − ␥ therefore representing the fraction of the resident water that is bypassed. According to this approach, the draining soil water NO3 − concentration exiting a specific soil layer is calculated as follows: Cout =
Vbi Cbi − Vfc Cin + Vin Cin + (1.0 − )Vbi Cbi Vout
maximum drainage rate sets an upper limit on the drainage that can take place in 1 day. Simulated NO3 − concentrations in the water draining from each layer, and NO3 − concentrations in the soil water of a layer following the drainage event were calculated on a daily time-step and compared to measured concentrations from WFDs and SCs, respectively.
(1)
when Vin > Vfc − (1.0 − )Vbi
3. Results
Cout = Cbi
3.1. Rainfall and irrigation
when
Total water input over the growing season to the lysimeter included 495 mm of irrigation and 251 mm of rain. Most of the rain occurred 130 days after planting (DAP) (Fig. 2). Depending on antecedent water content, irrigation applications of 14–22 mm were required for the WFD at 15 cm to respond, while irrigation applications of 20–36 mm were required for the WFD at 30 cm to respond.
Vfc − Vbi < Vin ≤ Vfc − (1.0 − )Vbi where Vin : volume of water entering (m3 ), Vfc : volume of water at field capacity (m3 ), Vbi : volume of water before irrigation/rainfall (m3 ), Vout : volume of water exiting (m3 ), Cbi : solute concentration before irrigation (kg m−3 ), Cin : solute concentration of entering water (kg m−3 ), Cout : solute concentration of exiting water (kg m−3 ). Corwin et al. (1991) observed that a mobility coefficient of 0.5 provided best overall fit between measured and predicted values for the loam soil used in their trial. The average ratio of all WFD to SC NO3 − concentration measurements across depths over the growth season yielded a value of 0.3. This value was therefore used as the mobility coefficient for the sandy clay loam used in this trial. Crop growth parameters for swiss chard were obtained from a trial conducted in close proximity to the lysimeter trial in the summer of 1996/1997 (Jovanovic and Annandale, 2000). Further calibration for N modelling, involving the estimation of crop N uptake factors was done using data from a preliminary trial conducted during the previous season (2007). Soil analysis results were used as inputs for the model, including organic carbon %, texture, soil pH(H2 O) and cation exchange capacity. Initial soil NO3 − levels were based on a visual estimate using the first four SC NO3 − measurements to account for uncertainty associated with beginning of season measurements. Guided by the inorganic N analyses done for the soil, NH4 + levels were initialized at 1/6 of NO3 − values. Finally, calibration was carried out to match simulated cumulative drainage with end of season measurements through adjustment of the drainage factor (0–1) and drainage rate (mm d−1 ), with the aim of ultimately assessing the ability of the model to simulate dynamic changes in draining and resident soil water NO3 − concentrations. The calibration yielded a drainage factor of 0.95 and a maximum drainage rate of 20 mm d−1 . For a layer, water in excess of FC can potentially drain to the next layer, and the drainage factor determines what fraction of that water will drain each day. The
3.2. Soil volumetric water content and response of WFDs Measured versus simulated profile volumetric water content (VWC) data to a depth of 90 cm is presented in Fig. 3. Although there was good agreement between measured and simulated data early in the season when plant water extraction was lower, the model under-estimated profile water content later in the season. Trends for measured and simulated values were, however, very similar. Measured and simulated VWC (), and WFD response at depths of 15, 30, 45 and 60 cm is presented in Fig. 4. While there were peri-
Fig. 3. Measured and simulated profile water content over the growing season (measurements are based on data from the capacitance sensors at five depths).
1782
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
Fig. 5. (a) Cumulative crop dry matter removed and (b) cumulative crop nitrogen removed over the growth season.
also coincided with times that high water contents were simulated (Fig. 4). 3.3. Cumulative aboveground dry matter production and nitrogen uptake Total aboveground dry matter (TDM) production ranged from 1600 to 2900 kg ha−1 per harvest and cumulative dry matter removed was slightly under-estimated by the model (Fig. 5a). Aboveground N mass ranged from 51 to 70 kg N ha−1 per harvest (Fig. 5b), and cumulative crop N removed was initially overestimated by the model but in good agreement for the final two harvests. 3.4. Drainage and leaching Over the growth period, 45 mm of cumulative drainage was measured from the lysimeter, with the first appearance of deep drainage occurring 150 DAP (Fig. 6). Despite calibrating the model to obtain a comparable final volume, the measured and simulated values differed notably through the growth season. Simulated
Fig. 4. Measured and simulated volumetric water content (VWC), and wetting front detector (WFD) response at depths of 15, 30, 45 and 60 cm (WFD responses are plotted at the VWC associated with a matric potential of −3 kPa).
ods of difference between measured and simulated VWC – which could be partly due to soil heterogeneity and sensor accuracy – the model was judged to perform reasonably well overall in simulating soil layer VWC. The WFDs were clearly observed to respond when spikes in VWC were measured by the automatic sensors, and
Fig. 6. Measured and simulated cumulative drainage (mm) over the growth season.
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
Fig. 7. (a) Measured and simulated cumulative NO3 -N leached and (b) drainage water NO3 − concentrations.
drainage commenced later but then occurred more rapidly in comparison to measured drainage. This may partly be attributed to the nature of drainage from a lysimeter, in which a positive pressure at the lower boundary is required to create free water for drainage (Flury et al., 1999). A total of 86 kg NO3 -N was measured to have leached from the soil profile (Fig. 7a). Measured NO3 − concentrations in the drainage water increased rapidly from 330 mg l−1 at 151 DAP to 1008 mg l−1 on 168 DAP and thereafter remained relatively constant at around 1000 mg l−1 (Fig. 7b). Simulated drainage concentrations ranged from 862 to 750 mg l−1 . Similar to drainage, NO3 -N leaching was initially under-estimated, then over-estimated through the midseason period, with the final end of season simulated cumulative NO3 -N leached (78 kg NO3 -N) in reasonable agreement with the measured value (Fig. 7a). 3.5. Soil water nitrate concentrations High soil solution NO3 − concentrations were observed at all depths at planting despite no fertilization having taken place since the previous season (Fig. 8). These high NO3 − concentrations are therefore attributed to mineralization occurring over a 4-month fallow period during which very little drainage took place. After planting, the removal of N from the system by an actively growing crop is observable in the measured data. In almost all cases, measured NO3 − concentrations from WFDs were less than those measured from SCs. This is consistent with lower solute concentrations found in the draining soil water phase due to incomplete mixing with the immobile soil water phase as has been observed in other experiments (Van Hoorn, 1981; Corwin et al., 1991). The addition of 10 kg N ha−1 7 DAP is observable by an associated increase in NO3 − concentration detected by the SCs placed at 15, 30 and 45 cm (Fig. 8). The effect of a second addition of 10 kg N ha−1 108 DAP is only observable in the SC and WFD at 15 cm. A third
1783
addition of 10 kg N ha−1 132 DAP does not result in a clear increase in SC NO3 − concentration at any of the depths. After an addition of 30 kg N ha−1 148 DAP, a sharp increase in NO3 − concentration followed by an immediate sharp decrease was observed in the SCs placed at 15 and 30 cm. An increase in NO3 − concentration for the WFD placed at 15 cm was also observed. The final application of 30 kg N ha−1 175 DAP did not cause clearly observable increases in NO3 − concentration in either the SCs or WFDs. As additions of fertilizer N were more clearly reflected at the beginning of the season when the crop did not yet have a fully developed root system, this NO3 -N ‘disappearance’ is therefore mostly attributed to rapid crop uptake. The onset of the rainy season 130 DAP clearly moved NO3 − down the soil profile, as observed from both the SC and WFD data. SCs placed at 45 and 60 cm showed an increase in NO3 − concentration after the onset of rain, with measurements demonstrating a pulse of NO3 − moving down the profile. A large increase in NO3 − concentration in the WFD placed at 60 cm 185 DAP is also consistent with the movement of a NO3 − pulse down the profile. Comparisons between measured and simulated data (Fig. 8) provide clear evidence that the SC concentrations reflect the concentrations in the resident soil water, while the WFD concentrations reflect those in the draining soil water. For both sets of comparisons, measured and simulated values showed similar trends to a depth of 60 cm, although simulated values did not fluctuate as much as the measured values. For the SCs at 80 and 100 cm, a sharp decline in NO3 − concentration can be observed after the onset of the rainy season (Fig. 9). This is after an initial increase in NO3 − concentration prior to 150 DAP. These data indicate that N is also moving past the 80–100 cm depth, as is confirmed by the leachate data collected at the base of the lysimeter. The initial increase in measured SC NO3 − concentrations at 80 and 100 cm was under-estimated by SWB-Sci. The rapid decrease in NO3 − concentration after 150 DAP was also under-estimated by the model, especially at the 100 cm depth.
4. Discussion Nitrate (NO3 − ) concentrations sampled from SCs were almost always higher than those sampled from WFDs for this particular soil and experiment. As measurements and comparisons were always made following a wetting event, this is a result of incomplete mixing between the resident and draining water phases. Good comparisons were observed between measured NO3 − concentrations from the SCs and simulated resident soil water concentrations, and measured concentrations from the WFDs and simulated soil water NO3 − concentrations draining from one layer to the next. This indicates that these samplers clearly sample different components of the soil water, and that the use of the Corwin et al. (1991) approach incorporated in a straightforward cascading soil water balance model with a daily time-step was effective in modelling the influence of incomplete mixing on solute transport. Furthermore, using WFD:SC NO3 − concentration ratios served as an effective estimate of the mobility coefficient. A major implication of this is that measuring and modelling can be used together to improve estimates of NO3 -N leaching losses. In using such an approach, a mechanistic crop N model (SWB-Sci in this case) is used together with data from WFDs and SCs to parameterise/calibrate and test the model for a specific site. Under field conditions, drainage data are rarely available to calibrate a model, as was done in this study. From the results presented in this paper, however, SC and corresponding WFD NO3 − concentration data can be used to estimate specific drainage parameters for the soil, and if
1784
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
Fig. 8. Measured nitrate (NO3 − ) concentrations from suction cups compared to simulated resident soil water concentrations (left) and measured NO3 − concentrations from wetting front detectors compared to simulated draining soil water concentrations (right) at depths of 15, 30, 45 and 60 cm.
SC and WFD NO3 − concentrations are well estimated by the model over time, then it can be assumed that drainage will also be accurately estimated. In addition, measured N concentrations together with water flux estimates obtained from a simple crop soil water balance model can also be used to estimate leaching. Although limitations in this dataset for model testing purposes are recognized, the ability of the model to estimate soil water, crop growth, N uptake and N leaching was judged to be adequate. Mechanistic crop N models aim to simulate key processes accurately.
Significant improvements have been made to the SWB-Sci model in order to achieve this. There is still some uncertainty, however, on how to best represent and test certain processes. For example a sharp decline in soil NO3 − concentrations was measured in the lower soil layers (80 and 100 cm) but not simulated by the model. Using data from instruments such as SCs and WFDs is proving useful in this regard. We recommend further work testing and applying the approaches proposed in this paper for a wide range of cropping systems a soil types to enhance their robustness in sup-
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
1785
doing so. Ultimately, it is envisaged that increased efforts to measure and model draining and resident NO3 − concentrations will lead to more efficient and less environmentally detrimental crop management practices. Acknowledgements We gratefully acknowledge the support of the Water Research Commission (Project 1516 ‘Development of an integrated modelling approach to prediction of agricultural non-point source (NPS) pollution from field to catchment scales for selected agricultural NPS pollutants’ and Project 1574 ‘Adapting the Wetting Front Detector to the needs of small-scale furrow irrigators and providing a basis for the interpretation of salt and nutrient measurements from the water sample’) and National Research Foundation of South Africa, and CSIRO and the Cooperative Research Centre for Irrigation Futures from Australia. References
Fig. 9. Measured nitrate (NO3 − ) concentrations from suction cups compared to simulated resident soil water concentrations at depths of 80 and 100 cm.
porting improved understanding of NO3 − dynamics and losses in agro-ecosystems. In addition to using SCs and WFDs to estimate leaching, basing adaptive management fertilization strategies on measured concentrations shows excellent potential. More work however still needs to be done before crop and soil specific guidelines based on measurements made from SCs and/or WFDs can be put into practice. Using predicted end of season total crop transpiration and N uptake to calculate the NO3 − concentration required in the soil water taken up by the roots to meet this total crop N demand could be a good way to start. Such an approach would help reduce over-fertilization, thereby maintaining crop yields while reducing N concentrations in the deep drainage leaving the root zone. Further research on such an approach is therefore recommended. Finally, access to the concentration data, especially resident water NO3 − concentration data, allows focus on improving N management from a crop production perspective, while access to the draining water NO3 − concentrations and fluxes allows focus on improving N management from an environmental perspective. 5. Conclusions Nitrogen leaching from agriculture can cost farmers dearly and pose a serious threat to receiving water bodies, but simple and effective ways of estimating these leaching losses are lacking. A diversity of approaches, ranging widely in levels of complexity, has been proposed to model solute concentrations in soil water. SWBSci, incorporating the functional solute bypass approach proposed by Corwin et al. (1991), was found to adequately simulate draining and resident soil water NO3 − concentrations that reflect the concentrations measured with WFDs and SCs, respectively. This work reinforces the value of combining monitoring and modelling to estimate solute leaching and proposes a pragmatic approach for
Addiscott, T.M., 1996. Measuring and modelling nitrogen leaching: parallel problems. Plant Soil 181, 1–6. Addiscott, T.M., Whitmore, A.P., 1987. Computer simulation of changes in soil mineral nitrogen and crop nitrogen during autumn, winter and spring. J. Agric. Sci. 109, 141–157. Annandale, J.G., Campbell, G.S., Olivier, F.C., Jovanovic, N.Z., 2000. Predicting crop water uptake under full and deficit irrigation. An example using pea (Pisum sativum cv. Puget). Irrig. Sci. 19, 65–72. Barbee, G.C., Brown, K.W., 1986. Comparison between suction and free-drainage soil solution samplers. Soil Sci. 141, 149–154. Clothier, B.E., Heng, L., Magesan, G.N., Vogeler, I., 1995. The measured mobile-water content of an unsaturated soil as a function of hydraulic regime. Aust. J. Soil Res. 33, 397–414. Coats, K.H., Smith, B.D., 1964. Dead end pore volume and dispersion in porous media. Soc. Pet. Eng. J. 4, 74–84. Corwin, D.L., Waggoner, B.L., Rhoades, J.D., 1991. A function model of solute transport that accounts for bypass flow. J. Environ. Qual. 20, 647–658. De Willigen, P., 1991. Nitrogen turnover in the soil-crop system; comparison of fourteen simulation models. Fert. Res. 27, 141–150. Flühler, H., Durner, W., Flury, M., 1996. Lateral solute mixing processes—a key for understanding field-scale transport of water and solutes. Geoderma 70, 165–183. Flury, M., Yates, M.V., Jury, W.A., 1999. Numerical analysis of the effect of the lower boundary on solute transport in lysimeters. Soil Sci. Soc. Am. J. 63, 1493–1499. Haines, B.L., Waide, J.B., Todd, R.L., 1982. Soil solution nutrient concentrations sampled with tension and zero-tension lysimeters: report of discrepancies. Soil Sci. Soc. Am. J. 46, 658–661. Ilsemann, J., van der Ploeg, R.R., Horton, R., Bachmann, J., 2002. Laboratory method for determining immobile soil water content and mass exchange coefficient. J. Plant Nutr. Soil Sci. 165, 332–338. Jovanovic, N.Z., Annandale, J.G., 1999. An FAO crop factor modification to SWB makes inclusion of crops with limited data possible: examples for vegetable crops. Water SA 25, 181–190. Jovanovic, N.Z., Annandale, J.G., Mhlauli, N.C., 1999. Field water balance and SWB parameter determination of six winter vegetable species. Water SA 25, 191–196. Jovanovic, N.Z., Annandale, J.G., 2000. Crop growth model parameters of 19 summer vegetable cultivars for use in mechanistic irrigation scheduling models. Water SA 26, 67–76. Jovanovic, N.Z., Annandale, J.G., Nel, A.A., 2000. Calibration and validation of the SWB model for sunflower (Helianthus annuus L.). S. Afr. J. Plant Soil 17, 117–123. Keating, B.A., Verburg, K., Smith, C.J., Probert, M.E., Gaydon, D., 2001. Assessing leakiness in Australia’s dryland farming systems. In: Proceedings of the International Congress on Modelling and Simulation, Australian National University, December, 2001, pp. 1811–1816. Litaor, M.I., 1988. Review of soil solution samplers. Water Resour. Res. 24, 727–733. Magid, J., Christensen, N., 1993. Soil solution samples with and without tension in arable and heathland soils. Soil Sci. Soc. Am. J. 57, 1463–1469. Paramasivam, S., Alva, A.K., Fares, A., 1997. Vadose zone soil solution sampling techniques to investigate pollutant transport in soils. Trends Soil Sci. 2, 115–136. Silkworth, D.R., Grigal, D.F., 1981. Field comparison of soil solution samplers. Soil Sci. Soc. Am. J. 45, 440–444. Stirzaker, R.J., 2003. When to turn the water off: scheduling micro-irrigation with a wetting front detector. Irrig. Sci. 22, 177–185. Stirzaker, R.J., 2008. Factors affecting sensitivity of wetting front detectors. Acta Hort. 792, 647–653. Tesfamariam, E.H., 2004. Modelling the soil water balance of Canola (Brassica napus L.) (Hyola 60). MSc(Agric.) Dissertation. University of Pretoria. Tesfamariam, E.H., Annandale, J.G., Steyn, J.M., Stirzaker, R.J., 2009. Exporting large volumes of municipal sludge through turfgrass sod production. J. Environ. Qual. 38, 1320–1328.
1786
M. van der Laan et al. / Agricultural Water Management 97 (2010) 1779–1786
Tillman, R.W., Scotter, D.R., Clothier, B.E., White, R.E., 1991. Solute movement during intermittent water flow in a field soil and some implications for irrigation and fertilizer management. Agric. Water Manage. 20, 119–133. Turner, G.A., 1958. The flow-structure in packed beds. Chem. Eng. Sci. 7, 156–166. Van der Laan, M., 2009. Development, testing and application of a crop nitrogen and phosphorus model to investigate leaching losses at the local scale. PhD Dissertation. University of Pretoria, Pretoria.
Van Hoorn, J.W., 1981. Salt movement, leaching efficiency, and leaching requirement. Agric. Water Manage. 4, 409–428. White, R.E., 1985. The influence of macropores on the transport of dissolved and suspended matter through soil. Adv. Soil. Sci. 3, 95–120. Yang, J., Greenwood, D.J., Rowell, D.L., Wadsworth, G.A., Burns, I.G., 2000. Statistical methods for evaluating a nitrogen simulation model, N ABLE. Agric. Sys. 64, 37–53.