Variability of topsoil properties at the southern coast of Finland and the number of soil samples needed for the estimation of soil properties

A total of 430 topsoil samples were collected from ten fields of the Viikki Experimental Farm, University of Helsinki. Particle size distribution, organic carbon content, pH(CaCl 2 ), exchangeable Ca, .Mg, K contents, plant available P (Bray 1), 1 M KCI extractable (Al+H) content and effective cation exhange capacity of the soils were determined. The coefficient of variation was used as indicator of the variability of soil properties within each field. The lowest coefficients of variation were observed for pH(CaCl2 ) and the highest for exchangeable Mg 1 M KCI extractable (Al+H) and effective cation exchange capasity. The results indicate that from 1 (pH(CaCl 2)) to 33 (exchangeable Mg) samples per hectare are needed from individual fields for strict level of accuracy in estimation of the soil properties. For determinationof soil type (according to clay content) and organic carbon content on average 8 samples, and for the plant available P (Bray 1) and exchangeable Mg and K contents 10 to 16 samples per hectare appear sufficient. Four samples suffice for a less stringent, lax accurate determination of all properties. The variability of soil properties is discussed from the viewpoint of agricultural advisory work and field experiments for agricultural research.


Introduction
As early as 1935 KIVINEN drew attention to the grat variability in the chemical properties of soil even within the space of 30 or 40 metres.This he found to cause wide variations in the yields of the reference variety in a field experiment.KAILA and RYTI (1951) studied soil samples taken at distances of 2 metres and distances of 25 centimetres, concluding that it is difficult to obtain really representative samples for the estimation of soil properties.In a review article BECKETT and WEBSTER (1971) noted about 80 studies dealing with the variability of the properties of agricultural and forest soils.The subject has since been re-investigated in connection, for example, with regulation of nitrogen fertilization of farmland (LINDEN 1979), forest management (QUES- NEL and LAVKULICH 1980) and prediction of timber yields (BLYTH and MACLEOD 1978).
The aim of this study was to clarify the variability in soil properties as these may affect the interpretation of results obtained in field experiments and the recommendation made in the course of agricultural advisory work.The number of soil samples needed for accurate estimation of the properties is calculated.

Materials and methods
The material of the study consists of topsoil samples collected from the agricultural area of the Viikki Experimental Farm (University of Helsinki).This farm is situated near the Gulf of Finland.The geography of the fields is rather flat.In the soil profiles there appears strata with considerable differences in the particle size distribution originating in the time of deposition.
After harvestings of the crops, in September 1979 and 1981, the fields were marked with lines 40 metres apart, along which soil samples (volume 2 litres) representing the plough layer were taken at 40-metre intervals.Ten fields were sampled as follows: About one third of the Viikki area of232 ha was sampled (Fig. 1).For the minorpart the fields number 54, 84 and 86 were at the time under fields experiments, no fertilization experiments.
The samples were kept on laboratory tables until they reached air-dry state, after which they were crushed to pass through a 2 mm-sieve.The properties of the soils were determined by the following methods: pH in 0.01 M CaCl, suspension, soil to solution ratio 1:2.5 (v/v), equilibration time 4 hours.The material was classified into four groups, where the pH(CaCl 2 ) value was 3.5-4.4,4.5-54, 5.5-64 and >6.5.organic carbon, by wet combustion with K 2 Cr 2 0 7 and H,S0 4 (cone.)and thereafter colorimetric determination.On the basis of the organic carbon content the mineral soils were classified into three groups of 1.7-3.4%, 3,5-6.9%, 7.0-11.5 %.The soils with organic carbon content 11.6-23.2% are mull soils.particle sizes were determined by pipette method (ELONEN 1971).Soils where the clay (< 2 /am) content is less than 30 % are non-clay soils.In the total material there were 179 samples of fine sand soils, 21 samples of finer fine sand, 226 samples of sandy clay soils.The number of soil samples in the different groups defined by clay content, pH(CaCl2 ) and organic carbon content are presented in Table 1.
the exchangeable cations were extracted with 1 M neutral ammonium acetate.Calcium and magnesium were determined by atomic absorption spectrophotometry (Varian Techtron 1 00), with interference Sr and exchangeable K by flame photometry (Lange).
the exchange acidity (Al+H) was displaced with 1 M KCI and titrated with 0.01 M NaOH.
the effective cation exhange capasity (ECEC) was determined as the sum of (Ca+Mg) and (Al+H)   extractable in 1 M KOI (KAILA 1971).
the plant available form of soil phosphorus was extracted with 0.03 M NaF+o.o2sM HCI by the Bray 1 test (BRAY and KURTZ 1945) modified by KAILA (1965) and determined by molybdenum blue method.
The mean (x), standard deviation (s) and coefficient of variation in per cent (v) of the soil properties were calculated for each of the ten fields.The number of samples needed for accurate determination of the soil properties was calculated according to SNEDECOR (1948) bv the equation n = where n = P number of samples needed, t 2 = square of Students t, v 2 = square of coefficient of variation, and p = allowable error in per cent.The number of samples needed was calculated both with t=s %, p=lo % (strict level of accuracy = n,) and with t=lo %, p=2s % (lax level of accuracy = n 2).

Results
Except in field 54 only one crop was grown in each field, so that fertilization was the same over the whole field.On field 54 two different crops were grown, in both sampling years.Within a field main reason for the variation should then be the inherent heterogeneity of soil properties and the cultivation history.The coefficient of variation of pH(CaCl 2 ) was between 5.4 and 10 %in the individual fields (Table 2).The difference between maximum and minimum pH(CaCl 2 ) values in a single field was at the lowest 1.0 unit (field  97) and at the highest 2.3 units (field 98).The distance between the extreme values in field 98 was about 170 metres.In the field 49 the pH(CaCl 2 ) values for the two adjacent points (distance apart 40 metres) with greatest difference in value were 6.1 and 4.9.Were the pH(CaCl 2 ) of the soil to be adopted as the indicator of the liming requirement, the minimum, maximum and mean values would indicate the addition of three different amounts of liming agents.
Because the pH scale is logarithmic the coefficient of variation of pH(CaCl 2 ) is not comparable to the coefficients for other soil properties.
The range in the organic carbon content was widest in field 49 from 1.8 to 14.6 % (Table 3) and the coefficient of variation was highest there too (Table 2).As high carbon contents were observed in field 94 as in field 49, but the material was concentrated near the mean and the coefficient of variation remained low.The change in carbon content in field 49 occurred gradually, unlike pH(CaCl 2 ).The great differences in organic carbon content cause differences in the water and nutrient retention capacity of the soil and can lead to uneven maturity of the crops.
The range of the clay content was especially wide in fields 54 and 86 (Table 3).In field 54 the clay content changed within a distance of 100 metres from 11 to 65 % and the fine sand content from 76 to 17 %.In field 86 the distance between minimum and maximum clay contents (7 and 62 %) was Table 2.The coefficient of variation (v %), the number of soil samples needed for the strict (nj) and lax (n 2) level of accuracy in determination of soil properties in the individual fields, and the average values (field 94 omitted).The change in the soil type from fine sand to heavy clay cannot be without effect on the cation content of the soil and on the fertilizer requirement.
The coefficient of variation in particle size distribution, in the amounts of clay, silt or fine sand fractions, were below 50 % except for field 54 (Table 2).In field 96 the coefficient of variation in clay content was low (11%), since there sandy clay samples accounted for 64 out of the total 66 samples.In the P (Bray 1) content, exchangeable Mg and K contents, (Al+H)  content and ECEC the coefficients of variation were over 30 % for almost all fields.Exceptionally high coefficients of variation were recorded for ECEC in three fields.The range of these properties was not widest in the same field for which the greatest coefficients of variation were observed.In the fields where the coefficient of variation in the exchangeable Mg was high there was also high in clay content.The coefficient of variation in the exchangeable Ca remained below 30 % extcept in fields 49 (31 %) and 94 (42 %).
The number of soil samples per hectare needed to satisfy the strict and lax (n 2) accuracy classes as defined above was calculated separately for each field. (Field 94 was omitted because of its small area).In general, n, for pH(CaCl 2 ) was less than one sample per hectare.For the determination of the soil type according to the clay content ni was between 2 and 16 samples and for n 2 about one sample.
For strict level of accuracy in determination of the organic carbon content the number of soil samples needed was between 2 and 22, and for lax level of accuracy about one sample per hectare.
Determinations of the exchangeable K and Mg contents with the strict critetion of accuracy demaded from 3 to 32 and from 5 to 33 samples per hectare, respectively.For the determination of the plant-available P content the number of soil samples demanded varied from 3 to 20 and for the exchangeable Ca content from one to five with the strict criterion of accuracy.Lax accurate determination of the nutrient contents could be saticfied by a collection of one (Ca) to four (K and Mg) samples.Because both low and high coefficients of variations were found for 1 M KCI extractable (Al+H) and ECEC, the number of soil samples needed for lax accurate determination varied from 0.1 to 32 per hectare.
For the determination of soil type and organic carbon content with the strict accuracy criterion on average 8 soil samples per hectare was found necessary.Correspondingly the determinations of plant available P and exchangeable Mg and K demanded 10 and 16 samples, respectively.

Diskussion
The 430 topsoil samples collected from the Viikki Experimental Farm represented mainly mineral soils.The mean pH(CaCl 2 ) was near the average value of Finnish mineral soils (SIPPOLA and TARES 1978), considering that pH(H 2 0) = 0.5 + pH(CaCl 2 ) (RYTI 1965).The exchangeable Ca and K contents were higher than the average values reported for Finnish mineral soils by KAILA (1973) and SIPPOLA and TARES (1978), while the exchangeable Mg content does not deviate from the values of the same authors, the ECEC's are likewise in good accordance with an earlier study (KAILA 1971).The plant-available P contents are higher than the contents reported by KAILA (1965) using the same method probably because of the heavy phos- phorus fertilization in the sixties and seventies.The drilling of fertilizers may also cause some differences in the nutrient contents of soil samples taken between rows and along the rows.This effect was probably slight in the present study, however, since samples were taken in the autumn after the harvesting of the crops (URVAS and JUSSILA 1979).
The variability in soil properties was studied in ten fields, revealing a wide range in all the soil properties in individual fields.The ranges obtained by KAILA and RYTI (1951) within 100 square metres and within 1 square metre were slightly narrower than in this material within 1 to 17 hectares.The size of the sampling area apparently has little effect of the ranges found in the soil properties (HEMINGWAY 1955).
The coefficients of variation observed by BALL and WILLIAMS (1968) for uncultivated and unfertilized soils in North Wales were almost the same as in this study for cultivated and fertilized soils.Likewise they reported the highest coefficient of variation for the exchangeable cation contents.
For the present agricultural advisory work in Finland about 1.5 soil samples are collected per hectare (KURKI 1982).In this study one to two samples were found adeqaute for strict accurate determination only of pH(CaCl 2 ).The number of soil samples should be decided according to the most variable property, which in this material was the exchangeable Mg content.Even for the lax accurate determination 4 soil samples per hectare were needed.LINDEN (1979) suggested collecting about 10 soil samples and for field experiments (LINDEN 1981) 14 cores per plot (108 square metres).
When field experiments are being laid out the determination of soil properties in advance is important, and the number of soil samples should rather be too high than too low.Several soil samples, each collected from a different sampling point, are more informative than a single sample made up of subsamples from different sampling points.Too little attention has thus far been paid to the density and mode of sampling.Hehtaaria kohti otettujen näytteiden määrä vaihteli 5.1-6.5.

Table
1 The number of soil samples in different classes of clay content, pH(CaCl 2 ) and organic carbon content in ten fields.
2 v % ri] n 2 v % n t n 2 v % n 2 n 2 n

Table 3 .
The mean value, standard deviation(x±s)and range for soil properties of individual fields.