Elsevier

Geoderma

Volume 108, Issues 3–4, August 2002, Pages 155-180
Geoderma

Marked differences between van Genuchten soil water-retention parameters for temperate and tropical soils: a new water-retention pedo-transfer functions developed for tropical soils

https://doi.org/10.1016/S0016-7061(02)00105-2Get rights and content

Abstract

All of the physical, chemical and soil water-retention data suitable for the derivation of a Pedo-Transfer Functions (PTF) for water retention for tropical soils (771 suitable horizons) were extracted from the IGBP-DIS soil database. The parameters θs, θr, α and n of the van Genuchten (vG) [Soil Sci. Soc. Am. J. 44 (1980) 892] equation were derived and compared with parameter values from two published data sets of temperate region soils. Thirty-five percent of the soils in the tropical (IGBP/T) database were classified as clays, compared with only 4% and 7% in the temperate databases. The IGBP/T soil bulk densities were significantly lower (p<0.001) for nine textural classes. For the more clayey textural classes, the value of α was higher for the IGBP/T soils, implying more large pores and more structure. For kaolinitic soils in the IGBP/T database, α was typically around 0.4 kPa−1 compared with 0.04 kPa−1 for montmorillonitic soils. Mineralogy is clearly important and should be included in PTFs (if available). The value of n for clay soils in the IGBP/T data set was significantly higher than for temperate soils, and θs was significantly higher for nine textural classes (p<0.001) reflecting their lower bulk density. θr was also significantly higher for all textural classes except sand. For clay, the mean value of θr was 0.27 m3 m−3, compared to only 0.11 m3 m−3 for temperate soils. For 17% of the IGBP/T data, θr exceeded 0.3 m3 m−3, a value often used as a fitting constraint. A tropical soil PTF was developed using multiple regression techniques. This predicted water-retention curves more reliably than either class PTFs or “soil class” PTFs, but was not reliable for low-density soils such as Andosols. The IGBP/T data set includes some groups of soils, e.g., Andosols and Ferralsols, whose properties are extremely different from those of most temperate soils. These differences emphasise the need to develop separate PTFs for tropical soils, or perhaps for specific groups of soils. There were few water-retention data from below 1.5-m depth and a complete lack of hydraulic conductivity data in the IGBP/T database. This is a major impediment to the modelling of water movement and uptake in deep-rooted ecosystems, for example, tropical forest and savannah.

Introduction

Parameters to represent the hydraulic properties of different soils are essential for a wide range of modelling studies, such as those to predict crop growth and yield, to evaluate agroforestry systems, and to represent the link between the soil, the vegetation and the atmosphere (SVAT models). The latter are vital components of General Circulation Models, describing the land surface–atmosphere interactions. The sensitivity of the global water cycle to the water holding capacity of the land surface has been demonstrated by Milly and Dunne (1994). Entekhabi et al. (1996) reviewed the mutual interaction between soil moisture and the atmospheric environment, and showed that the variability of both weather and climate are influenced by the soil moisture state. If this linkage between the soil and weather and climate is to be modelled successfully, sound estimates of soil hydraulic properties are required, in particular, the amount of water that the soil can store, and its hydraulic conductivity.

On a plot scale, soil hydraulic properties can be measured, but it is a time consuming process. For larger areas, the most widely used method to obtain these properties is the use of Pedo-Transfer Functions or PTFs (Bouma and van Lanen, 1987). These are generally empirical relationships that allow the hydraulic properties of a given soil to be predicted from more widely available data, usually texture (percent sand, silt and clay), bulk density and organic carbon (OC) content.

PTFs may be used to predict single values, for example, the water content at a particular matric potential (e.g., Pidgeon, 1972, Lal, 1979, van den Berg et al., 1997) or available water capacity (AWC, e.g. Batjes, 1996, van den Berg et al., 1997). Other PTFs have been developed to predict the parameters of equations, such as those of Brooks and Corey (1964) and van Genuchten (1980), which describe the water-release curve, e.g. Rawls and Brakensiek (1985), Saxton et al. (1986), Vereecken et al. (1989), van den Berg et al. (1997), Scheinost et al. (1997), Schaap and Leij (1998), Tomasella and Hodnett (1998) and Tomasella et al. (2000). The latter type of PTF is more suitable for modelling purposes as the equations describe the whole of the water-release curve. Until recently, most PTFs were derived through multiple regression techniques, but the neural network approach (e.g., Pachepsky et al., 1996, Schaap et al., 1998, Tamari et al., 1996) is becoming more favoured.

Most of the PTFs to predict Brooks and Corey (BC) or van Genuchten (vG) parameters have been developed using extensive databases for the soils of temperate regions. van den Berg et al. (1997) noted that the often empirical relationships to predict AWC derived for the soils of temperate regions “appeared to be inadequate for Ferralsols and related soils, which are dominated by low activity clays”. They derived a PTF specifically for Ferralsols and related soils. Only this PTF, and those of Tomasella and Hodnett (1998) and Tomasella et al. (2000) for Brazilian soils, have been derived exclusively using data for tropical soils.

Regardless of the methodology used to develop it, any PTF is likely to give less accurate or possibly even very poor predictions if used outside the range of soils from whose data they were derived. The PTF of Rawls and Brakensiek (1985) is valid for soils with a clay content of 5–60% and a sand content of 5–70%, but some kaolinitic tropical soils, particularly Ferralsols, can have clay contents of 70–90%. This might suggest, from a temperate soils viewpoint, that they are “heavy” clays, with a low permeability and a moderate to high available water capacity (AWC). However, most have a low bulk density (0.9–1.2 Mg m−3), are highly permeable because of their microaggregated structure, and have a low AWC. Correa (1984) found an AWC of 70 mm m−1 for high clay content Amazonian oxisols. The very low AWC was confirmed by field soil moisture observations by Hodnett et al. (1995) who warned that PTFs derived for the soils of temperate regions should be applied with great caution to these tropical soils.

Tomasella et al. (2000) showed that a PTF derived using solely Brazilian soil data and tested on an independent data set of Brazilian soils, gave markedly better predictions than PTFs derived using temperate soils data. Performance was evaluated using the approach described by Tietje and Tapkenhinrichs (1993). Within the test data set, some of the data were outside the range of textural validity of the “temperate soil” PTFs tested, but the new PTF outperformed the former, even when tested within the range of validity of the data. The evaluation of a PTF using data that fall outside its range of validity may seem inappropriate. However, models have been developed which require soil water-release data worldwide, but in the near absence of PTFs developed for tropical soils, there is little alternative but to apply established temperate soil PTFs, regardless of their validity or suitability. It is important, therefore, to evaluate how well the PTFs will perform when applied outside the range of the data that were used to derive them.

The silt content of the soils in the Brazilian data set used by Tomasella et al. (2000) were very low, when compared to that of temperate soils, and it was noted that the prediction errors for the temperate PTFs decreased with increasing silt content, while those of the “tropical” PTF increased. The better predictions of the Brazilian soil PTF even within the range of textural validity of the temperate PTFs also suggest that there may be differences between temperate and tropical soils caused by factors other than texture. Structure will certainly play a role in determining the water-release curve, and mineralogy may also have an influence. The strong weathering and leaching processes in large areas of the tropics (loss of Ca, Mg, Na and K, accumulation of Fe and Al) tend to create particular mineralogies and soil structure, which are less common in temperate regions. These weathering processes have been going on for long periods of time, uninterrupted by ice ages. In temperate regions, glacial action over large areas in the Quaternary period will have contributed to the larger amounts of silty soils.

The aim of this work is to derive vG parameters for a wider range of tropical soils and to compare these with published vG parameters derived from temperate region soils data sets. The texture and bulk density are also compared, and the effects of clay mineralogy on the vG parameters of tropical soils are examined. A further aim is to develop both class and continuous PTFs for tropical soils, which will be tested on an independent dataset of tropical soils.

There are two main types of PTF, “class” and “continuous”. A class PTF is used to predict the hydraulic properties of a textural class, for example, silty clay loam, or sandy clay. Wösten et al. (1995) describe class PTFs as “cheap and easy to use” because only the textural class has to be determined. However, they have limitations “because the approach only provides one average hydraulic characteristic for each texture class”, even though there may be a considerable range of characteristics within a single textural class.

A continuous PTF is used to predict the soil hydraulic characteristics from other, more readily available data. Wösten et al. (1995) make the key point that the indirect methods (i.e., PTFs) cannot exist without the direct methods (i.e., field sampling/lab measurements), “because only direct measurements create the database from which indirect methods are derived”. This is a strong argument for the development of more physically based, rather than empirical, methods to derive soil hydraulic properties on a large scale. However, these are still at an early stage of development. As mineralogy is not normally taken into account in PTFs, a third possible type of PTF is suggested: a “soil class” PTF. This could be derived for a major soil class on the basis that within a major soil class or group, the range of structure, mineralogy and texture might be expected to be narrower than for soils as a whole. However, there is still the problem that many soils, particularly cambisols, can show marked changes of texture with depth.

Section snippets

Methods

The IGBP-DIS soils database was obtained from ISRIC in Wageningen. This database contains data from 13,1472 horizons, mostly in the continental US. There are data from 4156 horizons outside the US, but the majority of these are from nontropical areas. Only data from the tropics were selected for this study although the definition of the tropics was relaxed slightly to include soils between approximately 25°N and 25°S. The question of soils within the tropics but in temperate climates due to

Results

The large number of clay soils and low numbers of soils with a high silt content are clearly shown in the textural triangle in Fig. 1. A comparison of the distribution of soil textural classes in the entire IGBP/T data set (all 771 horizons) and two temperate data sets is shown in Fig. 2.

The most marked difference in the distributions was in the percentages of samples in the clay class: 35.4% of the IGBP/T data set were clays, compared to only 4.3% and 7.1% in the S&L and C&P data sets. These

Discussion

The very low proportion of clay soils in the temperate data was somewhat unexpected. The notes beneath Table 1, Table 2, Table 3, Table 4, Table 5 in Carsel and Parrish (1988) indicate that the clay class was limited to “Agricultural soil, less than 60% clay”. This could explain the smaller proportion of soils in the clay class when compared to the tropical data set (which included all soils classified as clay). However, the percentage of clay soils in the S&L data set was even lower although

Conclusions

The IGBP/T data set used contained all of the tropical soils data from the IGBP-DIS data set that was suitable for developing a PTF for water retention. Overall, there is a dearth of tropical soils data. Although the IGBP/T data set is rather smaller than the temperate databases and may not be fully representative of tropical soils, in general, the bulk density and vG parameters for many textural classes were significantly different (p<0.001) to those of the temperate data sets with which

Acknowledgements

The IGBP-DIS database was made available by the International Soils Reference and Information Centre (ISRIC) at Wageningen, the Netherlands. This work was partially supported by the Forestry Research Programme of the UK Department for International Development.

References (30)

  • J.C. Correa

    Caracterı́sticas fı́sico hı́dricas dos solos latossolo amarelo, podzolico vermelho amarelo e podzol hidromórfico do estado do Amazonas

    Pesqui. Agropecu. Bras.

    (1984)
  • R. Lal

    Physical characteristics of soils in the tropics: determination and management

  • R. Lal

    Physical management of soils of the tropics: priorities for the 21st century

    Soil Sci.

    (2000)
  • F.J. Leij et al.

    Unsaturated Soil Hydraulic Database

  • L. Luckner et al.

    A consistent set of parametric models for the two phase flow of immiscible fluids in the subsurface

    Water Resour. Res.

    (1989)
  • Cited by (227)

    View all citing articles on Scopus
    View full text