Ya. A. Pachepsky

Pedotransfer functions: bridging the gap between available basic soil data and missing soil hydraulic characteristics

Water retention and hydraulic conductivity are crucial input parameters in any modelling study on water flow and solute transport in soils. Due to inherent temporal and spatial variability in these hydraulic characteristics, large numbers of samples are required to properly characterise areas of land. Hydraulic characteristics can be obtained from direct laboratory and field measurements. However, these measurements are time consuming which makes it costly to characterise an area of land. As an alternative, analysis of existing databases of measured soil hydraulic data may result in pedotransfer functions. In practise, these functions often prove to be good predictors for missing soil hydraulic characteristics. Examples are presented of different equations describing hydraulic characteristics and of pedotransfer functions used to predict parameters in these equations. Grouping of data prior to pedotransfer function development is discussed as well as the use of different soil properties as predictors. In addition to regression analysis, new techniques such as artificial neural networks, group methods of data handling, and classification and regression trees are increasingly being used for pedotransfer function development. Actual development of pedotransfer functions is demonstrated by describing a practical case study. Examples are presented of pedotransfer function for predicting other than hydraulic characteristics. Accuracy and reliability of pedotransfer functions are demonstrated and discussed. In this respect, functional evaluation of pedotransfer functions proves to be a good tool to assess the desired accuracy of a pedotransfer function for a specific application.

Fractal models for predicting soil hydraulic properties: a review

Modern hydrological models require information on hydraulic conductivity and soil-water retention characteristics. The high cost and large spatial variability of measurements makes the prediction of these properties a viable alternative. Fractal models describe hierarchical systems and are suitable to model soil structure and soil hydraulic properties. Deterministic fractals are often used to model porous media in which scaling of mass, pore space, pore surface and the size-distribution of fragments are all characterized by a single fractal dimension. Experimental evidence shows fractal scaling of these properties between upper and lower limits of scale, but typically there is no coincidence in the values of the fractal dimensions characterizing different properties. This poses a problem in the evaluation of the contrasting approaches used to model soil-water retention and hydraulic conductivity. Fractal models of the soil-water retention curve that use a single fractal dimension often deviate from measurements at saturation and at dryness. More accurate models should consider scaling domains each characterized by a fractal dimension with different morphological interpretations. Models of unsaturated hydraulic conductivity incorporate fractal dimensions characterizing scaling of different properties including parameters representing connectivity. Further research is needed to clarify the morphological properties influencing the different scaling domains in the soil-water retention curve and unsaturated hydraulic conductivity. Methods to functionally characterize a porous medium using fractal approaches are likely to improve the predictability of soil hydraulic properties.

Maximum compactibility of Argentine soils from the Proctor test;: The relationship with organic carbon and water content

Soil compaction is recognized as an increasingly challenging problem for the agricultural, horticultural and forest production in many climatic regions. The Proctor test provides a standardized method to study compactibility of disturbed soils over a range of soil water contents. The objectives of our study were: (a) to determine values of the critical water content for compaction and maximum bulk density from Proctor compaction curves for soils different in their properties; (b) to study the correlation between the maximum bulk density and readily available soil properties. Thirty soil samples were taken from six different locations in Argentina between 58 and 64°W and 34 and 38°S. The degree of saturation at maximum bulk density varied from 73.2 to 96.8%. Comparison of our data with data of two studies in USA showed that relationships between the maximum bulk density and the critical water content were similar to these studies. However, the slope of the relationship between the maximum bulk density and the organic carbon content was 50% less in our study as compared with the two others. The maximum bulk density was highly correlated with the organic carbon content and the silt content, the determination coefficient of the multiple linear regression, r2, was 0.88.

Use of soil penetration resistance and group method of data handling to improve soil water retention estimates

Abstract The accuracy of pedotransfer functions can be improved using more flexible equations and additional input variables. Penetration resistance as a parameter related to soil structure can be a useful additional input to pedotransfer functions. Our objectives were to see whether using penetration resistance can improve the accuracy of estimating water retention from soil composition and bulk density. To develop pedotransfer functions, we applied group method of data handling (GMDH) resulting in hierarchical polynomial regression networks or abductive networks. The advantage of GMDH is that it automates finding essential input variables to be included in pedotransfer functions and, unlike the artificial neural networks (ANN), presents an explicit form of the equations. We developed pedotransfer functions from data on texture, bulk density, penetration resistance, and water content at 0, −5, −10, −20, −100 and −1500xa0kPa in 180 samples of soils in New Zealand. Abductive networks were used to estimate water content at particular matrix potentials. The water content at −1500xa0kPa and the penetration resistance were the essential variables to include in pedotransfer functions along with bulk density and texture. The pore volume fractal dimension could be reliably estimated from the water content at −1500xa0kPa and penetration resistance. The variation coefficient rather than average value of penetration resistance was found to be a good predictor in some cases.

Scaling of soil water retention using a fractal model

A quantitative description of the spatial variability of soil hydraulic properties is increasingly important, especially in simulations of soil water regimes. The objective of this study was to use a fractal concept to quantify and simulate the spatial variability of water retention of fine textured soils over a wide range of soil matric potentials. To scale water retention data, we assumed fractal self-similarity of pore volume and derived an equation of the water retention curve in the form of the probability integral. Experimental data on water retention were obtained on sand-kaoline plates and above saturated salt solutions in the ranges of the soil matric potential from −50 to −1 kJ m-3 and from −140 to −20 MJ m-3, respectively, for chernozem soil (Typic Haploboroll, clay loam) sampled over a 12,000 m2 area and for dark chestnut soil (Ustolic Orthid, loam) sampled over a 1500 m2 area. The resulting formulae gave a good fit of water retention data. The spatial variability of water retention could be described by the spatial variability of the single dimension-less parameter. Pore fractal dimension could be considered constant over the areas of sampling for both soils.

Comparison of fractal dimensions estimated from aggregate mass-size distribution and water retention scaling

Relating soil structure to soil hydraulic properties is an important issue for understanding and managing soil functioning. Fractal models were applied to relate soil water retention and soil structure. One such model developed by Rieu and Sposito (1991a) predicts an equality of (i) soil matrix fractal dimensions derived from aggregate bulk density- aggregate size data and (ii) soil fractal dimensions derived from water retention. The objective of this work was to test the statistical hypothesis of such equality for model soil systems of packed soil aggregates. Typic Argiudol and Typic Argiaquol soils were sampled at eight locations that differed in long-term management practices. Soil water retention was measured at 12 levels of soil matric potential ranging from -1200 to -20 kPa. The aggregate bulk density-aggregate size data were obtained for seven ranges of aggregate sizes from 0.25 to 16 mm. The statistical hypothesis about the equality of mass fractal dimension as derived from the aggregate bulk density data and the scaling dimension of water retention could not be rejected in our study.

Indirect estimation of soil hydraulic properties to predict soybean yield using GLYCIM

GLYCIM, a mechanistic model of soybean (Glycine Max L.) growth and development, requires soil hydraulic parameters as input. These data are usually not readily available. The objective of this study alas to compare yields calculated with measured hydraulic properties to those calculated with hydraulic properties estimated from soil texture and bulk density. We reviewed estimation methods and chose two methods to estimate a soil moisture release function and two methods to obtain saturated hydraulic conductivity. Both methods use soil texture and bulk densit), as predictors. Soil water retention predicted bJ3 these methods correlated ,t,ell with measured soil water retention whereas the estimation of saturated hydraulic conductivity was poor. Soybean yields were simulated using GLYCIM with and without irrigation ,for seven locations in Mississippi. USA, using seven years of weather records. Simulated Isields lz’ere aJTected more by the method of estimating the moisture release curve than b>, the method of estimating saturated hJ,draulic conductivity. The average simulated yields from estimated properties tz‘ere higher than those from measured properties because estimated bllater retention provided more availuble water. Correlation betbtleen yields simulated using measured and estimated hydraulic properties was higher under non-irrigated conditions than ti,ith irrigation. Averaging I’ields over years lcith diflerent lveather conditions greatl_v improved the correlations. Published b?, Elserier Science Ltd

Effect of soil organic carbon on soil hydraulic properties

Publisher Summary The primary soil hydraulic properties that soil organic carbon affects are porosity, soil water retention, and hydraulic conductivity. This chapter discusses the effect of organic carbon or organic matter on these properties and explains the way these effects can be incorporated into pedotransfer functions (PTFs). The sensitivity of water retention to changes in organic carbon content decreases as the initial organic carbon content increases. A similar conclusion can be drawn from the equations presented for water retention of soils amended with the organic waste in soils in the United States, England, India, and Germany. In a study discussed in the chapter, the reduction of the effect of increasing organic carbon content on water content at −5 kPa with the increase in the original value of the organic carbon content was reported. Water retention of peat soils presents a limit case for the increase of organic carbon content in samples. Soil-survey databases contain data on soils in natural ecosystems and on agricultural soils showing similar responses of soil water retention to changes in organic carbon content. Modeling of the changes of organic carbon content in soils and related changes in ecosystem productivity attracts significant attention with regard to climate changes and management changes. Existing models lack the feedback effect of organic carbon content accumulation on water retention and saturated hydraulic conductivity. Results presented in the chapter can be used in those models to improve their predictive ability.

Convective-diffusive model of two-dimensional root growth and proliferation

Simulations of crop productivity and environmental quality depend strongly on the root activity model used. Flexible, generic root system models are needed that can easily be coupled to various process-based soil models and can easily be modified to test various hypotheses about how roots respond to their environment. In this paper, we develop a convective-diffusive model of root growth and proliferation, and use it to test some of these hypotheses with data on the growth of roots on potted chrysanthemum cuttings. The proliferation of roots is viewed as a result of a diffusion-like gradient-driven propagation in all directions and convection-like propagation downwards caused by geotropism. The finite element method was used to solve the boundary problem for the convective-diffusive equation. To test hypotheses, we wrote modules in a way that caused a test parameter to be zero, should the hypothesis be rejected. These modules were added or removed to test each hypothesis in turn and in various combinations. The model explained 92% of the variation in the experimental data of Chen and Lieth (1993) on root growth of potted chrysanthemum cuttings. For this dataset the following hypotheses were accepted: (1) root diffusivity (colonization of new soil) did not depend on root density, (2) there was no geotropic trend in root development, (3) potential root growth increased linearly with root density, (4) there were (at least) two classes of roots with different rates of growth and proliferation, and (5) potential root growth rate decreased with distance from the plant stem base.

Data mining and exploration techniques

Publisher Summary The data mining and exploration methods introduce algorithms that automate predictor and equation selections. This chapter describes three methods: artificial neural networks, group method of data handling (GMDH), and the regression tree that have recently been used in the pedotransfer function (PTF) development. Each of these methods has its advantages and disadvantages. For example, the advantage of regression trees is the transparency of results, whereas the advantage of neural networks is the ability to mimic practically any relationship. The disadvantage of all these techniques as compared to statistical regression is the heuristic element involved so that the rigorous statistical judgment is hard to make. The three techniques practically produce identical PTF accuracy. The database exploration is a useful step that may generate PTFs that are either sufficient for the intended application or may suggest further applications of more rigorous or more flexible PTF-building techniques.