Real-World Mathematics: Agriculture, Crop Yield, Irrigation, Sampling and Resource Planning

Application of Mathematics in Real-World Usage · Guide 20 · BTT Mathematics Hub

A field produces six tonnes per hectare. Another field produces seven. The simple average is 6.5 tonnes per hectare, but that may be the wrong farm-wide result if the fields have different areas. Twenty-five millimetres of water sounds small until it is spread across two hectares. Five sample plots can be measured precisely and still give a biased estimate if they were chosen from the healthiest corner.

Agriculture is an excellent setting for real-world Mathematics because nearly every important quantity is a rate tied to area, time or resource use. Yield is output per area. Rainfall and irrigation depth become volume when multiplied by area. Sampling estimates a large field from a small observed subset. Harvest loss changes gross biological output into net recovered output. Resource planning asks whether all active constraints can be satisfied at once.

Every field size, crop label, yield, rainfall total, water allocation and plant spacing below is fictional unless a source is explicitly cited. This is not agronomic, irrigation, pesticide, fertiliser, food-safety or farm-management advice. Real agricultural decisions depend on crop, soil, climate, regulation, engineering, local measurements and specialist guidance. The examples are designed to make the Mathematics auditable.

A hectare is an area unit, not a length

One hectare equals 10,000 square metres. A rectangular field 200 m by 150 m has area 30,000 m², or 3 ha.

If a field map is drawn at scale, convert linear dimensions first or square the scale factor for area. Multiplying a map area by a linear scale only once is dimensionally wrong.

Area is the natural denominator for many agricultural quantities: tonnes per hectare, plants per square metre and litres per square metre of water depth.

Yield multiplies by area to give production

Suppose a fictional crop yields 6 tonnes per hectare across 12 ha. Modelled production is 6×12=72 tonnes.

The units confirm the structure: t/ha × ha = t. If yield were instead 6 kg/m², the area would need to be expressed in square metres before multiplying.

A yield value is a rate over area. It does not by itself tell us total production until the represented area is known.

Farm-wide yield is area-weighted

Field A covers 4 ha at 5 t/ha. Field B covers 6 ha at 7 t/ha. Their productions are 20 t and 42 t, totalling 62 t over 10 ha.

Farm-wide yield is therefore 62/10=6.2 t/ha. The simple mean of 5 and 7 is 6.0, which gives equal weight to fields with unequal area.

The reliable calculation is total production divided by total area. This is the same weighted-average structure used in retail prices, sports subgroup rates and school marks.

A sample plot is evidence about a larger population only through a sampling design

The USDA National Agricultural Statistics Service describes objective yield surveys built from selected fields and sample units rather than simply inspecting convenient locations. The purpose of probability-based selection is to support inference beyond the sampled plants.

Suppose five equal-area teaching plots produce estimated yields 5.8, 6.2, 5.9, 6.4 and 6.1 t/ha. Their simple mean is 6.08 t/ha.

If these plots were selected representatively from a 12 ha field under a suitable design, 6.08×12=72.96 t might be used as a modelled production estimate. If the plots were deliberately chosen from the best-looking region, the same arithmetic would not justify the same field-wide inference.

Selection bias cannot be repaired by more decimal places

Imagine a field with a high-yield half averaging 8 t/ha and a low-yield half averaging 4 t/ha. The true equal-area field mean is 6 t/ha.

If every sample plot is chosen from the high-yield half, the sample mean may be very precise around 8 while remaining badly biased for the whole field.

Precision describes repeatability or uncertainty under a design. Bias concerns systematic displacement from the target quantity. Large samples from the wrong part of the population can be confidently wrong.

Stratification can preserve known area structure

Suppose 30% of a field is one soil zone with estimated yield 4 t/ha and 70% is another with 7 t/ha. The area-weighted estimate is 0.30×4+0.70×7=6.1 t/ha.

A simple average of the two zone means gives 5.5 t/ha and ignores the fact that the second zone occupies more area.

Stratified estimation is useful when the population is deliberately divided into meaningful subgroups and each subgroup receives the correct weight. The weights must describe the target population, not be chosen after seeing which result is preferred.

Sampling variability can be quantified under a stated model

For a classroom illustration, suppose plot-yield observations are independent with true standard deviation assumed known as 0.5 t/ha and a sample of 25 plots has mean 6.0 t/ha.

Standard error of the mean is 0.5/√25=0.1 t/ha. Under a normal model with known standard deviation, a 95% z-style interval is 6.0±1.96×0.1, or 5.804–6.196 t/ha.

This is conditional on the stated probability model and representative sampling. It is not a guarantee that every plot lies inside that interval, and it does not repair selection bias.

Yield and production are different forecast objects

USDA NASS distinguishes crop yield estimation from total production. Production combines yield with harvested area.

If estimated yield is 6.1 t/ha and expected harvested area is 100 ha, estimated production is 610 t. If area is later revised to 90 ha while yield stays 6.1, production falls to 549 t.

A production change does not therefore prove a yield change. The area term may have changed instead.

Gross biological yield and net harvested yield require a loss boundary

USDA’s objective-yield description notes that harvest loss is estimated and subtracted from biological gross yield to obtain net yield.

In a fictional example, gross field yield is 7.2 t/ha and measured harvest loss is 0.3 t/ha. Net recovered yield is 6.9 t/ha.

If loss were instead described as 4% of gross yield, the calculation would be 7.2×0.96=6.912 t/ha. An absolute loss and a percentage loss are different inputs and should not be mixed.

One millimetre of water over one square metre equals one litre

A water depth of 1 mm is 0.001 m. Spread over 1 m², volume is 0.001 m³. Since 1 m³ equals 1,000 L, the result is 1 L.

This identity makes rainfall and irrigation calculations convenient: millimetres × square metres = litres.

For hectares, 1 mm over 1 ha equals 10,000 L = 10 m³.

Irrigation depth becomes volume when area is known

Suppose a fictional plan applies 25 mm over 2 ha. Area is 20,000 m². Water volume is 25 L/m²×20,000 m²=500,000 L=500 m³.

The 25 mm number alone is not a volume. Doubling field area doubles the required volume at the same depth.

This is a geometry conversion only. It does not say whether 25 mm is agronomically appropriate for any crop or soil.

Rainfall depth can be converted through the same geometry

A hypothetical event delivers 18 mm uniformly over 3 ha. Volume incident on the area is 18×30,000=540,000 L=540 m³.

This is incident rainfall volume. It is not automatically the same as water stored in soil or available to plants. Runoff, interception, evaporation and infiltration are separate physical processes.

The Mathematics should therefore label the 540 m³ as rainfall arriving on the area, not “usable irrigation.”

Water balance is another conservation ledger

For a simplified storage model, ending water = beginning water + inflow − outflow − losses.

Suppose a teaching reservoir begins with 900 m³, receives 500 m³, delivers 700 m³ and has 50 m³ of other modelled losses. Ending storage is 650 m³.

If a physical measurement later reports 620 m³, the 30 m³ discrepancy signals missing flows, measurement error or an incorrect assumption. The balance equation identifies inconsistency but does not identify the cause by itself.

Water productivity is output divided by water consumed or used under a defined boundary

FAO’s WaPOR materials express agricultural water productivity in forms such as kilograms of production per cubic metre of water. The exact water denominator depends on the system and dataset.

For a fictional example, 12,000 kg of output is associated with 4,000 m³ of water under the declared boundary. Water productivity is 3 kg/m³.

If another field reports 4 kg/m³, that alone does not prove it has higher total yield or lower total water use. The ratio could rise because the numerator rises, the denominator falls or both.

Normalised productivity should be reported beside totals

Field A produces 10,000 kg using 2,500 m³, or 4 kg/m³. Field B produces 30,000 kg using 10,000 m³, or 3 kg/m³.

A has higher water productivity, while B has higher total production. Neither statement contradicts the other.

Normalisation helps compare efficiency-like quantities, but total resource use and total output remain important. A single ratio should not replace the ledger it summarises.

Plant density is a geometric packing approximation

In an ideal rectangular planting grid with row spacing 0.50 m and within-row spacing 0.25 m, each plant is assigned 0.125 m².

The idealised density is 1/0.125=8 plants/m², or 80,000 plants/ha.

This is a geometric count under perfect regular spacing and full planted area. Real headlands, missing plants, irregular boundaries and agronomic decisions change the actual count. The calculation is not a spacing recommendation.

Irregular field boundaries make gross area different from planted area

Suppose a 10 ha property contains 0.8 ha of roads, buffers and service areas not included in the planting zone. Net planted area is 9.2 ha.

At a modelled yield 6 t/ha, production based on planted area is 55.2 t, not 60 t.

The correct denominator depends on the yield definition. If yield is reported per harvested hectare, use harvested area rather than gross property area.

Crop counts from sample quadrats require an expansion factor

Suppose ten 1 m² quadrats contain a total of 74 plants. Mean density is 7.4 plants/m².

Under a representative-sampling assumption, multiplying by 10,000 gives 74,000 plants/ha.

If quadrats are 0.5 m² each, ten quadrats cover only 5 m², so the total count must first be divided by 5 rather than by 10. The sampled area belongs in the denominator.

A weighted crop estimate can use several field classes

Imagine three equal-method field classes: 20 ha estimated at 4.5 t/ha, 50 ha at 6.0 and 30 ha at 7.0.

Total production estimate is 90+300+210=600 t across 100 ha, giving weighted yield 6.0 t/ha.

The unweighted mean of the three class yields is 5.833… t/ha. The weighted result answers the whole-area question.

Forecast error should be kept separate from yield itself

Suppose four model forecasts are 5.8, 6.1, 6.0 and 6.4 t/ha, while realised values are 6.0, 5.9, 6.2 and 6.3.

Define error as realised minus forecast. Errors are +0.2, −0.2, +0.2 and −0.1. Mean error is +0.025 t/ha. Mean absolute error is (0.2+0.2+0.2+0.1)/4=0.175 t/ha.

A small mean error can coexist with larger individual misses because positive and negative errors cancel. Forecast evaluation therefore needs both bias-like and magnitude-like summaries.

Resource planning becomes a feasible-set problem

Consider a purely abstract land-and-water model with two plot types. Type A uses 2 land units and 3 water units; Type B uses 1 land unit and 4 water units. Available resources are 10 land units and 24 water units.

If x and y are numbers of A and B plots, constraints are 2x+y≤10, 3x+4y≤24, with x,y≥0 and whole if plots are indivisible.

The point (4,2) uses 10 land and 20 water, so it is feasible. The point (2,5) uses 9 land but 26 water, so it is infeasible. A plan must satisfy all active constraints, not merely the land limit.

An objective function turns feasibility into optimisation

Assign fictional output scores of 5 units per A plot and 4 per B plot. Objective is maximise 5x+4y under the previous constraints.

Checking integer feasible points gives, for example, (4,2) score 28, (3,3) score 27, (2,4) score 26 and (0,6) score 24. Under this small integer model, (4,2) is better than those alternatives.

The output score is invented. Real resource allocation requires agronomic, economic, ecological and legal inputs. Mathematics can optimise only the objective it is given.

Thresholds should be tested against uncertainty

Suppose a planning threshold requires estimated production of at least 60 t. A nominal yield estimate is 6.1 t/ha over 10 ha, giving 61 t.

If plausible yield under a deterministic scenario range is 5.7–6.3 t/ha, production range is 57–63 t. The threshold is not guaranteed across the stated range.

Calling 61 t “above target” is correct for the nominal scenario. Calling the target guaranteed would overstate the evidence.

A complete agricultural Mathematics report keeps stage and denominator visible

State whether a quantity is per square metre, per hectare, per plant, per cubic metre of water or total production. State whether the area is planted, harvested or gross property area. State whether yield is gross biological output, recovered output, forecast or observed estimate.

For sampling, record the target population and selection design. For water, record whether the number is rainfall incident on the area, delivered irrigation, stored water or a further modelled usable quantity.

Agricultural Mathematics becomes reliable when these boundaries travel with the number. The arithmetic is often straightforward; the hard part is keeping the represented system intact.

Practice: twenty agriculture Mathematics questions

1. Convert 4.5 ha to square metres.
2. A field is 250 m by 160 m. Find area in hectares.
3. Yield is 6 t/ha over 12 ha. Find production.
4. Four hectares yield 5 t/ha and six hectares yield 7 t/ha. Find total production and weighted yield.
5. Find the mean of 5.8, 6.2, 5.9, 6.4 and 6.1 t/ha.
6. Using that mean for a 12 ha teaching field, find estimated production.
7. A zone occupying 30% yields 4 t/ha and the remaining 70% yields 7. Find weighted yield.
8. Known σ=0.5 t/ha, n=25. Find standard error of the mean.
9. Using z=1.96 and sample mean 6.0, find the illustrative 95% interval.
10. Gross yield is 7.2 t/ha and absolute harvest loss 0.3 t/ha. Find net yield.
11. How many cubic metres is 1 mm of water over 1 ha?
12. Find water volume for 25 mm over 2 ha.
13. Find incident rainfall volume for 18 mm over 3 ha.
14. A store starts with 900 m³, receives 500, delivers 700 and loses 50. Find ending storage.
15. 12,000 kg output uses 4,000 m³ under the stated boundary. Find water productivity.
16. Spacing is 0.50 m by 0.25 m in an ideal grid. Find plants per m² and per ha.
17. Ten 1 m² quadrats contain 74 plants. Find mean density and expanded plants/ha.
18. Twenty hectares yield 4.5 t/ha, fifty yield 6.0 and thirty yield 7.0. Find total production.
19. For constraints 2x+y≤10 and 3x+4y≤24, test whether (2,5) is feasible.
20. Why can a precise sample mean still be a poor field-wide estimate?

Worked answers

1. 45,000 m².
2. 4 ha. Area is 40,000 m².
3. 72 t.
4. 62 t total; 6.2 t/ha weighted yield.
5. 6.08 t/ha.
6. 72.96 t.
7. 6.1 t/ha.
8. 0.1 t/ha.
9. 5.804–6.196 t/ha. Conditional on the stated normal model and known σ.
10. 6.9 t/ha.
11. 10 m³.
12. 500 m³.
13. 540 m³.
14. 650 m³.
15. 3 kg/m³.
16. 8 plants/m² and 80,000 plants/ha.
17. 7.4 plants/m² and 74,000 plants/ha.
18. 600 t.
19. No. Land use is 9≤10, but water use is 26>24.
20. Because precision does not remove selection bias. If sampled plots do not represent the target field, repeated or highly precise measurements can still estimate the wrong population quantity.

Sources and connected applications

For probability-based crop sampling and objective-yield field methods, see USDA NASS: Objective Yield and the NASS Yield Forecasting Program. For agricultural water-productivity definitions, see FAO WaPOR data and FAO AQUASTAT irrigation-water analysis. All field sizes, yields, water quantities and planning scores on this page are original teaching constructions.

Continue with Retail, Pricing, Discounts, Margins and Break-Even; Music, Rhythm, Frequency, Ratios and Sound; and Photography, Optics, Perspective, Resolution and Image Geometry. Return to the BTT Mathematics Hub.