Title: Water and nutrient movement related to soil productivity in an aggregated gravelly oxisol from Cameroon
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00103065/00001
 Material Information
Title: Water and nutrient movement related to soil productivity in an aggregated gravelly oxisol from Cameroon
Physical Description: vii, 160 leaves : ill. ; 29 cm.
Language: English
Creator: Anamosa, Paul R., 1954-
Genre: bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )
Statement of Responsibility: by Paul R. Anamosa.
Thesis: Thesis (Ph. D.)--University of Florida, 1989.
Bibliography: Includes bibliographical references (leaves 149-158).
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00103065
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 001563028
notis - AHH6754
oclc - 22711491

Full Text








During the course of this project I have been fortunate to receive

a great deal of assistance. The IFAS International Programs Office

provided most of my assistantship and travel fare to and from Cameroon.

In Cameroon, Eric van Ranst, Soil Science Department Chair, provided me

with access to vehicles and field technicians. Philip Mokoko and

Maurice Ndazame gave invaluable help to my efforts; aiding in the

management of the field project, translating French and the local

Dschang dialect to English, and advising me on matters of protocol as

well as cultural values.

I owe a great debt of gratitude to Dr. W. G. Blue, chairman of my

graduate committee, for his support of my field and laboratory

activities as well as his editorial review of this dissertation. I am

fortunate to have studied under his guidance and am appreciative of his

personal generosity and understanding.

I am grateful to Dr. P. Nkedi-Kizza, who was willing to join my

graduate committee mid-term and who provided new perspectives to my

objectives. He gave constructive guidance and criticism to my

laboratory experiments.

I would also like to thank the other members of my graduate

committee, cochairman Dr. J. B. Sartain, Dr. B. L. McNeal, Dr. P. E.

Hildebrand, and Dr. G. Kidder for the interest and feedback they

provided. Dr. Hugh Popenoe graciously substituted for Dr. Blue while he

was in Cameroon.

Lastly, I would like to thank those on the home front. My wife,

Frances, was joyfully willing to pull up stakes and move to Cameroon,

put up with late-night runs to the lab to check pumps, and gave constant

encouragement throughout the course of this ordeal. Our feline

housemates Ferguson and Abigail helped with typing the manuscript.



ABSTRACT.......................................................... iv


1. INTRODUCTION...........................................1

2. REVIEW OF THE LITERATURE................................ 3
Formation Processes............................... 4
Agricultural Productivity............................9
Research Topics................................... 12

3. SOIL CHARACTERIZATION................................ 15
Materials and Methods ........................... 15
Results and Discussion.......................... 17

SOIL COLUMNS........................................... 24
Introduction. ................................... 24
Materials and Methods ........................... 31
Results and Discussion .......................... 34

APPLICATION SCHEDULING............................... 61
Materials and Methods ........................... 63
Results and Discussion .......................... 66
Conclusions........................................ 94

Introduction....................................... 98
Materials and Methods .......................... 101
Results and Discussion........................... 106
Conclusions .................................... 137

7. OVERALL CONCLUSIONS....................................140
Introduction....... .............................140
Soil Characterization ......................... 141
Crop Response ................................. 142
Nutrient Leaching............................... 143

APPENDIX A SOIL PROFILE DESCRIPTION..............................145

APPENDIX B CROP COMPONENT YIELDS.................................147

REFERENCES .....................................................149

BIOGRAPHICAL SKETCH................................................159


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements of the Degree of Doctor of Philosophy



Paul R. Anamosa

August 1989

Chairman: W. G. Blue
Cochairman: J. B. Sartain
Major Department: Soil Science

Gravel decreases the water- and nutrient-holding capacities of

soil. Soils with gravel horizons (stone lines) are being increasingly

utilized for crop production in equatorial Africa. This study was

conducted to differentiate between the relative effects of water and

nutrient stress for crops grown on stone-line soils, and to determine if

preferential water flow and mobile/immobile water concepts should be

considered in describing nutrient and water behavior. The effects of

plant densities of maize (Zea mays L.) and bean (Phaseolus vulgaris L.)

and of split applications of plant nutrients were investigated for a

clayey-skeletal, oxidic, isohyperthermit, Typic Gibbsiorthox near

Dschang, Cameroon. The movement of soil nutrients was studied in soil

columns subjected to simulated rainy seasons. The nature of the porous

network of the soil was studied using miscible-displacement techniques

with tritiated water.

Increased splitting of mobile-nutrient applications (K, NO3, and

NH4) resulted in increased grain yields, but had no effect on stover

yields. Early-season moisture stress apparently decreased plant

emergence. However, high-density plantings yielded more grain and

stover than did similarly fertilized, low-density plantings. Thus, once

plants were established, grain yields were not adversely affected by

moisture stress. A 30-d delay in planting resulted in a 40% increase in

seasonal rainfall and 50 and 70% grain-yield reductions for bean and

corn, respectively.

Leaching of Ca, K, and Mg from 70-cm long soil columns was not

affected by rainfall regimes or fertilizer-application schedules,

although the distribution of Ca, K, and Mg in the columns indicated more

downward movement under higher rainfall. Leaching of K was negligible

under all treatments used in this study. Split applications of

fertilizer composed primarily of K, NO3, and NH4 resulted in greater

concentrations of Ca and Mg with depth.

Moisture-release curves showed that the soil drained nearly 30%

of total water content at 50-mbar tension, but still held 30% at 15-bar

tension. Miscible-displacement experiments indicated that, under

saturated conditions, the soil had a high dispersivity and held about

50% of it water in immobile-water regions.

Delays in planting to avoid early-season water stress result in

greater leaching losses and reduced grain yields. Splitting the

applications of mobile nutrients should increase their plant

availability later in the growing season. Gravel porosity and immobile-

water regions in the soil harbored highly mobile plant nutrients and

reduced leaching.


Chapter 1


In light of the present food-production crisis facing most

countries of sub-Saharan Africa, numerous policy priorities have been

proposed by academics and politicians to encourage the rapid development

of technology to improve Africa's food production capacity (Swindale,

1980; USAID, 1985; Mellor et al., 1987; lyegha, 1988). High on many

priority lists is the need for scientific and technological research

directed towards the development of efficient fertilizer utilization

practices specifically adapted for the low-fertility soils common to

tropical regions.

Shallow gravel horizons, frequently referred to as stone lines,

are common in soils throughout equatorial Africa. Stone-line soils are

generally considered to be agriculturally marginal; however, in a

continent where population growth is out pacing increases in

agricultural productivity, the development and utilization of marginal

lands for farming are increasing.

Stones in the root zone of a soil reduce root penetration and

water- and nutrient-holding capacities. These characteristics in turn

reduce root exploitation of the soil mass and increase both the

susceptibility of crops to water stress and the potential loss of

nutrients by leaching.

Several recent studies have indicated vesicular voids (pores) in

the gravel from stone-line soils (Muller and Bocquier, 1986; Amouric,

1986). The effects of porous gravel on soil-water behavior are not

easily inferred and depend on the porosity and pore-size distribution of

the gravel. In addition to possible storage of plant-available water,

the gravel porosity may also act as a sink/source for the storage of

leachable nutrients, and thereby, harbor nutrients from convective-water


The purpose of this dissertation was to assess several behavioral

characteristics regarding water and nutrient movements through a stone-

line soil from the western highlands of Cameroon. Specifically, the

objectives were:

1. To differentiate between the relative effects of possible

water and nutrient stresses on field crops grown on a stone-

line soil; and

2. To determine if preferential water flow and immobile-water

regions should be considered in describing nutrient-leaching

behavior in these soils.



Soils with gravel horizons are common on the hilly landscapes of

equatorial Africa. Commonly referred to as stone lines, these gravel

horizons were first discussed in the soils' literature in the mid 1930s,

and have experienced intermittent periods of scientific examination in

every decade since. Initial interests in morphology and formation

processes have given way to evaluation of aspects of agricultural


Owing to the limitations of slope and tillage, these soils are

generally considered to be agriculturally marginal (Hidlebaugh, 1984).

However, increasing population pressures in many of the regions where

they occur have necessitated their increased usage. The few published

studies evaluating agricultural behavior have focused on effects of the

gravel on root penetrability and water redistribution. Inferences

regarding appropriate-management practices for stone-line soils under

agricultural production have not been addressed.

The purpose of this review is to examine the scientific literature

de ling with various aspects of stone-line formation processes and

agricultural productivity, and to develop a consensus of the needs for

future research, specifically in the area of crop-management practices.

Formation Processes

The term "stone line" was originally proposed by Sharpe (1938) to

designate "a line of angular to subangular fragments which parallels a

sloping surface to a depth of several feet." Ruhe (1959, p. 223),

summarizing the definitions of several studies (Sharpe, 1938; De

Heinzelin, 1955; Parizek and Woodruff, 1957), defined a stone line as "a

concentration of coarser rock fragments in soils; in cross section it

may be a line, one stone thick or more than one stone in thickness, that

generally overlies material weathered in place from bedrock and that

usually is overlain by variable thicknesses of finer-textured sediment."

De Heinzelin (1955) objected to the term when used to designate the

gravel horizons common to equatorial African soils. He proposed instead

the term "nappe de gravat" (sheet of gravel), because it more

appropriately described the three-dimensional nature of the structure.

However, at present the term stone line is widely used in both the

English and French pedological literature.

The formation processes that create stone lines instill specific

morphological characteristics to the soil profile. It is these

morphological characteristics that have been used to develop hypotheses

concerning the formation processes. Pedologists working throughout

equatorial Africa on a variety of landscapes have developed two

different schools of thought concerning stone-line formation. These

were categorized as either autochthonous (same) or allochthonous

(different) with respect to the parent material of the stones and of the

underlying material (Collinet, 1969). The distinction between the two

categories involves whether the stones are residual from the underlying

parent material or were transported from elsewhere and then covered with

sediment. This distinction was the core of debate among pedologists

originally hypothesizing the formation processes. It is still a point

of contention, considering that allochthonous processes rarely exhibit

the transport of stones over distances greater than several hundred

meters (Riquier, 1969; Sdgalen, 1969; Fairbridge and Finkl, 1984).

The stone lines produced by the two widely accepted autochthonous

processes are relatively uncommon and show vast differences in

morphology. The reworking of soil materials by termites, resulting in

concentration of the finer above the coarser sediments, has been studied

throughout equatorial Africa (De Heinzelin, 1955; Nye, 1955; Sys, 1955;

Gennart et al., 1961). Variations exist among termite species and

geographical locations, but such stone lines generally consist of a

diffuse gravel horizon rarely exceeding 25% by weight of small, 2 to 7

mm, fragments of residual quartz and occasional ironstone nodules. The

gravel horizons range in thickness from 10 to 250 cm, and rarely exceed

depths of 300 cm.

Surface stone lines, frequently called "desert pavement," are

found in extremely arid climates that receive occasional torrential

rains. It is generally believed that these surface stones result from

the fracture of exposed bedrock due to large daily temperature

fluctuations. Sheet erosion during heavy rains then removes any

overlying soil, which either collects in crevices between the stones or

is washed away (Springer, 1958; Finkl, 1979).

By far the most common type of stone line in equatorial regions of

Australia, South America, and Africa is presently attributed to an

allochthonous process. However, several autochthonous-process theories

have been proposed and subsequently refuted. Sharpe (1938, cited in

Ruhe, 1956) and Ireland et al. (1939, cited in Ruhe, 1956) proposed a

theory involving surface creep, in which soil flowing slowly downslope

shears off resistant rock projecting up into the subsoil and carries the

rock along the bottom of the creeping mass. Ruhe (1956) and Parizek and

Woodruff (1957) rejected this theory. They concluded that the sheets of

gravel were originally surface deposits later covered by an over-lying

mantle. Ruhe (1959) later described in detail this theory, which

assumes the stones to be highly resistant residual parent material that

became concentrated on a developing erosional surface by the removal of

finer material with runoff water. Finer-textured sediment derived from

an upper-valley slope then is deposited on the sheet of gravel. This

process is autochthonous in nature, and can not explain soils with

multiple stone lines (Ollier, 1959).

An allochthonous process was first proposed by de Craene (1954)

and later applied to both quartz lines and gravel horizons by Collinet

(1969) and Riquier (1969). In its most basic form, the process begins

with the deposition of rock material from exposed escarpments (rock

outcroppings) onto sloping eroded surfaces. This material then is

covered by a fine colluvial mineral deposit. Therefore, both rock and

fine fraction are genetically different from the soil below the stone

line. The process can be repeated as long as a rock escarpment exists

above the erosional surface.

A similar process can lead to the development of thin, quartz-

stone lines. Quartz veins of geologic origin are frequently sandwiched

between layers of sedimentary rock. If near the surface, such rock may

be transformed into soil or saprolite, leaving the resistant quartz vein

intact. Where the quartz vein intercepts the earth's surface it

provides a source of quartz pebbles that then may be spread over the

soil depending on the slope of the land. If the surface is sloping, the

pebbles will be scattered downslope. If the surface is flat the pebbles

will form mounds or ridges that may run for a considerable length across

the landscape.

The allochthonous process requires winnowing (the movement,

deposition, and concentration of coarse material by wind and running

water), which in turn usually requires climatic instability so that

slopes may go through both erosional and stabilizing periods (Fairbridge

and Finkl, 1984). Such periods are attributed to torrential rains

during arid to semi-arid climatic phases within a normally humid era.

This pattern would allow for erosion of vegetatively bare surfaces

during intermittent heavy rains in an arid phase and subsequent slope

stabilization by vegetation upon return of the humid climatic phase.

Several independent lines of evidence suggest that the

pleniglacial age of the late Wisconsinan cycle was responsible for the

climatic conditions favorable for stone-line formation in the tropics.

Bruckner (1955) working in Africa, Bigarella and de Andrade (1965)

working in Brazil, and Finkl (1979) working in Australia have all

identified regional occurrence of common but discontinuous stone lines

dating from the late Wisconsinan period. The arid phases during the

Wisconsinan period were brought about by a combination of lower solar

radiation, disruption of major air-flow patterns, and extension of the

cold polar oceanic currents into low latitudes (Fairbridge, 1964;

Cailleux and Tricart, 1973).

Stone lines formed from the allochthonous processes of escarpment

retreat, rock-fragment deposition, and fine-fraction sedimentation

frequently have common physical characteristics (Fairbridge and Finkl,

1984). Such stone lines occur as slope deposits on the paleoslopes of

interfluves and residual pediplains (Ojanuga and Wirth, 1977).

Distances of transport usually range from several meters to several

hundred meters. The stones are angular to rounded, but are usually

similar within any singular continuous horizon (Ojanuga and Lee, 1973).

The stones are quite resistant to weathering either because they are

inherently durable such as quartz or because they consist of resistant

lateritic pseudomorphs (similar shape but different mineralogy) of the

original rock fragments (Ojanuga and Lee, 1973; Muller and Bocquier,

1986). Such pseudomorphs result from the natural weathering and

dissolution of rock parent materials, along with the precipitation of Fe

and Al minerals leached from overlying soil horizons rich in Fe and Al

oxides/hydroxides. Lateritic material in the stone line may also come

from pistolitic duricrust that forms with desiccation and hardening of

exposed surface soil resting on top of the fragmenting escarpment

(Frankel and Bayliss, 1966; Amouric et al., 1986). Further erosion and

retreat of the escarpment face causes fragments of the surface duricrust

to drop along with escarpment-rock fragments to the erosional plain


Agricultural Productivity

Little is known about the behavior of tropical stone-line soils or

the influence they exert on agricultural systems which they support.

The paucity of internationally available literature on soils of the

tropics, in general, is well known. The habit of national governments

to establish research stations on a region's best soils frequently

limits the generation of knowledge concerning hillside and

agriculturally marginal soils (Zandstra et al.,1981).

Lal, formerly of the International Institute of Tropical

Agriculture in Ibadin, Nigeria, has conducted several studies

investigating plant-root development and water availability on natural

and synthetic gravelly soils. Babaloa and Lal (1977a) evaluated the

effects of varying gravel concentrations on shoot growth and rooting

depth using soil/gravel mixtures in greenhouse pot studies. The weight

of corn shoots harvested after 21 d decreased by up to 50% as gravel

concentration increased from 10 to 75%. Rooting depth decreased only

slightly as gravel increased from 0 to 10%, but then decreased to 40% of

the non-gravel rooting depth as gravel increased to 25%. The rooting

depth decreased to 5% of the non-gravel rooting depth as the gravel

increased to 75%. Total root length was affected similarly. Shoot

weights increased by 20% as the depth to a 60%-gravel horizon increased

from 5 to 10 cm. Shoot tissue had a nonsignificant increase in

concentrations of N, P, and K as depth to the gravel horizon increased.

The researchers concluded that the gravel retarded rooting depth and

thereby decreased root exploitation of soil nutrients, resulting in

reduced nutrient uptake and consequent reduced overall growth.

Babalola and Lal (1977b) evaluated the effects of various gravel

sizes and mixtures and the effects of modifying a natural gravelly soil

on corn-seedling growth. Increased gravel size (4 to 8, 8 to 15, and 15

to 40 mm) decreased shoot weight, root weight, root depth, and overall

root length of 7-d-old seedlings harvested from soils of varying gravel

concentrations. Field studies were performed on a naturally occurring

gravel horizon following removal of the overlying 15 cm of surface soil.

Treatments included gravel horizon undisturbed; gravel horizon removed

and repacked at a lower bulk density; gravel horizon removed, sieved to

remove gravel, and repacked as only the fine fraction; and gravel

horizon removed and area repacked with the original surface soil. All

treatments with gravel had seedling emergence delayed 1 to 2 d. The

treatments did not produce differences in shoot height, shoot dry

weight, or root dry weight. In comparison to the undisturbed treatment,

reduction in bulk density, accomplished by removal and repacking of the

soil, increased root length and rooting depth by 50%. Removal of the

gravel and substitution of surface soil for the subsoil increased root

length and rooting depth by 80% over the undisturbed control. There

were no differences in root length, root depth, or shoot weight between

the subsoil without gravel and the replacement of subsoil by surface

soil. Roots in gravelly horizons exhibited an increased mean diameter,

stunted tips, and marked crookedness. Although the gravel had no impact

on dry-weight yields in this short 7-d trial, the stunted growth and

limited access of the roots to soil would probably have detrimental

repercussions on full-season, plant-yield components.

The influence of gravel on the moisture characteristics of the

whole soil results from the quantity of gravel, its arrangement in the

soil fabric, and its own hydrologic properties. Several researchers

have attributed the scarcity of information regarding water relations in

gravelly soils to difficulty in adapting standard laboratory and field

techniques to gravelly soil, which display a high degree of micro-

variability within repetitive samples (Reinhart, 1961; Hanson and

Blevins, 1979).

Experiments using drastically-disturbed and mixed-gravel soils

will be discussed and distinguished from those of naturally occurring

soils with gravel horizons. Miller and Bunger (1963) and later Unger

(1971a and 1971b) constructed soils with "pea gravel" horizons to study

water infiltration and redistribution. In all treatments of the three

studies, screens or special repacking techniques were used to prevent

soil from filling the interstitial spaces of the gravel horizons. These

studies showed that the gravel slowed downward percolation, and for all

practical purposes, prevented upward redistribution of soil water. The

behavior of these soils should probably not be extrapolated to soils of

the tropics with naturally occurring gravel horizons, in which a fine-

mineral fraction occupies the inter-gravel space and provides a

continuum of fine pores that can participate in the redistribution of

soil water.

Babalola and Lal (1977a) reported soil moisture-release curves for

the sieved, gravel-mixed soil used in their previously reported studies.

They showed an incremental decrease in soil-water content at tensions of

0 to 60 cm of water for each incremental increase in gravel

concentration from 0 to 75%. They concluded that, as gravel

concentration and, therefore, total solids increased, porosity and

consequently water-holding capacity decreased.

Ghuman and Lal (1984) studied differences in field-water

percolation and redistribution rates on a tropical Alfisol under

conventional plowing and no-till management. The soil had a naturally

occurring gravel horizon from the 10 to 80-cm depth that contained about

45% gravel by weight. Soil having an initial water content of 0.035

cm3/cm3 exhibited infiltration rates of 43 and 120 cm/h for the

conventional and no-till systems, respectively, upon application of 5 cm

of floodwater to the surface. The infiltrating water under both tillage

treatments reached the 30-cm depth before flood conditions ceased, at

which time the plots were covered to prevent surface evaporation.

Within 1 h the water had passed the 80-cm depth. The redistributing

soil water had stabilized after 5 h and the soil water content with

depth remained constant until cessation of observations at 48 h. Higher

initial soil-water contents resulted in slower infiltration rates. Even

under very dry conditions, the gravel horizon did not prohibit the

downward movement of infiltrating water.

Research Topics

The unique physical properties and generally unknown behavior of

tropical, stone-line soils lead to many questions regarding their

agricultural management. However, extrapolation of properties and

behavior of soils that simply contain stones can lead to the development

of management practices based on incorrect assumptions. Soil scientists

have historically witnessed the difficulty of transferring management

techniques developed for temperate soils to soils of the tropics

(Swindale, 1980). Research into the properties and consequent behavior

of a soil is generally considered the most sound approach for the

development of management practices (Dudal, 1980).

Vine and Lal (1981) concluded that gravel reduces volumetric-

moisture content, reduces nutrient-retaining capacity, and retards

plant-root development. The extent of such effects is related to the

properties of the gravel. The porosity of the gravel will influence the

degree of any reduction in soil-moisture content, and rainfall patterns

will influence the degree to which soil moisture becomes detrimental for

plant growth. Although Flint and Childs (1984) have demonstrated that

gravel can hold up to 40% of available water in temperate forest soils,

and Muller and Bocquier (1986) and Amouric et al. (1986) have

photographed voids in the gravel from tropical, stone-line soils from

both Cameroon and Senegal, the porosity and water-holding capacity of

tropical, stone-line soils have not been established. Babalola and Lal

(1977a and 1977b) and Ghuman and Lal (1984) made no mention of water

retention by the gravel in their studies of water relations in tropical,

stone-line soils.

Reductions in nutrient-retaining capacity result from volumetric

reductions in the soil's fine fraction with the increase in volume of

gravel. The fine fraction typically contains greater surface area and

organic matter, and consequent nutrient-retaining charge. However,

porous gravel may harbor weakly held mobile nutrients. Studies in soil

physics have firmly established the presence of immobile-water regions

in the fine porosity of soil aggregates (Kirda et al., 1973; van

Genuchten and Wierenga, 1977; Rao et al., 1980a).

Considerable evidence has shown that gravel contents above 10 to

20% by weight have deleterious effects on root development and soil

penetration (Babalola and Lal, 1977a and 1977b; Vine and Lal, 1981).

Regardless of gravel hydrologic properties, the gravel limits rooting

depth and, therefore, limits the volume of soil from which the plant can

extract immobile nutrients.

In light of the previously mentioned considerations, any research

effort with respect to the development of management practices may be

most productive if the research is designed to determine the combined

behavior of the processes and their combined effects on soil

productivity, instead of investigating separately the numerous

interdependent processes.



In the late nineteenth century, Russian earth scientists

introduced the concept of soils as independent natural bodies, each with

unique morphology resulting from a unique combination of climate, living

matter, earthy materials, relief, and age (Buol et al., 1980). Since

that time, characterization of soil morphological, physical, and

chemical properties has played a fundamental role in the development of

soil taxonomic and resulting classification systems (Soil Survey Staff,

1975). Data characterizing soil properties, and subsequent taxonomic

classification of the soil are useful tools in the development of a

unified concept of soil behavior in its natural environment (Sanchez et

al., 1982b). The purpose of this chapter is to describe the physical

and chemical properties, and to taxonomically classify, the soil used in

the experimentation discussed throughout this dissertation.

Materials and Methods

The field site was located 1 km north of the Leppo primary

school and 60 m west of the road passing from Dschang to Djuttitsa

through the Leppo quarter of the village of Bafou, in the Western

Province of the Federal Republic of Cameroon. Following the field

experiments (Chapter 4), a 2-m deep pit was excavated in the middle of

the site and the soil profile was described (Appendix A). Samples

from each horizon were taken for physical and chemical analysis. Soil

texture was measured by the pipette method (Gee and Bauder, 1986).

Bulk density, porosity, and moisture-retention characteristics were

determined from undisturbed soil samples in 5-cm long by 5-cm

internal-diameter cores (Klute, 1986). Mineralogy of the fine

fraction (<2 mm) was determined by x-ray diffraction following removal

of organic matter with hydrogen peroxide and removal of noncrystalline

material with ammonium oxalate in the dark (Kunze and Dixon, 1986).

Exchangeable Ca+2, Mg+2, K%, and Na* were extracted from the fine

fraction with 1 M NH4OAc at pH 7 and determined by atomic absorption

spectrophotometry. Exchangeable H' and Al+3 were extracted with 1 M

KC1 and determined by the titration procedure of Yuan (1959). Organic

matter was determined by the modified Mebius procedure (Nelson and

Sommers, 1982). Phosphorous adsorption isotherms were determined for

both fine-fraction (<2 mm) and gravel (>2 mm) samples of the Ap

horizon, using the method of Fox and Kamprath (1970).

Gravel was sieved from the fine fraction, washed, and then

separated according to its four predominant colors. Mineralogy of the

gravel separates was determined by x-ray diffraction of powder mounts

following pulverization with a ball mill. Porosity of the gravel was

determined by the Brunauer-Emmett-Teller (BET) method on a

Quantachrome AUTOSORB-6 surface-area unit. Gravel-particle density

was determined using a gas pycnometer (Danielson and Sutherland,

1986). The soil was classified in the USDA Soil Taxonomy (Soil Survey

Staff, 1975) system based on its morphological description and its

physical and chemical properties.

Results and Discussion

Geographical Location

The soil-profile description is presented in the Appendix. The

soil rests on a 12 to 16% convex slope. The surrounding landscape is

steeply dissected and supports numerous small fields cropped to mixed

cultures of corn (Zea mavs L.), beans (Phaseolus vulgaris L.), cocoyams

(Colocasia esculenta), coffee (Coffea arabica L.), and peanuts (Arachis

hypooaea L.). The soil is derived from a basal parent material of

basalt, along with surface deposits of volcanic ash. The profile is

well-drained and has a lithologic discontinuity between the Btc and 2BCt

horizons, where a gravel horizon meets a buried clayey horizon.

Soil climate at the weather station of the Institute of Agronomic

Research in Dschang, 8 km south of the field site, is udic

isohyperthermic. The soil's control section (50 to 100 cm) is dry no

more than 90 d/yr, and maintains a mean annual temperature greater than

22 OC with less than a 5 OC fluctuation from the warmest to coolest

temperature at a depth of 50 cm.

Physical and Chemical Properties

Selected physical properties are shown in Table 3-1. The gravel

content of the top 72 cm ranges from 33 to 72% by weight. The fine

fraction is dominated by clay and composed of kaolinite, quartz,

goethite, and gibbsite. Selected chemical properties are shown in Table

3-2. Contents of organic carbon and exchangeable bases are calculated

on the basis of the fine fraction only. Trace quantities of acidity

were extractable, but never exceeded 0.02 cmol(+)/kg of soil for any


Selected physical properties of the soil.

Bulk Fine fraction Clay minerals
Horizon Depth density Gravel Sand Silt Clay K Q GB GE

cm g/cm3 kg/kg -- kg/kg of <2 mm -- ---- % -----

Ap 0 11 0.88 0.335 0.211 0.355 0.435 55 15 15 15

Ac 11 22 1.00 0.588 0.111 0.403 0.486 60 10 15 15

Btc 22 72 1.46 0.720 0.112 0.337 0.551 60 5 20 15

2BCt 72 138 1.26 0 0.118 0.145 0.737 60 5 15 20

2CB 138 194+ 1.28 0 0.064 0.193 0.743 55 5 20 20

t K = Kaolinite Q = Quartz GE = Goethite GB = Gibbsite

Table 3-2. Selected chemical properties of the fine fraction (<2 mm)
of the soil.

Organic Extractable basest Extract. acidity Sum of pH
Horizon carbon Ca Mg K Na NH4OAct KCIt bases H20 KC1

g/kg ------- cmol (+)/kg fine fraction ---------

Ap 68.5 7.2 3.1 0.27 0.04 30.1 trace 10.6 5.33 4.76

Ac 57.0 4.5 2.6 0.13 0.03 22.9 trace 7.3 5.12 4.52

Btc 20.6 1.4 1.7 0.03 0.02 10.6 trace 3.1 4.98 4.85

2BCt 11.0 2.1 2.2 0.03 0.02 10.0 trace 4.3 5.52 5.43

2CB 7.5 3.6 2.4 0.05 0.05 8.4 trace 6.0 5.61 5.46

t extracted with 1 M NH OAc (pH 7.
f extracted with 1 M KC1.
less than 0.02 cmol (+)/kg.

Family designation: clayey-skeletal, oxidic, isohyperthermic,
Typic Gibbsiorthox


Table 3-1


The gravel is a composite of four visually distinguishable classes

composed of goethite, gibbsite, kaolinite, and an unidentified mineral

containing Mn (Table 3-3). The porosity of the gravel ranges from 0.13

to 0.32 mL/mL, with a natural-composite sample porosity of 0.2 mL/mL.

The moisture-release curve for the top 72 cm of soil is shown in

Fig. 3-1. The soil-water content exhibited no initial plateau at low

tension, thereby suggesting the presence of some very large pores that

are full when the soil is saturated, but which drain under relatively

low tensions. The soil lost nearly 0.2 mL of water per cm3 of soil

between saturation and 350-mbar tension (hypothetical field capacity).

It maintained 0.14 mL of water per cm3 of soil between 350-mbar and 15-

bar tension (hypothetical plant-available water).

Phosphorus adsorption isotherms are presented in Fig. 3-2. The Ap

horizon exhibits a strong affinity for P, and required nearly 500 ug P/g

soil (750 kg P/ha to a depth of 15 cm) to support a solution

concentration of 0.2 ug/mL. The gravel displayed a low affinity for P.

Taxonomic Classification

The Ap and Ac horizons constitute an umbric epipedon. The

epipedon has weak, medium, subangular-blocky structure that breaks to

moderate crumb. The color has a moist Munsell value and chroma darker

than 3.5. The organic-carbon content is greater than 2.5%, and the

depth of the epipedon is greater than 18 cm. Base saturation as

measured by 1 M NH4OAc at pH 7 is less than 50%.

The Btc, 2BCt and 2Cb horizons constitute an oxic horizon. This

horizon is at least 30-cm thick. The cation-exchange capacity using

NH4OAc (pH 7) is less than 16 cmol(+)/kg clay. There are no more than

Table 3-3. Physical and mineralogical characteristics of the gravel.

Color Predominant Pore Average Particle Bulk Porosity
minerals volume pore radius density density
mL/g nm g/mL g/mL mL/mL

Yellow Geothite (90)t 0.145 3.62 3.31 2.24 0.32
Pink Kaolinite (60) 0.184 14.40 2.55 1.74 0.320
(8) Gibbsite (20)
Red Gibbsite (80) 0.087 8.69 2.63 2.14 0.186
(75) Kaolinite (10)

Black Manganese 0.043 4.91 3.55 3.08 0.133
(5) oxides

Composite -- -- 2.71 2.18 0.195

t Approximate percentage of mineral content.
t Percentage content of natural-composite total

(by mass).


S05 0 0.526 8.20x10-2 log(T)
R2. 0.998**

. ...... ....... .. ........ ,
L 0.3

o 0.1

0 4

0 100 200 300 400 15000


Fig. 3-1. Soil moisture-release curve.

Ap Horizon
S = 1420 C939
800 =0.8
R2 =0.89*





S = 106 C1.30
R2 0. 91*

0.1 1.0 10.0

SOLUTION P, C (ug/mi)

Fig. 3-2. Phosphorus adsorption isotherms for the gravel and the Ap

trace quantities of weatherable primary aluminosilicates. The texture

is finer than sandy loam and the horizon has more than 15% clay, with no

or very few clay skins.

The soil is an Oxisol because 1) the oxic horizon in the top 2 m,

2) there is no plaggen epipedon, and 3) there is neither an argillic nor

natric horizon above the oxic horizon. The soil is in the Orthox

suborder, because it has no continuous phases of plinthite within 30 cm

of the surface, is not saturated with water at any time during the year,

has neither a torric nor an ustic moisture regime, and has less than 16

kg of organic carbon per square meter to a depth of 1 m. The soil is in

the Gibbsiorthox great group, by virtue of the presence of a horizon

within 1.25 m of the surface that contains 20% or more by volume of

gravel-sized aggregates that contain 30% or more of gibbsite. This

Gibbsiorthox is in turn Typic, because the gibbsitic gravel is within 50

cm of the surface and there are no mottles in the upper 1 m of the soil.

The particle-size class of the soil is clayey-skeletal, because

gravel makes up 35% or more by volume and the fine earth contains 35% or

more clay by weight. The mineral class is oxidic, because the soil

contains less than 90% quartz and less than 40% each of hydrated

aluminum (reported as gibbsite or bohemite) and iron oxides extractable

by citrate-dithionite, and the sum of the percentages of these two

mineral groups divided by the percent clay is greater than 0.2.

Therefore, the family designation for the soil is clayey-skeletal,

oxidic, isohyperthermic, Typic Gibbsiorthox.



A model is the representation of a form or process in an

alternative media. In modern science, chemical and physical processes

are modeled by representing behavioral processes with mathematical

relationships based on empirical and theoretical concepts. Models of

natural systems are frequently quite complex, because numerous

interrelated processes must be considered.

The ultimate goal in the conceptual development of a model is the

integration of mathematical relations that represent the true

mechanisms of the natural process. However, mechanistic approaches are

limited by insufficient understanding of processes and/or their

interactions. The limitations take the form of unverifiable

assumptions and exclusion of known but seemingly insignificant factors.

In lieu of mechanistic descriptions, processes may be lumped such that

the mathematical expression reflects the relation of several different

and detailed processes. Such a deterministic approach is advantageous

when the effects of a process can be modeled but the actual mechanisms

are unknown, or when a true mechanistic model requires extensive

characterization of the modeled media.

The value of a model lies in its ability to simulate the natural

process from measured or estimated parameters that characterize the

natural setting. Although concurrence of model-simulated and

independently derived parameters does not prove the correctness of the

model's underlying theoretical basis per se, overall confidence in the

model's theoretical basis is increased as concurrence continues to

exist under a variety of characterized conditions. Increased

confidence allows greater use of the model for purely predictive and

managerial purposes.

Numerous models have been proposed for describing solute

transport in aggregated porous media. Modeling solute transport in

aggregated or structured soils presents some unique problems due to the

complex three-dimensional nature of the inter-connected network of

irregularly sized and shaped soil pores. Attempts to model

displacement processes quantitatively have been based generally on the

convective-dispersive equation (Lapidus and Amundson, 1952),

ac/at = D a2C/az vo ac/az [4-1]

where C is the concentration (mg/mL), D is the dispersion coefficient

(cm2/day), vo is the pore-water velocity (cm/day), z is the distance

(cm), and t is time (days).

Adsorption of the solute to the porous

media may be considered using an adsorption coefficient derived from a

linear adsorption isotherm, defined by

S = KC [4-2]

where S is the sorbed solute concentration (mg/g), C is the equilibrium

solution solute concentration (mg/mL), and Kd is the adsorption

coefficient (mL/g) giving

R ac/at = D a2/az2 vo ac/az [4-3]

where R, the chemical retardation factor, is defined by
R = 1 + pKd/. [4-4]

where p is the soil bulk density (g/cm3) and 0 is the volumetric water

content (mL/mL). Eq. 4-3 can be rearranged to include the

dimensionless parameters:
T = vo t/L [4-5]
x = z/L [4-6]

P = vo L/D and [4-7]

C = C/C, [4-8]

where vo, t, z, and D have been previously defined and L is the column

length (cm), P is the Peclet number, T is the pore volumes of solution,
x is dimensionless distance, and C the ratio of effluent concentration

(Cb) to influent concentration (Co) to give the convective-dispersive
(CD) model,
R(aC/aT) = (1/P)(a2c/ax2) ac/ax [4-9]

The CD water-flow model has been used satisfactorily to simulate

nonadsorbed solute transport under laboratory and field conditions

(Nielsen and Biggar, 1961; Warrick et al., 1971). However, the model

has been relatively poor at simulating solute transport through well-

aggregated and structured soils (Green et al., 1972; Rao et al., 1974;
van Genuchten and Wierenga, 1976 and 1977).
Solutions of Eq. [4-1] predict nearly sigmoidal or symmetrical
concentration distributions (Coats and Smith, 1964; Gershon and Nir,

1969; van Genuchten and Wierenga, 1976). However, numerous
experimental studies have shown distinctly asymmetrical effluent curves

(Nielsen and Biggar, 1961; Biggar and Nielsen, 1962; Green et al.,

1972; van Genuchten and Wierenga, 1977). It was noted that this
asymmetry or tailing of effluent curves was more pronounced in

aggregated versus nonaggregated media and as solution velocities
increased. Coats and Smith (1964) hypothesized the existence of

regions of immobile water in small and dead-end pores. They modified

Eq. [4-1] to incorporate solute transfer by diffusion from mobile-

flowing water regions to stagnant immobile water regions, to give

0m (aCJaT) + Om (ac,/at)
= OmD, (a2Cm/az2) vmO, (aC,/dz) [4-10]


aim (ac,.Dt) = a(C, Cim) [4-11]
where 9m and 9im are the fractions of the soil filled with mobile and

stagnant water, respectively (cm3/cm3); Cm and Cim are the solute

concentrations (g/mL) in the mobile and immobile regions; vm is the
average pore-water velocity in the mobile region; Dm is the mobile-

water dispersion coefficient; and a is a mass-transfer coefficient


van Genuchten and Wierenga (1976) have extended this model to

account for solute adsorption to the porous media through the inclusion

of an adsorption coefficient in the retardation factor. To account for

the possibility of unequal distribution of adsorption sites between the

mobile- and immobile-water regions, f is defined as the fraction of
sites in the mobile region. Including these concepts in the model of
Coats and Smith (1964), van Genuchten and Wierenga (1976) derived

(0m + fpKd) aCm/at + [lim + (1 f)pKd] aci/at
= OmDm (a2caz2) vmOm (acwaz) [4-12]


[Oi, + (1 f)pKd] ac/at = a(C. C). [4-13]
The model may be described in terms of the dimensionless parameters

Peclet number, P; the mobile water partition coefficient, fl; and the

dimensionless, mass-transfer coefficient, w, defined as:

P = vmL/D, [4-14]

R = (0m + pfKd)/(e + pKd) and [4-15]
w = aL/q [4-16]

Additionally, the concentrations of the solutes in the two regions (Cm
and Cim) may be normalized with the original-solute pulse

concentration, CO by defining

c1 = CJ/C [4-17]


c2 = Cm/Co [4-18]

With these definitions of P, 0, w, c,, and c2, Eqs. [4-12] and [4-13]

fR (aC,/aT + (1 P)R aC2/aT = (1/P)(a2c1/ax2) ac,/ax [4-19]

(1 f)R aC2/aT = w(C, C2) [4-20]

The mobile-immobile model (MIM) (Eqs. [4-19] and [4-20]) contains

four dimensionless parameters; P, R, f and w. Agreement between model

simulation and experimental data is generally accepted as verification

of the conceptual basis of the model. However, experimental methods

are generally unavailable to measure f and w independently. When

experimental techniques are inadequate to measure parameters

independently, they are frequently estimated on the basis of a best-fit

of the model to experimental data (van Genuchten et al., 1977; Rao et

al., 1979; Nkedi-Kizza et al., 1983 ). van Genuchten (1981) has

developed a non-linear least-squares, curve-fitting computer program

that estimates MIM and CD parameters from miscible displacement

effluent data. Although such a technique is useful for parameter

estimation, it does not ensure process identification (Davidson et al.,

1980; Rao et al., 1980a).

Independently estimated model parameters for soil and synthetic

porous media have demonstrated slight deviations from those parameters

estimated from curve-fitting procedures based on the MIM. Rao et al.

(1980b) performed miscible-displacement experiments on fabricated media

consisting of mixtures of porous ceramic spheres, glass beads, and fine

sand. Parameters calculated by the MIM curve-fitting program were

compared to those experimentally measured or independently estimated

for the various mixtures (Rao et al.,1980a). Over a broad range of

pore-water velocities, close agreement was found between values

estimated by MIM curve-fitting to those independently determined.

Owing to the ease of utilization, unavailability of accurate

conclusive methods to determine some model parameters experimentally,

and otherwise general agreement between experimentally determined and

model-estimated parameters, the MIM has become a popular tool for

estimating soil-water behavioral characteristics. Seyfried and Rao

(1987) used the model in a study to examine the relative contributions

of soil-water characteristics to leaching in an aggregated tropical

Typic Dystropept derived from volcanic ash. Field studies monitoring K

movement were not successfully simulated by a simple convective-

dispersive water model (Seyfreid, 1986). Miscible-displacement

experiments on saturated soil columns and subsequent analysis of

effluent data by the MIM model estimated the mobile-water content at

about 55% of the total soil water. Low Peclet numbers and consequent

high dispersion coefficients indicated a high degree of preferential

water flow that bypassed large portions of the soil water.

Schulin et al. (1987) used the CD and MIM models to determine

behavior of water in undisturbed columns of soil containing about 55%

by volume gravel. Back-calculation of presented data indicated that

the gravel was not porous and contained no water. The MIM model

calculated the mobile-water content to be about 85% of the total water

present in unsaturated columns maintained at volumetric-water contents

of between 0.135 and 0.175 mL/mL for soils with total porosities

ranging from 0.25 to 0.30 mL/mL. Due to the low immobile-water

fraction, the CD model, which considers all water as mobile, was able

to estimate parameters capable of simulating the experimental BTC

nearly as well as the MIM model.

Several independently conducted studies have suggested that the

gravel resulting from mineral dissolution and precipitation in tropical

stone-line soils is porous (Amouric et al., 1986; Muller and Bocquier,

1986; Chapter 3.). Although the only apparent study on the

mobile/immobile-water content of gravelly soil indicates that the

immobile fraction is relatively small (Schulin et al., 1987), the

presence of porosity in gravel from tropical stone-line soils would

suggest that these soils may have a considerable immobile-water

fraction. The purpose of this study was to use the MIM and CD models

to evaluate tritiated water breakthrough curves from an aggregated

gravelly Oxisol, to determine if preferential water flow and immobile-

water regions should be considered when describing nutrient-leaching

behavior for this soil.

Materials and Methods

Column Preparations

Undisturbed soil columns were taken from unfertilized plots at

the previously described experimental site at the end of the growing

season. A soil-core sampler was constructed of a steel water pipe (102

mm i.d./ 114 mm o.d.) fitted with a sharp, hardened-steel, cutting edge

and a removable, threaded steel cap. An 80-cm length of PVC pipe was

inserted into the steel corer so that one end rested on a 1-mm wide

shelf at the base. The whole piece was held in place by tightening the

cap. The sampler was hammered into the ground until the top of the cap

was nearly level with the soil surface. The sampler was then lifted up

and out of the soil with a hydraulic jack. The PVC pipe full of soil

was removed from the sampler, sealed, and boxed for transport to the

laboratory in Gainesville. Excess PVC pipe was cut from the top of

each column so that the new end was 5 mm above the soil surface.

Approximately 5 mm of soil was removed from the bottom of the columns

and both ends were fitted with porous, fritted-glass plates (maximum

pore radius of 15um) and plexiglass end plates.

Miscible Displacement

The columns were held vertically and saturated from the bottom with

approximately 5 pore volumes of a degassed solution of 0.01 M CaCl.


The columns were then turned horizontally and the end plate in contact

with the surface Ap horizon was connected to an influent solution by a

three-way valve which allowed switching between tritiated and

nontritiated solutions of 0.01 M CaCl2 (Fig. 4-1). Effluent was

collected in a fraction collector from the other end of the column.

The 3H20 activity in the effluent fraction was monitored using liquid-

scintillation techniques. The resulting breakthrough curves were

fitted to the Convective-Dispersive (CD) and Mobile-Immobile (MIM)

transport models using the program CFITIM3 (van Genuchten, 1981), which

is based on a nonlinear, least sum of squares criteria for goodness of

fit. Boundary conditions assumed for the model analyses were constant

influent-solute concentration and a semi-infinite column.

Adsorption Isotherms

Adsorption isotherms for 3H20 were determined using a batch

technique similar to that described by Dao and Lavy (1978). Sieved (<2

mm) soil samples from each of the three soil horizons present in the

column, and a composite gravel sample (2 to 4 mm), were assayed. Moist

triplicate 4-g samples of each material were placed in a pre-weighed

10-mL plastic, screw-top centrifuge vial that had a 1-mm hole drilled

in the bottom. The vials were sealed and reweighed. Solution having

varying activities of 3H20 were injected into the basal hole until the

materials appeared near saturation. The vials were reweighed and then

placed on top of a glass marble resting on the bottom of a 30-mL

plastic, screw-top centrifuge tube. These larger tubes were then

sealed and set on their sides for 48 h to allow for equilibration of

the tritium throughout the sample. The 30-mL tubes were centrifuged at

Soil Column

3 Way Switches


S- U U
iip J



Fraction Collector

0.01 M CaCl2

Tritiated 0.01 M CaCl

Fig. 4-1. Schematic illustration of apparatus used in the
miscible-displacement experiments.


30 times the force of gravity, forcing the soil solution out of the

soil, and through the basal hole to be collected around the marble in

the bottom of the larger tube. The extruded solution was retrieved and

the 3H20 activity was measured. The soil sample was dried at 105 C

for 48 h and weighed. The adsorbed 3H20 was determined by subtraction

of the 3H20 activity in the extruded solution from the initial 3H20 in

the injected solution after accounting for the original water content

of the samples. Adsorption isotherms were constructed by plotting

adsorbed versus solution 3H20 activity. Linear-adsorption coefficients

were calculated using linear regression forced through the origin. An

overall soil-column retardation factor, R, was calculated using

weighted mean adsorption coefficients of the gravel and fine-fraction

samples from each horizon.

Results and Discussion

Description of Model Parameters

Information input into the non-linear, least-squares curve-

fitting program that optimizes dimensionless parameters for the CD and

MIM models consists of the observed tritium breakthrough curve (BTC),

which is composed of data pairs consisting of the pore volumes of

solution and the radioactivity of that solution relative to the

activity of the tritiated-pulse solution. The curve-fitting program is

capable of estimating the retardation factor, (R), Peclet number (P),

fraction of solutes in the mobile water region (f), dimensionless mass-

transfer coefficient (w), and tritiated-pulse volume (T). Confidence

in the predictive capacity of the model is improved as the number of

parameters which the model is required to predict decreases. The

curve-fitting procedure that estimates the model parameters from the

BTC, bases the parameter-selection process on the goodness of fit of a

model-predicted BTC with the observed effluent data. The model

calculates a 95% confidence interval for each estimated parameter;

however, the confidence interval measures the goodness of fit of the

estimated parameters to the effluent curve and does not involve any

consideration of random experimental error. Therefore, the final

estimation of soil-property parameters requires judicious

interpretation of the model-estimated parameters.

Several of the dimensionless parameters are measurable by

laboratory techniques. The retardation coefficient may be calculated

from an adsorption coefficient, KD, the water content, and the bulk

density. The tritiated-pulse volume may be measured during the

miscible-displacement process. Experimental methods to measure the

other three parameters, P, 8, and w, are generally unavailable.

The dimensionless parameters P and w are specific to the

particular conditions of the experiment from which they are derived.

The Peclet number relates [Eq. 4-14] the column's length and pore-water

velocity to the dispersion coefficient. Dispersion results from

physical mixing of soil water travelling at different velocities or

following different paths. The dispersion coefficient is an indicator

of soil-pore sizes and the pore-size distribution. Since the velocity

of water in a confined capillary is dependent on the capillary radius,

large capillaries can transport water more rapidly than smaller pores

under similar pressure gradients. The preferentially rapid transport

of water in large pores is called channelling, and results in solutes

travelling further and more rapidly than simple piston-displacement

concepts would allow.

The parameter f represents the fraction of solutes present in the

mobile region under equilibrium conditions. The mobile-water fraction,

0, may be calculated with Eq. 4-21:

0 = O/J = OR f(R-1) [4-21]

where 0 is the mobile-water fraction, Om is the mobile-water content, 8

is the total-water content, R is the chemical-retardation factor, and f

is the fraction of total adsorption sites in the mobile-water region.

The parameter f is typically approximated. Nkedi-Kizza et al.(1982)

argued that, since the surface area associated with a unit volume of

water in the small pores of the immobile region is probably much

greater than the surface area associated with a unit volume of water in

the mobile region, f may be approximated to be zero. However, NKedi-

Kizza et al. (1983) have also proposed equal distribution of the sites

between the two regions such that f = p and, therefore, 0 = #.

Seyfried and Rao (1987) proposed an intermediate approximation of f =

0/2. In all approximations, the severity of any error in the eventual

estimation of the mobile-water content is influenced by the value of R.

If there is almost no chemical adsorption or repulsion (R approaches

1.0), then the location of the sites becomes less important because the

value OR f(R-1) approaches both f and the mobile-water fraction, 0.

The dimensionless parameter, w, relates the mass-transfer

coefficient [Eq. 4-16] to column length and solution flux (volume per

time area). The mass-transfer coefficient is a lumped term including

both a tortuosity factor and a diffusion coefficient. Diffusion is the

transport of solutes from an area of high concentration to an area of

low concentration independent of any movement of the media. The

tortuosity of the media limits the exposure between a concentration

gradient. Both dispersion and diffusion will occur during water

transport through soil. The contribution of the dispersion process to

solute mixing is generally of greater magnitude than the diffusion

process, such that the diffusion process is frequently insignificant.

However, in soils that contain immobile-water regions, diffusion is the

only process that transports solutes into, through, and out of the

immobile regions. In such soils the magnitude of the diffusion process

becomes significant.

Parameter Estimation

Adsorption isotherms

The tritium-adsorption isotherms for the three column horizons

and composite gravel samples are presented in Fig. 4-2. A weighted

mean of the slope of the line obtained by plotting the adsorbed versus

solution concentrations of tritium was calculated considering the

depth and gravel content of each horizon. This column adsorption

coefficient, Kd, was applied to Eq. [4-4] with other column parameters

to calculate a retardation factor of 1.05 for both columns. This value

indicates that the tritium is slightly adsorbed to the soil and is

consistent with other values measured for soils of similar mineralogy

(Nkedi-Kizza, 1983; Seyfried and Rao, 1987).

Ap -11 cm

S 0.047C
2 **
r -.79**

Ac 11-22 cm





0 L



S =-0.013

4 2= 0.92*


0 10(-
0 10(





Btc 22-72 cm


S =0.031C

r2= 0. 93**



Fig. 4-2. Tritium adsorption isotherms for column horizons and
composite gravel.

S =0.052C

r 0. 94**

j 1


Column physical properties

Selected physical properties of the soil columns are presented in

Table 4-1. The saturated-water content, bulk density, and particle-

size distribution of the two columns exhibited only slight differences.

The Darcian flux and number of pore volumes applied to each miscible-

displacement experiment for both soil columns are shown in Table 4-2.

The solution flux varied from slowest to fastest by a factor of over

40. The tritium concentration in the column effluent was monitored

during both pulse injection and clearing.

CD model analysis

The parameters estimated by the CD model for the four

displacement experiments of Column I are shown in Table 4-3. The

tritium-pulse volume was held constant during each curve-fitting

process, but the retardation factor, R, and the Peclet number were

allowed to vary. The lowest retardation factor, R = 0.74, which was

estimated for the most rapid flux, Expt. I-i, implies chemical

repulsion of the tritium from some regions of the soil. Since the

retardation factor was measured (R = 1.05), the model-estimated lower

value of 0.74 is an indication of immobile-water regions that were not

in physical equilibrium with the mobile effluent, due to the short

residence time of the pulse in the soil. The CD model, which considers

all soil water to be mobile, was unable to describe the observed BTC of

Expt. I-1 when the value of R was fixed at 1.05 (Fig. 4-3). As the

experimental flux was decreased, the CD-estimated retardation factor

approached the measured value, and the CD model was able to simulate

Table 4-1. Dimensions and selected physical properties of
soil columns.

Soil or
column property Units Column I Column II

Surface area
Weight, oven-dry
Bulk density


Particle-size fractions
by mass
<2 mm
2-12 mm
12-75 mm
by volume
<2 mm
2-12 mm
12-75 mm

Particle density
<2 mm
2-12 mm
12-75 mm












t Intra-gravel porosity excluded.

Table 4-2. Set-up for tritium miscible-displacement
experiments on Columns I and II.

Column-Experiment Flux, q Pulse, T
no. t

(cm/d) (pore volume)

I-1 111 1.43
1-2 16.8 2.58
I-3 2.71 2.84
I-4 36.7 2.65

II-1 2.69 2.88
II-2 36.7 2.59

t Order of execution

CD water model optimized dimensionless parameters.

Experiment Flux Peclet number Retardation factor
no. q P R


I-i 111 1.4 0.74
(0.2)t (0.05)

1-4 36.7 1.0 1.02
(0.1) (0.06)

1-2 16.8 1.9 1.01
(0.2) (0.05)

1-3 2.71 4.0 1.12
(0.3) (0.02)

t Numbers in parenthesis (+) are 95% confidence intervals.

Table 4-3.

Expt. I-1
q 111 cm/d
0 Measured data points



1.4 0.74
- 0.82 1.05 (fixed)

0 2 4 6


Fig. 4-3. Measured and CD-simulated BTCs for Expt. I-1 with R
optimized or fixed at 1.05.





the observed BTCs (Fig. 4-4). At the slower flow rates, the pulse

resided in the column long enough to allow diffusion to bring the

mobile and immobile regions closer to physical equilibrium, thereby

masking the presence of an immobile-water region. Thus, at slow flux,

the conceptual assumption of the CD model, which considers all water to

be mobile, is falsely satisfied.

MIM model analysis

The dimensionless parameters estimated by the MIM model for the

four displacement experiments through Column I are presented in Table

4-4. The retardation factor for all of the trials was held constant at

the measured value of 1.05 during each curve-fitting process. The

Peclet number, B, and w were allowed to vary.

Although # is generally considered a constant for any given soil

sample, it showed a slight increase as the flux decreased. This

behavior is attributed to the inability of the model to distinguish

easily between the mobile and immobile regions when the flux is slow

enough to allow considerable diffusion between the two regions. This

implies that the best estimate of f is when the flux is infinitely

fast. Since the trial with the fastest flux exhibited the highest

degree of physical nonequilibrium (CD-model analysis), its MIM-model

analysis should yield the best estimate of f. Therefore, the BTC were

refitted to the MIM model holding # constant at 0.53 (Table 4-5).

In this study, f = 0.53 will be used as the value to approximate 0,

since R is very close to 1.0. The measured and MIM-estimated BTCs for

the four experiments from Column I are smooth, asymmetrical and show

Expt. 1-3

q = 2.71 cm/d
O Measured data points
-- 4.0 1.12

4.6 1.05 (fixed)



- I



Measured and CD-simulated BTCs for Expt. I-3 with R
or fixed at 1.05.





Fig. 4-4.


Table 4-4. MIM water model optimized

Experiment Flux, q P


I-1 111

I-4 36.7

I-2 16.8

I-3 2.71

t Numbers in parenthesis (+)





are 95%

dimensionless parameters.


1.05 0.53 0.20
(0.04) (0.05)

1.05 0.58 0.30
(0.07) (0.07)

1.05 0.61 0.38
(0.05) (0.08)

1.05 0.62 2.5
(0.07) (1.2)

confidence intervals.

Expt. I-1

q = 111 cm/d

SMeasured data points
MIM 2.9 1.05 0.53 0.20

0 2 4 6


Fig. 4-5. Measured and MIM-simulated BTCs for Expt. I-1 with R
fixed at 1.05.













Expt. 1-2

q = 16.8 cm/d
0 Measured data points
-- MIM 9.3 1.05 0.53 0.51

2 4 6


Fig. 4-6. Measured and MIM-simulated BTCs for Expt. I-2 with R
fixed at 1.05 and f fixed at 0.53.








Expt. 1-3

q 2.71 cm/d

0 Measured data points
P R # (J
MIM 8.1 1.05 0.53 2.80

0 2 4 6


Fig. 4-7. Measured and MIM-simulated BTCs for Expt. I-3 with R
fixed at 1.05 and f fixed at 0.53.

Expt. 1-4

q 36.7 cm/d
0 Measured data points
P R (J
- MIM 2.7 1.05 0.58 0.30

0 2 4 6


Fig. 4-8. Measured and MIM-simulated BTCs for Expt. I-4 with R
fixed at 1.05 and f fixed at 0.53.







Table 4-5. MIM water model optimized dimensionless parameters
with R and B fixed.

Experiment Flux, q P R B
no. fixed fixed

I-1 111 2.94 1.05 0.53 0.198
(0.55)t (0.052)

I-4 36.7 3.18 1.05 0.53 0.342
(0.24) (0.032)

1-2 16.8 9.28 1.05 0.53 0.510
(1.32) (0.036)

1-3 2.71 8.13 1.05 0.53 2.80
(3.52) (1.63)

t Numbers in parenthesis () are 95% confidence intervals.

very close agreement at both the fastest and slowest flow rates (Figs.

4-5, 4-6, 4-7, and 4-8).

The immobile-water fraction of 0.47 cannot be totally attributed

to the intra-gravel porosity. Since the volumetric-gravel content

(including intra-gravel porosity) of the column was 0.375 mL/mL, and 20

% of that was interior porosity, the intra-gravel porosity was only

0.075 mL/mL, or 14% of the total column porosity of 0.526 mL/mL. Even

if the intra-gravel porosity contains only immobile water, the

remaining immobile-water fraction (0.40) of the volumetric-water

content was associated with the aggregated fine-earth fraction of the

soil. Although the fine-earth fraction of the Ap and Ac horizons has a

weak-crumb structure, the Btc horizon has medium-sized, moderately

strong aggregates (Appendix A). Nkedi-Kizza et. al. (1983) have shown

that packed columns of sieved (2 to 4.7 mm), strongly aggregated, peds

from an Oxisol harbored over 50% of the total volumetric-water content

in immobile regions.

Schulin et al. (1987), using similar techniques, found an

equally small R = 1.12 but a f = 0.87 under unsaturated conditions.

Back-calculation of presented data indicated that the gravel was not

porous. The total porosity (0.255 mL/mL) was associated entirely with

the 20% by volume fine fraction. Extrapolation of their data to

saturated conditions, assuming the water volume between saturation and

the experimental water contents to be mobile, would yield a mobile

water fraction 0 = 0.92. Therefore, the difference in the volume of

the mobile-water regions of these two gravelly soils is more likely due

to differences in aggregate structure of the fine fraction than to

differences in gravel porosity.

The dimensionless parameters P and w are a function of

experimental conditions, and are used with their functional

relationships to determine soil-water properties. The Peclet number

relates the pore-water velocity and column length to the dispersion

coefficient (Eq. [4-14]). One theoretical exponential relationship of

the dispersion coefficient to the pore-water velocity is:

D = \ vmn [4-22]

where Dm is the hydrodynamic dispersion coefficient, X the

dispersivity, v, the mobile pore-water velocity, and n an empirical

constant. For most laboratory-displacement experiments involving

disturbed repackedd) soils, A is about 1.0 cm (van Genuchten and

Wierenga, 1986). For displacement experiments involving undisturbed

field soils, especially when aggregated, X can be one or two orders of

magnitude higher. The degree of dispersivity in a soil is increased as

the pore-size distribution in the mobile-water regions becomes broader.

The dispersivities of different soils are more easily compared if the

empirical constant, n, is assumed to be 1.0 and the equation is linear.

Schulin et al. (1987) and Russo (1983) determined dispersivities of

2.24 and 2.91 cm from soils containing 55 and 43% gravel by volume,

respectively, using a linear relationship. The Peclet numbers

estimated from the BTCs of Column I tended to increase with decreasing

flux, suggesting a nonlinear relation between dispersion coefficient

and pore-water velocity within the velocity range used in this study

(Fig. 4-9). The nonlinear plot provides a dispersivity of


12000 -



6000 -

4000 -

2000 -

Dm = 3.29 Vm3
R = 0.96*

100 200 300 400


Fig. 4-9. Relationship between MIM-estimated dispersion
coefficient, Dm, and mobile pore-water velocity, Vm.





3.3 cm2"n dn'1 with n = 1.3, which is of a magnitude similar to values

from the two previously mentioned studies on stony soils. Several

factors may be contributing to the high dispersivity of this soil.

Dispersion increases as the range of small to large pore sizes in a

soil increases, thereby providing a wide range of water velocities

within a single soil sample. Edwards et al. (1984) have shown that the

presence of non-porous gravel increases the total macropore volume at

the expense of the micropore volume. The large reduction in the water

content of the soil under slight tensions (0 to 50 mbar) indicates that

the soil possesses a considerable volume of large pores (Fig. 3-1).

Similarly, the retention of nearly 33% of the total soil water at 15

bars of tension indicates that the soil also contains a large volume of

very small pores.

The values of the dimensionless parameter w, estimated for each

experiment on Column 1, is presented in Table 4-5. It relates the

mass-transfer coefficient, a, to the column length and solution flux

(Eq. [4-16]). The mass-transfer coefficient is a lumped diffusion

parameter that relates the solute diffusion transfer to the molecular

diffusion coefficient, the mobile/immobile-water fraction, the

tortuosity, the radius of soil aggregates, and the solution flux. The

relationship between the mass-transfer coefficient and the solution

flux is shown in Fig. 4-10. The mass-transfer coefficient is not a

constant and has been shown to increase with solution flux using both

theoretical postulations and experimental methods (Rao, 1980a and

1980b; van Genuchten, 1985). The mathematical description of the

diffusion process employed by the MIM model has an underlying


S= 2.0x103 (q) + 8.7x10-2

r 2= 0. 99**




0.1 -L



FLUX, q (cm/d)

Fig. 4-10. Relationship between MIM-estimated mass-transfer
coefficient, a, and flux, q.


assumption of first-order exchange kinetics. However, conceptually the

assumption is only valid for dead-end pores with a neck of negligible

volume (Coats and Smith, 1964). van Genuchten and Dalton (1986) have

shown that first-order kinetics represent a very close approximation in

the case of radial diffusive exchange between the soil matrix and

hollow, cylindrical macropores but, for other pore geometries, first-

order exchange is only a crude approximation. Because a is a lumped

parameter, it depends not only on the pore-space geometry, solute

diffusivity and the magnitude of the immobile region, but also on the

changing solute concentration within the two regions. The increase in

the mass-transfer coefficient with the increase in solution flux is due

to the rapidity at which the concentration gradient reaches its extreme

(more diffusive force) as a solute front approaches and leaves a given

point in a soil column.

Estimation of Column II Parameters

An additional enhancement to the validity of the estimated soil-

column parameters lies in their transferability to a different sample

of the same soil. Model-simulated BTC parameters derived from

experiments using Column I were applied to the experimental data from

Column II (Figs. 4-11 and 4-12). The values of R and # were set at

1.05 and 0.053, respectively, and the values for P and c were derived

from the line:- and curvilinear relationships presented in Figs. 4-9

and 4-10. The derivations included consideration of slight differences

in column lengths and bulk densities (Table 4-1). The simulated BTCs

based on the parameters derived from Column I exhibit later break

Expt. 11-1
q 2.69 cm/d
0 Measured data points
P R # J
- MIM 10 1.05 0.53 2.4
( all fixed)


Measured and MIM-simulated BTCs for Expt. II-1.



0.8 -











Fig. 4-11.








Expt. U-2
q 36.7 cm/d
0 Measured data points
P R #
- MIM 4.3 1.05 0.53 0.30
(all fixed)

0 2 4 6 8


Fig. 4-12. Measured and MIM-simulated BTCs for Expt. 11-2.

through and a higher peak concentration than the experimentally

observed BTCs. The differences between the two estimated and observed

BTCs could be due to natural variability between the two soil samples.

Although the two columns contained nearly identical volumes of gravel,

the particle density of the gravel in Column II was higher. Since the

more dense gravel had less porosity (Table 3-3), less of the total

porosity of Column II was associated with the gravel and a greater

portion is associated with the remaining fine-earth fraction. However,

considering that all model parameters were independently estimated, the

MIM-model-generated estimates closely described the observed

asymmetrical BTCs.


The data from this study show that this soil, which contains a

strongly aggregated fine fraction and porous gravel, produced

asymmetrical BTCs for tritiated water. The degree of asymmetry

increased with increasing flow rates. The classical CD model was found

to be inadequate in describing water movement in this soil, due to the

inability of the model to account for diffusive mass transfer of water

into stagnant or immobile-water regions. The MIM model adequately

described water movement at all flow rates, and estimated that about

50% of the total-water content was in immobile regions. The high

immobile-water content and the relatively large dispersivity indicated

that, under natural field conditions consisting of short but intense

tropical rain storms, water transport in the larger soil pores could


carry small amounts of unadsorbed solutes beyond the root zones,

whereas considerable quantities of the solute could remain relatively

unaffected, harbored in immobile-water regions. Although this could

cause pollution of ground-water from nutrients and pesticides leaching

through macropores, the presence of immobile-water regions in this soil

will act as a source/sink for solutes that will be slowly released to




An integrated knowledge of the behavior of plant nutrients in soil

has been a major interest of agricultural scientists. Even after many

decades of concentrated research, fine-tuning of management practices

for agricultural soils still generally requires experimentation based on

trial and error. The basis of our remaining lack of knowledge lies in

the complex interrelations of the plant nutrients among themselves, and

with plants, soil, water, and the atmosphere. The seemingly simple

concept of a nutrient being "plant-available" involves complicated

physical processes regulating chemical speciation among three phases,

two of which are readily mobile, and the third of which changes

continuously and irregularly with depth (Addiscott et al., 1986). In

addition to comprehension of these processes, there is also a problem

with instrumentation and quantification of measurements for specific

events and objects (Harris and Hansen, 1975). It is no wonder that the

most reliable method for assessing the availability of a soil-borne

plant nutrient is still a field trial followed by analysis of the

resultant plant material (Melsted and Peck, 1977; Sumner, 1987). What

such a technique loses in its overall contribution to the knowledge of

individual processes, is offset by immediate knowledge regarding on-site

agricultural behavior.

High concentrations of stones in the rooting zone of an

agricultural soil impact both water-holding and water-movement behavior

(Epstein and Grant, 1966; Ghuman and Lal, 1984). The relative impact of

gravel on any one soil property is dependent on the amount and

properties of that gravel. Generally, for gravel with little or no

porosity, increased gravel concentrations in the soil increase bulk

density, decrease total volumetric water-holding capacity, increase

macroporosity, decrease microporosity, decrease saturated hydraulic

conductivity and change the proportion of water held at various tensions

when compared to values for the same soil minus the gravel (Edwards et

al., 1984). Logical inferences concerning the agricultural productivity

of gravelly (non-porous gravel) soils that may be deduced from these

characteristics include:

1. Decreases in water-holding capacity increase droughtiness

and, therefore, make crops more susceptible to water stress.

2. Decreases in total porosity increase the amount of water

transported by the remaining soil pores.

3. Increased water transport increases the potential loss of

nutrients by leaching.

The effect of porous gravel on soil-water behavior is not easily

inferred and depends on the porosity and pore-size distribution of the

gravel (Reinhart, 1961; Hanson and Blevins, 1979). Porous gravel has

been shown to hold considerable portions of plant-available water (Flint

and Childs, 1984). Soil with gravel of porosity similar to that of the


fine fraction may still limit a high quantity of the transported water

to the pores of the soil's fine fraction, due to the noncontinuous

and/or small size of the pores in the gravel. The potential loss of

soil nutrients by leaching will be influenced by the differential

effects of pore size and pore location on water transport and

channeling. Therefore, water in porous gravel may act as a sink for

leachable nutrients and thus may harbor nutrients from convective-water


Soils with shallow, subsurface gravel horizons are a common

occurrence in the western highlands of Cameroon. Although these soils

are typically not preferred by local farmers, increasing population

pressures have resulted in their increased utilization for food-crop

production. Scientific studies indicate that, depending on the

quantities and properties of the gravel, different practices are

required for effective agricultural management of these soils. The

purpose of this experiment was to develop a basic understanding of the

dynamics of water and nutrient availabilities for a soil with a shallow

gravel horizon, throughout a crop-growing season and using both locally

prevalent and modified management practices. The objective of this

experiment was to differentiate between the relative effects of possible

water and nutrient stress on corn and beans, by subjecting the crops to

combinations of plant densities and seasonal nutrient availabilities.

Materials and Methods

The experimental site was located 8 km north of the University

Center of Dschang campus in the Leppo quarter of the village of Bafou,


Western Province of Cameroon, Africa. The field, rented from a local

farmer, was on a 12 to 16% slope and contained a dense gravel horizon

from a depth of 22 to 72 cm. The depth, thickness, and location of the

gravel horizon were first determined by augering and later confirmed by

soil-pit sampling. The field had been planted to corn, peanut, and

cocoayam the previous year following a 3 to 5-yr fallow, and had

received no commercial fertilizers for at least the previous 5 yr.

Weeds and former crop residues were cut by hand, aligned in the furrows,

and buried under newly established ridges based on a 1-m row spacing.

A randomized, complete-block design with 4- by 12-m plots and four

treatment replications was composed of a 2 by 5 factorial consisting of

two planting densities (intra-row mix of corn Zea mays L. CIMMYT Z-290;

and red bean Phaseolus vulgaris L.) and five fertilizer schemes (Table

5-1). The two planting densities were (1) 30,000 corn plants/ha mixed

with 40,000 bean plants and (2) 45,000 corn plants/ha mixed with 60,000

bean plants. The fertilizer treatments consisted of a non-fertilized

control and four split-application treatments; all consisting of 400

kg/ha of a locally available 20-10-10 (N-P,20-K20) mixed fertilizer plus

248 kg/ha of triple superphosphate (TSP) (50 kg P/ha). The four

fertilized treatments had the 20-10-10 material applied (1) all

preplant; (2) one half preplant and one half after 8 wk; (3) one third

preplant and one third after 4 and 8 wk; and (4) one fourth preplant and

one fourth after 4, 8, and 12 wk. All of the fertilized plots had the

TSP applied preplant. The preplant fertilizers were first mixed

together, spread in a 33-cm band down the center of the ridge, and

spaded to a 10-cm depth. All subsequent applications were applied to

Table 5-1. Description of experimental design.

Design : Randomized Complete Block
4 blocks
2 x 5 factorial

Factor 1. Plant density (within-row mix)

Level 1. 30,000 pl/ha corn Zea mays L.
40,000 pl/ha bean Phaseolus vulqaris L.

Level 2. 45,000 pl/ha corn
60,000 pl/ha bean

Factor 2. Fertilizer application timing

Fertilizer Time after planting, wk
level 0 4 8 12

0 -

1 M+P

2 1/2 M + P 1/2 M

3 1/3 M + P 1/3 M 1/3 M

4 1/4 M + P 1/4 M 1/4 M 1/4 M

M = 400 kg/ha

P = 50 kg P/ha

20-10-10 (N-P20 -K 0)
to 80-17-34 kg N-P-K /ha

as triple superphosphate

the soil surface in a 33-cm band down the center of the ridge.

The field site was planted 22 and 23 Mar. 1986, following

initiation of the rainy season in early March and upon the advice of

local farmers. The plots were planted initially to 1.5 times the

desired densities. Plots were thinned to proper densities after 3 wk.

Two of the four blocks were stripped and replanted after 4 wk, due to

low plant densities in several plots of each block. Weeds were

controlled by hand cultivation every 4 wk. Beans and corn were

harvested 75 and 140 d, respectively, after planting. Corn grain and

stover were analyzed for N, P, K, and Ca contents.

Soil samples were taken from all of the high-density plots at

depths of 0 to 5, 5 to 10, 10 to 15, 15 to 25, 25 to 35, 35 to 45, 45 to

55, 55 to 65, and 65 to 75 cm, just after corn harvest. Soil samples

were analyzed for gravel content and for Mehlich I-extractable P, K, and


Results and Discussion

Plant densities in several plots from two of the four blocks

were below required levels (Table 5-2). Analysis of variance of percent

plant emergence for the first planting of the four blocks is presented

in Table 5-3. The percent emergence was apparently not affected by

either planting density or fertilizer treatment, but was affected by

block location (Table 5-4).

Several factors may have contributed to the lower plant stands.

There was an uncommon lull in seasonal rains during the first 4 wk after

planting. This caused afternoon wilting throughout the field. The

Table 5-2. Plant emergence percentages for plots falling below required
levels (< 66%).

Planting Fertilizer Crop
Block density schedule Corn Bean

-- Emergence, % --

3 low 0 61 58
3 low 4 48 46
3 high 2 51 45
3 high 4 53 55
4 low 1 70a 57
4 low 2 69a 56
4 high 0 49 72a
4 high 3 58 60

Adequate > 66% emergence

Table 5-3. Analysis

of variance for early-season plant-emergence

Source D.F. Corn Bean
---F value ----

Block 3 6.27** 8.39**
Density 1 1.68 < 1
Fertilizer 4 < 1 < 1
Density*Fertilizer 4 < 1 < 1
Error 27
Total 39

C.V. 12.4 12.3

**99% level of probability

Table 5-4. Comparison of percent emergence of corn and bean by

Factor / Level Crop
Corn Bean

--- Emergence, % ---
1 80.8a* 81.2a
2 78.7ab 76.3ab
3 65.1c 62.3c
4 70.9bc 69.7bc

Means in the same column followed by the same letter are not
significantly different at the 95% level of probability, as
determined by Duncan's Multiple Range Test.


experimental site was on a west-facing slope. Blocks 1, 2 and a portion

of block 4 were on a 12% slope. All of block 3 and much of block 4 were

on a 16% slope. All of the plots with inadequate densities were on the

16 % slope. The slope of the land could have affected plant

establishment in two ways. The steeper areas would have received less

direct morning, but more direct afternoon sunlight. In addition, since

the field was laid out according to the sloping surface area and not the

level surface area, the plots on the most sloping land had the least

amount of soil beneath them. Therefore, the most sloping land probably

had the highest evaporative demand but the least quantity of soil from

which to draw water. In relation to later discussions, it should be

kept in mind that the lower plant densities did not constitute a drastic

failure (the lowest density was still 75% of that required), though they

were lower than the design of the experiment allowed.

Due to inadequate plant densities in some of the plots, all of the

plants in the two affected blocks were removed and the area was

replanted on 23 and 24 April, following additional rain 4 wk after the

first planting. The replanting changed the experimental design of the

study (Gomez and Gomez, 1984). An F test of the error mean squares from

the analysis of variance (Table 5-5) for the two planting dates was

performed for grain, stover, and total dry-matter yields (Appendix B).

In all cases, the error mean squares were not different. Consequently,

the data from the two sites were pooled and planting date was added to

the experimental design as an additional factor with two levels.

Because the planting-date levels were not randomized within the blocks,

but were instead imposed over complete blocks, a whole-block error term

Table 5-5. Analysis
two planting dates.

of variance of grain and stover yield for the

Planting Error Error mean
period D.F. mean square square ratio F,, F99

Grain yield

First 9 40442 2.21 ns 3.18 5.35
Second 9 18301
Stover yield

First 9 411685 1.66 ns 3.18 5.35
Second 9 247991

ns not significantly different

replicationss nested within dates) became the appropriate error term to

evaluate the effects of planting date on yield components. However, the

whole-block error term did not have sufficient degrees of freedom (<6)

to constitute a valid F test (Gomez and Gomez, 1984; Montgomery, 1984).

Therefore, the whole-block error term was pooled with either the three-

way interaction term, or the subplot pooled-error term, on the condition

that the newly added error term was not different from the whole-block

error term at the 75% level of probability.

Differences in environmental conditions during the two time

periods when the crops were in the field are impossible to assess. One

of the more obvious differences was in the quantity and distribution of

rainfall (Fig. 5-1). Seedlings in the first-planting period experienced

considerable wilting due to the slow-starting rainy season. Seedlings

of the second-planting had frequent early rainfall and showed no

wilting. Both plantings experienced frequent mid- and late-season

rains; however, the second planting received more total water because

the rainy season peaked in August and September, after the first

planting had been harvested.

The differential effects of climatic factors on grain and stover

yields may be attributed to the seasonal partitioning of plant

photosynthetic and mineral resources into different yield components.

Corn plants continue to increase in total dry-matter accumulation

throughout the season, until near harvest. However, once past silking,

most of the increase is due to grain filling. The dry-matter content of

other plant components remains relatively constant during this period

(Fig. 5-2)(Hanway, 1962). Tropical maize, in general, including the


800 Second Planting



200 Frst Planting

0 14 28 42 56 70 84 98 112 126 140


Fig. 5-1. Cumulative rainfall for the first and second planting

Dry matter(g/m2)



1200 -

1000 -

800 -

600 -

400 -

200 -

0 -r----

Days after sowing

Fig. 5-2. Total crop and grain dry matter accumulation for Tuxeno-1
and Pioneer 3369A Zea mavs ,grown at Tlaltizapan, winter cycle
1974, at 80,000 plants/ha (from Fisher and Palmer, 1983).

line used in this study, CIMMYT Z-290, is late maturing, tall, leafy,

and less efficient in translocating to the grain photosynthates which

were previously deposited in the stems and leaves (Evans, 1975).

Although grain yields are intimately related to early-season plant

health, differences in grain and stover yields may be attributed to

differential early- and late-season environmental influences (Fisher and

Palmer, 1983).

The effects of these combined factors on corn grain, stover

(above-ground portion of the plant minus the grain), and total dry-

matter yields are presented in Table 5-6). The effect of the two

planting dates was large and significant on corn-grain yield, but

insignificant on stover yield. The second planting yielded only 35% the

amount of grain of the first-planting treatment, even though the stover

yields for the two dates were nearly identical (Table 5-7). The lack of

interaction between the effects of planting date and fertilizer

scheduling on stover yields indicated that the fertilizer schedule

affected the non-grain, plant dry-matter accumulation similarly over the

two crop-growth periods. The interaction between planting date and

fertilizer schedule on corn-grain yields reflects the environmental

effects that planting date had on this indicator of late-season

conditions. The differences in total rainfall for the two plant-growth

periods increased as the season continued (Fig. 5-1). This difference

may be used to explain differential effects on the yield components.

During the early part of each growing season the difference in

rainfall and in subsequent probable nutrient leaching were less

pronounced. If one estimates the evapotranspiration and effective

Table 5-6. Analysis of variance for corn grain, stover, and total dry
matter yields.
Plant component
Source D.F. Error term Grain Stover Dry matter
------- F value------
Date 1 Rep(Date) / Date*Den*Fert 131**a
Rep(Date) / Pooled error < Ia 16.3**a

Rep(Date) 2 Pooled error 3.14 1.62 2.51
Density 1 Pooled error 9.50** 30.5** 32.2**
Fertilizer 4 Pooled error 141** 36.9** 67.2**
Density*Fert 4 Pooled error 2.13 2.11 2.44
Date*Density 1 Pooled error 4.37 < 1 < 1
Date*Fert 4 Pooled error 22.8** < 1 1.34
Date*Den*Fert 4 Pooled error 1.29 < 1 < 1
Pooled Error 18
Total 39

C.V. 12.2 16.4 13.3

Date = Planting date
Rep = Replication
Den = Density = Planting density
Fert = Fertilizer = Fertilizer application schedule

a Mean square error pooled to increase degrees of freedom in order to
enhance validity of the F test.
** 99% level of probability

Table 5-7. Comparison of selected main-effect yield-component means.
Plant component

Factor / Level Grain Stover Dry matter
------------------- kg / ha ------------
Planting date
First 2060A* 3630A 5690A
Second 7558 3370A 41208

Low 1320B 3000B 4320B
High 1490A 4000A 5490A

4 3710a** 5650a
3 3860a 5560a
2 4060a 5890a
1 4500a 5900a
0 1360b 1530b

Planting date by fertilizer interaction
Fertilizer First Second
4 2760a 1130a
3 2470ab 918b
2 2680a 984ab
1 2070b 714c
0 322c 26d

Planting date by density interaction (90% level of probability)
Density First Second
Low 1920b 728a
High 2200a 782a
M n in tha came rnliimn inrfr th samP shhbheadino and followed by

the same
level of

uppercase letter, are not significantly different at the 95%
probability according to an F test analysis of mean square

Means in the same column, under the same subheading and followed by
the same lowercase letter, are not significantly different at the 95%
level of probability according Duncan's Multiple Range Test.



rainfall (actual rainfall minus evapotranspiration) during the two crop-

growth periods, it can be demonstrated that little or no leaching of

soil nutrients occurred in the first 8 wk of either season (described in

greater detail in Chapter 5). Availability of nutrients would have been

affected by the fertilizer-application schedule, but their possible

early-season leaching would not have been affected by the planting date,

because of the early-season dry period.

Stover yields, an early-season indicator, showed decreasing (but

not significantly different) yields as the application of fertilizer was

distributed over time, but no differences due to planting date.

However, the late-season indicator, grain yield, was affected by the

fertilizer schedule and planting-date interaction. Grain yields from

the first planting showed no differences among the split-fertilizer

schedules. The all-preplant, fertilizer-application treatments yielded

less grain than the split-application treatments, but still considerably

more than the non-fertilized control. The grain yields of the late-

planted corn showed greater separation of means and greater differences

in magnitude among the split-application schedules. The 4 by 1/4 split

schedule outyielded the one preplant application, the 3 by 1/3 split

schedule, and the 2 by 1/2 split schedule. The 3 by 1/3 and the 2 by

1/2 split schedules in turn yielded more grain than the all-preplant

schedule, all of which outyielded the non-fertilized control.

These differences indicate that the greater effective rainfall

during the later growing period caused more leaching and thereby reduced

plant availability of nutrients between application schedules. Due to

the low magnitude of these yields in relation to yields from the first

planting date, and the lower grain yields in relation to stover yields,

the differences in grain yields for the second planting associated with

fertilizer-application schedule did not translate into differences in

overall dry-matter yields. Dry-matter yields among the four fertilized

treatments showed no differences, although they all out-yielded the non-

fertilized control by nearly four-fold.

Effects of the three factors on nutrient uptake by the corn at

harvest are presented in Tables 5-8 and 5-9. Total uptake of N, P, K,

and Ca was greater for the first planting date than for the second date.

Second-growth-period uptake for each of the nutrients was a relatively

constant 75% of the values for the first growth period, which is

consistent with the differences in total dry-matter yields. There were

no differences among the fertilized treatments for uptake of any of the

four nutrients, although all of the fertilized treatments had higher

uptake than did the unfertilized controls. This information supports

the yield data, in that there were no differences in uptake among the

fertilized treatments, whereas uptake among all fertilized treatments

was much greater than for the unfertilized controls. The non-

significant planting-date by fertilizer-application interaction term

indicates that the later planting date decreased total nutrient uptake,

but that nutrient uptake between fertilizer treatments within the same

planting date was similar.

The plant-density factor was included in the experimental design

as a means to detect the effects of water stress on yields. The

utilization of high plant densities to induce stress, or early harvest

to reduce plant densities and reduce stress, are common tools used to

Table 5-8. Analysis of variance for uptake of N, P, K, and Ca by corn
dry matter.
Plant nutrient
Source D.F. Error term N P K Ca

-------- F value --------

Date 1 Rep(Date)a
Rep(Date)/Pooled errorb 6.24* 8.42** 3.50** 4.75*
Rep(Date) 2
Density 1 Pooled error 18.7** 10.8** 7.73* 22.5**
Fertilizer 4 Pooled error 41.8** 92.8** 16.2** 19.7**
Density*Fert 4 Pooled error 1.34 2.05 < 1 1.46
Date*Density 1 Pooled error < 1 < 1 < 1 < 1
Date*Fert 4 Pooled error < 1 2.01 < 1 < 1
Date*Den*Fert 4 Pooled error < 1 1 < 1 2.46
Pooled Error 18
Total 39

C.V 16.2 12.3 27.9 23.7

Date = Planting date
Rep = Replication
Den = Density = Planting density
Fert = Fertilizer application schedule

*95% and **99% level of probability
a Insufficient mean square error degrees of freedom for a valid F test.
b Mean square error pooled to increase degrees of freedom in order to
enhance validity of the F test.

Table 5-9. Comparison of means for uptake of N, P, K, and Ca from corn
dry matter.

Corn dry-matter nutrients
N P K Ca

Factor / Level

-------------- kg/ha ------------
















owed by
level of

* Means in the same column under the same factor heading foll
the same letter are not significantly different at the 95%
probability, as determined by Duncan's Multiple Range Test.

develop a qualitative understanding of field-crop behavior where more

determinate methods (irrigation) are unavailable (Frey, 1981). Overall,

water requirements increase with planting densities. In this field

study, the higher plant densities yielded more grain, stover, N, P, K,

and Ca than the lower plant densities (Tables 5-6 and 5-8). The higher

densities and consequent greater demand on soil moisture did not induce

sufficient stress to affect yield components. However, superior yield

production by all of the fertilized treatments (irrespective of the

density) over the unfertilized control is sufficient evidence to support

the hypotheses that ambient soil fertility and not water availability

limited production of non- or minimally-fertilized plants for this soil.

The larger grain and stover yields of the fertilized treatments versus

those for the non-fertilized controls resulted from much larger plants,

which would have required larger quantities of soil water.

The historical rainfall-distribution pattern for this area

suggests that moisture stress would most likely occur early in the

growing season. Early-season water stress has been shown to be less

detrimental to eventual grain yields than stress during silking or grain

filling (Denmead and Shaw, 1960; Claassen and Shaw, 1970; Grant et al.,

1989). The need to replant part of this experiment was most likely due

to the infrequent occurrence of rainfall during a 4-wk drought following

a seemingly normal to slightly-wetter-than-normal start of the rainy

season. The early-season stress experienced by the plants in the first

crop-growth period was insufficient to decrease grain yield for the

higher densities relative to the lower densities.

The insignificant planting-date by density interaction indicated

that yields responded similarly to both densities within the two crop-

growth periods (Table 5-6). It is interesting to note that, if the

level of probability for the F test were reduced to 90%, the planting-

date by density interaction would become significant for corn-grain

yields. However, the significant difference in grain yields among

densities is only for the first planting date, where the high-density

yield was greater than the low-density yield (Table 5-7). Water stress

would have affected the higher-density plots to a greater degree than

the low-density plots. This is not to say that water stress did not

occur, but only that it did not detrimentally affect grain or stover

yields. Grain yields from the two planting densities for the second

planting date were not different. This would indicate that nutrient

availability and not water stress limited yields for the second planting


Bean Yields

Bean yields between the two planting dates were also

differentially affected by extraneous conditions. Angular leaf spot

(Xanthomonas malvacearum E.F. Sm.) became very prevalent during the last

week before harvest of the plants in the first crop-growth period.

Although this probably had little effect on yields for the first crop-

growth period, it impacted the plants of the second crop-growth period

for 5 wk, and caused considerable premature leaf drop. Additionally,

drying facilities were inoperative and thus incapable of drying the

beans of the first harvest. They underwent some spoilage before

alternative drying facilities could be arranged.


The analysis of variance table for bean yields is presented in

Table 5-10. The effect of fertilizer on bean-grain yields was somewhat

peculiar. The beans were harvested after 75 d and, therefore, were

unaffected by the last (84-d) application of fertilizer for the 4 by 1/4

split. The second application in the 2 by 1/2 split and the third

application in the 3 by 1/3 split were applied at 56 d, which should be

about half way through the normal pod-filling period (Fig. 5-3).

Comparison of treatment means indicated that bean-grain yields increased

with the more numerous applications, even when one of the applications

occurred after the beans had been harvested (Table 5-11). Shading may

be the best explanation for bean-plant behavior in this mixed-crop

arrangement. The trend in bean-grain yields as affected by fertilizer

schedule is just the opposite of that for corn-stover yield. Maturation

of the bean plant, including pod filling, occurred simultaneously with

maturation of the corn stover tasselingg at 78 d). The increased

splitting of fertilizer applications that limited stover yields also

reduced the potential of the corn plant to shade the shorter beans.

Further evidence is the lack of a plant-density effect on the bean-grain

yields. Corn-grain yields in the high-density plots were higher than

for the low-density plots. The additional corn plants would have

provided more shade and consequently may have reduced the high-density

bean yields to levels comparable to those of the low-density bean plots.

Post-harvest soil samples were taken from the high-density plots

to monitor gravel content and discern end-of-season differences in

nutrient availability between fertilizer treatments for any given depth

of soil. The concentrations of gravel in the plots showed no

Table 5-10. Analysis of variance for bean grain yields.

Source D.F. Error term F value

Date 1 Rep(Date) / Pooled errors 51.1**
Rep(Date) 2
Density 1 Pooled error < 1
Fertilizer 4 Pooled error 31.3**
Density*Fertilizer 4 Pooled error < 1
Date*Density 1 Pooled error < 1
Date*Fertilizer 4 Pooled error < I
Date*Den*Fert 4 Pooled error < 1
Pooled Error 18
Total 39

C.V. 14.4

Date = Planting date
Rep = Replication
Den = Density = Planting density
Fert = Fertilizer = Fertilizer application schedule

*95% and **99% level of probability
a Mean square error pooled to increase degrees of freedom in order to
enhance validity of the F test.

Dry weight (g/m2),no.of nodes, pods(> 2.5cm)



Total biomass/
400 -

300 -

200 Pod nL

Seed (g/

0 n A- A

Leaf area index

ode no./m2


L ,

Days from emergence

Fig. 5-3. Key Phaseolus vulqaris component growth-accumulation
parameters for cultivar Porrillo Sint6tico planted at 25
plants/m2 at Palmira-CIAT (from Laing et al., 1983).

Table 5-11. Comparison of selected bean yield-component means.

Factor / Level Bean yield
Planting Date
Early 336A*
Late 223B

Fertilizer Schedule
4 355a**
3 328ab
2 300ab
1 265b
0 151c

* Means in the same column, under the same sub-heading and followed
by the same uppercase letter, are not significantly different at the
95% level of probability as determined by an F test of mean square

** Means in the same column, under the same sub-heading and followed
by the same lowercase letter, are not significantly different at the
95% level of probability as determined by Duncan's Multiple Range

Table 5-12. Analysis of variance for soil gravel percentage.

Source D.F. Error term F value

Date 1 Rep(Date) a
Rep(Date)+Pooled error < 1
Rep(Date) 2
Fert 4 Rep(Date)*Fert < 1
Date*Fert 4 Rep(Date)*Fert < 1
Rep(Date)*Fert 8
Depth(Fert) 40 Pooled error 9.99**
Date*Depth(Fert) 40 Pooled error 1.37
Pooled Error 80
Total 179

C.V. 36.7

Date = Planting date
Rep = Replications
Fert = Fertilizer scheduling
Depth = Depth of sampling
a Insufficient mean error square degrees of freedom for a valid F test.
b Mean error squares pooled to increase degrees of freedom in order to
enhance validity of the F test.
**99% probability of significantly different treatment means.

Table 5-13. Mean comparison of percent gravel associated with depths
for the fertilizer application schedule.

Depth Fertilzer application schedule
0 1 2 3 4

-- cm -- ------------------- % gravel -----------------------

0 5 36.6a 39.8a 42.Oa 29.7a 29.5a
5 10 30.8a 45.0a 34.6a 39.2a 39.1a
10 15 42.la 33.5a 41.6a 36.0a 40.la
15 25 59.4a 56.9a 52.4a 56.3a 60.5a
25 35 74.3a 68.8a 72.5a 71.8a 74.9a
35 45 74.3b 84.5a 78.9ab 80.3ab 84.5a
45 55 71.5a 79.4a 76.7a 77.5a 70.8a
55 65 66.8a 75.1a 70.4a 71.4a 65.3a
65 75 50.4a 59.la 60.1a 53.2a 54.0a

* Means in the same row followed by the same letter are not
significantly different at the 95% level of probability, as
determined by Duncan's Multiple Range Test.

Table 5-14.

Mean gravel content with depth.



I- 5
i- 10
) 15
S- 35
- 45
i 65
S- 75

Gravel content


* Means followed by the same letter are not significantly different at
the 95% level of probability, as determined by Duncan's Multiple
Range Test.

_. _

---- % -----

significant trends associated with experimental treatments (Table 5-12).

The gravel content showed differences between depths, but not between

fertilizer schedules (Table 5-13 and 5-14).

The split fertilizer applications were for the most part

applications of N and K, because most of the P and Ca applied were in

the triple superphosphate which had been applied preplant in all

application schedules (Table 5-15). Analysis of variance for the

effects of the experimental factors on Mehlich I-extractable P, K, and

Ca indicated that planting date and fertilizer schedule had an

insignificant effect on overall nutrient concentrations averaged over

all depths (Table 5-16). The effects of sampling depth on nutrient

concentrations were significant. Since depth was nested within

fertilizer treatment and our experimental interest was in the location

of nutrients as affected by fertilizer schedule, mean separations were

made to distinguish differences among fertilizer-application schedules

within each depth, instead of differences between depths among

fertilizer schedules. Clear patterns are difficult to discern. The

concentration of K from 5 to 25 cm in all of the fertilized plots was

less than for the unfertilized control (Table 5-17). This would suggest

that fertilizer application enhanced K uptake to an even greater extent

than the amount applied. Limiting of this effect to the top 25 cm is

most likely related to the large increase in gravel concentration at the

top of the Btc horizon at about 22 cm, and to the subsequent effect of

the gravel on root growth (Table 5-18).

The concentrations of Ca at the various depths showed no

discernable pattern for the fertilized treatments or the control (Table

Table 5-15. Relative nutrient concentrations associated with each
fertilizer-application schedule.
Fertilizer-application schedule
Element 0 1 2 3 4
----------- percent of total applied t --------------

N 0 0 100 0 50 50 33 33 25 25
P 0 0 100 0 87 13 83 8.5 80 6.5
K 0 0 100 0 50 50 33 33 25 25
Ca 0 0 100 0 87 13 83 8.5 80 6.5

t Preplant applications Each subsequent application

Table 5-16. Analysis of variance for concentrations of Mehlich I-
extractable soil P, K, and Ca after harvest.

Soil nutrients
Source D.F. Error term P K Ca

----- F value -----

Date 1 Rep(Date) a a a
Rep(Date)+Pooled error 3.84b
Rep(Date) 2
Fert 4 Rep(Date)*Fert 2.17 1.07 < 1
Date*Fert 4 Rep(Date)*Fert < 1 < 1 < 1
Rep(Date)*Fert 8
Depth(Fert) 40 Pooled error 20.3** 5.55** 46.2**
Date*Depth(Fert) 40 Pooled error 1.88** < 1 < 1
Pooled error 80
Total 179

C.V. 25.5 76.1 17.6

Date = Planting date
Rep = Replications
Fert = Fertilizer scheduling
Depth = Depth of sampling

a Insufficient mean error square degrees of freedom for a valid F test.
b Mean error squares pooled to increase degrees of freedom in order to
enhance validity of the F test.
*95 and **99% probability of significantly different treatment means.

Table 5-17. Effects of fertilizer-application schedule on Mehlich I-
extractable soil K and Ca concentrations.

Depth Fertilizer application scheme
0 1 2 3 4

-- cm -- Mehlich I-extractable soil nutrients (ug/g)


0 5 165a* 151a 175a 127a 136a
5 10 155a 83b 88b 118ab 65b
10 15 195a 67b 70b 88b 76b
15 25 116a 46b 50b 87b 71b
25 35 38ab 35b 44a 47a 33b
35 45 23a 29a 27a 50a 23a
45 55 17a 21a 19a 26a 17a
55 65 14ab 12b 14ab 19a 13b
65 75 11a 9a 11a 14a 22a


0 5 1800a 1830a 2180a 1920a 1660a
5 10 1870b 2130ab 2330a 1944ab 1914ab
10 15 2041a 1855a 1931a 1995a 1711a
15 25 1686a 1391a 1440a 1370a 1580a
25 35 1120a 1000a 1250a 1480a 1040a
35 45 730a 729a 712a 1032a 677a
45 55 505a 509a 548a 583a 470a
55 65 361a 384a 433a 419a 390a
65 75 295a 315a 356a 310a 322a

* Means in the same row followed by the same letter are not
significantly different at the 95% level of probability, as
determined by Duncan's Multiple Range Test.

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - Version 2.9.8 - mvs