computers and electronics in agriculturetal models. this enhancement is possible through a new...

Computers and Electronics in Agriculture xxx (2012) xxx–xxx

Contents lists available at SciVerse ScienceDirect

Computers and Electronics in Agriculture

journal homepage: www.elsevier .com/locate /compag

Toward delineating hydro-functional soil mapping units usingthe pedostructure concept: A case study

Mohammed Salahat a, Rabi H. Mohtar b,c,⇑, Erik Braudeau b,d, Darrell G. Schulze e, Amjad Assi b,c

a Natural Resources and Environment Department, The Hashemite University, Zarqa, Jordanb Qatar Environment and Energy Research Institute, Qatar Foundation, Doha, Qatarc Agricultural Biological Engineering Department, Purdue University, West Lafayette, IN 47906, USAd Institute for Research Development, IRD, Bondy, Francee Agronomy Department, Purdue University, West Lafayette, IN 47907, USA

a r t i c l e i n f o

Article history:Received 8 February 2011Received in revised form 23 April 2012Accepted 24 April 2012Available online xxxx

Keywords:Hydro-functional soil mapping unitsShrinkage curveWater potentialSoil mappingPedostructure parametersHydrostructural characterization

0168-1699/$ - see front matter � 2012 Elsevier B.V. Ahttp://dx.doi.org/10.1016/j.compag.2012.04.011

⇑ Corresponding author at: Qatar Environment anQatar Foundation, Doha, Qatar. Tel.: +974 44540517;

E-mail address: [email protected] (R.H. Mohtar).

Please cite this article in press as: Salahat, M., etComput. Electron. Agric. (2012), http://dx.doi.or

a b s t r a c t

Current soil maps contain qualitative attributes which are challenging to integrate into agronomic mod-els or decision support systems. In this paper, a new approach is proposed that will enable the use of soilmaps as a suitable information system of soils hydrostructural properties for agronomic and environmen-tal models. This enhancement is possible through a new methodology for characterizing the soil units onthe basis of the multi-scale soil–water functionality of the pedon which is a representative soil mappingunit delineated by pedologists according to the geomorphology. The hydro-functional mapping unitsshould represent soil mapping units that are theoretically homogenous from the perspective of the pedo-structure parameters of the pedon. For the purpose of this study, the Haggerty–Cox property, West Lafay-ette, Indiana, USA was selected. The soil data ‘‘soil mapping units and physical characteristics’’ of thestudy area were obtained from SSURGO. Due to time and cost constraints, the research was restrictedto three SSURGO soil units (SwA, Starks–Fincastle complex with 0–2% slope; MsC2, Miami silt loam,6–12% slopes; and HoA, Hononegah fine sandy loam with 0–2% slopes) to check their spatial homogeneityaccording to the pedostructure parameters (PS).

The discriminant statistical analysis for the area of study shows that the PS parameters are unique foreach soil mapping unit and can be effectively used to characterize these units and validate their delinea-tion. The resulting hydro-functional soil mapping units match SSURGO units but have additional physicalattributes. In general, subsurface horizons (B) are more distinguishable than surface horizons (Ap), andusing pedostructure parameters from both potential and shrinkage curves gives the highest distinguish-ing value. Results confirm that the pedostructure parameters are discriminant characteristics of thehydrostructural functioning of soils as delineated by pedologists. Having the hydro-functional soilmapping units will assist in the decision support system data analysis, which in turn, will allow forthe prediction of water and chemical transport and interactions.

� 2012 Elsevier B.V. All rights reserved.

1. Introduction

Current soil maps are the result of extensive, comprehensivesoil survey activities conducted by pedologists over many years.Soil maps consist of a collection of soil mapping units that are usedto delineate areas with similar soil properties. The mapping unitsare generated following well-established rules and approaches.However, these mapping units leave soil functional propertiesunattended (Bouma et al., 1999), which limits their use in agro-nomic models or decision support systems. Many researchers haveput great effort into enhancing and increasing standardization of

ll rights reserved.

d Energy Research Institute,fax: +974 44541528.

al. Toward delineating hydro-fug/10.1016/j.compag.2012.04.01

soil mapping units (Moore et al., 1993; Odeh et al., 1994; Voltzet al., 1997; Lagacherie and Voltz, 2000). For example, Mooreet al. (1993) enhanced soil maps and databases using terrain attri-butes (computed from high-resolution DEMs) to spatially distrib-ute estimated soil attribute data and were able to correlate aslope and wetness index with soil attributes measured at 231locations.

1.1. Hydropedology

Synthesis and quantification of disciplinary knowledge at thewhole system level, using process models of the agricultural sys-tem, are critical to achieving improved and dynamic managementand production systems that address environmental concerns andglobal issues of the 21st century (Ahuja et al., 2006). This fact is

nctional soil mapping units using the pedostructure concept: A case study.1

http://dx.doi.org/10.1016/j.compag.2012.04.011

mailto:[email protected]


http://www.sciencedirect.com/science/journal/01681699

http://www.elsevier.com/locate/compag


2 M. Salahat et al. / Computers and Electronics in Agriculture xxx (2012) xxx–xxx

also evident in investigating soils and soil–water interactions. Rel-evant work linking pedology, soil physics, and hydrology was re-viewed by Lin (2003) and given the name Hydropedology. Hepointed out that ‘‘when addressing the diverse soil and water is-sues at various spatial and temporal scales, it becomes clear thatbridging traditional pedology with hydrology and other relateddisciplines is necessary as well as synergic.’’ Hydropedology worksas a bridge to integrate the pedon and landscape paradigms withphenomena occurring at microscopic (e.g., pores and aggregates),mesoscopic (e.g., pedons and catenas), and macroscopic (e.g.,watersheds, regional, and global) scales.

Landscape water flux was proposed by Lin et al. (2006) as a uni-fying concept for hydropedology through which pedologic andhydrologic expertise can be better integrated. Source, storage, flux,pathway, residence time, availability and spatio-temporal distribu-tion of water in the root and deep vadose zones within the land-scape were included in the term ‘‘landscape water flux.’’ Severalknowledge gaps that hydropedology can address were discussedby the authors, and strategies were proposed to accomplish theintegration of pedology and hydrology.

1.2. Hydro-structural pedology

Recent progress and development in the understanding of thesoil–water medium and its hydrostructural characterization,behavior, configuration, and interactions, together with theadvancement in high-speed computing technologies and modeling,allow for a new approach to mapping soils that takes into accounttheir internal organization and functionality with respect to waterflow. During a project to establish a Spatial Information System forirrigated soils in Tunisia, Braudeau et al. (2002) applied a systemsapproach (SA) to mechanistically characterize and model the soilby referring to an old pedological map made in 1963, and introduc-ing the new concept of pedostructure and hydrostructural character-ization. In this concept, hierarchies within the soil organizationhave been considered and new descriptive variables of the soilstructure properties were defined and introduced into the equa-tions describing the physical soil characteristics. They showed thatby working within a SA framework, the functional levels of theinternal and external organization of the studied system, i.e., thesoil and its environment, can be investigated, defined and charac-terized. In this framework, characteristic variables, functions, andparameters for each scale, as well as the transfer function acrossscales, are acknowledged. The pedostructure concept is based onthis new ‘systems view’, and offers a physical description and char-acterization of the soil’s internal organization (Braudeau et al.,2004). The pedostructure is a representative volume of the soil-medium structure in the soil horizons. Braudeau et al. (2005)showed that the pedostructure characteristics can be used to iden-tify soil type according to their hydro-functional properties whichare compatible with the pedo-genetic classification.

The pedostructure concept permits the physical definition of thevariables and parameters describing the soil structure (arrange-ment of soil particles) and the soil organization (water and airrepartition relative to the solid surfaces of the structure), than theformulation of the physical equations describing the hydro-struc-tural soil properties, such as the shrinkage curve (Braudeau et al.,2004) and the water potential curve (Braudeau and Mohtar, 2004).

1.3. Hydro-structural pedology vs. hydropedology

There is an essential knowledge gap that hydropedology cannotaddress: the link between pedology (the science of soil organiza-tion) and soil water physics within the internal structure of the or-ganized soil medium (thermodynamic interactions between theinternal soil structure, the soil water, and the soil air). Braudeau

Please cite this article in press as: Salahat, M., et al. Toward delineating hydro-fuComput. Electron. Agric. (2012), http://dx.doi.org/10.1016/j.compag.2012.04.01

and Mohtar (2009) showed that a change of paradigm in soil sci-ence must occur to be able to take into account the internalhydro-structural characteristics of the soil medium in equationsdescribing soil water properties. These equations are currentlybased on the macroscopic behavior parameters of a black box(the REV, Representative Elementary Volume), which has no possi-bility of taking into account the internal hydrostructural character-istics of this black box, namely the pedostructure parameters.Using the SREV concept (Structural REV) defined by Braudeauand Mohtar (2009) instead of the REV allows modeling the hydro-structural functioning of the pedon at all its functional scales andhas led to the development of the multi-scale soil–water model Ka-mel� (Braudeau, 2006; Singh et al., 2012).

This research paper builds on this new soil water paradigm andtries to update the soil mapping units through characterizing theseunits using the hydrostructural properties of the pedostructures ofthe soil column (mainly Ap and Bt horizons). Using the pedostruc-ture hierarchy concept, the physical quantitative attributes (char-acteristic parameters of the hydrostructural soil–water behavior)will be added to the existing soil map. This addition to the soilmapping units will assist in the decision support system data anal-ysis, which, in turn, will allow for the prediction of water andchemical transport and interactions by using the soil–water modelKamel� which works specifically with these new attributes.

2. Materials and methods

Three data layers (study area map, landform map, and SSURGOmap) were used in order to characterize the hydro-functional soilmapping units (i.e., the primary ‘‘homogeneous’’ soil map unitsnested into geomorphologic map units), following the hierarchicalapproach used in Braudeau et al. (2002). Fig. 1 summarizes thisprocedure. The first step was to define the area of study withinthe watershed under investigation. The second step was to gener-ate the landform map obtained from a two foot contour map. TheTerrain Analysis System� (TAS) developed by Linday (2003) wasused for this task. The landform map shows topographic units thatdelineate the different landforms: foot slope, shoulder, back slope,or level. The third step was obtaining the SSURGO soil map andoverlaying it over the generated landform map to generate newsoil mapping units. The fourth and final step was to check for thehomogeneity of the generated soil mapping units by collecting rep-resentative soil samples then obtaining their pedostructure param-eters. Discriminate analysis was used to validate the results.

The details of generating the hydro-functional soil mappingunits are described in the section below.

2.1. Defining the areas of study – the first data layer

The study area shown in Fig. 2 was selected within the JordanCreek watershed (ID number 05120108030010) which is a smallsub-watershed of the Wabash River watershed. This sub-wa-tershed lies near West Lafayette, Indiana, close to Purdue Univer-sity. This site was selected because it includes significanttopographic relief, which addresses the need to have an area witha wide range of landforms.

2.2. Generating the landform Map – the second data layer

The second step was to generate the second data layer, which isthe landform map. A sixty centimeter contour map was obtainedfrom the Tippecanoe County GIS office (2006) and used to derive afive-foot landform map. The contour map was used to generate athree meters Digital Elevation Model (DEM) in ArcGIS�, which wasused as an input for TAS� to generate the landform map. The



Fig. 1. Schematic presentation showing the methodology of charcterization the hydro functional soil mapping units using nested system approach (according to thegeo-morphology.

M. Salahat et al. / Computers and Electronics in Agriculture xxx (2012) xxx–xxx 3

landform map contains topographic units that delineate the differ-ent topographic positions: foot slope, shoulder, back slope, or plain(Fig. 1). The filter tool found in TAS� was used to enhance the viewof the map.

2.3. Delineating the soil mapping units

The third step was to delineate soil mapping units within thestudy area by overlaying the SSURGO soil map on the landformmap obtained from the previous step (Fig. 1). These units will beconsidered while selecting the sampling sites and later will be inte-grated with the pedostructure parameters to generate the hydro-functional soil mapping units.

Since in our hierarchical approach challenge is to delineatenested functional levels of organization and do not accept overlap-ping between the different data layers, the output map was man-ually examined for overlapping areas. If an overlapping areaexists, further investigations including field visits should be madeto decide either to consider the overlapping as a new unit or toignore it and combine it with the adjacent functional mapping unit.


2.4. Characterizing the soil mapping units

The fourth and final step was to check for the homogeneity ofthe generated soil mapping units by collecting representative soilsamples with the help of the existing SSURGO soil map and thegenerated landform map then obtaining the pedostructure param-eters for them following the characterization procedures describedby Braudeau et al. (2004,2005), and Salahat (2006).

The SSURGO soil map used in this study is the one available on-line at the NRCS website (2006). Due to time and cost constraints,the research was restricted to three soil units out of the six conven-tional soil mapping units delineated in the study area. The threeunits used in this research are: SwA, Starks–Fincastle complex with0–2% slopes; MsC2, Miami silt loam, 6–12% slopes; and HoA, Hon-onegah fine sandy loam, 0–2% slopes. The study area is in the Tip-ton Till Plain physiographic region. The northern part of the studyarea is on a till plain overlain by loess, while the southern part is ona small terrace of Jordan Creek. Two map units, the SwA and MsC2(defined below), are on the till plain, while the HoA map unit areon the terrace.



Fig. 2. Aerial photo, SSURGO map, and sample locations.


The SwA map unit consists of a Starks–Fincastle complex on0–2% slopes. The Fincastle series is classified as a fine-silty,mixed, superactive, mesic Aeric Epiaqualf. Fincastle soils aresomewhat poorly drained and formed in 55.9–101.6 cm (22–40 in.) of loess overlying loamy glacial till. They are deep todense till and the depth to carbonates varies from 35 to 60 in.The Starks series is classified as a fine-silty, mixed, superactive,mesic Aeric Endoaqualf and is very similar to the Fincastle seriesexcept for the presence of water-worked, stratified material atthe base of the profile. These two soil series often occur so clo-sely associated that it is impractical to separate them duringmapping. Hence, they are mapped as a complex of two similarsoils.


The MsC2 map unit consists primarily of Miami silt loam on 6–12% slopes. The Miami series, classified as fine-loamy, mixed, ac-tive, mesic Oxyaquic Hapludalfs, consists of moderately well-drained soils formed in up to 18 in. of loess overlying dense, loamyglacial till. Depth to carbonates varies from 20 to 40 in. The Miamiseries occurs on the more sloping areas of the site, and the Ap hori-zon is frequently eroded.

The HoA map unit consists primarily of Hononegah fine sandyloam on 0–2% slopes. The Hononegah series, classified as sandy,mixed, mesic Entic Hapludolls, consists of very deep, excessivelydrained soils formed in sandy alluvial material underlain bywater-sorted sand and gravel at a depth of 20–40 in. Depth to car-bonates is typically 30–40 in.




2.5. Soil sampling

Finding the sample locations within each delineated soil map-ping unit was aided by using a shapefile generated using ArcMap�

9.1 software, ArcPad software, iPac, and a Global Position System(GPS) unit. Fig. 2 represents sample locations at the field. Oncethe required sample location was determined, undisturbed soilsamples that extended the entire length of the soil profile were ob-tained using a hydraulic soil probe mounted on a truck. The hori-zons in which the soil samples were obtained using the probewere identified following the nomenclature of the existing soil sur-vey. Two undistributed replicates were obtained for each samplinglocation corresponding to its horizons. One sample was taken fromthe surface horizon Ap and another sample was taken from themid-point of the B horizon of each soil mapping unit under inves-tigation. Each sample was kept in a 5-cm diameter x 10-cm highplastic sleeve and stored in the laboratory until the time of analy-sis. Seven Ap horizons and 7 B horizon samples were obtained fromeach soil type (Hononegah, Starks, and Miami) for a total of 42 soilsamples.

2.6. Laboratory measurements

Measurements of both soil characteristic curves were con-ducted separately for 21 Ap and 21 B horizon samples distributedas follows: seven samples from Hononegah (HoA), seven samplesfrom Miami (MsC2), and seven samples from Starks (SwA). Onelot of the 42 replicates was sent to the laboratory of IRD in Marti-nique equipped with the retractometer (Braudeau et al., 1999) forthe shrinkage curve measurements, the other was used to measurethe soil–water potential curve using the tensiometer (0–80 kPa). Amini tensiometer with a 2 mm diameter porous cup was used forthis purpose, and connected to a transducer that converts thewater tension into an electrical signal sent to a data logger moni-tored by LabView. The water tension and the weight of the samplewere recorded every 10 min. At the end of the measurement, thesoil sample was moved quantitatively to obtain soil oven dryweight.

For both soil analyses, stored soil samples were cut into two4.5 � 5 cm pieces for replication in case of non-valuable measure-ment. At the time of analysis, each sample was saturated withwater by placing it on a near saturated sand bath (�2 cm of waterhead). The samples were self-supporting and were not confined orconstrained in any way. After saturation for 48 h, the sample wasplaced on the corresponding device to continuously measure thesoil water potential and shrinkage curves upon air drying.

2.7. Pedostructure parameters extraction from the soil potential curveand the shrinkage curve together

A sample of measured pedostructure moisture characteristiccurves (water potential and soil shrinkage curves) is shown inFig. 3. The link between both curves is extended using the physicalequation describing the soil water potential curve given by Brau-deau and Mohtar (2004), Eq. (1). This equation relates the soilwater pressure, hma, read on the tensiometer and the macroporewater content, Wma, such as:

hma ¼ qwEma1

Wma þ r� 1

WmaSat þ r

� �ð1Þ

where Wma and hma are the water content and the water potential inthe macropores of the pedostructure (represented by the soil sam-ple); qw is the water bulk density; Ema is the potential energy of thesolid phase resulting from the external surface charge of the pri-mary peds, in joules/kgsoil; r is a part of the micro-pore water at


the surface of the primary peds, in kgwater/kgsolids; and WmaSat isthe macro pore water content at saturation (corresponding tohma = 0). Ema, WmaSat and r are the parameters of the potential curvewhich will be determined using experimental data (Braudeau andMohtar, 2004) as follows.

Considering the conditions of measurement of the soil waterpotential, h(W), and the soil shrinkage curve, V(W), all the statevariables of the pedostructure such as hma, Wma, hmi, Wmi, W, andh can be considered as state variables at hydro and thermodynamicequilibrium of the pedostructure for each value of the water con-tent W. Equations of the shrinkage curve V(W) and of Wma(W) atequilibrium are written such as (Braudeau et al., 2004):

V ¼ V0 þ Krewre þ Kbswbs þ Kstwst þ Kipwip ð2Þ

Wma ¼ wst þwip ¼ �1

kMln 1þ expð�kMðW �WMÞÞð Þ ð3Þ

where V0 is the pedostructure specific volume at the dry state indm3

soil=kgsoil; Kbs and Kst, and Kip the slopes of the linear shrinkagephases of the shrinkage curve (in kgwater=dm3

soil) due to evaporationof water types, namely: residual, basic, structural and interpedal(wre, wbs, wst, and wip). Parameters of Eq. (3) are kM and WM, the firstone being related to the curvature of the shrinkage curve and thesecond one corresponding, in value, to the micropore water Wmi

at saturation (WM = WmiSat).Therefore, the pedostructure parameters Wsat, r, Ema, KM, and

WM, were extracted by optimizing the fitting of the measuredwater potential curve with Eq. (1) (using Eq. (3) for Wma and con-sidering that WmaSat = Wsat �WM) using the Excel� solver tool. Theparticular water contents WC and WD corresponding, respectively,to the water content at 100 kPa and the water holding capacityaccording to Braudeau et al., 2005, were calculated from WM, andKM using equations given by these authors. They are used hereafterin the text as characteristic parameters instead of WM, and KM.

The second step is to use Wsat, WM, and KM obtained from thepotential curve fitting with Eq. (1), as values of these parametersfor the shrinkage curve (among the 8 needed) in the fitting of themeasured shrinkage curve with Eq. (2). However, most of theshrinkage curves measured in this study have a shape very differ-ent from the usual sigmoid and only Kbs, and V0, and sometimes,WN, and kN, can be determined, the other parameters (Kst, Kip, KL,WL) are undetermined in the fit.

Thus, the characteristic soil parameters which will be used inthis study are: Kbs, V0, Wsat, r, Ema, WC, and WD. The two last param-eters were preferred rather than WM, and KM as characteristics ofthe soil because of their empirical meaning which can be comparedwith the corresponding soil characteristics in the literature. In par-ticular, they can be compared with W100 (at 100 kPa) and WFC

(Field capacity) estimated from texture and organic matter contentusing pedotransfer functions.

2.8. Discriminant analysis

Discriminant analysis was performed on the results using SAS�

9.1 software. The analysis for the Ap horizons was separated fromthe B horizons. The shrinkage curve parameters (Kbs, V0, and WSat),potential curve parameters (r, Ema, WSat, WC, and WD), and thecombination of the two curves parameters (Kbs, V0, r, Ema, WSat,WC, and WD) were used separately for discriminating soil mappingunits. For each soil group the within-samples matrix of sums ofsquares and products (SSP matrix), W, and the analogous be-tween-samples SSP matrix, B, were formed. The latent roots andvectors of the matrix W�1B was then found and the discriminantscores obtained by multiplying the vectors by the original variates(the parameter values). The compilation of the scores in a table al-lowed an investigation of the uniqueness of the parameters. The



Fig. 3. Example of shrinkage curve and soil water potential curve, measured (dots) and simulated (lines) on two replicates of a soil sample.


discriminant analysis results table shows how many samples froma certain group are statistically grouped under the differentgroups, which evaluates the precision of the proposed methodol-ogy to discriminate samples and classifying them under the cor-rect groups.

3. Results and discussion

3.1. The pedostructure parameters obtained

For the sandy loam soils in this study area, the shrinkage curvemeasured was found to be flat or weakly swelling (Fig. 4) for mostof the samples. The shrinkage curve did not distinguish betweenthe different water pools and inflection points as shown by the typ-ical curve discussed by Braudeau et al. (2004). It is why the shrink-age curves could not provide more than V0, WN and Kbs ascharacteristic parameters that can be used for the discriminantanalysis. However, the obtained PS parameters from shrinkageand potential curves (Kbs, V0, WSat, WC, WD, r, and Ema) were usedin the final discriminant analysis.

Table 1 shows values of the soil PS parameters, and silt, clay andOM content for each soil type for the Ap and B horizons. Surface


horizons show higher pedostructural parameter values than sub-surface horizons, mainly due to higher organic matter content insurface horizons that overcomes the higher clay content in the sub-surface horizons. In general, Hononegah (HoA) soils show highervalues of r and Ema than Miami (Ms C2) and Starks (SwA) soilsdue to slightly higher organic matter content. Clay content is high-er in Miami and Starks than in Hononegah, which gives them high-er values of WC, WD, WSat, Kbs, and V0. Kbs is the only characterizingparameter that is not affected by organic matter content since it re-flects the rigidity of the soil structure.

In general, the values of the micropore water at the surface ofthe primary peds (r) and potential energy of the external surfaceof the primary peds (Ema) have the same trend because bothparameters are related to the total surface of their primary pedsand thus, to the soil structure division. They have higher valuesfor Ap horizons than B horizons (Table 1). Organic matter has thehighest influence on both parameters, since the higher the organicmatter the higher are r and Ema. The r is believed to be a structuralparameter. The Hononegah’s Ap horizon has higher r and Ema val-ues than the Miami or Starks Ap horizons, probably due to a higherorganic matter content. Miami and Starks have similar r valuesand similar organic matter content.



Fig. 4. Representative shrinkage curves (red) and soil water potential curves (tensiometric) curves (green) of the three types of soil, in the horizon A (left) and horizon B(right). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)


The water content at saturation (WSat) is directly related to thetotal porosity. In general, the values of Ap horizons are higher thanB horizons for all soil types due to higher organic matter content inAp horizons leading to a better structure.

The water content at point C (WC) on the shrinkage curve corre-sponding to the water content at 100 kPa showed higher values forAp horizons than B horizons. At 100 kPa tension, the macroporesout of the primary peds will be empty of water, while the primarypeds remain filled with water (Braudeau et al., 2005). The highervalues of WC for Ap horizons compared to B horizons can be attrib-uted to higher silt contents in Ap horizons than B horizons in thissoil landscape. Moreover, the value of WC increases with the in-crease of the clay content and Emi. WD is the water content at fieldcapacity. Field capacity status could be reached after gravitation-


ally draining a saturated soil for 48 h. This point is very importantbecause it reflects the best soil status to conduct agricultural prac-tices. The trend for WD is similar to what was discussed for WC. Sur-face soils have higher WD values than B horizons due to highorganic matter and silt contents. The HoA soil has the lowest val-ues for the Ap and B horizons.

Kbs, the slope of the linear basic phase of the shrinkage curve, isa structural feature of the assembly of aggregates that representsthe volume change ratio between the aggregated soil sample andthe primary peds. At the same time, the parameter is positivelycorrelated to organic matter. In general, the Ap horizons have low-er Kbs values than the B horizons. This is because Ap horizons havebetter structure due to the OM. This is very clear when studyingthe shrinkage curves because a large number of Ap horizons have



Table 1Soil pedostructural parameters, silt, clay and OM content for Ap and Bt horizons.

Soil type Sample no. Soil horizon Depth (cm) r Em Wc WD wsat Kb,s Ve OM Silt Clay(Kgwater=Kgsolids) (J/kgsoil) (Kgwater=Kgsoil) (Kgwater=Kgsoil) (Kgwater=Kgsoil) (dm3

soil=Kgwater) (dm3soil=Kgsoil) (%) (%) (%)

HoA I 471 Ap 5–15 0.009 0.9 0.246 0.346 0.420 0.25 0.69 3.8 48 6470 Bt 70–80 0.010 1.1 0.145 0.174 0.196 0.21 0.58 0.4 45 7

HoA lX 407 Ap 5–15 0.036 3.6 0.137 0.271 0.345 0.13 0.62 3.8 44 8408 Bt 75–85 0.014 1.2 0.088 0.156 0.177 0.10 0.52 1.8 35 10

HoA IV 404 Ap 8-18 0.026 2.2 0.157 0.247 0.309 0.21 0.66 4.3 41 13400 Bt 60-70 0.007 0.7 0.152 0.185 0.198 0.30 0.52 0.5 39 5

HoA X 399 Ap 5-15 0.071 8.0 0.157 0.253 0.329 0.01 0.66 5.0 44 5409 Bt 75-85 0.010 1.0 0.107 0.153 0.170 0.25 0.52 0.5 35 13

HoA VI 403 Ap 5-15 0.007 0.5 0.334 0.433 0.451 0.64 0.67 3.5 47 7405 Bt 90-100 0.011 1.4 0.159 0.188 0.195 0.34 0.51 2.3 50 25

HoA VII 402 Ap 7-17 0.013 1.1 0.145 0.248 0.248 0.31 0.56 3.6 45 7406 Bt 80-90 0.022 2.7 0.113 0.177 0.185 0.13 0.56 0.3 31 9

HoA XI 401 Ap 3-13 0.023 1.8 0.243 0.331 0.392 0.18 0.69 4.1 51 6472 Bt 40-50 0.015 1.5 0.207 0.249 0.290 0.31 0.63 1.0 37 5

Ms C2 I 432 Ap 5-15 0.034 5.1 0.256 0.339 0.341 0.23 0.64 3.6 67 17418 Bt 50-60 0.005 0.5 0.169 0.200 0.222 0.57 0.56 0.4 26 30

Ms C2 II 435 Ap 5-15 0.005 0.5 0.275 0.344 0.351 0.17 0.69 3.6 70 15436 Bt 53-63 0.010 1.1 0.226 0.243 0.278 0.49 0.58 0.3 60 27

Ms C2 III 437 Ap 7-17 0.007 0.7 0.236 0.299 0.299 0.81 0.69 3.4 68 15438 Bt 43-53 0.011 1.2 0.193 0.212 0.250 0.63 0.58 0.6 57 26

MS C2 IV 440 Ap 3-13 0.043 6.7 0.214 0.276 0.302 0.53 0.71 3.6 68 15439 Bt 95-105 0 008 0 8 0 163 0 193 0.212 0 44 0 55 0.3 45 28

MS C2 V 416 Ap 3-13 0.006 0.5 0.304 0.375 0.375 0.16 0.71 2.2 70 15417 Bt 43-53 0.003 0.3 0.176 0.200 0.219 0.51 0.56 0.5 60 26

MS C2 VII 414 Ap 3-13 0.013 1.3 0.245 0.368 0.368 0.29 0.63 3.3 40 24415 Bt 33-43 0.006 0.7 0.221 0.257 0.258 0.60 0.61 0.5 31 30

MS C2 IX 413 Ap 5-15 0.014 1.4 0.257 0.361 0.378 0.15 0.64 3.6 65 18412 Bt 40-50 0.017 1.9 0.218 0.267 0.275 0.61 0.58 0.5 47 22

Sw A1 419 Ap 5-15 0.006 0.5 0.287 0.371 0.389 0.46 0.67 3.8 72 12421 Bt 103-113 0.023 2.4 0.178 0.206 0.299 0.45 0.61 0.3 50 25

Sw A2 433 Ap 5-15 0.008 0.8 0.285 0.377 0.377 0.07 0.70 3.6 65 18434 Bt 63-73 0.006 0.6 0.177 0.222 0.262 0.34 0.60 0.4 37 27

Sw A3 410 Ap 5-15 0.025 2.7 0.191 0.298 0.338 0.58 0.72 2.8 71 12411 Bt 93-103 0.019 2.1 0.179 0.270 0.299 0.60 0.59 0.3 38 25

Sw A4 422 Ap 5-15 0.006 0.5 0.275 0.339 0.345 0.27 0.67 3.9 67 14423 Bt 53-63 0.011 1.2 0.197 0.210 0.245 0.56 0.60 0.3 40 28

Sw A5 428 Ap 3-13 0.007 0.6 0.277 0.403 0.453 0.23 0.69 4.0 70 14427 Bt 80-90 0.006 0.6 0.154 0.221 0.245 0.28 0.58 0.4 35 22

Sw A6 430 Ap 5-15 0.014 1.2 0.224 0.309 0.321 0.09 0.72 3.8 70 17429 Bt 55-65 0.013 1.6 0.230 0.247 0.274 0.45 0.59 0.3 62 25

Sw A7 431 Ap 3-13 0.012 1.3 0.319 0.384 0.389 0.28 0.71 3.7 68 16425 Bt 103-113 0.006 0.7 0.164 0.178 0.211 0.49 0.58 0.5 49 34

8M

.Salahatet

al./Computers

andElectronics

inA

griculturexxx

(2012)xxx–

xxx

Pleasecite

thisarticle

inpress

as:Salahat,M.,

etal.Tow

arddelineating

hydro-functionalsoilm

appingunits

usingthe

pedostructureconcept:A

casestudy.

Comput.Electron.A

gric.(2012),http://dx.doi.org/10.1016/j.compag.2012.04.011


Table 3Discriminant analysis for the potential curve using r, Ema, WC, WD, and WSat for the Apand Bt horizons.

GROUP/horizon Sw A Ms C2 HoA

Sw A Ap – horizon 5 1 1Bt – horizon 4 2 1

Ms C2 Ap – horizon 4 3 0Bt – horizon 1 6 0

HoA Ap – horizon 2 0 5Bt – horizon 1 0 6


minor shrinking, and the curve is almost linear. The standard devi-ation for Ap horizons is higher than for B horizons, which leads tothe conclusion that we have more extreme values in the Aphorizons.

For the last parameter V0, the Ap horizons showed higher valuesof the specific volume of the dry soil than the B horizons due to anabsence of structure in the later. For these silty soils, Kbs is a goodindicator of the soil structure. The lower the Kbs the lower theswelling capacity which means higher value for macropores vol-ume, hence, higher value of V0. For weakly structured soil, Kbs valuewill be high, V0 will be low which leads to low value of macroporosity.

3.2. Discriminant analysis

Discriminant analysis using SAS� 9.1 was performed on theparameters used in characterizing the soil mapping unit to investi-gate the uniqueness of these units. We refer to the soil mappingunits with their pedostructure parameters as hydro-functional soilmapping unit .The uniqueness of the characterization by the soilshrinkage curve, the soil water potential curve, and the two curvescombined, were investigated.

3.2.1. Discriminant analysis by soil horizonTable 2 shows the results of the discriminant analysis con-

ducted using shrinkage curve parameters alone for Ap and B hori-zon samples, respectively. The main parameters of the shrinkagecurve (Kbs, V0, and Wsat) successfully discriminates the soil mappingunits. For Ap horizons, six of the seven HoA samples are groupedunder HoA, while one sample is grouped under SwA. Four of theSwA samples are grouped under SwA, two under MsC2, and theseventh sample is grouped under HoA. For MsC2, the results arenot as good, only one MsC2 sample is grouped under MsC2, andthree samples are grouped under each of HoA and SwA. This isprobably due to relatively steep slopes and consequently greatererosion in the MsC2 map unit, resulting in more variability in thethickness of loess on these soils relative to the SwA map unit,and the possibility that in some areas the original surface horizonhas been eroded away and the current surface is actually the top ofthe Bt horizon. Subsurface horizons generally discriminate betterbetween the different soil types than surface horizons. Six HoAsamples are in the HoA group, all MsC2 samples are put underSwA, and four SwA samples are put under SwA. In general it isnoted that the discriminant analysis results are lower betweenMsC2 and SwA because these two soil type share the same parentmaterial and same landuse.

The results for using the potential curve alone in the discrimi-nant analysis are shown in Table 3. For the Ap horizons, five ofHoA, three MsC2, and five SwA are grouped under the correct soiltype. The results are better for the B horizons: six HoA and MsC2and four SwA are grouped in the correct group. The potential curvecorresponds mainly to the macropores, which is related to thestructure more than texture. Subsurface soils have more developed

Table 2Discriminant analysis for the shrinkage curve using Kbs, V0, and WSat for the Ap and Bthorizons.






structure than the Ap horizons due to less manmade interferences.This fact explains why better results were obtained for B horizons.

The results using the soil shrinkage and soil water potentialcharacterizing curves together are shown in Table 4. Using thetwo curves enhances the results, and the soil’s mapping unituniqueness is more distinguishable. Six HoA, five MsC2 and SwAare grouped under the correct soil type, for the Ap horizon. Subsur-face horizons showed better results with all HoA, six MsC2, andfive SwA in the correct group. It was expected that there wouldbe better results upon using the two curves together because theuniqueness found in both the micro and macropores are utilized.

3.2.2. Discriminant analysis combining surface and subsurfacehorizons

The results of discriminant analysis for shrinkage curve, poten-tial curve, and potential and shrinkage curves combining surfaceand subsurface horizons are shown in Table 5. Discriminant resultsfor shrinkage curve combining surface and subsurface horizons aregood but of lesser quality than results obtained after separating Apand B horizons (Table 5). Combining the two horizons for the anal-ysis might cause overlapping in the values for each parametersince each horizon has its own range. HoA is clearly discriminatedfrom the other two soils, while MsC2 is the least discriminated soil.Nine out of the fourteen HoA samples are grouped under HoA,seven MsC2 are correctly grouped, and six of SwA are groupedcorrectly.

When using the potential curve, ten of HoA, seven of MsC2, andeight of SwA are grouped under their own soil type. The results arealmost similar to those obtained by using the potential curve aloneand separating Ap and B horizons. When using both the potentialand shrinkage curves combining surface and subsurface horizons,eleven HoA, ten MsC2, and nine SwA are grouped correctly (Table5). These results are better than using the shrinkage or potentialcurves alone but not as good as compared to separating Ap fromB horizons. Miami and Starks soils have mixed results because theyshare the same parent material, and development environment.The main difference between them is the drainage class. Miamisoils are better drained than Starks. The Hononegah soils havehigher percent sand and are developed over sandy sediments.Overall, results of discriminant analysis and PS characterization re-flect the functionality of the pore space, which corresponds to thedevelopment of soil structure.

Table 4Discriminant analysis for the potential and shrinkage curves for the Ap and Bthorizons.







Table 5Discriminant analysis for shrinkage curve, potential curve, and potential and shrinkage curves combining the two horizons Ap and Bt.

Discriminant analysis Group Sw A Ms C2 HoA

For shrinkage curve using Kbs, V0, and Wsat combining the two horizons Ap and Bt Sw A 6 6 2Ms C2 4 7 3HoA 4 1 9

For potential curve using r, Ema, WC, WD, and Wsat combining the two horizons Ap and Bt Sw A 8 4 2Ms C2 5 7 2HoA 3 1 10

For potential and shrinkage curve combining the two horizons Ap and Bt Sw A 9 4 1Ms C2 3 10 1HoA 2 1 11


3.3. The hydro-functional soil mapping units vs. the SSURGO soilmapping units

The resulting hydro-functional soil mapping units are similar tothe existing soil mapping units obtained from SSURGO at least forthe three soil units tested (Figs. 1 and 2). The key difference be-tween the hydro-functional soil mapping units and the SSURGOsoil mapping units is that the hydro-functional units were definedby physically based quantitative attributes that describe the soilstructure and soil–water interactions which generate all the phys-ical properties of the soil medium. The fact that the three soil typesused in this work had approximately the same boundaries afterperforming the four-step system hierarchy approach describedshowed that; (1) SSURGO soil data set consists of a detailed secondorder soil survey in which microrelief is already taken into consid-eration; and (2) the hydro-structural soil properties matched wellwith the holistic approach used by pedologists when they definedthe soil mapping units delineation and their qualitative character-ization. It is expected that the hydro-functional mapping unit willbe different than the existing mapping unit if a less detailed soilmap was used. Upon using the SSURGO map with a scale of levelfour (for example), some of the micro-relief is not captured andthus the mapping unit does not show all the details within it.

Overall, the hydro-functional soil mapping units agreed to ahigh extent with the SSURGO soil mapping units. Still, furtherworks should be made in this way by testing more soil units forestimating the uniqueness of the correspondence between the hy-dro-functional characterization and the existing soil mapping unitdelineation.

4. Conclusions

This research shows the potential for coupling the pedostruc-ture concept into existing SSURGO soil maps to overcome their lackof comprehensive quantitative attributes which put limitationsupon their use directly in agronomic models, hydrologic models,and precision agriculture or decision support systems.

The work demonstrated four steps of a hierarchical systems ap-proach to generating hydro-functional soil mapping units thatcomprise physically based and quantitative parameters extractedfrom continuously measured water potential and soil shrinkagecurves. Due to time and cost constraints, the research was re-stricted to three soil units, but should be extended to all mappingunits of the zone in study.

In spite of this limitation, it was found that the hydro-functionalmapping units match the existing soil mapping units of the largescale SSURGO map of which the soil data set consists of a detailedsecond order soil survey in which microrelief is already taken intoconsideration.

The discriminant analysis results reveal the potential of usingthe PS parameters including parameters of the shrinkage curveand the water potential curve together to differentiate and charac-terize the different hydro-functional soil mapping units. Addition


of these hydro-structural parameters to the attributes of the soilmapping units makes them hydro-functional and potentially dy-namic with the water cycle modeled by a soil water model likeKamel.

The results of this research are promising and can be consideredas the first cornerstone in developing a new approach of physi-cally-based characterization of the existing soil mapping unitswhich takes into consideration the hydrostructural properties ofthe soil medium according to the pedostructure concept and thehierarchical system approach. The delineated soil units based onthis approach can be termed as hydro-functional soil mappingunits.

Acknowledgments

This work was supported by the Hashemite University, thepedology laboratory of IRD in PRAM (Agricultural Research Centerof Martinique), and Purdue University.

References

Ahuja, L.R., Liwang, M., Dennis, J.T., 2006. Trans-disciplinary soil physics researchcritical to synthesis and modeling of agricultural systems. Soil Science Society ofAmerica Journal 70, 311–326.

Bouma, J., Stoorvogel, J., van Alphen, B.J., Booltink, H.W.G., 1999. Pedology, precisionagriculture, and the changing paradigm of agricultural research. Soil ScienceSociety of American Journal 63, 1763–1768.

Braudeau, E., Costantini, J.M., Bellier, G., Colleuille, H., 1999. New device and methodfor soil shrinkage curve measurement and characterization. Soil Science Societyof America Journal 63, 525–535.

Braudeau, E., Loukil, A., Zidi, C., Derouiche, C., Niang, M., Hachicha, M., Belouchette,H., Decluseau, D., Mtimet, A., 2002. Le Sirs-Sols du périmètre irrigué de Cébala(Basse vallée de la Mejerda, Tunisie). Un système d’information pédologiquespécialisé pour la modélisation agronomique. La gestion des périmètres irriguéscollectifs à l’aube du XXIe siècle, enjeux, problèmes, démarches. P. Garin, P.Y. LeGal, Th. Ruf (éditeurs scientifiques), Montpellier, France. Pcsi, Cemagref, Cirad,Ird, Montpellier France, Colloques, p. 280.

Braudeau, E., Frangi, J.P., Mohtar, R.H., 2004. Characterizing non-rigid aggregatedsoil–water medium using its shrinkage curve. Soil Science Society of AmericaJournal 68, 359–370.

Braudeau, E., Mohtar, R.H., 2004. Water potential in non-rigid unsaturated soil–water medium. Water Resources Research (40), W05108. http://dx.doi.org/10.1029/2004WR003119.

Braudeau, E., Sene, M., Mohtar, R.H., 2005. Hydrostructural characteristics of twoAfrican tropical soils. European Journal of Soil Science 56, 375–388.

Braudeau, E., 2006. Kamel Simile version, Agence pour la Protection desProgrammes, Paris, IDDN.FR.001.390019.000.S.P.2006.000.31500.

Braudeau, E., Mohtar, R.H., 2009. Modeling the soil system: bridging the gapbetween pedology and soil–water physics. Global and Planetary Change Journal67, 51–61.

Singh, J., Mohtar, R.H., Braudeau, E., Heathmen, G., Jesiek, J., Singh, D, 2012. Fieldevaluation of the pedostructure-based model (Kamel�). Computers andElectronics in Agriculture. http://dx.doi.org/10.1016/j.compag.2012.03.001.

Lagacherie, P., Voltz, M., 2000. Predicting soil properties over a region using sampleinformation from a mapped reference area and digital elevation data: aconditional probability approach. Geoderma 97, 187–208.

Lin, H., 2003. Hydropedology: bridging disciplines, scales, and data. Soil ScienceSociety of America Journal 2, 1–11.

Lin, H., Bouma, J., Pachepsky, Y., Western, A., Thompson, J., van Genuchten, R., Vogel,H., Lilly, A., 2006. Hydropedology: synergistic integration of pedology andhydrology. Water Resources Research 42, 1–13.

Linday, J., 2003. Terrain Analysis System (Online), The University of Manchester.<http://www.sed.manchester.ac.uk/geography/research/tas/>.


http://dx.doi.org/10.1029/2004WR003119

http://dx.doi.org/10.1029/2004WR003119


http://www.sed.manchester.ac.uk/geography/research/tas/



Moore, I.D., Gessler, P.E., Nielsen, G.E., Peterson, G.A., 1993. Soil attribute predictionusing terrain analysis. Soil Science Society of America Journal 57, 443–452.

Natural Resources Conservation Service, 2006. Soil Survey Geographic (SSURGO)Database (online). <http://www.ncgc.nrcs.usda.gov/products/datasets/ssurgo/>(accessed 15.12.06).

Odeh, I.O.A., McBratney, A.B., Chittleborough, D.J., 1994. Spatial prediction of soilproperties from landform attributes derived from a digital elevation model.Geoderma 63, 197–214.


Salahat, M.A., 2006. Field scale soil characterization using pedostructural properties.Ph.D. dissertation, Purdue University, IN, USA.

Tippecanoe County GIS office, 2006. GIS and Mapping (Online). <http://www.tippecanoe.in.gov/mits/division.asp?fdd=29-88> (accessed 15.12..2006).

Voltz, M., Lagacherie, P., Louchart, X., 1997. Predicting soil properties over a regionusing sample information from a mapped reference area. European Journal ofSoil Science 48, 19–30.


http://www.ncgc.nrcs.usda.gov/products/datasets/ssurgo/

http://www.tippecanoe.in.gov/mits/division.asp?fdd=29-88

http://www.tippecanoe.in.gov/mits/division.asp?fdd=29-88


computers and electronics in agriculturetal models. this enhancement is possible through a new...

Documents