Although these studies demonstrate well the importance of numerical modelling in the design and management of irrigation and fertigation systems for various crops, most studies involving salinity and nitrate leaching are based on either an analysis of hypothetical scenarios, or are carried out for annual crops. Hence, there is a need to carry out modelling studies for perennial horticultural crops such as mandarin, using experimental results from field studies involving modern irrigation systems such as drip. The objectives of the present investigation were to evaluate water, salt , and nitrate movement in soil below young mandarin tree using HYDRUS-2D, and to evaluate various irrigation and fertigation strategies for controlling deep drainage and nitrate leaching, whilst maintaining soil salinity below the threshold for mandarin. This approach will help us understand the best irrigation and fertigation management practices to be adopted in future practical applications, with the goal to increase root water and nutrient uptake.The field experiment was conducted at the Dareton Agricultural and Advisory Station , located in the Coomealla Irrigation Area, 3 km from Dareton and 10 km from Went worth in New South Wales . The research station forms part of the Sunraysia fruit growing district of NSW and Victoria located in the Murray Darling Basin. An experimental site with an intensive fertigation system,hydroponic growing systems consisting of various mandarin varieties budded onto a number of root stock varieties , was established in October 2005. The trees were planted at a spacing of 5 m -2 m.

The actual monitoring and measurements were initiated in August 2006. The trees were managed and fertilized following current commercial practices, although the amounts of applied fertilizer varied. The soils of the site are alkaline , with red sandy loam from the surface to 90-cm depth, and loam below . The total organic carbon content is very low in the first 30 cm, and below 0.25% in the remainder of the root zone. The climate is characterized as dry, with warm to hot summers and mild winters. The total rainfall during the experimental period from 21 August 2006 to 20 August 2007 was 187 mm , which was slightly below average for the area. Potential evapotranspiration is normally high and equal to 1400 mm per year. Mild frost conditions occur during the winter months. Weather data were collected from an automated weather station located within the research station.The HYDRUS-2D software package was used to simulate the transient two-dimensional movement of water and solutes in the soil. This program numerically solves the Richards’ equation for variably-saturated water flow, and advection–dispersion equations for both heat and solute transport. The model additionally allows specification of root water uptake, which affects the spatial distribution of water, salts and nitrate between irrigation cycles. The solute transport equation considers the advective–dispersive transport in the liquid phase, as well as diffusion in the gaseous phase. The theoretical part of the model is described in detail in the technical manual and in Šimu˚ nek et al. .The nitrogen balance for the mandarin crop was evaluated for two fertigation strategies. First, the fertigation pulse was applied at the beginning of each irrigation event . Second, the fertigation pulse was applied near the end of each irrigation event . It is a common practice that irrigation water is initially and at the end free of fertilizer, to ensure a uniform fertiliser application and flushing of the drip lines . Therefore, fertigation applications were simulated to either start one hour after irrigation started or to end one hour before irrigation stopped.

Nitrate management strategies also include a judicious manipulation of irrigation and N fertilizer applications, and increasing or decreasing the frequency of applications. These interventions should improve N uptake by plants and reduce N leaching out of the plant root zone . The evaluated scenarios are described in Table 2. Scenario, S1, illustrates the impact of applying the same volume of water in small irrigation events . Scenarios S2 and S3 then represents the reduction of the irrigation volume application by 10% and 20%, respectively. Scenarios S4 and S5 are based on decreasing the nitrogen application by 10–20%, respectively, while scenarios S6 and S7 represent a combined reduction in irrigation and fertigation by 10–20%, respectively. Five scenarios were executed, in which irrigation was reduced during the second half of the crop season, i.e., between January and August, by 10%, 20%, 30%, 40%, and 50%, respectively.The water contents measured weekly by EnviroSCAN at different depths at a horizontal distance of 10 cm from the dripper, and corresponding values simulated by HYDRUS-2D during the entire growing season are illustrated in Fig. 5. The measured water contents remained similar at 10 and 80 cm cm, fluctuated between 0.1 and 0.2 cm3 cm 3 at 25 and 50 cm, and stayed higher than 0.2 cm3 cm 3 at 110 cm soil depths throughout the growing season, indicating a favourable moisture regime in the crop root zone. However, the simulated water contents were lower than the measured values during the initial period at a depth of 10 cm and during the mid period at a depth of 110 cm. The simulated values matched the measured values more closely at soil depths of 25 and 50 cm, which is the most active root zone for water and nutrient uptake for citrus . However, the profile average water distribution matched well.

The MAE between weekly measured and simulated moisture content values across all locations varied from 0.01 to 0.04 cm3 cm 3 , indicating a good agreement between the two sets of values . Slightly higher temporal MAE values during the mid-season agreed well with the variation shown in Fig. 5. Similarly, the MAE values at 10, 25, 50, 80, and 110 cm soil depths at a 10 cm lateral distance from the dripper also revealed that the variation between measured and simulated water contents remained between 0.02 and 0.04 cm3 cm 3 . However, the differences were slightly higher at 10 cm depth as compared to greater depths . Higher variations at the surface depth are to be expected because this part of the soil profile is influenced by soil evaporation, which peaks in day time and is low at night time,ebb and flow table while the assumption of a constant atmospheric boundary flux for daily time steps in the model deviated from the actual transient conditions existing at the surface boundary. Other studies also showed a similar magnitude of variations between measured and predicted water contents.Comparison of simulated electrical conductivities of soil solution with weekly measured values at different depths are shown in Fig. 6. Despite of low irrigation water salinity and low initial soil salinity , the measured ECsw increased in the soil with the onset of irrigation at all depths, except at 150 cm where the increase in salinity occurred only after December 2006. Subsequently, a decreasing trend was observed in ECsw later in the season. The higher amount of irrigation compared to ETC. and an significant amount of precipitation during this period resulted in a reduction in soil solution salinity. On the other hand, the model over-predicted ECsw at a depth of 25 cm from October to December 2006 and under-predicted it at a depth of 100 cm during the same period. However, at a depth of 150 cm, simulated values remained constant till January 2007, indicating a delayed response. The increase in simulated ECsw values was delayed at 100 and 150 cm depths as compared to measured values. Both set of values matched well at a depth of 50 cm and the profile average of ECsw also showed a close match. It is significant to note that irrigation with good quality water in our study led to the development of significant levels of measured ECsw . However, the ECsw values remained below the threshold of salinity tolerance of orange throughout the season . The MAEs between weekly measured and simulated ECsw in the soil ranged from 0.08 to 0.76 dS m 1 , which are acceptable for a complex and highly dynamic soil system, with the exception of a few divergent values obtained between mid October and December . The disagreement in ECsw values during this period was correlated with corresponding fluctuations and low values of water contents, especially at soil depths of 10 and 25 cm and this variability was transferred to the ECsw values. Differences between measured and simulated ECsw values at 50 cm depth were relatively higher than at other depths . The mean MAE at 25, 100, and 150 cm depths ranged from 0.19 to 0.36 dS m 1 , showing a good agreement with the measured values at these depths. The spatial distribution of ECsw in the soil profile at various dates is depicted in Fig. 7. It can be seen that salts remained restricted to roughly the upper 50 cm of the soil profile until December . The salts mass was later pushed deeper due to high rainfall . The downward movement of salts continued in February and March , because in March the amount of irrigation was higher than ETC.

It is pertinent to note here that the ECsw distribution under the dripper remained lower as compared to the adjoining soil at all times, because a continuous water application in this region pushes the salts towards the outer boundary of the wetting front. The drainage flux during and after March transported salts vertically downwards, thereby making the soil directly beneath the dripper relatively salt free by the end of the season. Applying additional water at the end of the season could be a strategy to create a salt free root zone which may encourage vigorous root development, and assist the plant growth in the ensuing season.Comparison of weekly measured and daily simulated nitrate– nitrogen concentrations at different depths in the soil profile is illustrated in Fig. 8. Over-prediction was observed at a depth of 25 cm from October to November 2006, which coincided with similar over-prediction for salinity. Similarly, both measured and simulated values matched well at a depth of 50 cm, while a delayed response in predicted nitrate contents was observed at lower depths. However, a fairly good correspondence was observed between profile averaged NO3–N contents. The temporal MAE values for NO3–N ranged from 0.1 to 1.97 mmol L 1 . Similar differences between measured and HYDRUS-2D simulated values were also reported in another study involving simulations of nitrogen under field cropped conditions. Additionally, MAE at a 25 cm depth had a higher value L 1 than at greater depths L 1.The reason for differences in ECsw and NO3  –N values may be partially due to the fact that model reports point values, whereas the SoluSAMPLER draws in solution from a sampling area of a certain volume, the size of which depends on the soil hydraulic properties, the soil water content, and the applied suction within the ceramic cup . Hence the measured parameters considered in modelling may not represent the inherent spatial variability of the soil. In addition, while a homogeneous soil environment is assumed by the model, the field site could be far more heterogeneous and anisotropic. Also, the model simulations considered only a 2D movement of nitrogen and the nitrification process, while more complex nitrate processes were not taken into account. Ramos et al. documented numerous factors influencing the correspondence between measurements and simulations of water contents and solute concentrations in the soil under drip irrigation conditions and these factors are relevant also for the present investigation. These factors, including those mentioned above, may modify the error in the simulated NO3–N values. The simulated movement of nitrate–nitrogen in the soil under a mandarin tree at various dates is shown in Fig. 9. Nitrate fertigation increased the nitrogen content in the soil with time, as is evident from an increasing size of the concentration plume below the dripper as the season progressed. This indicates that the plant was not able to take up all nitrogen added through fertigation, and thus nitrogen built up in the soil over time, leading to a maximum concentration values in January . Ultimately, nitrogen started moving downwards after late January, when there was high rainfall and total water additions exceeded ETC. Alva et al. also detected greater variations in NO3–N concentrations in the 0–15 cm depth horizon, as compared to greater depths in a field experiment involving citrus.