Principles of soil classification and the future of the South African system

Abstract

Humans classify their environment to create order, make it understandable, aid recollection and to communicate. The nature of these classifications is not always understood, because they are learnt from an early age. Building on these principles provides a sound basis for any scientific classification. This paper explores these principles, those of the USDA Soil Taxonomy, the World Reference Base for soil resources, and the South African Soil Taxonomy. Knowledge should be ultimate aim of soil classification. A hierarchical system with four levels is proposed for the South African Soil Taxonomy. This can easily be achieved by adding a higher level, proposed to be called a Soil Group, to the current three levels (form, family, and phase). The South African Soil Taxonomy must guard against too many taxa, because humans have a limited ability to comprehend numerous taxa. The distinguishing criteria between taxa should be more clearly defined, while at the same time guarding against becoming too data hungry. The classification should not shy away from intergrades. The object being classified (soils) is a natural system and intergrades will necessarily occur. It is proposed that these should be classified as intergrades, rather than trying to artificially separate natural soil bodies.

Keywords:

• soil classification
• USDA Soil Taxonomy
• World Reference Base

Introduction

Classification is a very human trait. Humans classify the objects around them to create order, thereby making it more understandable, to aid recollection and to communicate this environment, thus creating the environment for gaining knowledge. As such the classification process follows the knowledge tree stating that data is used to extract information and information is used to gain knowledge (Drucker 1990). Drucker (1990) further states that knowledge changes the individual or how the individual acts. Knowledge can therefore not be taught, but has to be learnt. Classification is consequently a language, unique for each discipline and sometimes a plethora of languages for each discipline. This language has to be learnt to enable communication in the scientific community.

Classification systems must, however, be upgraded from time to time to reflect the increase in the knowledge base and to address the increased demands placed on the interpretation thereof (Krasilnikov 2002).

The aim of this paper was to review the principles of soil classification, as well as the history and principles of the World Reference Base for Soil Resources, the USDA Soil Taxonomy, and soil classification in South Africa and to present principles for the upgrade of soil classification in South Africa.

Classification breaks natural continuums into discrete classes to improve communication among humans (Ibáñez et al. 2008). Classification therefore necessarily involves uncertainty (Roth 2005). At the most basic level categories of maximal distinction, or a mental image, are constructed. The latter implies cognitive bias. Humans, however, prefer to classify in levels. Think, for example, of the classification: food – fresh produce – fruit – apples – red apples. Five levels are applied in this example with increased focus and decreased variation in each class. According to Miller's rule, humans favour 7 ± 2 classes, with a maximum of 15 (Miller 1956). Miller's rule has been proven to hold true in a wide variety of disciplines (Wong and Richmond 1993, Tichý et al. 1999, Devillers et al. 2007).

Classes at the basic level are defined to be maximally distinct. This can be done by defining a prototype (the best or modal example) of a specific category. A prototype is necessarily generalised. Membership of a prototype is determined by the perceived similarity of a member to the prototypical category (Ibáñez et al. 2008). Distinction from a prototype is therefore based on exclusion.

Soil classification can be no different. It should follow the knowledge tree. Soil data should be used to construct information, which can be used to gain or infer knowledge. Knowledge should therefore be the ultimate aim, but it should be based on data. Additional to the objectives of classification listed above, soil classification also aims to reflect on the diversity of soils, their internal and external relations, it serves to inventorise the natural resource, and provide information on its protection and use (Dobrovski and Trofimov 1996, as cited by Krasilnikov 2002).

Principles of (soil) classification

Very few natural systems are ordered in such a way as to enable an entirely natural classification. The periodic table (Mendelejew 1869) is probably the only example of a natural classification system. Soil classification systems can be grouped into nominal (single level) systems, tables (two dimensional), or reference bases (Krasilnikov 2002). Nominal systems and tables eventually grade into taxonomies as new hierarchies are added.

Soils represent a diversity of individuals, with populations too numerous to comprehend. Some soils differ only in minor respects and are difficult to distinguish, whereas others may differ widely (van der Eyk et al. 1969). Soil classification aims to reduce the number of individuals and thus bring comprehension at the expense of detailed information. Archetypes (prototypes) are the point of departure for most soil classification systems, followed by diagnostic properties, which are used for differentiation. In most landscapes a limited number of archetypical soils are found, with intergrades between these. In other landscapes these intergrades might themselves be archetypical, whereas the archetypical soils of the first example occur as intergrades (Krasilnikov 2002). This implies that archetypical soils of different regions or countries would differ. Coupled with the fact that soil science is a relatively young science, it is probably the cause for the proliferation of soil classification systems internationally.

Creating exclusive archetypes in soil classification leads to disaster (Smith 1986). Few soil properties are exclusive to a specific soil and those soils that fall out after excluding two or more soil properties are very difficult to classify. The SA Soil Taxonomy (Soil Classification Working Group 1991) is largely driven by conceptual soil genetic classes, defined by qualitative morphological criteria. In comparison, the USDA Soil Taxonomy (Soil Survey Staff 2003) and WRB (WRB 2007) use quantitative soil properties as differentiating criteria.

The problem of intergrades

The differentiating criteria are selected to represent obvious boundaries. These boundaries are, however, not always obvious (Witty and Arnold 1987). The differentiating criteria defined for soil classification are also not necessarily relevant for land-use requirements. For example, the chemical and fertility requirements of crops do not fit soil taxon boundaries.

According to Krasilnikov (2002) the Russian soil classification defines a qualitative central archetype, implying some variability around the central concept. The USDA Soil Taxonomy quantitatively defines the borders of the archetype. Borders are fuzzy in the first method, whereas objects might be artificially separated by the second. These two approaches were combined by AFES (1998) in defining a strict quantitative central concept, but not for the borders of the concepts. An object is then referred to a specific class based on its similarity to the central concept. Intergrades are referred to two or more classes. Ibáñez et al. (2008) proposed that soil intergrades be addressed similarly to hybrids in botany by using the ‘x’ (cross between two soils) or ‘notho-’ prefix (Editorial Committee 2000).

The object being classified

What do we classify – the profile or the pedon? Or do we classify our idea of the object (Haskett 1995)? Refer, for example, to the statements by some pedologists that ‘this is a humic landscape’ or ‘this soil will behave like...and should therefore be classified as...’. AFES (1998) defines a soil mantle (three-dimensional soil body; the pedon of the USDA Soil Taxonomy), a soil profile (vertical change in a specific property) and a solum (sequence of horizons). The solum is the object of classification in AFES (1998). Krasilnikov (2002) argues that the latter is the logical object of classification, because our cognitive structure is overlain onto the soil profile. USDA Soil Taxonomy (Soil Survey Staff 2003) classifies the pedon: a soil body with a rough area of 1–10 m², whereas the WRB (WRB 2007) classifies any material within 2 m from the Earth's surface, with a horizontal area ranging from 1–10 m². The SA Soil Taxonomy classifies the profile or pedon as being representative of the polypedon.

Classification structure

Krasilnikov (2002) defines an idealised classification structure with four levels (five levels, if one includes a level ‘0’, which would be all soils). The first level is termed collective and is a group of common archetypes, enabling a more rapid classification and sometimes indicating a hierarchy in importance of soil properties. The second, generic level defines the archetypical soils. On the third, specific level soils are defined based on their deviation from the archetype. The fourth, variative level defines quantitative variations in some soil properties.

Diagnostic criteria

Soil classification serves soil survey (Smith 1986) and soil surveys are done with land use in mind (Witty and Arnold 1987). Although the USDA Soil Taxonomy uses soil properties with decreasing importance on lower levels (Witty and Arnold 1987), the selection of soil properties for use as classification criteria are subjectively selected with land use in mind. The implication is that this principle is subconsciously and not explicitly applied in soil classification systems. The soil depth criterion of the SA Soil Taxonomy is an example (Soil Classification Working Group 1991). This limit fits the auger length, which in turn fits the human body structure. We must therefore strive to distinguish between taxonomic practices, utilitarian bias, subconscious cognitive rules (tacit knowledge) and scientific proof. Evidence of subconscious cognitive rules in the USDA Soil Taxonomy is encouraging (Ibáñez and Ruiz-Ramos 2006).

Land-use requirements for cash crops vary only slightly but are increasing with increased crop production intensity. For example, a soil quality for annual cash crops in the western Free State is an impermeable clay layer. The depth requirement for this layer increased from 1 200 mm down to several metres. Depth of tree roots (both agroforestry and the natural biome) can exceed 5 m. More extreme variation is evident in environmental interpretations.

Stability of scientific value is also important. The challenge is to classify without perceptions and predictions of the rate of soil-forming processes. These processes form part of the natural nature of the system. This led soil classification to focus on subsoil properties as soil erosion can affect topsoil properties (Smith 1986).

Krasilnikov (2002), however, warns against delving too deep into the ontological problems of classification theory, while ignoring the peculiarity of soil classification, but also against delving too deep into the narrow problems of soil science.

World Reference Base

History

A resolution was passed in 1960 by the International Soil Science Society to publish a world soil map (FAO 1971–1981). In the same year the United States Department of Agriculture published the Seventh Approximation (Soil Survey Staff 1960). The FAO world soil map legend (FAO 1971–1981) was the precursor to the World Reference Base for Soil Resources (WRB 2007). In 1998 the WRB was recommended as a soil correlation system for the International Soil Science Society, while the European Commission selected the WRB as a correlation system for harmonized soil maps and databases for Europe (Jones et al. 2005).

Early soil classification systems were based on the recognition of soil-forming processes, rather than utilising quantifiable soil properties. The properties resulting from soil-forming processes are, however, more easily quantifiable than the soil-forming processes. Both the WRB and USDA Soil Taxonomy systems are therefore based on quantitative characteristics that form the definition for diagnostic horizons, properties and materials. In doing so, the soil forming processes are de-emphasised, but remain the underlying philosophy of the WRB classification.

The WRB was designed as an easy means of communication among scientists to identify, characterise and name the major soil types of the world. It is therefore not meant to replace national soil classification systems, but rather to be a tool for better correlation between national soil classification systems. It also aims to help in improving national soil classification systems and should therefore be wide enough to stimulate harmonisation and correlation of national classification systems. It should also serve as a communication tool for global soil databases and inventories.

Principles

Classification in the WRB is based on soil properties measurable and observable in the field that are used to define diagnostic horizons, properties and materials. The diagnostic characteristics used are related to soil-forming processes. It is recognised that an understanding of soil-forming processes contributes to a better characterisation of soils, but that they should not be used as differentiating criteria. The diagnostic features selected are used for their significance to soil management. Climatic parameters are not applied so that classification is not subordinated to the availability of climate data. The nomenclatures used are traditional terms, are easily introduced and should avoid confusion.

Some problems

The FAO map legend and eventually the WRB adopted terms from USDA Soil Taxonomy, but gave different definitions and/or criteria to some of these. Further, some terms adopted in the WRB were later modified in USDA Soil Taxonomy, but not in the WRB. Some terms from the USDA Soil Taxonomy are used differently at the various taxonomic levels. All of these led to a divergence of nomenclature between the WRB and USDA Soil Taxonomy.

Structure

The WRB utilises taxonomic units based on diagnostic horizons, properties and/or materials. These are used to define 32 reference soil groups (defined by a key) on the first level. Reference soil groups are differentiated on primary pedogenetic processes that produce characteristic soil features, except where special soil parent materials are of overriding importance. The key for the reference soil groups follows the sequence: (1) soils with thick organic layers, (2) soils with a strong human influence, (3) soils with limited rooting because of shallow permafrost or stoniness, (4) soils influenced by water, (5) soils set by Fe/Al chemistry, (6) soils with stagnating water, (7) accumulation of organic matter and a high base status, (8) accumulation of less-soluble salts or non-saline substances, (9) soils with a clay-enriched subsoil, and (10) relatively young soils or soils with little or no profile development.

The second level is defined by qualifiers, which are noted as prefixes and/or suffixes, in priority sequence and unique for each reference soil group. Qualifiers are differentiated according to secondary soil-forming processes that affect primary soil qualities. The qualifiers therefore emphasise soil features that are important for land use and management.

USDA Soil Taxonomy

History

To set up and improve the USDA Soil Taxonomy took 1 000 work years and over 25 calendar years up to 1975, the year of publication. The aim was to include all soils, and thus by implication produce a universal classification system (Guthrie 1987). The polypedon is the classification unit and a collection of polypedons forms a soil series (Witty and Arnold 1987).

Knowledge of soils grew with increasing soil-related solutions to land-use problems (Guthrie 1987). Soil classification therefore serves soil survey (Smith 1986), implying that the diagnostic criteria must serve the mapping procedures and will change with significant changes in mapping requirements (driven by land-use requirements), mapping techniques and technology, for example hydraulic augers and digital soil mapping.

Land-use requirements are strongly linked to agriculture because dry-land cropping and agroforestry place more demands on soil information. Rangeland on the other hand does not have similar requirements. Although land use drives classification, new soils are only identified when the identification facilitates mapping. Mapping only makes sense if the soil properties of the map units are quantifiable (in the field or laboratory), repeatable and related to genesis but not genesis itself. Subsoil properties are preferred to avoid human-impacted changes to the topsoils. Application of these principles resulted in the definition of 15 000 series up to 1987 (Witty and Arnold 1987).

Several critical decisions were taken in the development of the USDA Soil Taxonomy. For example, in the early format of the USDA Soil Taxonomy, terminology from several languages complicated development and it was decided to stick to the classical languages (Guthrie 1987).

Structure

The USDA Soil Taxonomy has six levels, with 12 Orders at the highest level. It defines nine topsoil horizons, 19 subsoil horizons, and 26 and 33 diagnostic properties for organic and inorganic soils, respectively (Soil Survey Staff 2003). Definitions of the diagnostics served to create uniformity in the classification (Guthrie 1987) and serve as the vocabulary of the soil classification language. In spite of the large number of definitions (for the 28 diagnostic horizons and 59 characteristic diagnostic properties), several soil properties are also defined by lengthy descriptions that can probably be replaced with definitions.

The USDA Soil Taxonomy is property based and concepts are discussed as guidelines, but not as criteria. Properties include horizons and soil morphological properties. The use of definitions for horizons and characteristic diagnostic properties limit descriptions of diagnostic criteria and enhance communication. This supports an accurate natural and concise classification of the pedon. The topsoils are anthropic (man-made), folistic (organic carbon rich, not saturated and very low bulk density of <0.1 Mg m³), histic (organic carbon rich, saturated and very low bulk density of <0.1 Mg m³), melanic (dark, acid and high organic carbon of >6%), mollic (less dark, less organic carbon and less bases), ochric (doesn't meet these criteria), plaggen (manured) and umbric (defined by nine exclusions). The technique used for excluding properties to define the umbric does not support development of a unique soil entity.

The system keys out Orders with unique characteristics and to a degree a template is created. Vertisols are a good example of the template created by the key. A limitation of the Order level is the use of climate because it implies that all soil properties are in phase with the current climate and at least suggests that climate is the dominant soil-forming factor controlling the development of soils and soil properties.

Suborders are mainly keyed out on the soil water regime. The first criterion applied is the duration of saturation with water and then plant-available water. This line of thinking is supported by the use of the water control section and the temperature control section, indicating that suitability for crop production was important in designing the system.

Accentuating diagnostic properties is applied for example in combining soil water regime with soil chemistry in the Aridisols. Saline, duripans (silcrete), gypsum and lime properties are used in this sequence to key out soils. Another example is the use of the nature of organic carbon in the suborders of the Histosols.

At Great Group, Subgroup, and Family levels the key seems to be driven by classification convention logistics rather than soil properties. In principle the creation of levels are just another key system and the result shows in these levels.

Series, at the sixth level, is a pragmatic grouping and therefore an easily identifiable soil body with unique characteristics.

The keying principle, used by the USDA Soil Taxonomy, makes the naming of soils easy and accurate. However, it depends on the identification of a single soil property or group of properties that is unique to that group that does not occur in other soil individuals. Interrelationship of soil properties and lack of knowledge of the phases of formation creates a complex system to work with. The challenge is to identify individuals, for example Vertisols and Oxisols, which are distinct and can serve as vehicles that transport information and, second, to develop a system that identifies them accurately and repeatedly, and preferably by lay persons.

The six-level structure of the USDA Soil Taxonomy supports a system that has maximum scientific value on the highest level and is user-friendly on the lowest level. The highest level of criteria selection for the USDA Soil Taxonomy was controlled by the level of understanding of soil genesis. Soil properties and factors of soil formation related to the inferred soil-forming processes were used, whereas soil properties with decreasing importance were applied on lower levels. Soil properties subject to change, for example soil depth and mineralogy, were only applied at Family level (Witty and Arnold 1987). In total 90 definitions of formative elements and unique adjectives makes it possible to interpret the properties of all soil names on all levels excluding the series level.

The format of USDA Soil Taxonomy supports repeatable user-friendliness by thoroughly defining the diagnostic criteria. The use of [the USDA] Soil Taxonomy has effectively placed all soil scientists on an equal footing with regard to their ability to classify a soil (Ditzler and Ahrens 2006: 142). The key system of the classification supports consistent application of criteria to define diagnostic horizons and features (Ditzler and Ahrens 2006).

Soil classification in South Africa

A brief review of the history, development, and proposals for future development is also given by Laker (2000, 2003).

The SA Soil Taxonomy can be described as a morphogenetic system, similar to that proposed by Kubiëna (1953). The soil classification systems in New Zealand (Cutler 1983) and Australian (Isbell 2002) use a similar approach.

Beater (1957, 1959, 1962, 1970) first described soils in association with the geological formations of the KwaZulu-Natal Coastal Belt. Horizons were recognised, profiles described, and summaries of the soil properties recorded. He defined soil series that could be regarded as a collection of soil profiles of similar morphology and were distinguished by recognisable differences in underlying geology within the relatively narrow geographical region of the coast belt. De Villiers (1964, 1965) developed the concept of soil series further, describing variations in soil properties within physiographic and climatic regions. Conceptual genesis, mainly interpreted from morphology with some chemical and mineralogical data profiles, was given prominence thus linking soil properties with the factors of formation. MacVicar (1965) extended the soil series concept by analysing and describing soils of a single underlying geology formation over a climate and time sequence.

MacVicar et al. (1965a) compiled the first soil classification prototype for South Africa, based on the first soil series list, which included soil profile descriptions and analyses of modal sites (MacVicar et al. 1965b). Additional soil surveys, notably the Key Areas Surveys (Loxton 1962, MacVicar and Loxton 1967, Roberts 1969, Verster 1971, Verster 1973) as mentioned by MacVicar (1978) and the Soil Classification Projects of the Highveld and in the north-western Free State (Loxton, Hunting and Associates 1970a, 1970b, 1970c) added to the understanding of soil distribution. Soil profile morphology was the central concept in the recognition of soils, with strong associations to the underlying geological formation and physiographic region, implying a mode of soil genesis through the climate–time relationship.

In the absence of a formal soil classification system, soil series were recognised as a collection of individuals, essentially similar in differentiating characteristics (morphological properties) and genetic horizons. Grouping of similar soil profiles with similar morphology around a central genetic concept appears to have played the major role in soil recognition.

Soils of the Tugela Basin (van der Eyk et al. 1969) set the initial principles and practice used in South African soil classification. Soils of the Tugela Basin describes the ‘sphere of pedological influence’ as the material at or near the surface, darkened by organic matter that supports an active biotic community and merges imperceptibly with hard rock. The arbitrary chosen limits, particularly the horizon definitions and the depth limit, give expression to the classifiable soil body. Properties that can be objectively described or to which numerical values can be assigned were considered suitable for distinguishing between soils. The use of covariant criteria was preferred in the transfer of information. This classification created classes accommodating natural soil bodies as opposed to single-value classifications (e.g. pH classes). In creating classes, the major practical objective is stated as that where different people working independently will identify a profile in the same way.

Van der Eyk et al. (1969) defined diagnostic horizons. These were developed from the bottom upwards, but were designed to function from the top downwards. In designing the classification, groups of like profiles considered to represent different classes were identified first and then placed in the higher classification class. However, in applying the classification, profiles were identified first as a member of the higher class (soil form) and then of the lower class (soil series). It was stated that laboratory analyses should not be required for classification. The classification functioned as a key, but it was more than a key because it revealed genetic relationships.

In applying the criteria for the classification of diagnostic horizons, all soils that have the same kind and number of diagnostic horizons must meet (1) the specifications of definitions, (2) occupy a specific position in the profile relative to other horizons, and (3) occur within 1 200 mm (48 inches) of the soil surface. The latter criterion is described as an arbitrary chosen limit. The O, A, B, C, R, G, and A2 (now known as an E horizon) master horizons are recognised. Transition horizons (illustrated in a text sketch as A3 and B1) are assumed to be thin and were not assigned prominence in the classification. The classification is thus a hierarchal system with the lower classes defined by the limits set for higher classes. It thus departs from the initial concept of the soil series, which was a grouping of similar individuals. These principles and definitions were subsequently adopted with only minor alterations in the Binomial Classification (MacVicar et al. 1977).

Van der Eyk et al. (1969) cite literature that gives valuable insight into the thinking, construction and sources of the classification. They also present profile descriptions and analytical data of modal profiles. However, the use of geographical series names tends to retain the association of soil series as a collection of individuals associated with those places rather than as defined in the Binomial Classification.

The Binomial Soil Classification System (MacVicar et al. 1977) is a two-class system with identification of the soil form first and then the series. The procedure for identifying soils is established by demarcating master horizons, identifying diagnostic horizons, establishing the soil form, identifying the series differentia and establishing the soil series (van der Eyk et al. 1969, MacVicar et al. 1977). Brief definitions of master horizons, including an A2 (E horizon) and G horizon and their sequence are given. Master horizons may differ widely in their properties. The definitions of diagnostic horizons include and exclude certain properties. Classification is made possible only by the literal application of the definitions. The soil form and series is established through soil form – series schedules. However, there does not seem to be a defined order in the keying out of soil forms. Soils with unique topsoils seem to key out first, then soils with subsoil water saturation, followed by structured soils, soils with apedal subsoils, and then the young soils.

In the SA Soil Taxonomy (Soil Classification Working Group 1991) the identification procedure is essentially similar, although extended definitions to those of the Binomial System (MacVicar et al. 1977) for certain horizons were made. Provision for certain undifferentiated material (young alluvia, raw sand, man-made deposits and rock) were made, whereas transition horizons take on an altered significance. Definitions include certain properties and exclude others, are phrased in terms of soil properties and are as concise as possible. Definitions give expression to the concept. Soil forms were largely retained without alteration, whereas the soil family was introduced below the soil form to replace the Binomial System soil series. In the identification procedure an additional step was introduced, namely to determine the textural class of the A horizon. Although a three-category system was recognised, the practical information transfer was largely achieved through the two higher categories, thus retaining the essential two-category nature of the system where the essential soil interpretation evaluations are performed.

The Soil Classification Working Group (1991) states that the diagnostic criteria will neither account for anomalous cases, nor take care of all eventualities. Anomalous cases, not comfortably accommodated in the classification, should be classified by relying on personal judgement, common sense, and should be backed by an understanding of the concepts and intentions behind the definitions of diagnostic horizons and materials. The diagnostic horizon concept follows the definitions. Increased knowledge, large-scale adoption, by numerous users, and in varied geographical settings led to these concepts becoming so vague that the intention of the original authors is no longer well understood. The divergent interpretations in personal judgements among pedologists are also increasingly defeating the purpose of the classification. The SA Soil Taxonomy proposed the development of soil series, but except for one attempt (Turner 2000) did not come to fruition.

The soil form specifies the kind and sequence of diagnostic horizons and materials present and, in some cases, also the general nature of the underlying material. The soil family is defined by a narrower range of variation of soil properties or recognises properties not used to define the soil form. The key to soil forms follows a sequence of soil forms. A logical sequence key is not apparent and is not discussed in the text. The key to the soil forms seems to first key out soils with unique topsoils, followed by wet soils (in the sequence wettest to driest), then structured soils, young soils, aridic soils, very young soils and, lastly, anthropogenic soils.

Proposed principles for soil classification in South Africa

The principles proposed here are based on an analysis by the authors of the strengths and weaknesses in the SA Soil Taxonomy (Soil Classification Working Group 1991).

Strengths

The strongest point of the SA Soil Taxonomy is that the soil forms represent natural soil bodies. The SA Soil Taxonomy aims to identify individuals and then presents criteria to distinguish between these. The system is morphology based and very few, if any, analyses are required to classify most soils. Soil classification is therefore quick and cheap. The predictability of the behaviour of the soil forms makes it applicable to a wide variety of land interpretations. It is widely used in the agronomic sector, but is also suited for hydrological interpretation (van Huyssteen et al. 2005, Le Roux et al. 2010, Kuenene et al. 2011, van Tol et al. 2011), environmental impact assessments, wetland delineation (DWAF 2005) and engineering interpretations (Jones 1977, Partridge et al. 1993). The system therefore has widespread national acceptance and utilisation.

Weaknesses

The diagnostic criteria include concepts, conventions, material and methods, structure and precedence often with several of these intermixed in one criterion. This is illogical and leads to confusion. A good example is the widespread confusion on the classification of red and yellow-brown apedal B vs neocutanic B horizons. The 1 500 mm depth limit is outdated, no longer valid and often ignored. Soil surveys in the semi-arid production area are currently routinely done to a depth of 3 m. Similarly, interpretation for soil hydrology requires soil information to a depth of several metres (Le Roux et al. 2010). Diagnostic criteria for the classification of and differentiation between unspecified and unconsolidated material with signs of wetness should be better defined. Consideration should also be given to the combination of these two diagnostic horizons. Their properties are, however, different to that of G horizons (van Huyssteen et al. 2005). The practical application of master horizon classification is ambiguous and may lead to circular arguments. In practice, selective interpretation of the master and transitory horizon concept and hence its demarcation can be used to manipulate the resultant classification. More information can be added through the use of qualifiers, as applied in the WRB (2007) and USDA Soil Taxonomy (Soil Survey Staff 2003). Lack of a policy or procedure for new diagnostic horizon and soil-form identification will result in a proliferation of soil forms. The number of soil forms is becoming counter-productive from a classification perspective. Care should therefore be taken before new soil forms are added, while at the same time considering the reduction of others, for example the podzol soil forms. The diagnostic criteria employed are not in line with those employed by the WRB or USDA Soil Taxonomy. The presence of an abrupt transition in the Katspruit and Kroonstad soil forms are not captured at the form or family level. The differentiating criteria and naming of soil Families are clumsy. The classification of soil families can easily be replaced by qualifiers, enhancing user-friendliness. Some classifications can only be made after costly analyses are made and are not adhered to in practice. The organic carbon content for the organic O and plasticity index for the vertic A are pertinent examples. In the organic O and G horizons reference is made to long periods of saturation with water without any indication of what ‘long’ means. Similarly, one of the criteria of the soft plinthic B horizon states that mottles related to a fluctuating water table is indicative of the conceptual morphological nature not defined in any publication. Some of the calculations based on analyses are difficult, counterintuitive and clumsily explained.

Principles

The principles applicable to the classification system should first be defined before criteria can be established. If criteria are applied before principles are laid down, the danger exists that it can become a personal and subjective idea, not to be confused with tacit knowledge. Principles and diagnostic criteria should therefore preferably be based on sound (published) peer-reviewed research. Most of the decisions taken on soil classification criteria for the 1977 and 1991 versions were based on tacit knowledge built on visual observations and are therefore interpretative and subjective. Whilst these observations may have served the purpose of the time, they must now be subjected to strict review to validate their present significance. Written and traceable archived reports must form the basis of the classification. These reports must adhere to good scientific principle, must contain a minimum information content that has received acceptance by the scientific community through publication or repeated and traceable application

All diagnostic criteria in the SA Soil Taxonomy need to be scientifically described and referenced. Reference to texts, research data or tacit knowledge will establish an essential record of the understanding thereof.

The SA Soil Taxonomy is a morphologic conceptual system. Subconsciously it was, at least partially, developed into a soil quality-based system. Soil qualities such as drainage, fertility, physical activity and environmental accumulation can be attributed to soils. For example, red apedal B horizons indicate good drainage, whereas G horizons indicate poor drainage; humic A and vertic A signify fertile soil, vertic A are physically active, and organic O and G horizons are poorly drained.

User-friendliness

User-friendliness impacts on the applicability of the classification system. The value of a system is probably not affected by the degree to which it is user-friendly, but the impact is. Custom-made national soil classification systems only accommodate the variety of soils experienced in the country. The SA Soil Taxonomy is an example, because it does not include frozen soils and others not present in the country. National systems are therefore limited to the country of origin and are therefore expected to classify soils well only in countries with similar factors of soil formation. This is the case with the SA Soil Taxonomy that classifies Australian soils with some degree of accuracy.

Classification systems vary in their ability to effectively communicate essential information. Not only does the classification system influence the conveyance or transfer of information, but it also influences the quality of communication. Pedotransfer is best done by morphology because it is the only visible relationship. Soil qualities are best transferred by genetic history also linked to morphology.

A limitation born out of the previous statements is that soil classification systems focus on the symptoms or properties of soil genesis, which are related to the past, rather than the causes, which are used to predict future behaviour. This is present in the SA Soil Taxonomy where focus is placed on the identification of a soft plinthic B horizon while omitting the slowly permeable horizon underlying the soft plinthic B horizon. Several identifiable layers are known to occur below soft plinthic B horizons and all have significant implications. These include soft carbonate horizons (sandy, wet, arid conditions), pedocutanic and prismacutanic horizons (dry, semi-arid conditions), G horizons (wet, semi-arid to subhumid conditions) and sandstone (semi-arid to subhumid conditions). These horizons are presently not diagnostic and hence ignored, resulting in the loss of this information for interpretation.

It is not easy to classify boundary conditions without laboratory support. User-friendly systems tend to steer away from defining or addressing boundary conditions. The USDA Soil Taxonomy, a property-based system, takes this head on and requires laboratory analysis for classification (Soil Survey Staff 2003). However, because measurements are only valid for the sample analysed, the transfer of properties and qualities is done with morphology.

User-friendly terminology can lead to scientific limitations. An example is the term ‘signs of wetness’ (Soil Classification Working Group 1991) used to define redoximorphic features (Soil Survey Staff 2003). It complicates the communication of science as soil ‘with signs of wetness’ can be dry in the dry season and soil ‘without signs of wetness’ can be wet in the wet season. This is an example of mixing classification, naming and description of an entity.

Soil classification systems of the world vary between using numeric, soil property, morphological, and other features as their basis (Krasilnikov 2002). The morphological approach of the SA Soil Taxonomy makes it possible to apply it effectively in the field. However, some of these morphological properties are often not easily visible, for example gleyed features in organic and vertic soils.

The use of uniform or aligned criteria should be pursued as far as possible, while at the same time guarding against repetition in the definitions thereof, as is the case in Soil Survey Staff (2003). The natural approach of the SA Soil Taxonomy with its use of morphological criteria supports user-friendliness. However, the use of a natural approach tends to over-emphasise the conceptual pedogenesis. Alhough some horizons (vertic A) have well-defined pedogenetic concepts, including their formation and behaviour in land use, others horizons (neocutanic B) lack these concepts. A problem with conceptual pedology is that the predicted behaviour can override the concept, implying that behaviour is conceptualised. This cannot be the case in a property-based system. Using soil morphology as the carrier of soil data, including both soil properties (vertic A horizons are clayey) and soil qualities (red apedal B horizons are freely drained) is a major advantage.

Structure

The SA Soil Taxonomy has two levels with 73 taxons at the highest level, compared to six levels with 12 taxons or orders in the USDA Soil Taxonomy. The WRB has a less formal structure of 36 taxons and two levels. The SA Soil Taxonomy effectively defines the entities occurring in the country, but tends to proliferate the number of soil forms defining these entities, while limiting the number of soil horizons required to define each soil form. The structure of the SA Soil Taxonomy supports this proliferation of soil forms, while limiting the number of additional diagnostic horizons required to define each soil form. A good example is the recognition of the Lichtenburg soil form (orthic A/red apedal B/hard plinthic B) without defining additional diagnostic horizons.

The manner in which classification criteria address soil attributes differs. Systems can simply state the criterion or first state the exclusions. The latter tend to focus on the features absent before it defines what is present. Examples are the criteria defining mineral soil material as saturated with water for less than 30 days... (Soil Survey Staff 2003), which focuses the reader on saturation, instead of a lack thereof. The SA Soil Taxonomy uses many negative statements in its terminology (non-red/red colours, non-bleached/bleached horizons, does not qualify as...; Soil Classification Working Group 1991), giving the impression that the entity where the property is absent is most prominent. Users of the classification system are distracted by these statements.

Classification systems use terminology to define very specific values. These terms become the nouns (both common nouns and proper nouns) and adjectives used to classify and name the entity. USDA Soil Taxonomy is structured for the use of acronyms with very specific meanings. However, the limited use of terminology leads to long descriptions of cracks, linear extensibility and redoximorfic features in Alfisols (Soil Survey Staff 2003). The SA Soil Taxonomy should steer away from this, but at the same time define terminology well enough to allow for the consistent application of diagnostic criteria to prevent ambiguity.

All classification systems have conventions that ought to guide the classifier each time to the same answer. These conventions are not always well defined in most soil classification systems. In the SA Soil Taxonomy conventions are built into (sometimes implied) diagnostic criteria making the structure tedious. Examples are does not qualify... and occur under…horizons, which implies the criteria must be tested and that if it is identified in abnormal positions it is not diagnostic. These should preferably be controlled by classification conventions, rather than as diagnostic criteria.

Discussion

Soil classification systems should be rejuvenated regularly to ensure that they remain currently relevant. As such, classification systems should represent the knowledge that is continuously being generated. For example, the USDA Keys to Soil Taxonomy are periodically upgraded when new information becomes available (Soil Survey Staff 2003). This can easily occur every two years. The WRB follows a similar approach. The South African classification system was only upgraded once in 1991, 14 years after its first publication in 1977 (Soil Classification Working Group 1991). Attempts to review the system are currently ongoing.

Classification systems communicate a message using a name. Variation is inherent in all taxons and can therefore not be limited. The user of the classification system must learn the degree of permitted variation from definition criteria of the taxon. The SA Soil Taxonomy has good interpretative value for most cash and permanent crops as well as for environmental assessment (DWAF 2005). These interpretations flow from the intrinsic nature of the system since it inter alia uses soil properties associated with the soil water regime (van Huyssteen and Ellis 1997, van Huyssteen et al. 2004, 2005, 2007, Le Roux et al. 2010).

Concepts are impossible to apply accurately in the field and/or laboratory, but frequently occur in soil classification systems. Buol et al. (1997) state that conventional wisdom has at times and in places frozen soft, tentative hypotheses into hard dogma, preventing acceptance of new ideas and concepts leading to rigor mortis. The albic horizon and albic materials (Soil Survey Staff 2003) and the E horizon (Soil Classification Working Group 1991) are conceptualised to be zones of depletion of colloidal material including silicate clay, sesquioxides and organic carbon. In practice this is not strictly correct. Profile analyses of South African soils classified with an E horizon (higher Munsell colour value) indicate that 50% of the E horizons have more clay than the overlying A horizon, 40% have more sesquioxides and 20% have more organic carbon (Land Type Survey Staff 2007).

Well-defined concepts of entities are good vehicles to convey the message of variation and uniqueness. However, it does not solve the classification of boundary conditions. Definition of precedence through a key is therefore necessary. For example, the boundary between soft plinthic B and soft carbonate B horizons in the SA Soil Taxonomy is vague because the soft plinthic B may be calcareous and the latter may have mottles. Furthermore, the criterion ‘dominating the morphology’ specified for the soft carbonate B horizon is also vague and should be quantified.

Any soil depth criterion for classification should be based on the depth of biological activity and therefore the depth of impact of soil on vegetation (Fanning and Fanning 1989), because the aim to classify natural soil bodies can be restricted if the classification depth is limited artificially. Knowledge of natural soils also increased dramatically. Well-developed ‘diagnostic’ soil horizons can occur under solid rock layers implying that macrobiology is not the only part of soil formation. Microbiology, on the other hand, can occur several kilometres deep in the Earth's crust (Pfiffner et al. 2006, Borgonie et al. 2011). By implication soil morphology should therefore be a reliable indicator of soil depth. Application of soil classification information for environmental (Le Roux et al. 2010) and wetland evaluation (DWAF 2005) requires the absence of depth limitations. There is thus a discrepancy between soil classification for agronomic and environmental purposes. The question therefore arises if observation limits (soil auger reach) should be used as diagnostic classification criterion.

All soil classification systems of the world probably claim to be natural. The basic design of a soil classification system can focus on mapping soil qualities, properties (chemical, physical or morphological) or concepts. Morphological classification serves the pursuit of natural classification. The development of the SA Soil Taxonomy is clearly a hydrological driven approach because oxidation and reduction morphology play an important role. This is contradicted by the approach by geochemical approaches (Fey 2010). This implies that a concept drives the natural character. The USDA Soil Taxonomy aims to make soil surveys and to transfer knowledge gained to other maps using natural entities (Witty and Arnold 1987). The choice of differentiating criteria was intended to group soils with similar genesis, but genesis itself was not in the definitions. Instead, it was one step removed from the definitions. Although the thrust to classify is driven with soil formation in mind, the end products vary. The perception is that soil properties can be mapped (Witty and Arnold 1987) but soil qualities are evaluated for land use. The implication is that soil qualities can only be calculated using soil properties. This places the emphasis on selecting soil properties controlling soil qualities that influence land suitability. Soil property-based systems are suitable for this purpose. However, although soil morphology is difficult to quantify, it is probably most suitable for interpretation of many soil qualities. Soil morphology is also the only soil property that can transfer information when interpolating between soil survey observations or extrapolating soil survey data to regions outside the area of survey. While conceptual classification can be too vague for soil surveys of high quality, soil morphology-driven systems relate soil morphology to soil properties.

Revisions of the SA Soil Taxonomy should therefore consider (1) reducing the number of soil forms, (2) restricting the adoption of new soil forms, (3) developing stricter (more and better) differentiating criteria to distinguish between diagnostic horizons, (4) adding a higher category above the current soil form category, (5) prefix and/or suffix qualifiers should be introduced to describe the soil families, (6) remove ‘unspecified’ designations because it does not add information and only serves to confuse, and (7) preferably use basic cation saturation as a differentiating criterion for humic A horizons. It gives the same result, is more logical, and would better align with the WRB and USDA Soil Taxonomy.

Conclusions

The classified object (soil) must have extrapolation value to transfer data and information for the sake of gaining knowledge. Communication, integral in the transfer of information, can be improved further by adopting descriptive terminology, possibly in augmenting current geographic soil form names. Adding a higher level of classification is proposed to increase the taxonomic levels to four. Soil forms are already grouped informally by many soil scientists in South Africa. The number of soil forms should further be limited. Humans do not possess the ability to comprehend a vast number of taxa. The classification should not avoid classifying intergrades. The object (soil) being classified is a natural system, without natural classes. Forcing intergrades into taxa has the danger of splitting naturally occurring soil bodies, while at the same time ignoring the transitory nature of the soil being classified.

Acknowledgements

Funding by the National Research Foundation, the University of the Free State and the Agricultural Research Council is gratefully acknowledged.