Combining soil map and soil analysis for improved yield prediction

10
CATENA vol. 15, p. 529-538 Braunschweig 1988 COMBINING SOIL MAP AND SOIL ANALYSIS FOR IMPROVED YIELD PREDICTION A.O. Ogunkunle & P.H.T. Beckett, Oxford Summary 1 Introduction The choice of agricultural land use or management can be based entirely on a soil map or on results of soil analysis or on a combination of the two. This paper presents the kind of comparison that should form the basis for choos- ing among these alternatives. It is illus- trated with the dry matter yield of bar- ley (Hordeumvulgare cv. Julia) in topsoil samples. The results clearly indicate the supe- riority of the combination of mapping and analysis over either of these for crop yield prediction. Thus, subdividing the soil of the study area of 250 ha into 2, 3 and 4 mapping units reduced the un- described variability from 100% to 93, 73 and 66% respectively. There was no further benefit from mapping into 5 and 6 units. Regressions of yield on values of 1, 2 and 3 soil properties reduced the undescribed variability to 62, 60 and 58& respectively. There was little bene- fit from including more soil properties in the regression. Regressions on 2 and 3 soil properties within each of 2 mapping units reduced the undescribed variability to 50 and 48%, while for 3 units it was reduced to 45 and 34°,/0 respectively. ISSN 0341-8162 @1988 by CATENA VERLAG, D-3302 Cremlingen-Destedt, W. Germany 0341-8162/88/5011851/US$ 2.00 + 0.25 It is common experience that some soil maps are used extensively and some are not. Many factors can be responsible for this and these factors may differ from one situation to another. Obviously, a soil map that provides adequate answers to the current problems of the user will find more use than one that does not. As an inventory of soil resources, soil sur- vey information must be accurate to be in great demand by users. Generally, the accuracy of soil maps for a given area in- creases with the number of delineations they contain. Hence a map with 6 units is expected to be more accurate than one with 4 units for the same area and by the same soil surveyor. However, if the extra accuracy from separating two more map- ping units is proportionally outweighted by the additional efforts and costs, it is better to have 4 rather than 6 units. Soil survey has many uses including agriculture, engineering, recreational and urban development. In each of these, soil survey information aids decision mak- ing so that better decisions are made than could have been made without it. But soil survey is expensive, so where funds are limited and soil survey ser- vices are not easily obtained, the analysis of some easily determined soil proper- ties may serve the immediate need of a farmer (arable land, e.g. corn), i.e. crop CATENA--An Interdisciplinary Journal of SOIL SCIENCE--HYDROLOGY~EOMORPHOLOGY

Transcript of Combining soil map and soil analysis for improved yield prediction

CATENA vol. 15, p. 529-538 Braunschweig 1988

COMBINING SOIL MAP AND SOIL ANALYSIS FOR IMPROVED YIELD PREDICTION

A . O . O g u n k u n l e & P.H.T. B e c k e t t , O x f o r d

Summary 1 Introduction

The choice of agricultural land use or management can be based entirely on a soil map or on results of soil analysis or on a combination of the two. This paper presents the kind of comparison that should form the basis for choos- ing among these alternatives. It is illus- trated with the dry matter yield of bar- ley (Hordeum vulgare cv. Julia) in topsoil samples.

The results clearly indicate the supe- riority of the combination of mapping and analysis over either of these for crop yield prediction. Thus, subdividing the soil of the study area of 250 ha into 2, 3 and 4 mapping units reduced the un- described variability from 100% to 93, 73 and 66% respectively. There was no further benefit from mapping into 5 and 6 units. Regressions of yield on values of 1, 2 and 3 soil properties reduced the undescribed variability to 62, 60 and 58& respectively. There was little bene- fit from including more soil properties in the regression. Regressions on 2 and 3 soil properties within each of 2 mapping units reduced the undescribed variability to 50 and 48%, while for 3 units it was reduced to 45 and 34°,/0 respectively.

ISSN 0341-8162 @1988 by CATENA VERLAG, D-3302 Cremlingen-Destedt, W. Germany 0341-8162/88/5011851/US$ 2.00 + 0.25

It is common experience that some soil maps are used extensively and some are not. Many factors can be responsible for this and these factors may differ from one situation to another. Obviously, a soil map that provides adequate answers to the current problems of the user will find more use than one that does not. As an inventory of soil resources, soil sur- vey information must be accurate to be in great demand by users. Generally, the accuracy of soil maps for a given area in- creases with the number of delineations they contain. Hence a map with 6 units is expected to be more accurate than one with 4 units for the same area and by the same soil surveyor. However, if the extra accuracy from separating two more map- ping units is proportionally outweighted by the additional efforts and costs, it is better to have 4 rather than 6 units.

Soil survey has many uses including agriculture, engineering, recreational and urban development. In each of these, soil survey information aids decision mak- ing so that better decisions are made than could have been made without it. But soil survey is expensive, so where funds are limited and soil survey ser- vices are not easily obtained, the analysis of some easily determined soil proper- ties may serve the immediate need of a farmer (arable land, e.g. corn), i.e. crop

CATENA--An Interdisciplinary Journal of SOIL S C I E N C E - - H Y D R O L O G Y ~ E O M O R P H O L O G Y

530 Ogunkunle & Beckett

Subject Linked to

Lime requirement P & K fertilizers

N Fertilizer

Drainage of Land capability

Soil analyses alone Soil analyses, but modified according to the prevailing yields on the farm and how the straw is to be disposed of Recent crops, and of broad soil textural classes e.g. sandy, silty or heavy; 'black puff chalk soils' Soil series or group of similar series or series divided according to land gradient

For farmers' extension leaflets the Soil Survey groups soil series into broader classes, e.g. all series developed on clay into three groups

Tab. 1: Extension officers' advice to farmer in South England.

yield prediction. Even where soil survey services are easily obtained, the farmer's decision on the best land use and/or management could be better made by a combination of the soil map informa- tion with results of soil analysis. Thus, three alternatives are open to the farmer as basis for choice of appropriate land use:

(i) soil map, and the level of detail of mapping,

(ii) soil analysis and the properties to analye for;

(iii) a combination of (i) and (ii)

No doubt the decision to choose any of these alternatives will be aided by knowledge of the advantages and dis- advantages of (i) and (ii). For instance, although a soil map is relatively expen- sive to produce, the information it pro- vides is long term. Conversely, because ad hoc soil analysis only provides so- lution to immediate problems, separate analyses must be made for every new problem that the farmer may face. It is also possible that a soil map of broad delineation supplemented by ad hoc soil

analysis may be more useful than a more detailed soil map alone.

These comparisons are not trivial. It may be shown, for example, that in some countries in Southern England (Berkshire, Buckinghamshire, Oxford- shire) some of the advice that farmers receive from extension officers is linked to soil taxonomic class, some is linked to soil analysis and some to both (tab.l). Field trials reveal the same dichotomy. Thus BOYD & D E R M O T (1964) exam- ined a very large number of field tri- als with main crop potatoes throughout England and Wales and compared their responses to P and K fertilizers. They found that classification (at the series level or a little above) described 19 and 33% of the total vaiance in responses to P and K. Linear regression of yield on P and K described 36 and 27% of the to- tal variance. Within-class regressions on P and K described 44 and 46%, if the regressions were of the same form and 55 and 51°/0 with different regression in different soil classes.

Similarly in Oklahoma, A L L G O O D & GRAY (1978), compared in a study the effect of soil properties and the class criteria in Soil Taxonomy for predict-

CATENA An interdisciplinary Journal of SOIL SCIENCE HYDROLOGY~GEOMORPHOLOGY

Yield Prediction by Soil Map and Soil Analysis 531

2 mapping units (m.u.) 3 m.u. 4 m.u. 5 m.u. 6 m.u.

2.1 Coarse sandy: 3.1 Loamy coarse 4.1 Loamy coarse 5.1 Loamy coarse 6.1 Loamy coarse freely drained; sand to c.s.1.; sand to c.s.1.;* sand to c.s.1.; sand to c.s.l.: summit sites, freely drained; freely drained; freely drained; freely drained;

plateau plateau plateau plateau

4.2 Coarse sandy 3.2 Sandy loam loam to sandy to sandy clay clay loam; freely loam; variable to somewhat poorly drainage; drained; upper slope transitional and springline sites 4.3 Silt loam to

silty clay loam : somewhat poorly to poorly drained baekslope

5.2 Coarse sandy loam to sandy clay loam; freely drained: Plateau edge and upper slope

5.3 Clay loam to silt loam; some- what poorly to poorly drained; springline

2.2 Silty clay 3.3 Silty clay 5.4 s.l;* to to clay; to clay; poor 4.4. Silty clay s.c.l; poor poorly drained; to very poor to clay; poor internal footslope sites, internal drain- to very poor drainage;

age; footslope internal drainage; footslope. footslope

6.2 Loamy c.s. to c.s.1. freely drained; plateau edge

6.3 Loam to silt loam; freely to slightly poorly drained upper springline

6.4 Silt loam to silty clay loam; slightly poorly to poorly drained: springline.

6.2. Silty clay to clay; poor

5.5. Silty internal drainage clay to clay; footslope. poor to very poor internal 6.6. Clay; very drainage poor internal footslope drainage;

footslope

Mainly derived from Lower Green sand formation

Transitional

Mainly derived from Kimmeridge formation

c.s.1. = coarse sandy loam; si.1 = silt loam; siccl. = silty clay loam.

Tab. 3: The description of the mapping units showing their relationship to each other on the landscape.

ing crop yield. In terms of production indices for wheat, sorghum and cotton lint, they reported R 2 values of 0.83, 0.90 and 0.76 respectively for the regression model, and 0.78, 0.85 and 0.63 respec- tively for the classification model. They concluded that either model can be used in making productivity index predictions. COSTIGAN & McBURNEY (1983) ex- plored the causes of variation in the yield of cabbage and lettuce between two sim- ilar soil series. They reported that the main factor of variation was the amount of native potassium which was inade- quate in the poor site. WEBSTER et al. (1977) related the yield of sugar beet

in Britain to soil drainage, subsoil tex- ture and rainfall and selected classes at the subgroup level. They concluded that before proceeding to develop a classifi- cation, it is important th ascertain by similar procedures, which distinguishing criteria would produce the most useful classes. Although, strictly, this conclu- sion applies if soil maps or classifications are made for crop yield prediction alone, it shows the need for guidance on the ba- sis for choosing between alternative land uses or management systems.

Mapping units are defined on the more permanent morphological proper- ties, while most of the properties ob-

CATENA An Interdisciplinary Journal of SOIL SCIENCE HYDROLOGY GEOMORPHOLOGY

532 Ogunkunle & Beckett

1-1

2 -un i t s

4-units

S-uni t s

1-2

I

I _ 3 . : t •

• • o J (

_ • . , [

- units

6-untts

o a,2 o.5 o,~ )okm

Q sampling point

Fig. 1: Successive mapping of the area.

tained through soil analysis are chemi- cal properties, which may change very easily with management, and crop yield is a function of both groups of proper- ties. This paper therefore reports an ex- ploratory study on how the farmer can decide, which of the three alternatives given earlier (i.e. soil map, soil analysis or their combination) should be used as a basis for predicting the most appropri- ate agricultural land use or management. For simplicity it is illustrated by early growth of barley under uniform man-

agement. Nevertheless, the design of the study and the way the results were anal- ysed are applicable to other agricultural land uses.

2 Materials and methods

The study area of 250 ha in the NE of Oxfordshire lies across the contact be- tween the coarse feruginous sands of the Lower Greensand formation and the un- derlying Kimmeridge clay formation in which a discontinuous outcrop of fine

CATENA An Interdisciplinary Journal of SOIL SCIENCE--HYDROLOGY~EOMORPHOLOGY

Yield Prediction by Soil Map and Soil Analysis 533

sand-silty bed overlies silty clays and clays. The area slopes sharply from the Lower Greensand plateau to the Kim- meridge footslope, with a line if weak springs at the junction.

2.1 Soil mapping

The area was first delineated into the units by auger examinations using soil texture, natural drainage class, geology and physiography as differentiating char- acteristics (tab.3). It was then remapped in 3, 4, 5 and 6 mapping units by pro- gressively narrowing the range in these characteristics. At each stage of mapping the whole soil continuum was remapped rather than subdividing the mapping units of the preceeding stages; hence boundaries at successive stages do not coincide perfectly (fig.l). These various stages of mapping are equivalent to semi- detailed (2 units), detailed (3 units), very detailed (4 units), intensive (5 unit) and very intensive (6 unit) scales. The 2-unit map showed that the area is dominated by two series, the Shrivenham and the Evesham series (AVERY 1973). These belong to Arenic Paleudalf and Vertic Ochraqualf respectively in the USDA Soil Taxonomy (SOIL SURVEY STAFF 1975) and Nitosol and Luvisol in the FAO system (FAO 1974).

2.2 Sampling

Soil samples were collected from 36 loca- tions in the study area (fig.l.2). Follow- ing ISBELL's procedures for preliminary survey of soil nutrient status (CRACK & ISBELL 1970, ISBELL & G I L L M A N 1973) only the topsoils (0-15 cm) were sampled. For the purpose of demonstrat- ing the methodology of choice between alternative basis of land use selection,

this sampling was considered satisfac- tory.

2.3 Soil analysis

The samples were analysed for pH, ex- tractable P and K (M.A.KK 1973) and for particle size (TOMLINSON et al. 1977). Texture was included in the labo- ratory properties even though it was one of the properties used for defining the mapping units, because it was observed that the variation in soil texture within the mapping units was still substantial. As can be seen from tab.3, no unit lay wholly within one textural class.

2.4 Yield determination

Two subsamples were taken from each of the 36 soil samples; they were sieved (with 12.5 mm sieve to break clods of clay-textured soils) and cropped to spring barley (Hordeum vulgate c.v. Ju- lia) in the greenhouse. Six plants were sown in 14 cm pots. The crops received no fertilizer and the soil was held at field capacity by a wick of 2 cm cotton tape that passed through the base of each pot into a trough of water beneath. This treatment minimized the effects of any artificial contrasts in permeability due to the structures of the sieved soils, and it also eliminated any contrasts between heavy and light textured soils to drought or waterlogging.

All the pots were harvested 60 days later, when most plants had reached the five-leaf stage that precedes tillering. The harvested shoots were dried and weighed. Yield was calculated as the mean yield of dry matter (9/6 plants) from two pots for each sampling point.

CATENA An Interdisciplinary Journal of SOIL SCIENCE H Y D R O L O G Y ~ E O M O R P H O L O G Y

534 Ogunkunle & Beckett

Source df SS MS

Between map units 2 13.01 6.51 Within map units (error) 33 28.77 0.87

Total 35 41.78

Tab. 2: A one-way analysis of variance of yield on 3 map units.

2.5 Analysis of data

As indicated above, the study was de- signed to compare the success of a soil map or soil analysis at predicting crop yield. More specifically it was to mea- sure the fraction of the total variability of th 36 yield values (total variance) that was not described or predicted (i.e. the error variance) by soil maps of 2-6 map- ping units, by multiple linear regressions on 1-6 soil properties, or by multiple re- gressions on 1-6 soil properties within each of 2-5 mapping units. The compar- ison requires three different approaches:

(a) Success of mapping alone: A one- way analysis of variance (ANOVA) was used to estimate success; tab.2 is an example of the ANOVA for 3 mapping units. There is, how- ever, a complication; for example, if all the sampling points were to be completely randomly distributed over the 3 mapping units and the yields and values of soil proper- ties were determined at each point, the regression SS (or MS) would be zero, (i.e. no yield variation for re- gression to account for). But the between-unit SS (or MS) will not be zero, it will have a value higher than zero. Thus, the comparison of the predictive power of mapping and regression in terms of the er- ror SS (i.e. undescribed variabil-

(b)

ity may favour mapping over re- gression. Thus, an adjustment to the calculation of between-unit SS that makes the two more compara- ble will lead to more valid results, even though complete randomisa- tion very rarely (if at all) operates in practice. So the between-unit SS was calculated as follows, using the ANOVA for 3 mapping units (tab.2 as an example :)

Corrected between-unit MS = (6.51 - 0.87) = 5.63

Corected between-unit SS -- 13.01 × 5.63/6.51 = 11.25

Thus, 41.78-11.25 73.1% (instead of 41.78 - -

28.77 68.86%) 41.78 o r

is the fraction of the total variability that the soil has failed to describe. The reasoning is that the only statis- tic available for estimating by how much the between-unit MS could have been overestimated is the error MS. So the between-unit MS is ad- justed by subtracting the error MS, and the corrected MS is then used to obtain the corrected between-unit SS. Similar calculations were per- formed for the various stages of mapping as shown in fig.1.

Success of analysis alone: multiple linear regressions of yield on 1-6 soil properties were determined by a stepwise procedure. This pro- cedure first determines the regres- sion on all the six properties after which the least significant property is dropped. This process continues until the regression has been calcu- lated on the last single property re- maining. The error sums of squares

CATENA--An Interdisciplinary Journal of SOIL SCIENCE HYDROLOGY -GEOMORPHOLOGY

Yield Prediction by Soil Map and Soil Analysis 535

No. delineation 2 delineations 3 delineations 4 delineations 5 delineations 6 delineations

Y=411g n=36

1. Y=182.8g n=17

2. Y=228.2g n=19

1. Y=13 ?.6g n=12

2. Y=123.3g n=12

3. Y=155.1g n - 1 2

1. Y=110.2g n=10

2. Y=93.4g n=9

3. Y=75.8g n=7

4. Y=131.6g n=10

1. Y=77.3g n=7

2. Y=84.6g n=8

3. Y=92g n=9

4. Y=529g n=4

5. Y= 104.2g n=8

1. Y=67.4g n=6

2. Y=109.3g n - 1 0

3. Y=77.2g n - - 8

4. Y=47g n=4

5. Y=52.5g n=4

6. Y=57.6g n--4

Tab. 4: Total yield of young barley within mapping units with increasing number of delineations.

at each stage were obtained from regression ANOVA.

(c) Success of soil mapping and soil analysis; similar stepwise regres- sions were calculated for each of the 2 and 3 mapping units (fig.1 and 1.2). For maps with more than 3 units the degrees of freedom (df) in some units were too few, hence the calculation of regressions within unit was not possible. For instance, for the map with 4 units, one unit had 7 observation points (i.e. n=7 and n-l=6, tab.4), with 6 soil prop- erties in regression, there is no df for error, i.e. error df=0. Even with fewer properties in regression the small number of observations makes the estimate that may be ob- tained less reliable. The situation is even worse for maps with 5 and 6 units, where some units contain 4 observations. This is because the number of observation points per unit decreased with increase in the number of units, while the number of soil properties in regression re- mained the same.

For success of soil mapping and soil analysis, the errors SS for the units for each map were pooled, e.g. for 3 soil properties in each unit of the 2-unit soil map.

% undescribed variability - Error SS for regression error SS for regression of yield on 3 soil + of yield on 3 soil properties in unit 2.1 properties in unit 2.2

Total SS

The Standard deviation was calculated for each value of undescribed variability. Taking the % undescribed variability as 1-R 2, its variance (i.e. the variance of the undescribed variability), is approxi- mately given as Var. (1-R 2) = 4R2(1-R2)2

n

(KENDAL & STUART 1973), where n is the total degree of freedom.

3 Results and discussion

As contrived, the study ignored the ef- fects of subsoil on crop performance. This tends to underestimate the influ- ence of mapping units, which are usu- ally defined on both surface and subsoil properties. Besides, many field crops ex- plore depths greater than 0-15 cm for water and nutrients. The study also used a single species as test crop and did not

x 100

CATENA An interdisciplinary Journal of SOIL SCIENCE HYDROLOGY~EOMORPHOLOGY

536 Ogunkunle & Beckett

Clay Silt Coarse pH Available Ex K sand P

1 5 6 2 3 4 One mapping units

Two mapping units

Three mapping units

2.1 1 5 2 6 4 3 2.2 1 4 6 3 2 5

3.1 1 2 3 6 4 5 3.2 5 6 1 3 2 4 3.3 3 2 6 5 1 4

I * From stepwise regresions.

T a b . 5: The relative importance* of six soil properties as predictors of yield of young barley (1 is most significant).

100-

80"

> 60-

u

t+O-

a. 20-

1 mapping unit {no mapping)

2 mapping units

3 mapping units

'0' ~I ~ '2' '3' ' 4'

N_ ° of soil properties in the regression

° 5 ' ' 6 '

Fig . 2 : Fraction of yield variability not described by mapping, soil analysis or their combination.

CATENA An interdisciplinary Journal of SOIL SCIENCE HYDROLOGY -GEOMORPHOLOGY

Yield Prediction by Soil Map and Soil Analysis 537

include the response of soils to manage- ment inputs (e.g. fertilizer application) or to non-agricultural uses and employ dry matter rather than grain yield of the test crop. However, even in situations where the assumptions made in this study do not hold, the techniques of comparison being proposed would still be valid.

The results (tab.5) follow BOYD & D E R M O T T (1954) in finding that the soil properties that determine yields within one soil unit are not the same as those that determine yields within an- other soil unit.

It is clear (fig.2) that within the cir- cumstance implied by the condition of the trial, there would be no advantage in defining and mapping more than 3 units. Indeed if the farmer's advisor could esti- mate soil texture (clay and coarse sand) accurately by auger examination and perform analys on extractable P (tab.5), there would be no advantage in mapping beyond the level of 3 units. This result has two implications: First, that combin- ing mapping and analysis is superior to either of them for yield prediction, and second that there is a stage of combina- tion where prediction is optimum; after this the benefit derived is not worth the extra effort.

The benefit (time and cost) of this combination over either mapping or analysis alone is very obvious. This re- sult indicates that even for a small area it might be more beneficial to have a simple soil map (few mapping units) and analyse for a few relevant and easily de- termined soil properties than either very detailed mapping or extensive soil anal- ysis.

Further studies ( O G U N K U N L E 1979) have shown that results of simi- lar studies with undisturbed soil cores at a central site gave the same conclusions

as those drawn from field crops adjacent to the sites from which the undisturbed cores were collected. This study thus illustrates very simply the kind of com- parison that should be made as a basis for selecting the most appropriate agri- cultural land use or management or even before commissioning a special purpose soil survey.

Indeed, by treating the distinguishing criteria of the soil units in alternative schemes of soil mapping as the basis for 2 or 3-way analysis of variance, similar trials could be used to identify the opti- mum number of delineations in the form of, for instance, capability or suitability classes. This falls in line with the sug- gestion by WEBSTER et al. (1977): "It seems that a better classification (than at the soil series level) for site selection and advice in beet growing could be de- vised by appropriately combining values of soil properties. The relative impor- tance of these factors or indeed of any other, and whether there are any criti- cal values of any of them will be deter- mined".

Acknowledgement

We thank Dr. EH.C. Mariott, Mr. H.L. Wright, Dr. R. Webster and Dr. H.C. Dawkins for advice on statistics and the trustees of Oxford Preservation Trust for permission to work in the study area. We are also grateful for the financial support by the Nigerian Institute for Oil Palm Research (NIFOR) for Ogunkunle dur- ing the period of the study.

References

ALLGOOD, F.R. & GRAY, F. (1978): Utiliza- tion of soil characteristics in computing produc- tivity ratings of Oklahoma soils. Soil Science 125, 359-366.

CATENA An Interdisciplinary Journal of SOIL S C I E N C E ~ Y D R O L O G Y -GEOMORPHOLOGY

538 Ogunkunle & Beckett

AVERY, B.W. (1973): Soil classification in the soil Survey of England and Wales. Journal of Soil Science 24, 324-338.

BOYD, D.A. & DERMOTT, W. (1964): Fertil- izer experiment on main crop potatoes. Journal of Agricultural Science Cambridge 63, 249-259.

COSTIGAN, P.A. & MeBURNEY, T. (1983): Variation in yield between two similar sandy loam soils. 11. Causes of yield variation in cab- bage and lettuce crops. Journal of Soil Science 34, 639-647.

CRACK, B.J. & ISBELL, R.F. (1970): Studies on some solodic soils in north eastern Queensland. 1. Morphological and chemical characteristics. Australia Journal of Experimental Agriculture and Animal Husbandry 10, 334-341.

FAO (1974): Soil map of the world. The legend. Food and Agriculture Organization, Rome.

ISBELL, R.F. & GILIMAN, G.P. (1973): Stud- ies on some deep sandy soils in Cape Yorke Peninsula, North Queensland. 1 Morphological and chemical characteristics. Australian Journal of Exprimental Agriculture and Animal Hus- bandry 13, 81-88.

KENDALL, N.G. & STUART, A. (1973): The advanced theory of statistics. Volume 2, Griffin, London, 355.

M.A.F.F. MINISTRY OF AGRICULTURE, FISHERIES AND FOOD (1973): Fertilizer recommendation for agricultural and horticul- tural crops. Bull. No. 209, 88-89.

OGUNKUNLE, A.O. (1979): The optimisation of classification for soil survey. D.Phil. thesis, University of Oxford.

TOMLINSON, P.R., BECKETT, P.H.T., BAN- NISTER, P. & MARSDEN, R. (1977): Simpli- fied procedure for routine soil analysis. Journal of Applied Ecology 14, 253-260.

SOIL SURVEY STAFF (1975): Soil Taxonomy. A basic system of soil classification for making and interpreting soil survey. USDA Handbook 436, Washington, D.C.

WEBSTER, R., HODGE, C.A.H., DRAYCOTr, A.P. & DURRANT, M.J. (1977): The effect of soil type and related factors on sugar beet yield. Journal of Agricultural Science Cambridge 88, 455-469.

Addresses of authors: A,O. Ogunkunle', P.H.T. Beckett Soil Science Laboratory, Agricultural Science Building Oxford, England

* Present address: Department of Agronomy, University of Ibadan Ibadan, Nigeria

CATENA-- An Interdisciplinary Journal of SOIL SCIENCE HYDROLOGY~EOMORPHOLOGY