Creating and testing a geometric soil-landscape model in dry steeplands using a very low sampling density.

Introduction

Approach to soil-landscape modelling in steeplands

The aim of this study was to determine whether a predictive geometric soil-landscape model could be created for 29 soil properties on seasonally dry greywacke steeplands using a very low soil sampling density (one profile per 360 ha).

Semi-automated systems for predicting soil distribution or soil properties usually relate vegetation, landform, geology, or climate characteristics to soils in a typical area and then use these characteristics as `indicators' of soil map units present in broadly similar landscapes elsewhere. For example Skidmore et al. (1991, 1996) used an expert system to infer the most probable soil present from vegetation, slope gradient, and slope curvature information, after having first established soil-vegetation-landscape relationships on the ground. McKenzie and Ryan (1999) correlated landform, climate, and airborne gamma radiometric data with soil properties and accounted for between 40% and 80% of the sample variance of 3 soil properties measured in a larger area. Moore et al. (1993) used terrain attributes of catenas to predict values of 2 soil properties. Such approaches have been formalised in programs such as Expector (Comer et al. 1997). In the greywacke steeplands study reported here we postulated that the use of landform, vegetation, or geological `indicators' to predict soil distribution was not useful at the output scale required (c. 1:50000) because: (1) the dominant landforms are angular ridges with a repeated pattern of long planar slopes, sharp crests, and narrow valley floors (Fig. 1) and provide little indication of the soils on them; (2) the vegetation pattern of short tussock, tall tussock, weed species (e.g. Hieracium species), and scrub (e.g. Rosa rubiginosa) has been largely induced by burning and grazing of variable intensity, and is not a good guide to soil patterns and trends; and (3) the geology is relatively uniform. Instead we considered that the geometry of the landscape itself, and inferred associated microclimate effects, are likely to be the dominant determinant of the soil variation, a proposition which is supported by earlier investigations of Cuff (1973), McIntosh et al. (1981), and McIntosh and Hunter (1997) in similar terrain. If this proposition is valid, it should be possible to predict soil patterns and trends of individual soil properties over an entire mountain range (dispensing with traditional soil map units), from a small number of sampling sites, provided that sampling sites are carefully stratified to encompass the predicted main sources of variation. At other (larger) scales (c. 1:10 000) soil properties are known to be related to landform character and vegetation (Lilbume et al. 1998; Hewitt and Lilburne 1999) and at still larger scales to plant species or associations of species (McIntosh and Allen 1993; McIntosh et al. 1995). However, in these rangelands land use decisions are usually made for very large fenced management units that can be identified on 1:50 000 maps. Therefore, the focus of this study was to find a means of predicting soil properties on the major landscape units identifiable at this scale, and scale-related variation was not examined. Further work (Landcare Research, unpublished data) tested the accuracy of GIS-interpolation techniques for predicting soil properties at randomly selected points on the Benmore range, on an adjacent range, and over a similar mountain range 250 km distant.

[Figure 1 ILLUSTRATION OMITTED]

Although traditional ground-based soil survey and mapping techniques have been successfully used in steeplands of the main New Zealand ranges (e.g. Ives and Cutler 1972; Ives 1974; Wilde 1974; McIntosh et al. 1992; McIntosh et al. 1995), they are not only slow and expensive, but results cannot easily be extrapolated to other areas because relationships of soil properties to landscape factors such as aspect and altitude are qualitative and largely based on the personal judgement of the pedologist concerned. This problem was highlighted by Hudson (1992) who strongly advised that more soil-landscape models be published to aid dissemination of knowledge, and emphasised the limitations of soil map units as units of information transfer. The soil map unit can be unsuitable for predicting values of individual soil properties (single factors) because the properties may have large variance (Webster 1977). On the steeplands studied here, soil map units have the additional limitation of being inappropriate for predicting values of soil properties within large landscape units in which microclimate variation causes soil trends to be gradual (McIntosh et al. 1981; McIntosh and Hunter 1997, fig. 6) rather than stepped.

We chose as our study area the 26 000 ha Benmore Range, South Canterbury, New Zealand, because of its relatively uniform greywacke parent rocks (Mutch 1963; Gair 1967) and because its soils are representative of the drier South Island ranges. In addition vehicle access is adequate, farmers were cooperative, and the altitude range of 400-1800 m is typical of pastoral leasehold land in the South Island. The drier soils of the South Island ranges with less than 700 mm mean annual rainfall are not only important in area [they cover about 780 000 ha, of which 400 000 ha have greywacke parent rocks (New Zealand Soil Bureau 1968)] but they also support fragile ecosystems prone to degradation by fire, overgrazing, pests (e.g. rabbits), nutrient decline (O'Connor and Hams 1991; McIntosh 1997), and weed invasion (Hunter 1991). An understanding of the pattern of soil nutrients and properties throughout the landscape can help predict environmental risks and trends, can help identify resilient and vulnerable land, can assist farm management, and can be used to identify land suitable for new land uses such as plantation forestry and vineyards. New resource mapping in this type of terrain is required because current soil resource data in the New Zealand Land Resource Inventory (National Water and Soil Conservation Organisation 1975-1979) and the National Soils Database (McDonald et al. 1988) are insufficiently detailed both for onsite farm management and for national resource studies such as determining carbon content of steepland soils for the national C budget assessments.

Characteristics of study area

The Benmore Range is at latitude c. 44 [degrees] 30' south. It is bounded by the Mackenzie Plains to the west and north, Lake Benmore to the east and the Ahuriri Plains, and the Ahuriri River to the south. The climate at Tara Hills on the Ahuriri Plains has been summarised by New Zealand Meteorological Service (1984). Rainfall is spread evenly through the year. Mean air temperature is 9.3 [degrees] C, ranging from 1.5 [degrees] C in July to 15.9 [degrees] C in January. Monthly sunshine hours range from 47% to 53% of possible. Mean daily windrun is 242 km. November, December, and January are the windiest months and have a mean daily windrun exceeding 300 km. The strongest winds are from the north-west. Basement rocks of the Benmore Range are predominatly greywacke and argillite of medium induration (Mutch 1963; Gair 1967). Soils were previously mapped at a scale of 1:253440 by New Zealand Soil Bureau (1968), and detailed studies in selected areas were reported by McIntosh et al. (1981, 1995).

Methods

Establishing the soil-landscape model

Because altitude and aspect are strong drivers of soil variation in mountainlands, as they control the microclimate, we decided to locate sites, and to sample, in a structured and stratified manner that would take into account the effect of these drivers, and allow statistical analysis. On the basis of rainfall and landform character, assessed from aerial photographs and field observations, we divided the Benmore Range landscape into 4 land systems, named Glenbrook, Glencairn, Ben Omar, and Upland Plateau (Fig. 2). In general terms the Glenbrook land system on the west side of the range is the wettest (c. 700 mm mean annual rainfall) with good vegetation cover and stable soils. The Glencairn land system on the east side of the range is drier (c. 550 mm rainfall) with reasonable vegetation cover and generally stable soils, except on some sunny slopes. The Ben Omar land system in the south and east is dry (c. 500 mm rainfall) with many rock outcrops, steeper slopes, and more signs of active erosion. In each of these land systems rainfall probably increases with altitude, to values above those stated. The Upland Plateau land system, on the rolling and undulating tops of the range at 1300 1850 m, is a subalpine herbfield and tussockland, becoming increasingly stony above 1700 m. This land system was not included in the present study.

[Figure 2 ILLUSTRATION OMITTED]

In the Glenbrook, Glencaim, and Ben Omar land systems we sampled profiles by horizon in pairs on upper and middle backslope sites on 4 aspects approximating to magnetic N, S, E, and W, and at low, medium, and high altitudes (c. 650 m, 985 m, and 1260 m). Sampling positions are indicated in Fig. 2. Most sampling was undertaken in spring, when soils were moist. The overall sampling strategy was factorial (3 land systems x 3 altitudes x 4 aspects x 2 landscape positions), giving 72 sampling sites in total. Profiles were described and all horizons sampled. Twenty bulked 0-7.5 cm samples were also taken at each site, using a 2.54-cm-diameter soil corer. To complete the altitude coverage, steeplands at very high altitude (c. 1620 m) were sampled on 4 aspects. These extra sites were not included in the statistical analysis but were used to establish trends and to enable derived maps to have complete coverage of all slopes.

Testing the soil-landscape model

To test the predictive power of the soil-landscape model we described soils at 22 new locations, at sites chosen using the same altitude and aspect criteria used in the primary survey (i.e. fixed altitudes and predetermined aspects).

Analytical procedures

Weight per unit volume of the [is less than] 2 mm fraction of the A horizon was calculated by oven-drying a known volume of A horizon at 110 [degrees] C, and estimating stone content in the field. The soil properties measured were A horizon pH([H.sub.2]0), total C, total N, P retention, exchangeable Ca, Mg, K, and Na, thickness, stone content, and field moisture content. Samples from 0-7.5 cm were analysed for the same chemical soil properties and also for pH(Ca[Cl.sub.2]). Subsoil horizons were analysed for P retention and pH. Soil analyses were by the methods of Blakemore et al. (1987).

Analysis of variance of soil properties in relation to landscape was by GENSTAT (Lawes Agricultural Trust 1993). There was no formal replication; therefore, aspect x altitude combinations, and aspect x land system, and altitude x land system sums of squares were conservatively tested against aspect x altitude x land system interaction sums of squares, which were also used for calculating the relevant standard errors of the means (s.e.), standard error of the difference between means (s.e.d.), and least significant differences between means (l.s.d.). Values for single factors at the 22 sites were statistically compared with values measured in the 72-site survey by the method of Altman and Bland (1983). The variability between the results was defined as the difference between the predicted value and the observed value, divided by the mean of the two. The significance of the difference was calculated using a t-test, values of t [is greater than] 2.09 indicating a significant (P [is less than] 0.05) difference between the 2 values obtained. In field terms, values of t [is less than] 2.09 were taken to indicate that the soil-landscape model was probably a useful predictor of a soil measurement at a comparable site.

Results and discussion

Pedological trends

There were significant effects of altitude and aspect on many topsoil characteristics (Table 1), particularly those likely to be related to soil moisture status (C and N) and to greater leaching and weathering at higher altitudes and on shady slopes (pH, P retention, exchangeable K, and exchangeable Ca). Slope position had no significant effect on soil properties in 0-7.5 cm soils (concentration basis) and few effects in A horizon samples (concentration basis) but was there was a significant effect of slope position on A horizon depth. Slope position had a significant effect on several properties expressed on an area basis, as to be expected from the significant effect of slope position on A horizon depth; sites on lower slope positions had greater amounts of C, N, and exchangeable Ca and Mg on an area basis.

Table 1. Significance of main effects of altitude, aspect, and position for selected soil attributes

                                                        Slope
Attribute                         Altitude   Aspect    position

                                          0-7.5 cm soils

pH([H.sub.2]O)                     <0.001    <0.001     >0.05
pH(Ca[C1.sub.2])                   <0.001    <0.001     >0.05
C (%)                              <0.001    <0.001     >0.05
N (%)                               0.011    <0.001     >0.05
P retention (%)                    >0.05     <0.001     >0.05
Exch. Ca (cmol(+)/kg)              <0.001     0.01      >0.05
Exch. Mg (cmol(+)/kg)              >0.05     >0.05      >0.05
Exch. K (cmol(+)/kg)               <0.001    >0.05      >0.05
Exch. Na (cmol(+)/kg)               0.001    >0.05      >0.05

                                   A horizon, concentration basis

pH([H.sub.2]O)                     <0.001    <0.001      0.035
C (%)                               0.01     <0.001     >0.05
N (%)                              >0.05     <0.001     >0.05
P retention (%)                    <0.001    <0.001     >0.05
Exch. Ca (cmol(+)/kg)              <0.001    >0.05      >0.05
Exch. Mg (cmol(+)/kg)              >0.05     >0.05      >0.05
Exch. K (cmol(+)/kg)                0.004    >0.05      >0.05
Exch. Na (cmol(+)/kg)              >0.05     >0.05      >0.05

                                      A horizon properties

A horizon depth (cm)               >0.05      0.032      0.027
Stones in A horizon (%)            >0.05     >0.05      >0.05
Water in A horizon (mm)            >0.05      0.008      0.002

                                     A horizon, area basis

C (t/ha)                           >0.05      0.001      0.033
N (t/ha)                           >0.05      0.001      0.017
Exch. Ca (kg/ha)                   <0.001    >0.05       0.003
Exch. Mg (kg/ha)                   >0.05     >0.05       0.048
Exch. K (kg/ha)                     0.05     >0.05      >0.05
Exch. Na (kg/ha)                   >0.05     >0.05      >0.05

                                       Profile properties

Lowest subsoil pH([H.sub.2]O)      <0.001    >0.05      <0.001
Highest subsoil P retention (%)    <0.001     0.026     >0.05
Total soil depth (cm)               0.033    >0.05       0.001

For those properties for which the main effects of altitude or aspect were significant at P [is less than] 0.001 (Table 1), values for different aspects and altitudes are shown, and in order to show trends over the whole altitude range, results for the 4 samples from very high altitude sites have been included (Table 2). Soils on W and N aspects (broadly described as sunny), which are exposed to the dessicating north-westerly winds that are characteristic of the region, generally have similar values of soil properties for sites in the low to high altitude range. Soils on E and S (shady) aspects also have similar values. The soils on shady aspects have approximately twice the C concentration, and 2-3 times the C on an area basis, of the soils on sunny aspects. P retention is higher on shady aspects only at medium and high altitudes, where weathering is probably at a maximum; at lower (drier) altitudes only S aspect soils have higher values than other aspects. Similar contrasts between soil properties on sunny and shady slopes were noted by McIntosh et al. (1981) and McIntosh and Hunter (1997).

Table 2. Trends of soil properties showing variation with altitude and aspect, Benmore Range Low, 650 m altitude; medium, 985 m altitude; high, 1260 m altitude; very high, 1620 m altitude. For each value n = 6, except for `very high' values, where n = 1

                                               Aspect
Property                Altitude         Sunny           Shady
                                      W       N     E       S

                                      0-7.5 cm soils

pH([H.sub.2]O)          Very high   5.30    5.46    5.03    5.28
                        High        5.70    5.75    5.61    5.15
                        Medium      5.80    5.94    5.49    5.11
                        Low         6.21    6.40    6.16    5.81
pH(Ca[Cl.sub.2])        Very high   4.46    4.54    4.38    4.38
                        High        4.73    4.72    4.61    4.33
                        Medium      5.12    5.24    4.90    4.51
                        Low         5.54    5.70    5.50    5.03
C (%)                   Very high   1.84    2.39    3.50    3.44
                        High        2.00    1.81    3.30    7.38
                        Medium      2.65    2.18    3.72    5.81
                        Low         1.92    2.04    2.10    4.03
N (%)                   Very high   0.12    0.15    0.21    0.18
                        High        0.14    0.12    0.21    0.48
                        Medium      0.18    0.16    0.24    0.40
                        Low         0.15    0.16    0.16    0.33
P retention (%)         Very high   23      26      33      66
                        High        23      28      29      42
                        Medium      22      14      25      34
                        Low         11      10      13      20
Exch. Ca (cmol(+)/kg)   Very high   0.78    1.05    1.46    0.85
                        High        1.88    1.65    2.18    3.53
                        Medium      4.69    5.23    4.73    5.23
                        Low         4.48    6.07    5.84    6.97
Exch. K (cmol(+)/kg)    Very high   0.22    0.24    0.24    0.13
                        High        0.49    0.42    0.35    0.49
                        Medium      0.89    0.92    0.75    0.79
                        Low         0.82    0.89    0.77    1.02
Exch. Na (cmol(+)/kg)   Very high   0.03    0.04    0.04    0.00
                        High        0.02    0.03    0.02    0.03
                        Medium      0.04    0.06    0.06    0.06
                        Low         0.04    0.05    0.05    0.05

                                         A horizon

A horizon
pH([H.sub.2]O)          Very high   5.27    5.85    5.27    5.30
                        High        5.71    5.74    5.47    5.24
                        Medium      5.64    6.06    5.64    5.22
                        Low         6.14    6.31    6.21    5.71
C (%)                   Very high   2.39    2.01    1.93    2.10
                        High        2.00    2.04    3.59    4.99
                        Medium      2.15    1.66    3.44    4.68
                        Low         1.63    1.65    1.89    3.09
N (%)                   Very high   0.15    0.15    0.14    0.12
                        High        0.13    0.12    0.22    0.31
                        Medium      0.15    0.13    0.21    0.32
                        Low         0.13    0.13    0.15    0.25
P retention (%)         Very high   30      28      31      20
                        High        25      24      47      51
                        Medium      24      25      35      48
                        Low         13      12      15      22
Exch. Ca (cmol(+)/kg)   Very high   0.56    1.35    0.70    0.16
                        High        2.25    1.88    1.46    1.62
                        Medium      3.72    4.10    3.76    2.84
                        Low         4.48    6.08    5.74    6.83
C (t/ha)                Very high   58.2    21.6    30.3    5.4
                        High        26.1    26.7    75.2    95.6
                        Medium      28.5    24.4    56.2    87.8
                        Low         28.3    18.9    26.2    55.7
N (t/ha)                Very high   3.55    1.57    2.12    0.30
                        High        1.79    1.60    4.57    5.75
                        Medium      2.02    1.93    3.50    5.77
                        Low         2.30    1.53    2.13    4.56
Exch. Ca (kg/ha)        Very high   272     290     220     8
                        High        590     502     618     551
                        Medium      1033    1181    1081    986
                        Low         1682    1338    1668    2555

                                         Profile properties
Lowest subsoil          Very high   5.50    5.56    5.69    5.76
 pH([H.sub.2]O)         High        5.44    5.24    5.32    5.44
                        Medium      5.75    6.17    5.85    5.40
                        Low         6.22    6.65    6.33    5.97
Highest subsoil         Very high   34      36      66      23
 P retention (%)        High        47      48      48      47
                        Medium      34      21      42      60
                        Low         18      13      25      33

Values which show marked and consistent trends can be effectively illustrated as `radar diagrams' and trends of 2 soil properties meeting these criteria are displayed in this way (Fig. 3). Values for C in the A horizon increase steadily with increasing altitude on the shady aspects (S and E) to a maximum of 96 t/ha at high altitude, but values are uniformly low (15-30 t/ha) on sunny aspects (N and W). This 3-fold difference in A horizon carbon contents on different aspects is attributed to moist cool conditions favouring C accumulation on shady aspects; in contrast, seasonally very dry, hot, and windy conditions on sunny aspects not only favour organic matter mineralisation but also wind erosion. At very high altitudes this trend on shady slopes is broken; these soils are eroded and the trend for C accumulation to increase with altitude is disrupted by the effect of erosion and topsoil loss. On W aspects at this altitude, soils are moister than at lower altitude and the A horizon has a C content similar to that in E aspect soils at medium and high altitude. In contrast, trends of exchangeable Ca, expressed as kg/ha (Fig. 3), are remarkably regular, and probably reflect increasing leaching with increasing altitude.

[Figure 3 ILLUSTRATION OMITTED]

Soils on the Benmore Range are chiefly Brown, Pallic, and Recent soils (McIntosh et al. 1995). It was the intention of this study to bypass the necessity to classify, define, and map soil units, and instead to plot individual soil properties. However, as P retention values in subsoils are used to define soil orders in the New Zealand Soil Classification (Hewitt 1998), maximum P retention values in subsoils (Table 2) can be used as a mapping tool. Brown soils, which have P retention values of 30% or more in the Bw horizon (Hewitt 1998) occur at medium and high altitude (above about 800 m) except on northerly aspects, where they occur above 1100 m. Pallic and Recent soils occur below these limits. Thus, contrasting aspect causes similar soils to be `displaced' by an altitude difference of about 300 m.

Effectiveness of the soil-landscape model as a predictor of soil properties

In general, soil values based on only one laboratory measurement appeared satisfactory as predictors (Table 3). Parameters based on several measurements (e.g. A horizon values on an area basis), which incorporated field measurements of A horizon thickness, stoniness, and weight of the [is less than] 2 mm fraction, were less satisfactory, presumably because they contained more measurement error and/or propagation of variability. Exchangeable Mg was not predicted satisfactorily on a concentration or area basis and factors other than altitude and aspect (e.g. vegetation or grazing intensity) may have influenced values.

Table 3. Comparison of predicted and actual values of soil properties

Soil property             Mean    Variability   s.e.d.   t-value

                                      0-7.5 cm soils

pH([H.sub.2]O)            5.77         0.067    0.083     -0.74
pH(Ca[Cl.sub.2])          5.04         0.083    0.090      1.28
C (%)                     3.08         0.331    0.218      0.75
N (%)                     0.219        0.336    0.016      0.81
P retention (%)           21.2         0.553    2.49      -3.12(*)
Exch. Ca (cmol(+)/kg)     4.87         0.478    0.496      1.54
Exch. Mg (cmol(+)/kg)     0.938        0.628    0.126     -3.49(*)
Exch. K (cmol(+)/kg)      0.735        0.660    0.104      2.65(*)
Exch. Na (cmol(+)/kg)     0.046        0.675    0.007     -1.11

                             A horizon, concentration basis

pH([H.sub.2]O)            5.81         0.088    0.109     -0.02
C (%)                     2.62         0.512    0.287     -0.93
N (%)                     0.191        0.046    0.019     -1.14
P retention (%)           24.0         0.474    2.42      -1.67
Exch. Ca (cmol(+)/kg)     4.42         0.605    0.569     -0.54
Exch. Mg (cmol(+)/kg)     0.843        0.808    0.145     -3.3(*)
Exch. K (cmol(+)/kg)      0.475        0.950    0.096      1.60
Exch. Na (cmol(+)/kg)     0.052        0.435    0.004      0.23

                                  A horizon properties

A horizon depth (cm)      18.8         0.406    1.63      -1.90
Stones in A horizon (%)   26.3         0.585    3.29       2.12(*)
Water in A horizon (mm)   53.8         0.645    7.39      -0.12

                                  A horizon, area basis

C (t/ha)                   39.7        0.425    3.59      -2.29(*)
N (t/ha)                   2.91        0.391    0.242     -2.57(*)
Exch. Ca (kg/ha)           1360        0.679    197       -1.35
Exch. Mg (kg/ha)           160         0.796    27.2      -3.94(*)
Exch. K (kg/ha)            278         0.958    56.8       0.784
Exch. Na (kg/ha)           18.6        0.644    2.55      -0.391

                                  Profile properties

Lowest subsoil             5.86        0.076    0.094      1.46
 pH([H.sub.2]O)
Highest subsoil            33          0.527    3.69       0.343
 P retention (%)
Total soil depth (cm)      59.2        0.454    5.74       2.95(*)

(*) Significant difference between predicted and measured values (t > 2.09).

Using a digital terrain model, and mathematical interpolation between sites, the trend table (Table 2) can be used to produce single-factor maps with application to resource inventories (Lilburne et al. 1998). For example, preliminary calculations (McIntosh et al. 1998) showed that the carbon contained in the topsoils of the Benmore range was approximately 1 million tonnes, a value which is about one-third of that calculated from previously available information (Tate et al. 1997, as discussed by McIntosh et al. 1998). This difference of values shows that if dependable national resource inventories are to be developed, pedological information should be gathered by survey methods designed for the output desired. Further work has tested the accuracy of GIS-interpolation of the soil-landscape model here described over the entire landscape of the Benmore Range, and has applied the model and GIS-interpolation to a nearby mountain range and a range 250 km distant (Landcare Research, unpublished data).

Conclusions

Several conclusions can be drawn about the effectiveness of geometric soil-landscape models.

(1) In steeplands of relatively uniform geology and simple and repeated landforms it is possible to model soil-landscape relationships over a 26 000 ha area and show trends of several single factors by carefully stratifying sample sites in relation to landscape geometry (altitude and aspect), using a sampling density of 1 site per 360 ha.

(2) The soil-landscape model can be used to predict values of many individual soil properties including topsoil C content and nutrients over an entire mountain range, without using the traditional intermediate steps of defining soil map units or classifying soils. Prediction is most accurate for properties based on one laboratory measurement. Where several field measurements are used to predict values (as is the case for kg/ha values for topsoil characteristics), predicted values are less accurate probably because of the cumulative effect of errors from field measurements and variability.

(3) The methodology described for constructing geometric soil-landscape models is most likely to be useful for resource surveys in areas of relative uniform geology where altitude and aspect, through their influence on microclimate, are the main influences on soil variation, and broad-scale (e.g. 1:50 000) output is required. The methodology has been shown to improve the accuracy of carbon pool figures used in national resource inventories.

Acknowledgments

We thank the farmers of the Benmore Range for cooperating with the soil survey; K. Giddens for soil analyses; L. Lilburne for discussions on project design and methods; H. Wallace for soil preparation and technical help; M. Kingsbury for field assistance; and Dr A. E. Hewitt and 2 anonymous referees for commenting on a draft of the manuscript. This research was supported by the New Zealand Foundation for Research, Science and Technology, Contract CO9626.

References

Altman DG, Bland, JM (1983) Measurement in medicine: the analysis of method comparison studies. The Statistician 32, 307-317.

Blakemore, LC, Searle, PL, Daly, BK (1987) Methods of chemical analysis of soils. New Zealand Soil Bureau, Scientific Report No. 80.

Corner, R J, Cook, SE, Moore, GA (1997) Expector: a knowledge-based soil attribute mapping method. ACLEP Newsletter 6(2) (June 1997), pp. 9-11. CSIRO Land and Water, Canberra.

Cuff JRI (1973) A study of the influence of aspect on the nutrient requirements and soil chemistry of a selection of Hurunui steepland soils in South Canterbury. Unpublished MAgrSci Thesis. Lincoln College, Canterbury, New Zealand.

Gair HS (1967) Sheet 20. Mt Cook. 1st Edn Geological Map of New Zealand, 1: 250 000. Department of Scientific and Industrial Research, Wellington, New Zealand.

Hewitt AE (1998) New Zealand Soil Classification. Landcare Research Science Series 1.

Hewitt AE, Lilburne LR (1999). Mapping soil carbon using Expector in the South Island high country, New Zealand. Abstract, 3rd Conference of the Working Group on Pedometrics of the International Union of Soil Science, Sydney, September 1999.

Hudson BD 1992 The soil survey as paradigm-based science. Soil Science Society of America Journal 56, 836-841.

Hunter GG (1991) The distribution of hawkweeds (Hieracium spp.) in the South Island, indicating problem status. New Zealand Mountain Lands Institute Review 48, 21-31.

Ives DW (1974) Soils of part Mowbray catchment, South Canterbury, New Zealand. New Zealand Soil Survey Report 23.

Ives DW, Cutler EJB (1972) A toposequence of steepland soils in the drier high-country yellow-brown earth (dry-hygrous eldefulvic) region, Canterbury, New Zealand. New Zealand Journal of Science 15, 385-407.

Lawes Agricultural Trust (1993) `Genstat 5, release 3, reference manual.' (Clarendon Press: Oxford).

Lilburne LR, Hewitt AE, McIntosh PD, Lynn IH (1998) GIS-driven models of soil properties in the high country of the South Island. In `Proceedings SIRC 98--10th annual colloquium of the Spatial Information Research Centre'. University of Otago, Dunedin, New Zealand, 16-19 November 1998. pp. 173-180.

McDonald WS, Giltrap DJ, McArthur AJ (1988) Revised SPG1 Database System manual (V1.2). New Zealand Soil Bureau Laboratory Report SS16. Department of Scientific and Industrial Research, Wellington.

McIntosh PD (1997) Nutrient changes in tussock grasslands, South Island, New Zealand. Ambio 26, 47-51.

McIntosh PD, Allen RB (1993) Soil pH declines and organic carbon increases under hawkweed (Hieracium pilosella). New Zealand Journal of Ecology 17, 59-60.

McIntosh PD, Loeseke M, Bechler K (1995) Soil changes under mouse-ear hawkweed (Hieracium pilosella). New Zealand Journal of Ecology 19, 29-34.

McIntosh PD, Hunter GG (1997) Soils and land use issues in the Mackenzie hill country. Landcare Research Science Series No. 19.

McIntosh PD, Backholm G, Smith J (1981) Soil variation related to landscape and vegetation features in North Otago hill country. New Zealand Journal of Science 24, 225-244.

McIntosh PD, Allan BE, Dreaver KR (1992) Soils and land use potential on Tara Hills high country research station. DSIR Land Resources Scientific Report No. 24.

McIntosh PD, Lynn IH, Webb TH (1995) Soil map of Glencairn Station, Omarama, 1:25 000. Landcare Research, Lincoln, Canterbury, New Zealand.

McIntosh PD, Lynn IH, Lilburne LL, Kingsbury M Giddens K (1998) Is the soil carbon of dry South Island mountain ranges being accurately assessed? New Zealand Soil News 46, 13-15.

McKenzie NJ, Ryan PJ (1999) Spatial prediction of soil properties using environmental correlation. Geoderma 89, 67-94.

Moore ID, Gessler PE, Nielsen GA, Peterson GA (1993) Soil attribute prediction using terrain analysis. Soil Science Society of America Journal 57, 443-452.

Mutch AR (1963) Sheet 23. Oamaru. 1st Erin Geological Map of New Zealand, 1:250 000. Department of Scientific and Industrial Research, Wellington, New Zealand.

National Water and Soil Conservation Organisation (1975-1979) New Zealand Land Resources Inventory Worksheets. Ist Edn 1:63360. National Water and Soil Conservation Organisation, Wellington, New Zealand.

New Zealand Meteorological Service (1984). Summaries of climatological observations to 1980. New Zealand Meteorological Service Miscellaneous Publication 177.

New Zealand Soil Bureau (1968). Soils of South Island, New Zealand. New Zealand Soil Bureau Scientific Report No. 27.

O'Connor KF, Harris PS (1991). Biophysical and cultural factors affecting the sustainability of high country pastoral uses. In `Proceedings of the international conference on sustainable land management, Napier, Hawkes Bay, New Zealand'. 17-23 November 1991. (Ed. PR Henriques) pp. 304-313. (Hawkes Bay Regional Council: Napier, New Zealand)

Skidmore AK, Ryan P J, Dawes W, Short D, O'Loughlin E (1991) Use of an expert system to map forest soils from a geographical information system. International Journal of Geographical Information Systems 5, 431-445.

Skidmore AK, Watford F, Luckananurig P, Ryan PJ (1996) An operational GIS expert system for mapping forest soils. PE&RS, Photogrammetric Engineering and Remote Sensing 62, 501-511.

Tate KR, Giltrap DJ, Claydon J J, Newsome PF, Atkinson IAE, Taylor MD, Lee R (1997) Organic stocks in New Zealand's terrestrial ecosystems. Journal of the Royal Society of New Zealand 27, 315-335.

Webster R (1977) Quantitative and Numerical Methods in Soil Classification and Survey. Clarendon Press, Oxford.

Wilde RH (1974) Soils of the Camp Stream catchment, Canterbury, New Zealand. New Zealand Soil Survey Report No. 15.

Manuscript received 26 April 1999, accepted 16 November 1999

P. D. McIntosh(A), I. H. Lynn,(B) and P. D. Johnstone(C)

(A) Landcare Research, Dunedin, New Zealand; present address: Forest Practices Board, Hobart, Tasmania, Australia.

(B) Landcare Research, Lincoln, Canterbury, New Zealand.

(C) AgResearch, Invermay Agricultural Research Centre, Mosgiel, New Zealand.