Documente Academic
Documente Profesional
Documente Cultură
Table 3. Medians of variables that are significantly different (P<0.05) among plan
and profile curvature classes within the backslope
Median
Variables Horizon Significance
Convex (1) Plane (2) Concave (3)
Plan Curvature
Ca B-4 1 vs.3 12.7 13 14.7
OC B-2 1 vs.3 0.6 - 0.5
BS-7 B-3 1 vs.3 74 - 83
BS-7 B-4 1 vs.3 77 80 86.5
MotlDepth - 1 vs.2 46 65 50
Profile curvature
SiCL B-1 2 vs. 3 - 94.3 93
SiCL B-3 2 vs.3 - 95.4 93
Clay A-2 2 vs.3 - 28.5 25.3
Clay B-4 1 vs.3 32.9 - 33.5
Source: Young and Hammer, 2000 (-) no data
Figure 1. Regression tree for the prediction of soil profile depth (m) in the Bago-
Maragle, Australia (McKenzie and Ryan, 1999)
8 Jurnal Ilmu Tanah dan Lingkungan Vol 5 (2) (2005)
The A horizon is the most area (relative elevation less than 3.8)
biologically active layer where most has deeper profile depth (2.37m).
soil C is stored. Its thickness is McKenzie and Ryan (1999) stated that a
indicative of biological activity, but it tree with six terminal nodes was
also provides clues to the location of optimal for predicting soil profile
erosion and deposition on the hillslope. depth and this model accounts for a
Soil depth provides and indication of modest 42 % reduction in variance
water storage pools, overall compared to a null model (i.e. single
productivity, and longer-term erosion, mean).
deposition, and soil building processes. A regression tree model
Over all, these variables provide a predicting total phosphorus (T ha-1)
basis for hypothesizing the spatial summed over the top 1.0 m of soil
distribution of process zones and profile is presented in Fig. 2. McKenzie
landscape function. and Ryan (1999) reported that a tree
Figure 1 shows tree-based with five terminal nodes to be optimal
model for soil profile depth. and it accounts for 78.5 % of the
Regression tree provide (i) information variance. The rad. K less than 0.84
of predictors and their contribution to accounting for a large proportion of
the response variance. Contribution the variance (i.e. 46.9) identifies
decreases from top (order 1) to down; basalt and depositional areas with high
(ii) terminal nodes, that provide level of phosphorus (i.e. 13.01 Tha-1).
predicted value namely number in Higher potassium (15.25 T ha –1) in this
rectangular and variance showing area may be found on sites with high
homogeneity of predicted data for relief (> 38 m). In high relief, areas
given area. Variance or more common having downslope gradient less than 7
in regression tree called deviance close % tend to be zone of sediment
to zero means predicted data is more accumulation and have a large total
homogeny. Number in circle is also phosphorus value (23.97 T ha-1). In
predicted data but it then subject to these areas, higher total phosphorus
further decomposition. (15.04 T ha-1) may be found at plan
Slope, position of landscape curvature less than 0.8.
(CTI), relative elevation, and climate
(temperature) are predictor variables Quantitative relationship between soil
for profile depth prediction at Bago- properties and landscape attributes at
Maragle, Southern New South Wales, intermediate scale
Australia (Figure 1). Comparing to Table 5 lists the model for
others, slope is the most important predicting some soil properties within
predictor as shown by regression tree pedoderm of Trangie Cowal and Gin
where it is on first order. Average Gin Hills. A pedoderm is a mappable
profile depth from whole area is 2.10 unit of soil, which is at either the
m. The gentle area (slope less than 5 land’s surface, or partially or wholly
%) has deeper profile depth (3.20 m) buried by other pedoderm. Pedoderm
and accounts for larger variance represents one period in which soil
(1.02). The CTI is the most important formation has taken place;it was once,
variable in gentle area, where lower and may still be, at the land’s surface
area (CTI more than 10.3) has highest (McKenzie and Austin, 1993).
profile depth (4.85 m). Relative Within Trangie Cowal pedoderm,
elevation is the most important clay content at 0.10 m can be
splitting variable in the steep area predicted by the use of gilgai
(slope more than 5 %), where lower occurrence and landform. Geomorphic
Sulaeman et al. Modeling soil-landscape relationship 9
Figure 2. Regression tree for the prediction of total soil phosphorus in the upper
1.00 m (T ha-1) in the Bago-Maragle, Australia (McKenzie and Ryan, 1999)
two areas. In sloping-high area, erosion deposition of soil and solute material is
carrying surface soil is dominant. dominant.
Conversely, in low-plain area,
Gilgai (i)
Gilgai absent 0.0
Gilgai present 16.1
Landform (j)
Crest/slopes 0.0
Depressions 9.7
Figure 3. Regression tree for the prediction of A-horizon depth (in m) of tuff soil in
Lampung Province, Indonesia
model (e.g. McKenzie and Ryan, 1999). These tools facilitate workers to
The combination of generalized linear develop and demonstrate their idea.
models and generalized additive Second, workers more realize
models is superior to tree-based that in nature all variable co-relate to
procedure but all were acceptable for each other. As a result, they use soil
practical application. The advantage of quality as predictor instead of soil
tree-based model is the capacity to characteristic. For instance, Gessler et
model non-additive and non-linear al (2000) used CTI instead of slope to
relationship in relatively simple way estimate soil properties. CTI is not soil
being useful for pedologic data where characteristic but land quality. CTI is
interactions between the response useful in integrating topographic
variable and environmental variable that is a guide to water and
explanatory variables are often sediment movement in particular
conditional on other explanatory landscapes. In addition to that, in
variables. In contrast, the generalized nature land is not flat (two dimension)
model is able to use continuous (e.g. as on map, but 3-dimension. Hence,
CTI) or nominal (e.g. rock type). As a prediction will more reliable if 3-
consequence, predictions are more dimension is used. Terrain attributes
realistic because they portray soil provide a quantification of landform
variation as being gradual or shape, connectivity and adjacencies
discontinuous. The response variable of that define external landscape
a generalized linear model can be geomorphometry and water flow
continuous or binary (McKenzie and patterns.
Ryan, 1999). Third factor that contributes to
Since 1990-an, the method of the advent of soil-landscape modeling
soil-landscape modeling has is the abundant, readily available data
increasingly developed due to some from previous study. Using this,
factors. First factor is the advent of workers can try to correlate one soil
computer technology. Nowadays characteristic to other soil
software and hardware have been characteristics or to soil qualities. This
created. For soil research particularly study is helpful in determining whether
pedological study, one may use GIS one variable is dependent variable or
software such as ArcView, ArcInfo, independent variable. This information
MapInfo, IDRISI, ILWIS, etc. To analyze is used as basis of later studies.
soil data, one can use statistical
software such as S-plus, SPSS, Systat, SUMMARY AND CONCLUSION
SAS, etc. Some GIS software (e.g. 1. Landscape position significantly
IDRISI 32) even has provided module determine the characteristic of the
for statistical analysis. Computer soil
peripherals also have developed. One 2. Prediction based on 3-dimension is
may use digitizing table to input analog more reliable than 2-dimension and
map. Also, one can use big- more accurate in detail scale
sophisticated printer to print maps. 3. General linear and tree-based
Overall, without these tolls, it is model can be used to develop soil-
impossible for Gessler et al (2000), landscape modeling
McKenzie and Ryan (1999) and other 4. By general linear model technique,
workers to develop three-dimension slope, flow accumulation and
landscape and build mathematical compound topographic index may
model of soil-landscape relationship. be used as predictors for A-hor
depth; flow accumulation and
Sulaeman et al. Modeling soil-landscape relationship 13
upper mean slope and CTI are the Research and Development Centre,
best predictors for soil depth; flow Indonesia.
accumulation and CTI is the best Gessler, P.E., O.A. Chadwick, F.
predictor for Cmass; microrelief and Chamran, L. Althorse, and K.
landscape position is the best Holmes (2000) Modeling soil-
predictor fro clay content; and landscape and ecosystem
landscape position is the best properties using terrain attributes.
predictor for ESP. Soil Sci. Soc. Am. J. 64:2046-2056.
Graybill, F.A. (1976) Theory and
5. By tree-based technique, slope,
Application of the Linear Model.
CTI, relative elevation and
Duxury Press, North Scituate,
temperature is best predictor for Massachusetts.
soil profile depth; rad K and relief Hoosbeek, M.R. and R.B. Bryant (1992)
are best predictor for total soil Towards the quantitative modeling
phosphorus of pedogenesis-a review.
Geoderma 55: 183-210.
AKNOWLEDGEMENT Hoosbeek, M.R., R.G. Amudson, and R.B.
We would like to thank to the Bryant (2000) Pedological
Indonesian Department of Agriculture modeling.. In: Handbook of Soil
thru PAATP Project, Agency for Science. M.E. Sumner (editors).
Agricultural Research and Development CRC Press, Boca Raton, pp. E77-
for funding this paper. E116.
Jenny, H. (1941) Factors of Soil
REFERENCE CITED Formation. McGraw-Hill Co., New
Amin, T.C, Sidarto, C. Santoso, and W. York.
Gunawan (1994) Geology of Lagacherie, P. and M. Voltz (2000)
Kotaagung quadrangle, Sumatera. Predicting soil properties over a
Geological research and region using sample information
Development Centre, Indonesia. from a mapped reference area and
Buol, S.W., F.D. Hole, R.J. McCracken, digital elevation data: conditional
R.J. Shouthard (1997) Soil Genesis probability approach. Geoderma
and Classification. 4th edition. The 97: 187-208.
Iowa State University Press, Ames. Laslett, G.M. and A.B. McBratney (1990).
Burhan, G., W. Gunawan, Y. Noya, and Further comparison of spatial
U. Subiyatna (1993) Geology of the method for predicting soil pH. Soil
Menggala quadrangle, Sumatera. Sci.Soc.Am.J. 54: 1553-1558.
Geological Research and Mangga, S.A, A. Amiruddin, T. Suwardi,
Development Centre, Indonesia. S. Gafoer, and Sidarto (1994)
Dijkerman, J.C. (1974) Pedology as Geology of the Tanjungkarang
science: the role of data model quadrangle, Sumatera. Geological
and theories in the study of natural Research and Development Centre,
system. Geoderma 11: 73-93. Indonesia.
Draper, M.R. and H. Smith (1981) Applied McKenzie, N.J., K.R. Smettem, and A.J.
Regression Analysis. John Wiley Ringrose-Voase (1991) Evaluation
and Sons. New York. of methods for inferring air and
Eswaran, H. and J. Kimble (2003) Land water properties of soils from field
quality assessment and monitoring: morphology. Aust. J. Soil Res.
the next challenge for soil science. 29:587-602.
Pedosphere 13 (1): 1-10. McKenzie, N.J. and D.A. McLoad (1989)
Gafoer, S., T.C. Amin, and R. Pardede Relationship between soil
(1994) Geology of Baturaja morphology and soil properties
quadrangle, Sumatera. Geological relevant to irrigated and dryland
14 Jurnal Ilmu Tanah dan Lingkungan Vol 5 (2) (2005)