Pore rigidity in structured soils—only a theoretical boundary condition for hydraulic properties?

Abstract

The quantification of water fluxes and the identification of site-specific hydraulic properties are widely required for various research, economic or environmental purposes. Pore rigidity, as one of the major boundary conditions, is always assumed to exist. Under in situ conditions the validity of this assumption is questionable and strongly depends on climate, land use, soil type and management. Changes of water content or water potential will change porosity for structured or homogeneous soils irrespective of geological origin and clay mineralogy. Shrinkage curves show volume loss with decreasing matric potential due to drying and this was more pronounced the less aggregated and the wetter the soils. Water retention curves as well as the hydraulic conductivity vs. matric potential curves differed all the more from the origin with increasing drying; the less rigid were the pores. The smaller the structural shrinkage range and the more pronounced the proportional shrinkage, the greater the difference between the measured and the corrected curves concerning the water retention and the hydraulic conductivity/matric potential. The shrinkage-induced volume change can result in up to 12% volume loss which is interpreted as water loss if pore rigidity is assumed. Underestimating the occurrence of shrinkage may result in incorrect data for modelling hydraulic functions. These incorrect hydraulic functions also introduce a high uncertainty in the forecast of changes in hydraulic site properties due to climatic change. Non-rigidity of soils should therefore be more intensely investigated in future research.

Key words:

• Rigidity of pores
• water retention curve
• hydraulic conductivity
• shrinkage
• soil aggregation
• climate change

INTRODUCTION

Soil properties  are the result of different kinds of external and internal physical, chemical and biological stresses, which can be described as static or dynamic, short- or long-term, affecting pedogenetic processes, but altogether they alter soil functions. Anthropogenic effects most obviously alter the soil quality parameters, because the changes in pore continuity, total pore volume and depth depletion of pore connectivity may already cause smaller air-filled porosity, smaller saturated but relatively higher unsaturated hydraulic conductivity and reduced gas diffusivity, less positive redox potential values apart from a retarded or even mostly prevented accessibility of the exchange places at the particles surfaces compared to the potential, i.e., capacity values (Krümmelbein et al. 2013). It is, however, mostly overlooked that not only these well-defined processes are summed up under the term “in situ soil degradation” and these altered properties defined as given for further soil research or modelling approaches, but a further step of soil dynamics must be considered and quantified as soils react according to the previously applied physical, chemical or biological stresses as dynamic systems, which furthermore complicates all modelling approaches.

The Darcian law, for example, describes the one-dimensional water flux assuming one-dimensionality, laminar flow, inert properties and complete pore rigidity. These restrictions may limit the applicability of the Richards equation to coarse textured, e.g., sandy soils (if no turbulent flow occurs), while silty, loamy and clayey textured substrates or organic substrates like peat in fact would not fulfill these boundary conditions irrespective of the actual land use systems (Hartge and Horn 2014). Internal processes like accumulation of organic substances, long-term fertilization or loosening of the topsoil horizons due to biological activities can all strengthen or weaken the interparticle and the intraaggregate bindings, as can be derived amongst others from the findings of Watts and Dexter (1998); Markgraf and Horn (2007); Barré and Hallett (2009); Gregory et al. (2009); Holthusen et al. (2010); Zhang et al. (2013). Soil formation processes like weathering of soil minerals, clay migration and podsolization are also examples of soil strengthening or weakening which in all cases make the presumptions of a given pore rigidity more uncertain. Soil management, like plowing, puddling, or pulling roots out, results in weaker soil systems, while soil compaction increases the internal soil strength and makes the soil mostly more rigid (Horn and Smucker 2005; Horn et al. 2007). Soil degradation as the result of inadequate land use and soil management also alters the internal soil strength, varying from complete rigid (heavily overcompacted to stress beyond all former stress applications) to a completely weak non-rigid system as a result of, for example, tillage under too wet conditions in combination with soil shearing.

Swell shrink processes are the most effective for soil and pore strengthening by inducing aggregate formation and rearrangement of particles. Soil strength is increased the more often and longer that contracting water menisci occur, while only a few but very intense drying processes are mostly not capable of rearranging particles in a more rigid way and fail to gain smaller entropy (Horn 1994). If soils are rewetted during rainfall or access capillary rise, the primarily horizontally anisotropic rearrangement of particles results in an additional formation of rectangular cracks and strengthening of these smaller aggregates (Peng and Horn 2007; Peng et al. 2007). Tuller and Or (2003) proposed a framework for swelling soils which combined even physico-chemical processes with pore scale geometrical, hydrostatic and hydrodynamic considerations in order to derive hydraulic properties of these soils. Based on more than 300 shrinkage curves with varying origins, Peng and Horn (2013) defined six different shrinkage types in soils under various land use systems and climatic conditions. Furthermore, amongst others, Baumgartl (2003) described the parallelism of the pattern of mechanical stress strain curves and hydraulic stress (matric potential) and volume change behavior and subdivided the curves in the recompression and reshrinkage ranges, the precompression or preshrinkage stress followed by the virgin compression and proportional shrinkage ranges. He concluded that, irrespective of the strength origin, the rigidity boundary conditions can be classified.

The three-dimensional (3-D) volume change is well described by the decrease in volume of water (moisture ratio ϑ m³ m⁻³,) compared with the decrease in total soil volume (void ratio e, m³ m⁻³) (Bronswijk 1990). Usually, four characteristic shrinkage zones are distinguished, which indicate different structural rigidities of the soil as it desiccates: structural, proportional (normal, basic), residual and zero shrinkage (Groenevelt and Grant 2001). These zones can also be identified visually from the shrinkage curve and are more precisely defined by recently introduced mathematical procedures generated to distinguish these zones (Groenevelt and Grant 2004; Peng and Horn 2005). Only within the structural shrinkage range is the volume loss negligibly small during drying, while especially in the proportional shrinkage zone the high volume change occurs over a wider matric potential range followed by smaller values with more intense drying.

Thus, for all these zones, excluding the initial, i.e., rigid structural shrinkage range, the soil volume changes and affects not only the ratio of soil mass per volume but may also lead to a modified representative elementary volume (REV) of the soil samples. Hendrickx and Flury (2001) amongst others assumed rigid pore systems when classifying the preferential flow mechanisms at pore and Darcian scales regarding macroscopic heterogeneous and/or homogeneous flow processes. Smiles (2000) stated that soil volume for non-rigid soils generally decreases with water loss as a function of the rate and completeness of drying causing rearrangements of soil particles and aggregates. Hence, new pore systems and soil properties always have to be defined as soon as the previous mechanical or hydraulic strengths are exceeded (e.g. Dörner et al. 2010a). These processes are not restricted to a defined matric potential or mechanical stress but can generally occur at all stages as long as particles can be moved or pore diameters are still wider than the soil particles, and/or the friction between the particles is smaller than the generated mechanical or hydraulic forces.

Garnier et al. (1997a) showed that swelling and shrinkage modifications of soil structure influence the movement of water and solutes in the soil matrix, which consequently results in a more complex definition of transport process phenomena for water and solutes in non-rigid soils compared with those in rigid ones. Dörner and Horn (2009) demonstrated not only the strong effect of predrying intensity at given aggregate types on the gas and water permeability but also the dependency of the volume change directions on the aggregate type. Gräsle (1999) described the non-rigidity of disturbed soil samples and also predicted an overestimation of hydraulic conductivity due to the negligence of soil deformation resulting from changes in hydraulic stress state. However, shrinkage-induced changes in the pore system can also result in the persistence of water-filled pores of various diameters which would result in remaining higher hydraulic conductivity values assuming “rigid” pore systems. Stange and Horn (2005) showed the effect of non-rigidity of soils on hydraulic functions and developed corrected model equations for homogenized substrates. Gebhardt et al. (2012) described the shrinkage effects on volume changes while Dörner et al. (2010b) determined significant changes in the pore system during drying for volcanic ash soils. Seyfarth et al. (2012) finally developed a new 3-D laser technique to quantify the matric potential dependent volume change.

With respect to the ongoing debate about climate change and the need for more reliable model predictions, consequences of future intensified drying on changes of hydraulic properties must be considered to improve predictions on water balances under various management systems. So far, the modification of model input parameters with time due to the processes described above is completely missing. Furthermore, information on the “mechanical and/or hydraulic site histories” is missing although it must be investigated in order to define an additional input parameter which is needed to extrapolate the data for further matric potential boundary conditions.

In the present paper, we therefore want to verify the following hypotheses: (1) current soil hydraulic models are only reliable in the preshrinkage, i.e., structural shrinkage range and depend on the in situ hydraulic history, and (2) exceeding the structural shrinkage range results in a new 3-D volume reduction altering both retention curve and hydraulic conductivity versus matric potential relations.

MATERIALS

Soils

Disturbed and undisturbed soil samples were taken from various soil profiles with different geological origins in Northern Germany and Southern Chile under forest, arable and pasture conditions. Representative soil horizons have been sampled and classified according to IUSS/ISRIC/FAO (2006). Table 1 gives information about the soil types and some physical and chemical soil properties. The investigated German soils are labeled with abbreviations according to the German soil classification system to help to define the origin of the data (Boden 2005).

Table 1 Physical (grain size distribution, structure, and bulk density ρt) and organic matter of all investigated sites/soil horizons. Soil types are defined according to IUSS/ISRIC/FAO (2006) and texture are defined according to the German classification system (Boden 2005). Wacken, Ritzerau, Hildesheim are locations in Germany; Cudico and Pelchuquin are sites in Chile. The sampling depth corresponds to the number attached to the abbreviation; a soil depth of 6 cm was always collected

			Depth			pt	Clay	Silt	Sand	Organic matter
Soil type	Site	Horizon	(cm)	Abbreviation	Soil structure	g cm^−3	< 2µm	2–63µm	63–2000µm	%
Eroded Stagnic Luvisol derived	Ritzerau/Germany	Ap	15–31	SS-LL(15, 25)	coherent-subangularblocky	1.59	7.4	27.4	65.2	1.7
from periglacial cover beds		App*	35–41	SS-LL(35)	platy	1.73	4.3	24.5	71.2	0.9
above Weichselian glacial till		2Btg	50–56	SS-LL(50)	subangular blocky	1.68	20.4	38.6	41	0.2
(loam and sand)		3Bw	75–81	SS-LL(75)	blocky-prismatic	1.77	8.1	28.4	63.5	0.05
Haplic Gleysol from fluvioglacial Weichselian deposits	Ritzerau/Germany	Cl	40–46	GGn(40)	coherent-blocky	1.46	34.9	57.7	7.4	0.07
Haplic Stagnosol derived from periglacial cover beds above Saalian glacial till	Wacken/Germany	2Cg	50–56	SSn(50)	coherent-(blocky)	1.21	64.9	15	20.1	0.15
Haplic Chernozem derived from loess	Hildesheim/Germany	Ah	40–46	TTn(40)	subangular blocky	1.45	19.2	78.7	2.1	1.9
Rheic Histosol derived from peat above Saalian glacifluvial sands	Wacken/Germany	Hl	30–36	HNn(30)		0.81				>20
Drained Histic Gleysol derived of peatclay above Saalian glacifluvial sands	Wacken/Germany	Cl	40–46	HN-GH(40)	coherent	1.08	39.9	34.1	26	25
Humic Ultisol derived from	Cudico/Chile	Ap	5–11	US	subangular blocky	1.05	42.7	37.2	20.1	18
old volcanic ashes		Ap		UnS-1.1	coherent	1.10	42.7	37.2	20.1	18
		Ap		UnS-0.8	coherent	0.80	42.7	37.2	20.1	18
Haplic Andosol derived from volcanic ashes	Pelchuquin/Chile	Ap	5–11	A5	granular	0.69	35.3	59.5	5.2	14
		Bw	40–46	A40	subangular blocky	0.53	19.3	75.5	5.2	4

*plough pan

Stagnic Luvisol

Luvisols derived from glacial till dominate the Weichselian deposit region in Schleswig-Holstein in northern Germany. These soils are highly fertile and usually support high yields (on average 9–10 Mg wheat (triticum aestivum) ha^–1 yr) under intensive agriculture. The clay content of the investigated eroded Stagnic Luvisol (abbreviated as SS-LL according to the German Soil Mapping system, horizon sequence: Ap–2Btg–3Bw) derived from periglaical cover beds above glacial till (glacial loam and sand) varies between the soil horizons (due to accumulation of clay by illuviation as well as lithological discontinuities) but never exceeds 21%. The organic matter content in the top 30 cm shows 1.7%, while in deeper soil layers it falls to less than 0.2%. The samples were taken in 15 and 25 cm (Ap horizon), 35 cm (plough pan), 50 cm (Btg horizon) and 75 cm (Bw horizon) depths.

Haplic Gleysol

The investigated Haplic Gleysol (abbreviated as GGn according to the German Soil Mapping system, horizon sequence: Ap–Cl –Cr) represents the basin clay areas (tongue-like basins) in the Weichselian deposit region in Schleswig-Holstein in Northern Germany. This soil is characterized by high clay contents with illite as the dominating clay mineral, and a high water table which also results in a coherent soil structure. Because these soil horizons have never been dried out more intensely, such coherent structure is very weak. We sampled the Cl horizon in 40 cm depth, characterized by the oxidized iron (Fe)-cutanes at the pore walls. The organic matter content is negligible small (< 0.1%).

Haplic Stagnosol

The investigated Haplic Stagnosol (abbreviated as SSn according to the German Soil Mapping system, horizon sequence: Ah–Cg–2Cg) are derived from periglaical cover beds above Saalian clay-rich glacial till and characterize the central part of Schleswig Holstein but can be also found in Lower Saxony in plains in Northern Germany. They are mostly used as pasture land but also as arable sites (drained). Because of the very high clay content of 65% and coherent structure the subsoil has a small hydraulic conductivity. The samples were collected at a depth of 50 cm (2Cg horizon).

Haplic Chernozem

The investigated Haplic Chernozem (abbreviated as TTn according to the German Soil Mapping system, horizon sequence: Ap–Ah–C) is derived from loess and situated in the Hildesheimer Boerde (loess basin). The soil has very high water and nutrient storage capacities, and pronounced biological activity in the topsoil. The arable site was artificially drained. Samples were taken from the deeper Ah soil horizon (30–50 cm) which shows a very rigid subangular blocky structure. The clay content is about 20%.

Rheic Histosol

The investigated Rheic Histosol (abbreviated as HNn according to the German Soil Mapping system, horizon sequence: Hl–Hr–2Cr) is developed from peat above Saaleian sand deposits under very moist conditions, i.e., a high groundwater table. They are typical for approx. 6% of the area in Schleswig Holstein and are mostly used as pasture land. We sampled the continuously water-saturated Hl horizon which can be also defined by a high organic fiber decomposition rate. The samples were collected at 30 cm depth.

Histic Gleysol

Histic Gleysols (abbreviated as HN-GH, horizon sequence: Hl–2Cl–Cr) derived from peat above fluviolimnic deposits (claypeat) and are used as pasture land after drainage. We sampled the oxidized Cl horizon which was developed by the former clay sedimentation in water. Because of the capillary fringe effects with respect to continuous water saturation, the aggregation is less pronounced and the bulk density is low. The clay content is about 40% and the organic matter content is about 25%. The samples were collected at 40 cm depth.

The following two soil types were sampled in Chile, in order to also include shrinkage phenomena with very different parent material (volcanic ash material) under both natural but also completely homogenized conditions.

Humic Ultisol

The Humic Ultisol (abbreviated as US, or for the disturbed profile as UnS) was formed from old-volcanic ashes in Chile, collected in the vicinity of Valdivia/Chile and presents deep profiles. The soil textural classes are clay loam to clay between 0–35 cm depth, and clay in deeper horizons (Bt horizon > 35 cm depth). The soil structure is relatively well developed in the profile, presenting fine subangular blocky aggregates (0–35 cm) and a prismatic structure in the Bt horizon (> 35 cm). The organic matter content is 18%. We sampled the Ap horizon. In order to describe the effect of soil structure in those deeply weathered volcanic ash material, both undisturbed (US) and disturbed (repacked) soil samples (UnS-1.1, UnS-0.8) were used. More details about the site are provided in Dörner et al. (2010a).

Haplic Andosol

The Haplic Andosol presents a deep profile (> 160 cm) with a well-developed A horizon (0–27 cm depth, granular structure defined as A5), a Bw horizon with a subangular blocky structure (27–39 cm depth and a less-developed Bw horizon (defined as A40) (39–200 cm, subangular blocky to massive structure). We selected the A and B horizon because of the differences in amount of organic matter (A5: 14%, A40: 4%) and allophane (A5: 9%, A40: 13%). More details about the site are provided in Dörner et al. (2010b).

METHODS

From each soil horizon, in total six undisturbed soil core samples (V = 471 cm³, diameter = 10 cm, height = 6 cm) were taken to determine the water retention curves, the shrinkage curves and the resultant changes in the hydraulic conductivity to matric potential function (for detailed information see Hartge and Horn 2009).

Additionally, disturbed samples were taken to determine general soil properties as well as to pack soil samples from the Humic Ultisol. The refilling of the homogenized material to the identical bulk density occurred with a slight compression, but as the bulk density was relatively low, the needed energy was negligibly small (Dörner et al. 2010a). The particle density was measured by the pycnometer method (Blake and Hartge 1986); organic matter was determined by dry combustion at 1200°C and coulometric carbon dioxide (CO₂) measurement (Blume et al. 2011). Soil texture was measured by the combined sieve (> 63 µm) and pipette method (Hartge and Horn 2009).

Both the water retention and the shrinkage characteristics of the undisturbed soil core samples were determined after complete saturation by capillary rise and drained to fixed matric potential values. At the end of complete saturation no further surface trimming was carried out and the wet samples were directly placed on the ceramic plates. The drainage steps used were –3, –6, –15, –30 and –50 kPa (ceramic suction plates) and 300, 600 and 1500 kPa (pressure chambers and 1000 MPa (equivalent to pF7 in order to get the corresponding water content i.e. oven drying at 105°C for 16 h). After each drainage step, the weight of every sample as well as the volume decrease due to shrinkage was measured. Volume change due to shrinkage can occur vertically or horizontally, isotropic or anisotropic. The vertical height change was determined by a digital measuring caliper with an accuracy of 0.05 mm at eight fixed points at the sample surface. The total height change was then calculated from the arithmetic mean of these eight values. The change of sample height multiplied by the surface area gives the vertical volume change.

The horizontal volume change can be either quantified by direct measurements or determined from the documentation of anisotropic volume change. As the measuring procedure is rather time consuming and requires picture analyses taken after each drying step in order to obtain insight into the maximum volume change and degree of isotropy of the shrinkage, we mainly stuck to the assumed isotropic volume change behavior. This latter assumption is in agreement with those tests carried out by numerous researchers like Bronswijk (1991a, 1991b); Garnier et al. (1997a, 1997b); Braudeau et al. (1999, 2004) and Boivin (2007). With these data, the corresponding water contents, actual sample volumes and bulk density values were calculated and the volume-dependent change in water content equivalents according to the new total volume were derived. In a few samples the horizontal shrinkage was directly derived from corresponding picture analyses.

The water retention curve was calculated and the effect of shrinkage on changes in the soil volume corrected via the bulk density increase. The dry bulk density was determined as the ratio of dry soil weight (after drying at 105°C for 16 h) to the actual (post-shrinkage) sample volume.

Saturated hydraulic conductivity was measured under non-steady state flow conditions with a falling head permeameter (Hartge and Horn 2009). The hydraulic conductivity vs. matric potential ratio was determined for undisturbed samples in a transient flow experiment with an evaporation method where two microtensiometers and two time domain reflectometry (TDR) probes were horizontally inserted into the sample from the side through holes in the cylinder at a vertical distance of 3 cm. The sensors continuously recorded the matric potential and water content changes while the initially saturated sample progressively dried out from the sample surface. The hydraulic conductivity coefficient was calculated from average changes in mean matric potential and water content in small time intervals according to Becher (1970). For more detailed information see also Peth (2004) or Hartge and Horn (2009).

The measured hydraulic conductivity/matric potential (k/Ψm) relationship was compared with the calculated curves based on the retention function (van Genuchten 1980) assuming: (1) no shrinkage, i.e., a rigid pore system, (2) only vertical shrinkage, (3) isotropic shrinkage and (4) anisotropic shrinkage.

RESULTS

The shrinkage curves of the Stagnic Luvisol samples are shown in Fig. 1a. The sand-enriched loamy horizons SS-LL (15–75) show the smallest shrinkage at all depths compared to the other samples (not shown in total) but the four shrinkage ranges can still be differentiated. Samples taken from the ploughed topsoil layer of SS-LL (15, 25) are more sensitive to volume changes while, with increasing bulk density and/or stronger aggregation in the subsoil, shrinkage decreases. With increasing clay content (> 30%) very pronounced shrinkage curve ranges can be detected (Fig. 1b). It becomes obvious that the moisture ratio range for the structural shrinkage differs with soil aggregation which depends on the previous drying history (see the general description of the sites). The Ah horizon of the Chernozem with subangular blocky structure is characterized by a large structural shrinkage range, i.e., it shows a very rigid pore system followed by a small proportional shrinkage and residual shrinkage range, while the Gleysol samples and especially those of the Stagnosol show a very pronounced proportional but only a small structural shrinkage behavior. Both curves show similar residual shrinkage patterns. Volcanic ash soils like the Haplic Andosol and the Humic Ultisol (where the mineral allophane is present and dominates in the first one) under long-term pasture have a wide structural shrinkage range although they have a very low initial bulk density, while the homogenized samples at a given bulk density of 1.1.or 0.8 g cm^–3 show a pronounced proportional shrinkage with drying (Fig. 1c). The shrinkage curve of the repacked soil at 1.1 g cm^–3 is mostly identical to the undisturbed but less dense soil volume. Although the void ratio is the highest in the Haplic Andosol at 40 cm depth, it is the most rigid over a moisture ratio range from approximately 2.8 to 1.4; identical texture classes do not result in the same shrinkage curve pattern (compare Ultisol A40 with Chernozem Ah). Although the Ultisol (A40) had a much smaller bulk density value than the Ah, the remaining void ratio at complete drying was still much higher. The Histosol and the Histic Gleysol (peatclay) samples show a slightly different shrinkage pattern, as the residual and the zero shrinkage ranges are missing, which can be caused by the organic carbon (Corg.) loss due to oxidation with enhanced drying and oxygen accessibility at the organic sufaces (Fig. 1d).

Figure 1 (a) Shrinkage curves of the various soil horizons of the Stagnic Luvisol (SS-LL) derived from glacial till in Northern Germany. The abbreviation is described in detail in Boden (2005). (b) Shrinkage curves of selected soil horizons of the Gleysol (GGn), Stagnosol (SSn) and Chernozem (TTn). The samples come from Northern Germany. (c) Shrinkage curves of selected soil horizons of the Haplic Andosol (A) and a Humic Ultisol (US), both under long-term pasture. Furthermore, the homogenized samples of the Humic Ultisol at a given bulk density of 1.1 or 0.8 g cm^–3 (UnS-1.1 or UnS 0.8) are shown compared to the initial structured sample (US) at the same depth of 5 cm. All samples were collected in Chile. (d) Shrinkage curves of selected soil horizons of the Rheic Histosol (HNn) and the Histic Gleysol (HN-GH), collected in northern Germany. Numbers following abbreviations define the soil depth (cm).

The changes in the void ratio vs. matric potential relation (i.e., the shrinkage curve) depict obvious effects of initial bulk density and texture as well as of soil type dependent structure (Fig. 2). Soils with lower initial void ratio values (or higher values for the bulk density at defined values of the particle density) are more rigid than those with higher values for the void ratio. The Stagnosol (SSn) undergoes a more pronounced change in the void ratio during drying at matric potential values smaller than –20 kPa. The samples with clay contents < 25% [SS-LL (75) and TTn (40)] are mostly rigid while with increasing clay content the matric potential induced shrinkage increases [SSn(50)]. If the clay content exceeds 35%, e.g., in GGn (40) and in SSn (50), an intense void ratio decline already begins at approximately –10 kPa matric potential. Shrinkage is higher the less pronounced the structure formation, as can be derived from the Ah horizon of the Chernozem with a subangular blocky structure compared to the SS-LL samples. Even at initially greater values of the void ratio, its volume reduction with drying intensity is smallest. Predrying intensity at given soil texture results in altered soil strength, which can be derived from the shrinkage curves of the Gleysol (GGn) and of the Stagnosol. At comparable bulk density values but increasing clay content [TTn(40) and GGn (40)] the shrinkage is more pronounced the smaller the aggregate strength. At more negative matric potential values, we see the nearly parallel pattern of the proportional and residual shrinkage curves of the Gleysol and Stagnosol, with the highest volume loss. It is also evident that the exceeding of the internal soil strength due to more negative matric potential results in nearly identical, now initial (i.e., proportional) shrinkage behavior. The complete shrinkage curve, however, can be subdivided into distinct structural and proportional shrinkage ranges. The Gleysol (GGn), for example, shows a structural shrinkage range up to – 10kPa matric potential followed by a proportional and a second structural shrinkage range which ends in a new proportional shrinkage. The Chernozem is mostly rigid and shows only two pronounced structural ranges and one narrow proportional shrinkage range. Such separation in distinct shrinkage ranges can be caused by a complete volume decrease of the bulk soil in time and a contemporary formation of coarser macropores, while the single aggregates themselves also shrink and cause such stepwise decline.

Figure 2 Relationship between void ratio and matric potential for the Stagnic Luvisol (SS-LL) Haplic Gleysol (GGn), Haplic Chernozem (TTn) and Haplic Stagnosol (SSn) collected at various sites in Germany for defined depths with varying soil structure.

The difference between the water content, assuming rigidity, and the shrunken soil volume depends on the soil strength and the hydraulic history of the analyzed soils (Fig. 3a and b). We found for the soils with higher clay content and/or less pronounced soil aggregation an increased risk for the “misinterpretation of retention data” as soon as the soil samples dried out to more than –20kPa. The Ah horizon of the Chernozem with a subangular blocky structure showed a strong rigidity with negligibly small errors in water content calculations, assuming rigidity, while the actual shrinkage-induced changes in the available pore space approached even 12% at pF 4.2 for samples with smaller aggregate strength (e.g., cohesive structure) and increasing organic matter content. The higher the bulk density and the more pronounced the soil aggregation, the smaller is the difference between the predicted to actual water content (smaller than 1%; not shown for the SS samples). The high values for the peat samples with more than 10% shrinkage at –1500kPa matric potential underline the effect of soil structure, texture and bulk density, as well as of the geological parent material, on the non-rigidity of soils.

Figure 3 (a) Difference between the predicted and the actual water content due to shrinkage as a function of matric potential. (b) Difference between the predicted and the actual water content due to shrinkage as a function of matric potential for a Haplic Andisol (A) and Humic Ultisol (US), both under long-term pasture. The number following the abbreviation defines the soil depth (cm). Furthermore, the homogenized samples of the Humic Ultisol at a given bulk density of 1.1 or 0.8 g cm^–3 (UnS-1.1 or UnS 0.8) are shown compared to the initial structured sample (US) at the same depth of 5 cm.

With increasing bulk density the differences are reduced, but at the same bulk density of approximately 1.45g cm^–3 the shrinkage-induced volume loss (calculated as water loss) is greater in the GGn, with up to 4% due to a weaker soil structure (coherent to blocky caused by a less intense predrying in combination with a higher clay content) as compared with the TTn samples with subangular blocky structure. The same trend can also be seen for the Haplic Andosol and the Humic Ultisol where the water loss due to shrinkage is higher the smaller the proportional shrinkage (Fig. 3b). For the homogenized and at nearly identical bulk density refilled samples, shrinkage-induced volume loss is at a very high level and is higher the smaller the initial bulk density.

In order to substantiate the assumption that soils generally shrink during the determination of the retention curve, we document one site out of all measured samples to show this enormous effect. The Stagnosol with coherent structure (Fig. 4) shows intense changes depending on the kind of shrinkage included in the calculations. The retention curve without considering volume loss has the highest air capacity (pores > 50 µm) and plant-available water volume. Based on the measured height changes and assuming that the shrinkage occurs only vertically, both air capacity and the amount of plant-available water decrease slightly. However, compared with the “rigid” system, the actual 3-D shrinkage causes a further decline in air capacity and plant-available water as shown by the anisotropic curve. The assumption of an isotropic shrinkage for these samples finally results in no coarse pores and the lowest plant-available water content compared with all other options. Table 2 shows the corresponding van Genuchten parameter data sets derived from fitting the retention functions.

Figure 4 Effect of soil shrinkage on the pattern of the water retention curve of the Stagnosol. The differences in the water contents at a given matric potential hPa (i.e., pressure head cm) depending on the shrinkage intensity and direction are obvious. “Vertical” means that only the vertical soil shrinkage due to drying was considered, while isotropy is based on the vertically taken measurements and transferring them to three-dimensional isotropy. Anisotropy describes the actual shrinkage effects in all directions.

Table 2 van Genuchten parameters of the Stagnosol (SSn50) under various rigidity boundary conditions

Parameter	Without	Vertical	Isotropic	Ansiotropic 3-D
theta_r [cm^3 cm^−3]	0.22	0.29	0.41	0.41
theta_s [cm^3 cm^−3]	0.51	0.51	0.51	0.51
alpha [1 cm^–1]	0.018	0.011	0.002	0.015
n [-]	1.102	1.124	1.233	1.216
m [-]	0.092	0.110	0.189	0.178
l [-]	0.5	0.5	0.5	0.5
Ks [cm s^–1]	5.15E-06	5.15E-06	5.15E-06	5.15E-06

Ks, hydraulic conductivity (cm/s); 3-D, three-dimensional.

If for the same Stagnosol, the van Genuchten parameters in addition to the hydraulic conductivity values are applied to predict the hydraulic conductivity/matric potential ratio, the following trends can be derived: (1) for the SSn, the curve patterns differ depending on the matric potential dependent volume loss, and (2) the comparison between the measured k/pressure head ratio and the theoretical curve patterns based on the van Genuchten parameters results in a best fit assuming isotropy (Fig. 5). This approach underlines the effect of non- rigidity but, as it still includes the drying-induced changes of the referred soil volume, it is still not correct but only an indirect way to show the influence of type and intensity of shrinkage on conductivity curves.

Figure 5 Hydraulic conductivity/matric potential (K/Ψm) relation of the Stagnosol derived from clay-rich till. The differences in the curve patterns depend on the measured shrinkage intensity and dimensions. “Vertical” means that only the vertical soil shrinkage due to drying was considered, while isotropy is based on the vertically taken measurements and transferring them to three dimensions. Anisotropy describes the actual shrinkage effects in all directions. The circles define actual measurements which coincide with the predicted K/Ψm curve assuming isotropy.

DISCUSSION

The determination of hydraulic properties, such as the water retention curve and the hydraulic conductivity vs. matric potential relations, are essential for many research questions dealing with mass balances under various management systems or climatic conditions. They will become even more essential for reliable water balance predictions as a key parameter for climatic change scenarios.

Wösten et al. (1999) developed the so-called hydraulic properties of European soils (HYPRES) database which allows determination of hydraulic properties based on texture data. Vereecken et al. (1992) stated that pedotransfer functions (PTFs) are not evaluated with respect to specific applications. The criteria presently used to assess the goodness of fit for a PTF do not provide information about the performance of these functions for applications such as the prediction of the downward water flux below the soil root zone or of the soil moisture deficit during the growing season. They conclude that > 90% of the variation in the simulated moisture supply capacity was caused by estimation errors in the hydraulic properties. Hartge et al. (1986) dealt with the derivation of the water retention curve from texture analyses and concluded that within a range of 3–5%, reliable data can be obtained. The German Soil Survey Guidelines (Boden 2005) differentiate between the air capacity and plant-available water for the various texture classes as a function of bulk density. Because the differences from actual values were even too high, Riek et al. (1995) classified these values as a function of the parent material and geological origin. Thus, if we relate our findings to those of the German Soil Survey Guidelines, we have to conclude that a differentiation based on bulk density is not an appropriate classification system.

The determination of the retention curve and the saturated hydraulic conductivity is commonly a routine procedure and is conducted in most laboratories worldwide. The determination of the hydraulic conductivity vs. matric potential curve is more time-consuming and expensive and will be therefore often either completely ignored or only estimated from retention data using the van Genuchten approach (1980). Peng et al. (2009, 2011) showed that the pore rigidity depends on the predrying intensity and that the biological exudates modify the rigidity of interparticle bondings and induce soil shrinkage. They also pointed out that exceeding this value, which is synonymous with the mechanical precompression stress value (Baumgartl 2003), always exerts stresses causing volume change and a non-rigid pore system response, which makes application of any hydraulic model more difficult and mostly unreliable (Horn 1994). The obtained results of this study also are a proof that a “memory effect” of the hydraulic history occurs. The site properties, e.g., Gleysol as a very wet and therefore never strengthened soil profile or the peat soils, all were mostly non-rigid and underwent a severe shrinkage, while the Chernozem as a soil type exposed to drier climatic conditions showed hardly any intensive shrinkage-induced volume loss.

Janssen et al. (2006) concluded that the pore rigidity depends on the predrying history; they also described a memory effect even for paddy soils after several decades of puddling. Irrespective of this partial (= matric potential-dependent) soil strength, they determined a severe volume loss during desiccation on ceramic plates. Lennartz et al. (2009) validated such findings also under in situ conditions and concluded that the original soil strength is only a quasi-dynamic equilibrium but it causes the formation of new cracks as soon as further drying, desiccation due to water uptake by plant roots, etc. takes place. The latter coincides with a further volume loss. The general approach of our research and the obtained results are also in agreement with the findings of Gregory et al. (2010), who estimated relative hydraulic conductivity values from the water release characteristics of a shrinking clay soil. They stated that if a mostly constant decline in porosity due to drying occurs, the constancy of the hydraulic conductivity with drying is realistic. This differentiation between the shrinkage patterns of the various soil samples with different hydraulic history is reconfirmed by the aggregate types and their varying mechanical or hydraulic strength (Horn and Smucker 2005). This requires detailed measurements of shrinkage behavior during the determination of retention functions. Surprisingly, such data are rarely measured although this could be easily done and thus would help to reduce uncertainty in water balance modeling. They are also urgently required in order to improve the reliability of the climate change scenarios, which is currently limited because of the described conceptual shortcomings. Even if we only consider actual scenarios, the question must be answered to what extent such incorrect prediction of the hydraulic properties will occur and how far soil management and climatic conditions alter these general findings, too.

Baumgartl et al. (1998) analyzed the maximum drying intensity in the subsoil (> 60 cm) under arable soil and climatic conditions in Germany at the end of the growing season, for more than 300 experimental years. They found that the matric potential value ranged mostly between –20 and –40 kPa and was slightly more negative in silty compared to sandy and clayey soils and in areas with less precipitation. Under forest or pasture, drying intensity was in most cases less negative irrespective of the texture. If we furthermore add data about the mechanical precompression stress, which defines the maximum former mechanical or hydraulic stress applied, we found for some sites (where both tensiometer readings and mechanical strength measurements were carried out) a good agreement between the hydraulic properties and the mechanical strength. Thus, the assumption of a completely rigid pore system does not hold true. Based on our findings, it can be stated that rigid soils mostly do not exist but the rigidity depends on the hydraulic, mechanical and physicochemical site history. A link of measured retention curve data with hydraulic soil functions without including the shrinkage-induced alterations in the pore system always results in unreliable α, n and m values as well as in predicted false unsaturated hydraulic conductivities values. Thus, the easy use of hydraulic properties (documented in databases) should not sidetrack modelers to an interpretation of the data as constant values neglecting the temporal dynamics of conductivity functions depending on site, land use and climatic boundary conditions, as this inevitably results in misunderstanding and computational error of model outputs (e.g., the water balance).

Finally, we state that an intense further evaluation of the interrelations between soil water related forces, soil deformation and hydraulic functions is needed in order to develop more reliable databases including reduction factors for hydraulic conductivity as a function of the predrying intensity/or structural shrinkage moisture ratio range. Recent advances in non-invasive methods and modeling techniques (Richards and Peth 2009; Peth et al. 2010) could significantly contribute towards this goal by providing a more detailed understanding of the coupled mechanical-hydraulic relationships in soils.

CONCLUSIONS

The assumption of pore rigidity and the corresponding retention curve are only valid for the structural shrinkage range but fail to predict the actual in situ conditions as soon as the proportional shrinkage range starts. This statement holds true on a wide range of soil types and land use systems.

A more intense consideration of the aggregation and the predrying processes is needed to obtain a more reliable prediction of in situ hydraulic properties.

Organic fiber in peaty material can result in an altered rigidity due to strengthening effects, which also may result in smaller differences for the hydraulic conductivity matric potential ratio. More intense research must be started on this.

Extrapolating hydraulic parameters derived from current databases to predict water fluxes based on conditions described in climate change scenarios will most probably result in incorrect water balance forecasts.