Abstract

Introduction

Dryland salinity is on the increase in the upper catchments of the central and northern river valleys of New South Wales (Murray Darling Basin Ministerial Council, 1999). This could adversely affect irrigated schemes further downstream, if there is insufficient leaching. In order to determine the potential impact and long term sustainability of irrigated-farming systems we need to know the spatial distribution of soil and suitability and effect of the water quality being used for irrigation. Important also is the soil-water balance, which needs to be modeled in order to provide estimates of potential soil salinity accumulation as affected by the current quality of irrigation water. Worst case scenarios can also be applicable.

The SaLF model (Shaw and Thorburn, 1985) is based on the assumption that soil leaching or deep drainage is related to the soil hydraulic conductivity, which in turn is influenced by the amount of clay (%), clay mineralogy (defined by Cation Exchange Capacity/Clay %) and exchangeable sodium percentage (ESP). Once these soil properties and water quality and quantity parameters have been determined the empirically-based model can be used to estimate the average root-zone EC_e at steady-state using progressively saline water. Spatial extension of the model can be achieved by interpolation in order to predict areas of risk. Disjunctive Kriging (DK) can be used to estimate the conditional probability that an indicator variable exceeds a critical tolerance level (Yates et al., 1986a). Wood et al., (1990) used DK to estimate the conditional probability for soil EC_e.

In this paper, we combine the SaLF model predictions of EC_e with DK to assess the current status and potential threat of soil salinity using data from soil and water surveys in the lower Namoi valley of northern NSW, Australia. We tested different simulations based on the application of water of variable quality (i.e., EC_w 0.4, 1.4, 4.0 and 0.0 dS/m). We interpret the results in terms of best management strategies to forestall the threat of soil salinity to irrigation.

Materials and methods

Study area

The study area is approximately centred on the township of Wee Waa located in the lower Namoi valley in northern New South Wales, Australia (Figure 1). The area includes one of the largest grazing properties in the district (i.e. “Boolcarrol Farm”). It also includes the cotton-growing areas of Myall Vale, Doreen Lane and “The Gardens” as illustrated in Figure 1. We selected this area because dryland salinity is prevalent in the upper parts of the Namoi catchment.

Physiography

Stannard and Kelly (1977) carried out a reconnaissance soil survey of the lower Namoi valley and identified eight physiographic units including the 1) clay plains, 2) prior stream formations, 3) the low dissected floodplains and, 4) Pilliga Scrub complex. The dominant clay plains are generally uniform in topography except where dissected by current streams. The prior stream formations occur mostly in continuous belts of elevated and undulating land: the uppermost materials of are a coarser texture than the clay plains. The dissected low floodplains, which lie adjacent to the Namoi River between Narrabri and Wee Waa, occupy depressed positions relative to the clay plains. The greatest contribution of these coarse-textured materials is the result of dissection of the Pilliga Scrub by the course of prior stream formations. The low dissected lands associated with the river are uniformly textured (Stannard and Kelly, 1977).

Figure 1. Location and physiography (after Stannard and Kelly, 1977) of lower Namoi valley study area.

Soil and water data

The soil data used for this study were from the Edgeroi data set (McGarry et al., 1989), and soil survey data for the area west of Edgeroi. The Edgeroi data consist of 210 sampling sites arranged in a systematic equilateral triangular grid with an approximate 2.8 km spacing. Additional sites were used at the Australian Cotton Research Institute (19 sites), and the I.A. Watson Wheat Research Centre (17). Detailed transects were also taken at several locations.

At each location a core was recovered and sampled for laboratory analysis at depths of 0.0-0.1, 0.1-0.2, 0.3-0.4, 0.7-0.8, 1.2-1.3, and 2.5-2.6 m. The samples were air-dried and ground to pass a 2-mm sieve. The soil was analysed for exchangeable cations (mmol(+)/kg) based on Tucker’s (1974) method using a mechanical leaching device (Holmgren et al., 1977); silt and clay (%) were determined using the pipette method (Coventry and Fett, 1979).

The soil survey data for the area west of the Edgeroi consists of 125 sites, which were selected using a stratified simple random sampling design with site spacings ranging from 2 to 10 km (Figure 3). To be consistent with the Edgeroi data, soil samples were collected from four layers and depths of 0.0-0.1, 0.3-0.4, 0.7-0.8, 1.2-1.3 m. The clay and silt content and effective cation exchange capacity (ECEC) were determined using the same methods as described for the Edgeroi data set.

Data on current water quality were obtained during a reconnaissance water survey in the Namoi valley. Two samples were collected from along the Namoi River: one from Collins Bridge (near Wee Waa) and the other in the township of Narrabri. Another sample was taken from the Mooki River.

At Collins Bridge and at Narrabri water salinity measurement (EC_w) was 0.569 dS/m and 0.366 dS/m, respectively. The EC_w sample from the Mooki River (tributary of the Namoi) was 0.759 dS/m. These values suggest that by the time the water reach Narrabri and Wee Waa (avg. 0.468 dS/m) the Namoi River water and other tributaries have diluted salts received from the Mooki River. As a result the water available to irrigators is of good quality.

Figure 2. Location soil sampling sites in the lower Namoi valley study area.

SaLF Modeling

In order to determine the salinity risk associated with irrigated farming systems we chose three progressively saline water qualities for simulation using the SaLF model. These were EC_w values of 1.4, 4.0 and 9.0 dS/m. The first value is predicted to be the salinity of water in the Namoi valley and available for irrigation in the year 2100 (Murray-Darling Basin Ministerial Council, 1999). The other two values of salinity are used for irrigation in Colorado and Tunisia, respectively (Rhoades et al., 1992).

Each of the 329 sites were inputted to the SaLF program, including the attributes of clay content and ECEC at four depths (i.e. 0-0.1, 0.3-0.4, 0.6-0.7 and 1.2-1.3m). A value of exchangeable sodium at a depth of 1.2m (i.e. beyond the rootzone) was also entered. An EC_w of 0.468 dS/m was used for estimating current salinity status of the study area. In the SaLF model, we assumed that average annual rainfall and irrigation water were 584 mm and 600 mm, respectively. Estimates of average root zone EC_e at steady-state were then calculated.

Non-linear kriging

Geostatistical methods of spatial interpolation are not new to the soil science community. Various methods – ranging from linear (ordinary kriging, co-kriging and regression kriging, see Triantafilis et al., 2001a) and non-linear methods (e.g., indicator and disjunctive kriging) have been used. The non-linear methods make predictions using the conditional probability of distribution of the underlying random field producing an error variance for each estimate (Journel, 1983).

DK estimates the conditional probability that a measured indicator variables exceeds a critical tolerance level (Yates et al., 1986a). The conditional probability can be used in management decision-making to determine the level at which some form of management action needs to be implemented. Two types of information are required: the first is the critical level at which the variable becomes a threat (i.e. the cut-off); the second is the probability level that spurs management action (Yates and Yates, 1988). Many examples of its use exist in soil science (Yates et al., 1986b: Yates, 1986; Yates and Yates, 1988; Webster and Oliver, 1989; and Finke and Stein, 1994). DK is presented in detail in Rivoirard (1994).

Results and discussion

Estimates of soil salinity

Figure 3a shows the frequency distribution of estimated soil EC_e using the current water quality (EC_w 0.4 dS m^-1). This suggests EC_e in the root-zone would not accumulate to levels deleterious for crops predominantly used in the current irrigated farming system (i.e. Dolichus lab lab, wheat and cotton). This result is consistent with the prevailing soil salinity values in the lower Namoi valley around Wee Waa. With respect to the use of water of EC_w 1.5 dS/m, average EC_e will have been 1.79 dS/m. Figure 3b shows that sensitive leguminous crops such as Dolichus Lab Lab may require some management in order to ensure its continued use.

Figure 3c shows frequency distribution of soil EC_e if water with quality of EC_w of 4.0 dS/m was applied. Approximately 60 and 5 % of prediction sites would require some form of management to enable inclusion of legumes (as a rotation crop) and wheat production as part of the irrigated cotton farming system, respectively. If water quality of EC_w 9.0 dS/m was applied, more than half the sites would require management to enable wheat cropping whilst all sites would require management to enable production of leguminous crops. With respect to cotton production about 30 % of the sites would need to be managed in some way to reduce salt build up in the root-zone.

The cut-off values of EC_e at steady state considered in this study included 2, 4, 6 and 7.7 dS/m. These correspond with salinity values of sensitive crops, most crop species (Bowers and Wilcox, 1965), wheat and cotton (Mass and Hoffman, 1977).

Figure 3. Blob plots and frequency distribution showing concentration of EC_e as predicted using SaLF when EC_w of a) 0.4, b) 1.5, c) 4.0, and d) 9.0 dS/m was simulated.

Spatial distribution of cut-off EC_e values for crop production

Figure 3 also shows patterns of spatial distribution of EC_e (SaLF estimates) using EC_w values = 0.4, 1.5, 4 and 9 dS/m. It is apparent that responses in the Edgeroi and the Wee Waa data sets are different. This is most obvious on either side of Spring Plains Road. To the northeast, soil EC_e is generally higher as compared with the area to the southwest and around the township of Wee Waa. The reason for this is attributable to the generally sandier and coarser textured sediments associated with the prior stream channel of the Namoi River, which flowed in a northwesterly direction, parallel with Spring Plains Road. As a consequence less soluble salts such as sodium have accumulated in these areas and the soil is relatively well drained compared to the clay plains to the northwest of Spring Plains Road (Triantafilis et al., 2000 and 2001b and c).

Spatial distribution of conditional probability

In order to map the risk areas associated with these levels of EC_e, the spatial distribution of conditional probability produced by DK was plotted along with the frequency distributions for a few cut-off values: 2 dS/m sensitive crops, 4 dS/m most crops (Bowers and Wilcox. Isatis (Geovariances, 1994) was used to carry out DK. The results are shown in Figures 4-7. Figure 4 shows the spatial distribution of conditional probability using DK, where soil EC_e exceeded 2 dS/m if water with quality of EC_w 1.5 dS/m was applied. The white patches on the map indicate lowest risk (i.e. conditional probability < 0.2) whilst the progressively darker gray scale areas indicate higher probabilities (darkest shade: conditional probability > 0.8).

Figure 4. Spatial distribution of conditional probability that soil EC_e at steady state exceeds 2 dS/m if EC_w = 1.5 dS/m using DK (disjunctive kriging).

Much of the area surrounding the township of Wee Waa exhibits low conditional probability that the critical value for legume crops would be exceeded. This is also the case for “The Gardens,” south of Spring Plain Road and “Cumberdeen,” west of Burren Junction, north of Narrabri, and the area around Pilliga. The areas of highest risk are those associated with the clay plains to the north of the Kamillaroi Highway and Spring Plains Road and around Burren Junction. The continued use of sensitive crops such as legumes in cotton farming systems would require some form of soil, crop or irrigation management. This map was similar to the scenario where EC_e exceeds 4 dS/m when application of EC_w of 4 dS/m was simulated (Figure not shown).

Figure 5 shows the result when EC_e exceeded 4 dS/m if EC_w of 9 dS/m (i.e. worst case scenario) was applied. The result achieved here was similar to that when EC_e exceeded 2 dS/m if EC_w of 4 dS/m was simulated (Figure not shown). The conditional probability was only low in the area southeast of Wee (i.e. < 0.2). This suggests that for most crop species some form of management would be required in order to enable their use in irrigated farming systems where water quality was 4 dS/m.

Figure 5. Spatial distribution of conditional probability that soil EC_e at steady state exceeds 4 dS/m if EC_w = 9.0 dS/m using DK (disjunctive kriging).

The results achieved if EC_e cut-off value of 6 dS/m was considered along with the application of 9 dS/m of water is shown in Figure 6. As with the previous cut-off value, much of the area is characterised by high risk (i.e. conditional probability > 0.6). The areas associated with the clay alluvial plains to the north of the Kamillaroi highway between Narrabri and Wee Waa and Spring Plain Road had the largest risk (i.e. >0.8). This was particularly the case north of the Kamillaroi Highway between Narrabri and Wee Waa and northeast of Spring Plain Road. It would be likely therefore that on “Togo” and “Auscott”, the two largest corporate irrigated cotton farms in the district, some form of irrigation, crop or soil management would be required to continue the use of wheat in the irrigated cotton farming system.

Figure 6. Spatial distribution of conditional probability that soil EC_e at steady state exceeds 6 dS/m if EC_w = 9.0 dS/m using DK (disjunctive kriging).

Figure 7 illustrates the results achieved when the EC_e cut-off value was 7.7 dS/m and EC_w of 9 dS/m was applied. It is clear, that the area surrounding Wee Waa and “The Gardens” is at lowest risk (i.e. <0.2). It is also evident that a corridor of low conditional probability (i.e. between 0.2-0.4) occurs adjacent to the Kamillaroi Highway from Myall Vale in the east all the way to Bugilbone in the west. This avenue is consistent with the location of the prior stream channel shown in Figure 2. Another area of low risk is associated with the Pilliga Scrub south of “Cumberdeen” and the township of Pilliga. In these areas, however, it could reasonably be expected that deep drainage would be high and hence any saline water applied for irrigation may end up recharging any good quality groundwater’s underlying these parts of the landscape.

With respect to the areas where the conditional probability was high (i.e. > 0.8), it is apparent that at “Togo” and “Auscott” as well as some areas west of Doreen Lane and north of Bugilbone and southwest of Burren Junction, management would be required to enable germination of the crop. The reason why these parts of the district continually receive the highest conditional probabilities is because the ESP was greater than 6 %. This is particularly the case at “Togo” and “Auscott” which are associated with the fourth alluvial fan of Namoi alluvium (Triantafilis et al., 2001b and c), and where the subsoil tends to be strongly sodic (i.e. ESP > 9 %). As a result, salts would generally tend to accumulate more readily than in other parts of the landscape.

Figure 7. Spatial distribution of conditional probability that soil EC_e at steady state exceeds 7 dS/m if EC_w = 9.0 dS/m using DK (disjunctive kriging).

General Discussion and conclusions

This paper provides an account on how the Salt and Leaching Fraction (SaLF) model in conjunction with a non-linear method of kriging (DK) was implemented to identify salinity risk for various water quality scenarios in the lower Namoi valley. The results suggest the use of water of good quality (EC_w = 0.4 dS/m) and currently drawn from the Namoi River does not pose as a threat to the irrigated-cotton farming systems as it may not result in an increase in soil salinity. If water of slightly lower quality (EC_w = 1.5 dS/m) is used there may be some cause for concern about the viability of growing Dolichus Lab Lab and other sensitive leguminous crops in the valley. This is particularly the case for the irrigation farms located in the clay alluvial plains, where large amounts of sodium are stored. As a consequence this is likely to impede deep drainage and hence cause salts to accumulate in the root zone.

Further, if water of even lower quality is used (EC_w = 4 or 9 dS/m), the worst case scenarios eventuate. Potential threat of soil salinity increases dramatically to levels where some management would be necessary in order to continue with the inclusion of wheat in the irrigated cotton farming system. The area to the northeast of Spring Plains Road is particularly vulnerable. Conversely, the areas associated with the prior-stream channels and the Pilliga Scrub (which runs along the southern border of the study area), potentially have a lower risk of accumulation of salts. However, it is expected that deep drainage is higher and the saline water may contaminate the groundwater reserves. This would reduce management options to farmers who could use the good quality groundwater as currently exists in the valley.

Acknowledgements

The Cotton Research and Development Corporation and the Australian Cotton Cooperative Research Centre (CRC) provided the core funding for this research. Additional monies have been obtained through the Australian Federal Governments, Natural Heritage Trust in collaboration with a community-based project submitted by the CRC for Sustainable Cotton Production and the University of Sydney on behalf of the Coordinating Committee of Namoi valley water users association. The cooperation of all landholders in the lower Namoi valley that allowed unrestricted access to their farms to collect the soil samples is also acknowledged.

References

Bowers, C.A., and Wilcox, L.V. (1965). Soluable Salts. Methods of Soil Analysis, Monograph 9, 933-951. American Society of Agronomy, Madison Wisconsin.

Coventry, R.J. and Fett, D.E.R. (1979). A pipette and sieve method of particle size analysis and some observations on its efficacy. CSIRO Australia Div. of Soils, Div. Rep., No. 38. CSIRO, Australia.

Deutsch, C.V. and Journel, A.G. (1992). GSLIB:Geostatistical Software Library and Users Guide. Oxford University Press. New York, Oxford. pp 314.

Finke, P.A. and Stein, A. 1994. Application of disjunctive co-kriging to compare fertiliser scenarios on a field scale. Geoderma, 62, 247-263.

Holmgren, G.G.S., Juve, R.L. and Geschwender, R.C. (1977). A mechanically controlled variable rate leaching device. Soil Sci. Soc. of Am. J. 41:1207-1208.

Geovariances (1994). Isatis - The geostatistical key. Geovariances. École des Mines de Paris.

Journel, A.G. (1983). Non-paramatic estimation of spatial distributions. Mathematical Geology, 15, 445-468.

Maas, E.V. and Hoffman, G.J. (1977). Crop salt tolerance-Current assessment. Journal of Irrigation and Drainage, Division IR2, 115-134.

McGarry, D., Ward, W.T. and McBratney, A.B. (1989). Soil studies in the Lower Namoi Valley - methods and data 1: The Edgeroi Data Set. CSIRO, Australia, Division of Soils.

Murray-Darling Basin Ministerial Council. 1999. The Salinity Audit of the Murray-Darling Basin. Murray-Darling Basin Commision. Canberra.

Rendu, J-M., 1980. Disjunctive kriging: comparison of theory with actual results. Mathematical Geology, 12, 305-321.

Rivoirard, J. 1994. Introduction to Disjunctive Kriging and Non-Linear Geostatistics, Clarendon Press Oxford.

Rhoades, J.D., Kandiah, A. and Mashali, A.M. (1992). The use of saline waters for crop production. FAO irrigation and Drainage Paper 48, pp 133.

Shaw, R.J. and Thorburn, P.J. (1985). Prediction of leaching fraction from soil properties, irrigation water and rainfall. Irrigation Science, 6, 73-83.

Stannard M.E. and Kelly, I.D. (1977). The irrigation potential of the lower Namoi valley. Water Resources Commission, New South Wales, Australia.

Triantafilis, J., Laslett, G.M. and McBratney, A.B. (2000). Calibrating an electromagnetic induction instrument to measure salinity in soil under irrigated cotton. Soil Science Society of America Journal, 64, 1009-1017.

Triantafilis, J., Odeh, I.O.A. and McBratney, A.B. (2001a). A comparison of five geostatistical models to predict soil salinity from electromagnetic induction data across irrigated cotton fields. Soil Science Society of America Journal, 65, 869-878.

Triantafilis, J., Ward, W.T., Odeh, I.O.A. and McBratney, A.B. (2001b). Creation and interpolation of continuous soil layer classes in the lower Namoi valley. Soil Science Society of America Journal, 65, 403-413.

Triantafilis, J., Ward, W.T. and M^cBratney, A.B. (2001c). Continuous land suitability assessment in the lower Namoi valley, NSW Australia. Australian Journal of Soil Research, 39, 273-289.

Tucker, B.M. (1974). Laboratory procedure for cation exchange measurements in soils. CSIRO Division of Soils, Technical Paper No 23. (CSIRO, Australia).

Wood, G., M.A. Oliver and Webster, R. 1990. Estimating soil salinity by disjunctive kriging. Soil Use and Mngt. 6:97-104.

Webster, R. and Oliver, M.A. (1989). Optimal interpolation and isarithmic mapping of properties. VI. Disjunctive kriging and mapping the conditional probability. Journal of Soil Science, 40, 497-512.

Yates, S.R. and Yates, M.V. (1988). Disjunctive kriging as an approach to management decision making. Soil Science Society of America Journal, 52, 1554-1558.

Yates, S.R., Warwick, A.W. and Myers, D.E. (1986a). Disjunctive kriging 1, Overview of estimation and conditional probability. Water Resources, 22, 615-622.

Yates, S.R., Warwick, A.W. and Myers, D.E. (1986b). Disjunctive kriging 2, Examples. Water Res. 22, 623-630.

Yates, S.R. (1986). Disjunctive kriging 3, Co-kriging. Water Resources, 22, 1371-1376.