The kūmara patch: a trial that cannot answer its own question

The core problem: with one plot per lime rate, the lime dose is completely tied to the plot. Any plot difference - position, variety, shade, history - is therefore also a 'lime effect' as far as this design can tell.

Specific confounders and their direction

1. Starting pH already differs by dose. Plots 1–2 started at 5.1–5.2 and plots 3–4 at 6.3–6.5. So 'no lime' went to the most acidic plot and 'most lime' to the least acidic one. Plot 4 was pushed from an already-optimum pH up towards the range where micronutrients (B, Zn, Mn) become less available - which is a plausible reason for its lowest yield of 3.2 kg. The pH change caused by lime is tangled with the pH the plot started at.
2. Variety. Plots 1 and 4 grew Owairaka Red; plots 2 and 3 grew Beauregard, a longer-season variety. The two highest yields (5.8, 6.0 kg) are exactly the two Beauregard plots, so variety alone could produce the apparent '200 g/m² is best' pattern.
3. Water and drainage. Plots 1–2 are on the lower slope near the tile drain. Kūmara are sensitive to waterlogging and tuber rot, so the lower plots may under-yield for reasons unrelated to lime - which is consistent with plot 1's 4.1 kg despite receiving no lime.
4. Herbicide history and shade. Plots 3–4 were sprayed with glyphosate; residual effects could suppress early growth, though the two highest yields are in those plots, so this is a smaller effect. Plot 4's afternoon shade from 2:30 pm cuts photosynthesis during the highest-light part of the day and is a plausible contributor to its 3.2 kg.
5. No replication, crude harvest measure. One plot per treatment means no estimate of variability; weighing 'all tubers' mixes many small tubers with a few large ones and does not measure marketable size or dry matter. The two lost tipu slightly reduce plot 3's plant density.

Judgement: the trend 4.1 → 5.8 → 6.0 → 3.2 kg is equally consistent with variety, position and shade. The data cannot separate lime from those, so the conclusion is unsupported even though the numbers look tidy.

Design principles: replicate, randomise, hold everything else constant

1. One variety, one planting date. Use a single variety (say Owairaka Red) throughout, and plant all plots on the same day at the same spacing. If variety is itself of interest, treat it as a second factor with every lime rate tested on both varieties - but double the plots, not the confounders.
2. Many small plots, randomised within blocks. Divide the māra into blocks that each contain the same soil type and light exposure, and within each block randomly assign lime rates. Four blocks × four rates = 16 plots of at least 1 m × 2 m. Randomisation stops any one rate always sitting at the wet end.
3. Always include the 0 g/m² control in every block, so the comparison is 'lime versus no lime in the same conditions', not 'plot 1 versus plot 4'.
4. Equalise water and light. Irrigate all plots to the same soil moisture (a tensiometer or simple hand-feel check), and choose a site with no afternoon shade, or record shade and exclude affected plots from the analysis.
5. Measure pH before and after. Take a bulked soil sample per plot before liming and again at harvest, so the actual pH shift is known, not assumed. Note that lime reacts over months - applying in September for a November planting may mean the pH has not equilibrated within one season.
6. Measure yield properly. Harvest per plant, record tuber number, marketable mass (e.g. tubers >100 g) and total mass separately, and dry a subsample to constant weight for dry matter. Weigh without knowing which treatment each plot received (blind scoring) to avoid bias.
7. Analyse as means with spread. Report each rate as a mean of four blocks with a range or 95% CI; only call a difference real if the ranges do not overlap.
8. Keep a treatment log. Record the dog, the herbicide, anything unusual - and use fencing.

Why pH 5.1 is limiting

• Aluminium and manganese solubility rise sharply at low pH. At around pH 5, Al³⁺ becomes soluble enough to damage root tips. Kūmara store yield in roots, so restricted root growth directly limits tuber size.
• Phosphorus becomes less available. In acid soil, phosphate binds strongly to Al and Fe oxides and to clay, so P is present but not plant-available. Raising pH towards 6.0–6.5 releases some of that bound P - often the largest single effect of liming.
• Calcium and magnesium supply. Lime is calcium carbonate (and dolomite adds Mg). At pH 5.1, Ca availability is low and the soil is more likely to be Mg-deficient.
• Molybdenum and microbial activity. Mo, needed for nitrate reductase, is less available in acid soils, and the bacteria and fungi that mineralise organic matter and cycle nitrogen work better near neutral pH. More mineralised N and better root growth together mean a bigger canopy and more tuber bulking.

Why 400 g/m² without a test is risky

• Liming is slow-acting and cumulative. 400 g/m² of garden lime can shift pH by roughly 0.5–1.0 unit depending on soil buffering; on a light, sandy soil it can overshoot past 7.0, where boron, zinc, manganese and iron become less available - a different set of deficiencies replacing the first.
• Kūmara also tolerates a range; pushing an already-6.5 plot (plot 4) higher is the opposite of what is needed, and plot 4's 3.2 kg is consistent with that risk.
• The season matters: apply lime months before planting so it can react, and retest pH before deciding on a second application.
• The defensible recommendation is a soil test first, then the smallest lime rate that brings pH into the 6.0–6.5 range.