European Soil Bureau ¾ Research Report No. 4
Definition and Use of Functional Soil Horizons as Keys in Spatial Land Information Systems
282
European Soil Bureau ¾ Research Report No. 4
Summary
Local soil management, which combines economical production with ecological protection of soils requires specific and detailed information about the soil continuum. Precision Farming in particular is a challenge to Soil Informatics for developing Soil Information Systems (SIS). Tools are needed which support the most effective, safe and efficient transfer of soil and land characteristics to the integrated evaluation and sustainable use of land. A Three-dimensional Rule-based Continuous Soil-modelling system (TRCS) has been constructed which uses horizon classes as the main carrier for information transfer and integrates the local knowledge of soil experts.
J. Lamp1)
M. Ameskamp2)
1)Universität Kiel, Institut f. Pflanzenernährung und Bodenkunde,
Olshausenstrasse 40-60, D-24118 Kiel; Germany
2)Universität Kiel, Institut f. Informatik und Praktische Mathematik,
Olshausenstrasse 40,
D-24118 Kiel, Germany
The evaluation of the intra-class variances and discriminant analysis of horizon classes at the main and sub horizon, as well as at the texture level – based on a profile database of German soils (ca. 10,000 horizons) – shows that the definition of horizon classifications should
· not be oversophisticated pedogenetically;
· include geogenetical information about the origin of and human impacts on erosion and compaction of horizons;
· be localized and adapted to the soil conditions of interest.
The definition of horizon designations was applied to the soils of Holstein (North Germany) for optimal prediction of ecological ‘horizon transfer functions’.
Challenges of local soil management
Soils, central parts of land, have come under severe dual stress: increasing populations require more and better food, and manifold degradations endanger or destroy the non-renewable natural resource. Unavoidably, soils have both to be protected and used more intensively. This dilemma can partly be solved with the help of Soil Information Systems (SIS) which extend in four directions. The land user and protector, separated or as one person, urgently need more information about the
· soil-factor relationships. Soil genetics can forecast future processes of soils induced by local and global changes of ecosystems.
· soil attribute space. Increasingly, data about physical, chemical and biological aspects of soils needs analysis and evaluation.
· soil functions. Soils are not only the main crop production factor, but they also affect many alternative land uses by their functions in heat, water and chemicals transfer, as well as in traffic, housing and recreation.
· soil distribution. With increasing precision, the geo-referenced occurrence of the above items should be inventorized and disseminated to various land users.
There are not only degradation risks in soil and land use, but also consequences and challenges. Detailed soil surveys and assessments in drinking water recharge areas, focal sites of alternative land use, show clearly that optimized soil management varies significantly at local level. Differences in water storage, humus binding or denitrification potentials, for example, cause soil-specific risks or inevitable nitrate and pesticide leaching. Therefore, detailed functional soil maps are a very cost-effective prerequisite for land under intensive and competing use in Europe. In these areas, the annual value of a claim for agroproduction quantity versus drinking water quality, say a hundred Ecu per hectare, may exceed the total survey cost. Of course, the intensity and scale of surveys depend very much on the lateral variability of the soils, but also on the smallest units of land management, traditionally the fields.
As much as a decade ago, a new approach for local soil management was initiated for conditions in North America (Robert and Anderson, 1987) and Europe (Lamp and Knoop, 1988). Differential Global Positioning Systems linked to a computer and ‘yieldometer’ as well as to sensors on farm machinery are state-of-the-art and offered by competing vendors to farmers (Robert et al., 1992, Olesen, 1995). Even though the reliability of electronic devices has still to be adapted to the rough conditions of farming, this approach, called Precision Farming, is now capturing the interests of farmers.
Accordingly, a new functional soil unit has been defined as follows (Lamp, 1986 ):
· The pedocell is the smallest agrotechnical soil unit which can be treated as unique in local soil management. Its lateral size is defined by point errors of surveying, positioning, steering and recording as well as by the width of tractor lanes (mostly 12-24 m). The lower vertical boundary is determined by the effects on and from the roots of main crops (down to approximately 1.5 m). Based on a detailed farmland inventory, monitoring pedocells are selected which represent all main soil and management units and from which mixed samples are taken by replicates at different soil depths in order to obtain representative analyses.
Local soil management depends on detailed information and is thus confronted with two key questions: Is conventional soil science willing and able to take up these challenges, making a shift from semi-detailed surveys (often 1:25,000 maps) to more specific and detailed inventories, both in the attribute- and the geo-space? How can Soil Information Systems (SIS) help to capture and transfer soil and land characteristics for the evaluation of land qualities at local scale?
Soil Informatics: goals of a new discipline
These questions have to be answered by Soil Informatics, a new discipline between soil and computer science. An important concept is the conservation of information when characteristics (survey data) are transmitted to the assessment of soil qualities for sustainable land use (FAO, 1983, Smyth and Dumanski, 1993). A basic rule is to delay the inevitable loss of information due to classification until the final steps within this knowledge transfer. The main topics of Soil Informatics – methodology and control of the information transfer process – are guided by criteria of
· effectiveness (predicting the most relevant soil functions),
· safety (minimum information loss, best possible correlations) and
· efficiency (the most simple and practical methods in relation to effort).
How can we survey soil variation in detail, and transfer information for local soil management, both for better crop production and environmental protection?
One often encounters the misconception that soil variation becomes less prominent at the local scale. Investigations in soil regionalization and geostatistics have shown that each scale has its own sources, methods and interpretations of soil variation, but the relative magnitude is roughly the same. Stepping from the regional to the local scale of variation will increase the need to represent soils not as crisp, sharply bordered pseudo-objects or as association of objects (poly-pedons or complex map units), but to treat soils as they are defined theoretically in textbooks of pedology: as four-dimensional (4D) continuous systems in time and space (e.g. Schroeder, 1994). For inventory purposes only, this system can be reduced to a 3D soil continuum .
A straightforward approach would be to describe, analyze and map soils in a 3D geo-space by spanning at each observation point an attribute-space of m continuous variables whose values have to be determined by intensive analyses. Based on probability theory, geostatistics offer variogram analyses and Krige-interpolation in order to plan and evaluate point surveys and generate isoline maps of variables describing the soil continuum. Variograms from surveys in North Germany show repeatedly that the efficient distances which account for half of the relevant semivariance between the nugget effect and the sill are mostly between 40m and 80m for young and old morainic soilscapes, respectively (Otte, 1988). The geostatistical approach may be appropriate for sites of scientific ecosystem research, but it faces problems in practical applications for local soil management. Even soil analyses at a 50m grid, which still do not meet all steerable soil variation within fields of many soil landscapes, will not be accepted by farmers due to economic restrictions. Therefore, an approach requiring less effort has been developed which aims to make use of available landscape and soil data and to take advantage of the knowledge that soil experts have accumulated through many present and past inventories. The ability in contextual thinking of human experts still exceeds that of computers by orders of magnitude.
In acting as a tool for local soil inventories, a Three-dimensional Rule-based Continuous Soil Modelling System (TRCS, Ameskamp, 1997) was created. Its functions are explained by Ameskamp and Lamp (1998), but the characteristics are summarized as follows:
· Three-dimensional: Soils are treated as they are defined in textbooks. 3D modelling stresses the mental concepts of pedologists about soil covers. In contrast to conventional soil mapping, early and irreversible information losses are avoided.
· Continuous: The Fuzzy set theory is applied to soil continua in oder to represent gradual changes within the often ´diffuse` contents of the soil cover by the membership of horizons.
· Rule-based: The model aims to accept and use the expert knowledge of soil surveyors (loc.cit). In TRCS, rules can be defined interactively in order to link landscape features, as given in standard data sets, to models of the soil cover.
· Horizon-based: Continuous variables defining the soil continuum are not available at appropriate distances for applying direct interpolation techniques. Even experienced soil surveyors find it difficult to think within the soil geo-space in more than a limited number of additional attribute dimensions. Both reasons forced the use of soil horizons as the main carrier of information.
Spatial models of the solid soil are essential for landscape balances which take into account not only matter iputs and outputs of profiles, but also the effects of lateral translocations caused by water and tillage erosion as well as soil interflow. They are also important for the monitoring and prevention of ecologically risky substances, like phosphate, nitrogen and pesticides, and are a prerequisite to 3D dynamic modelling. Lateral translocations play major roles in most sloping European landscapes but levelled soils derived from fluvial and marine stratified sediments show significant lateral matter transfer, too. Horizons open the ‘understanding’ of soils to expert modelling, but imply a rather early classification. Therefore, horizon concepts will be discussed in detail and solutions offered to avoid or minimize information losses.
Concepts of horizon taxonomy
Traditionally, soils are described by horizons: horizontally stratified layers within the soil continuum, which specify the genetical status of chronal developments within and of soils.
The early designations A - B - C, originated by Dokuchaev, represent the effects or products of soil processes by three horizons: A signifies an ongoing period of topsoil humus accumulation, B summarizes transformations and translocations of minerals in the subsoil, and C is the unchanged parent material. Since then, various systems of designating horizons have been developed at international and national levels as summarized by Bridges, (1990, also ISSS 1985, Fitzpatrick, 1993). The German federal soil surveys use a system – more extended than others – of 15 main symbols (capital letters) combinable with 25 pre- and 31 suffixes (small letters and numbers). Preferably, it describes relative differences of attributes within soil profiles. Following the diagnostic chain of the survey manual and classification system (Arbeitsgruppe Boden 1994, WG on Soil Informatics 1985), values of basic field attributes, complemented by laboratory analyses, define horizon designations, and combinations of these the pedogenetical soil ‘classes’, ‘types’, ‘subtypes’ and ‘variations’. Subdivided by geogenetical criteria of texture stratification, ´local soil forms` are used at the lowest level to describe the contents of semi-detailed soil maps. Thus, horizons play a basic role in the information transfer of German soil surveys.
A background for horizon taxonomy is given by the extended chain of pedogenesis and pedofunction (Fig. 1; Schroeder and Lamp, 1976): based on the identity of factors and eco-components in a feedback ecosystem, a bundle of factors steer a combination of processes (isogeneous soils) which manifest themselves in similar attributes (isomorphic soils). These effect the same functions within the ecosystem (isofunctional soils) and extend mainly at the same location (isotopical soils). By this chain, all soil taxa defined from the different views should in principle coincide. The departures from this ideal which often occur in practice can in turn be used to optimize the system.
282
European Soil Bureau ¾ Research Report No. 4
Figure 1: Extended chain of pedogenesis and pedofunction
282
European Soil Bureau ¾ Research Report No. 4
The four principles of soil classification – affinity, similarity, functionality and neighboorhood -–characterize the most important views on soils. They have entered TRCS as follows.
Affinity: This principle relates soil attributes to external eco-components, especially the pedogenetical factors climate, organisms, relief, parent material and time. Jenny (1942) was the first to formalize this view and to describe the cause-effect relationships under ceteris-paribus conditions: litho-, bio-, topo-, and chronosequences provide an idealized description of the pattern of soils in the landscape. Thus, pedogenetics was founded by Dokuchaev and extended by more than three generations of pedologists and soil surveyors. It is summarized in many textbooks, publications and map reports, and as the paradigm of soil surveys it is still the basis for making maps. Relief and vegetation data especially can easily be captured in the field, and by including parent material where geological maps are available, these sources are used to map soils indirectly, thus reducing the efforts of soil augering. Soil indicators, especially topsoil colour and structure which can be sensed remotely, increase the prediction power. The US Soil Survey uses colour air photos with great advantage, but also in North Germany the survey effort can be reduced with the aid of stereo-images to 40% (without quality loss in mapping: Jakob, 1981). Therefore, TRCS uses panchromatic images and base topographic maps, including 1m elevation isolines, and the soil rating data which are available for all agrarian landscapes of Germany. These standard sources are transferred via fuzzification into the landscape model which is related to the soil model by means of rules about the factor-soil relationships. In order to include the local pedogenetical knowledge, these rules are interactively defined by soil experts and constitute the soil-landscape model.
Similarity is based on assumptions about internal relationships of soil attributes which can be investigated by various techniques of multivariate inference, ordination and cluster analyses. If soil taxa are treated as independent (class) variates, the relative intra-class-variance of special soil attributes can be evaluated and interpreted as a morphometric measure of the classification quality. Profile databases can be used to answer these questions and another about the depth function of horizon taxa which will help to set up and improve the taxonomic soil model as used in TRCS.