The Gold Standard of Soil Science | Composite Samples, Unequal Contributions and Hidden Spatial Variation

Suppose a school receives supplied records for four equal-sized garden zones. The question could be the average concentration across the garden, the location of an unusual patch, or whether the zones should be investigated separately. Those questions need different information.

A composite sample combines contributions from several places into one mixed sample. A discrete sample retains a particular place or sampling unit. Mixing can reduce the number of laboratory analyses, but it also removes some information about the original spatial pattern.

The University of Minnesota Soil Testing Laboratory advises separate samples for substantial areas that differ and distinguishes lawn from garden sampling. That is a useful starting principle: define an area whose result has an interpretable purpose.

Do not turn that into a universal instruction to mix every convenient handful. A sampling plan needs a target area, depth, timing and measurement method. If the question concerns an unusual feature, quietly blending it into a broad background sample may remove the evidence needed to investigate it.

Contents · Next chapter

Use four fictional zone concentrations: 10, 10, 10 and 90 milligrams of X per kilogram of dry soil. Assume each zone is internally uniform for this paper exercise. Take equal dry-soil masses from the four zones and mix them completely, with no loss or transformation of X.

The composite concentration is (10 + 10 + 10 + 90)/4 = 30 mg/kg. For four contributions of 0.10 kilograms each, X masses are 1, 1, 1 and 9 milligrams. The mixture contains 12 milligrams in 0.40 kilograms, giving the same 30 mg/kg.

The mass balance explains the average. It also exposes its scope: the result describes the mixture under the stated assumptions. Extending it to the whole garden requires a defensible relationship between the sampled contributions and the target garden soil.

Contents · Previous chapter · Next chapter

Compare Garden One, with zone values 10, 10, 10 and 90, against Garden Two, with values 30, 30, 30 and 30. Equal-contribution composites both report 30 mg/kg. The first garden has a concentrated contrast; the second is uniform in the supplied model.

A composite value alone cannot reconstruct which pattern existed. Neither extra decimal places nor repeating the analysis on the same mixed jar restores the zone labels that mixing removed.

This is not an argument that composites are inherently poor. It is an argument for matching the sample to the question. A representative composite can serve an average-condition enquiry. A spatial investigation needs retained spatial information.

A careful report might say: “The equal-mass composite was 30 mg/kg; this analysis does not establish the maximum concentration or its location.” That statement is more informative than calling the entire garden uniform. Where a decision concerns individual areas, maintain those identities or use an appropriate design that can resolve the relevant differences.

Contents · Previous chapter · Next chapter

Change only the material added to the mixing container. Take 0.10 kilograms from each of A, B and C, but 0.30 kilograms from D. The mixture now contains 1 + 1 + 1 + 27 = 30 milligrams of X in 0.60 kilograms of dry soil.

Its concentration is 50 mg/kg. The source concentrations have not changed; the contribution weights have. Zone D now provides half the mixed mass, rather than one quarter.

The general mass-balance expression is composite concentration = sum of each concentration times its contributed dry mass, divided by total contributed dry mass. An unweighted average is the special case in which all those masses are equal.

Equal-looking wet handfuls do not necessarily contain equal dry-soil masses. Moisture and sampling geometry can change the contributions. For an actual protocol, follow the specified collection and preparation method; do not improvise a correction after location and mass records have been lost.

The repair is to state the intended weights before collection. “One scoop from each zone” is not a complete quantitative description unless the method makes those scoops appropriate for the target being estimated.

Contents · Previous chapter · Next chapter

Suppose a field has two internally uniform zones. Eighty percent of its area has X concentration 10 mg/kg; the remaining twenty percent has 90 mg/kg. An area-weighted mean of the concentration values is 0.8×10 + 0.2×90 = 26 mg/kg.

Equal samples from one point in each zone, averaged without weights, give 50 mg/kg. That gives the small zone the same influence as the large one. The sampling allocation and the target area are mismatched.

Now make the distinction more precise. At the same sampling depth, suppose the first zone has dry bulk density 1.0 and the second 1.5 in consistent units. Relative dry-soil masses are then 80×1.0 = 80 and 20×1.5 = 30. The mass-weighted concentration across that layer is (80×10 + 30×90)/110 = 31.82 mg/kg.

The values 26 and 31.82 answer different questions. One averages local concentration across area; the other expresses total X mass divided by total dry-soil mass. Neither should be labelled simply “the field average” without its basis. Inventory calculations need area, depth, density and units, not concentration alone.

Contents · Previous chapter · Next chapter

Suppose a profile has an upper 10-centimetre layer at 10 mg/kg and a lower 20-centimetre layer at 40 mg/kg. Assume constant dry bulk density, the same footprint and conservative mixing across both layers.

The 0–30-centimetre mass-weighted concentration is (10×10 + 20×40)/30 = 30 mg/kg. Averaging the two layer concentrations equally gives 25 mg/kg and ignores the lower layer’s double thickness.

If bulk densities differ, the weights need depth times dry bulk density, along with area when footprints differ. If the question concerns only the surface layer, including deeper material changes the sampled quantity rather than improving the surface estimate.

Keep depth intervals written as boundaries, such as 0–10 and 10–30 centimetres, with the reference surface defined. A label saying “deep sample” does not support an exact comparison next year.

The root-zone water-storage guide uses related layer weighting for a different physical quantity. The transferable habit is to identify what each layer contributes. It is not permission to use the same averaging equation for every soil measurement.

Contents · Previous chapter · Next chapter

Our X example assumes that its mass is conserved and its reported concentration is recovered consistently by the measurement method. Many soil results are more complicated. Extraction methods can measure operationally defined fractions, and mixing can change chemical conditions.

Soil pH is a particularly useful warning. It is a logarithmic measure related to hydrogen-ion activity under specified conditions, and soils have buffering behaviour. Averaging two pH numbers is not a general prediction of the pH of the resulting soil mixture. Even averaging corresponding solution hydrogen-ion concentrations would not account fully for interacting, buffered soil systems.

Likewise, an average particle-size percentage does not reproduce intact soil structure. Combining material destroys the original arrangement of aggregates and pores. A jar cannot preserve the undisturbed profile merely because its composition came from that profile.

Before applying a weighted formula, ask what is additive, what the denominator represents and whether the laboratory method supports the assumed relationship. A mass balance for a deliberately conservative teaching tracer should remain distinct from a prediction about an actual nutrient extraction, pH or hydraulic property.

Contents · Previous chapter · Next chapter

Imagine mixing one composite thoroughly and dividing it into three labelled laboratory portions. Results are 29.8, 30.0 and 30.2 mg/kg. Their close agreement supports a narrow statement about consistency among those analysed portions under the method.

It does not give three independent observations of the garden’s spatial variability. All portions came from the same combined material. Treating them as three separate field samples would exaggerate the independent evidence about the place.

Now imagine collecting three independently designed field composites, each using a suitable sampling pattern for the defined unit. Those may reveal field-sampling variation as well as analytical variation. Their interpretation still depends on whether the design and collection actually support independence and representation.

Keep the hierarchy in the record: field or management unit, sampling occasion, composite, subsample and analytical repeat. Each level can answer a different uncertainty question. A large number of laboratory readings cannot automatically compensate for a very limited field design.

Contents · Previous chapter · Next chapter

If the purpose is to detect a small unusual area, sample design becomes central. Consider an intentionally simplified model in which a patch occupies ten percent of the accessible area, and each sampling point is an independent uniform random location.

The chance that ten points all miss the patch is 0.9 to the power ten, approximately 0.349. The chance that at least one point hits it is therefore about 0.651, or 65.1%. Ten points do not make a hit certain.

These figures are a geometry-and-probability exercise. Real sampling may involve spatial correlation, access constraints, known features and a different target definition. A systematic layout or targeted investigation is not the same as ten independent random locations.

There is a second limitation: hitting the patch does not guarantee recognising it after mixing. Combining one high contribution with many low ones can dilute its concentration, and detectability also depends on the method. Keep sampling coverage and analytical detection as separate stages. Suspected contamination needs an appropriate professional investigation rather than an improvised garden-testing plan.

Contents · Previous chapter · Next chapter

Begin with a map and a sentence defining the decision. Identify which areas belong together and which should retain separate labels. Record the depth interval, date, relevant management history and the method requested by the laboratory.

For a supplied-data learning exercise, retain every original zone value and contribution weight before calculating the composite. Then write the strongest statement supported by those records and one question they cannot answer.

For actual sampling, agree the appropriate protocol with the laboratory or relevant specialist. This article’s invented masses, zone counts and depth examples are not a collection specification. Different intended tests can require different handling and depths.

A useful evidence check asks whether another person could identify the represented area from the label and records. If the answer is no, a precise result may remain difficult to interpret. Preserve the connection from place to sample to preparation to measurement to conclusion.

Contents · Previous chapter · Next chapter

Check one: a smaller unusual contribution

Mix 0.30 kilograms of soil at 20 mg/kg with 0.10 kilograms at 100 mg/kg. X masses are 6 and 10 milligrams. The composite is 16/0.40 = 40 mg/kg. The simple average, 60 mg/kg, gives the smaller contribution too much weight.

Check two: recover an omitted contribution

Three equal dry-mass contributions at 10, 20 and an unknown concentration produce 30 mg/kg. The unknown is 60 mg/kg because the concentrations must total 90. This inference requires equal weights and the conservative-mixing model. Without them, the missing value is not determined by this equation.

Check three: repair a replication claim

A report says: “Three tests from our one mixed jar prove the whole garden is uniform.” Replace it with: “The three tested portions gave similar results; the composite does not resolve the original spatial variation.” Then identify the field information needed to investigate uniformity.

For a final transfer check, change the target from the whole garden to its upper five centimetres. Explain why the existing 0–30-centimetre mixture cannot simply be relabelled. The missing information is physical, not a calculator setting.

Contents · Previous chapter · Next chapter

Is a composite sample a mistake?

No. It can be useful for a defined, appropriately sampled unit and question. Its limitations follow from the information combined and the intended interpretation.

Does a low mean exclude a high local value?

No. A mean alone does not specify the maximum or spatial pattern. The two invented gardens illustrate that loss of information.

Can I average any two laboratory numbers?

Only after identifying the quantity, method, weights and target. An arithmetic mean is not a universal model for the physical result of mixing.

Which route develops this further?

For the physical context, return to The Gold Standard of Soil Science. For measurements whose denominators and layers matter, read gravimetric moisture and root-zone water depth and infiltration, storage and drainage.

For field interpretation, continue with soil compaction and moisture–density mathematics, gardening and living soil or geotechnical engineering and ground uncertainty, according to the decision. The Gold Standard of Statistics develops sampling and interpretation more broadly.

For learning across subjects, use The Gold Standard of Education and How X Works.

The University of Minnesota Soil Testing Laboratory’s lawn and garden sampling guidance supports defining and separating sampling areas. The University of Maryland Extension’s sampling for agricultural production explains management units and representative sampling. The NRCS soil pH introduction supplies chemical context. These institutional guides have their own purposes and local conditions; no foreign sampling depth or treatment rule is prescribed here for Singapore sites.

Contents · Previous chapter · Return to Soil Science