This is the routine sampling schedule — the planned rounds that happen whether or not anything interesting is going on. It establishes the seasonal envelope and the baseline that event samples are measured against.

How the frequencies are set. Every determinand is placed in one of four bands according to how much it moves in this particular source type:

• Dynamic — changes within hours to days. Sampled frequently, and these are the determinands the event programme exists for.
• Seasonal — changes over weeks to months. Monthly sampling resolves it adequately.
• Stable — set by geology or aquifer mineralogy and effectively fixed. Sampled a few times to establish the value and confirm it is not drifting, then left alone. Repeatedly re-measuring a stable determinand is the most common way an investigation budget gets wasted.
• Screening — a one-off or occasional look to confirm something is absent. Only worth doing if a plausible pathway exists, but then genuinely worth doing.

Phases. The programme is staged. Phase 0 costs nothing and often answers half the questions. Phase 1 establishes the baseline. Phase 2 runs alongside it and captures events. Phase 3 is treatability work that only makes sense once the range of the water is known.

The compliance minimum under the 2026 Rules is shown separately for reference. It is a floor for an operating supply, not a characterisation programme.

Routine sampling tells you what the source usually does. It almost never tells you what the source does at its worst, because the worst conditions occupy a small fraction of the year and a monthly grab has a correspondingly small chance of landing in one.

Three questions have to be answered about peak conditions, and they are separate questions:

• How bad does it get? The peak value sets the design case. Miss the peak and the plant is designed for the average, which it will meet comfortably right up until the first real storm.
• How long does it stay there? Duration decides whether the plant can ride through on storage, reduce output, or has to treat the peak. A two-hour spike and a five-day elevation are entirely different design problems at the same peak value.
• How often does it happen? Frequency decides whether the peak is a design condition or an emergency response condition.

The affordable way to answer all three. Duration and frequency come from continuous records — stage, rainfall, turbidity — which are cheap. Magnitude and composition come from laboratory samples, which are not. So use continuous surrogates to define the shape of every event in the record, and use a small number of well-placed laboratory samples to calibrate what those surrogates mean chemically. Sampling every event with a full suite is both unaffordable and unnecessary.

Sample spacing is geometric. Concentration changes fastest early. Uniform spacing wastes samples on the recession and misses the rising limb entirely, which is where first-flush organics and microbial load usually peak — often ahead of the turbidity peak, not with it.

Antecedent conditions matter as much as the rainfall. Record dry days before the event, rainfall depth and intensity, and season with every event. The first significant rain after a long dry spell is frequently the worst event of the year and is nowhere near the largest.

Events are not only rainfall. For a stratifying lake the design event is autumn turnover, which releases dissolved manganese, iron, ammonia and taste and odour compounds accumulated in the anoxic hypolimnion. For a bore the equivalent is sustained drawdown during a long dry period. This tab adapts to the source type.

How samples and readings are actually obtained. There are three distinct approaches and they are not interchangeable:

• Manual grab sampling. A sampler visits, takes readings on site with portable instruments, fills bottles and sends them to the laboratory. This is the base case for a new source and will carry most of the programme.
• Automatic grab sampling. A fixed sampler with a carousel of bottles, triggered by level or turbidity rather than a clock. It collects the same kind of sample as a person would, at times nobody could reasonably attend. It is how an event sequence gets collected on a fast catchment.
• Continuous instrumentation. A permanent installed probe logging at short intervals, typically every minute, producing hundreds of thousands of readings a year. This is a different class of data, not a faster version of grab sampling.

Be realistic about which of these a new source will get. Most source investigations are grab sampling only, because capital is unavailable, the site has no power, or nobody will own the instrument maintenance. Assume that as the starting point and treat permanent instruments as a case that has to be argued.

Where continuous instrumentation earns its place is the duration and frequency questions. How long a peak lasts and how often it recurs cannot be answered by grab sampling at any affordable density. A permanently installed level or turbidity record answers both across the whole investigation, and the laboratory samples then only have to establish what those readings mean chemically.

Check what already exists before installing anything. Regional council flow, stage, rainfall and sometimes turbidity records are permanently installed, continuous, already paid for, and usually cover decades. That is a better data set than anything a two-year programme can build, and it costs nothing.

The table below ranks each measurement by whether a permanent instrument is worth it for this source, and states what to do instead when the answer is no.

The full determinand list, with every inclusion decision shown and overridable. Determinands the tool has excluded are still listed — you can see what was left out and why, and put any of it back.

Reading the table. The status column shows whether a determinand is in the programme automatically, in because a risk flag triggered it, out because it does not apply to this source type, or overridden by you. The frequency column is the routine sampling frequency; the event column shows whether it is collected during an event, and at every event sample or only at the key ones.

Total and dissolved. Where a metal appears twice, that is deliberate. Total and dissolved fractions answer different questions. Total sets the load the plant has to remove; dissolved sets whether it can be removed by filtration at all or needs oxidation first. Field filtration through 0.45 micron at the point of collection is what makes the dissolved result meaningful — filtering in the laboratory a day later gives a number that describes the sample bottle, not the source.

Measured on site means exactly that. The sampler carries the instruments and takes the reading at the point of collection: a calibrated multiparameter meter for pH, dissolved oxygen, redox potential, temperature and conductivity, a portable nephelometer for turbidity, and for groundwater a flow cell so the water is never exposed to air before it is read. These are not permanently installed instruments and they are not laboratory results. They are listed at zero cost because once the kit is bought there is no per-sample charge, but the kit does have to exist, be calibrated, and be carried to every visit.

These determinands cannot be sent to a laboratory in any meaningful sense. They begin changing the moment the sample is exposed to air. A laboratory pH on a groundwater sample is frequently a full unit different from the pH in the aquifer, and every carbonate and corrosion calculation built on it is wrong.

Calculated values are not sampled. Specific ultraviolet absorbance, the ionic balance check and the potassium-40 correction are arithmetic on results you already have. They appear in the list because they are worth doing and easy to forget, but no bottle is collected and no analysis is ordered.

Hold times. The hold time column is the practical limit for a defensible result. It governs sampling logistics more than anything else — microbiological samples at 24 hours are the reason an autosampler cannot cover the whole programme.

A build-up of the programme cost from unit rates, split into laboratory analysis, field labour, travel and freight. All unit rates are editable in Settings. These are planning figures in New Zealand dollars excluding GST — get a schedule of rates from the laboratory before committing.

Where the money usually goes. On a surface water investigation the event programme normally dominates, and within it the specialist determinands — protozoa, cyanotoxins, particle size distribution, pesticide screens — dominate again. A small number of line items typically account for most of the cost, which is good news: the savings are concentrated in the same place.

Cutting cost without gutting the data set. The levers listed below are ranked by how much they save relative to what they cost you in information. Reducing the frequency of a determinand that does not change is free. Dropping the event programme to save money defeats the purpose of the investigation and will cost far more later in plant that does not work.

The collect-and-hold approach is worth reading twice. Bottles are cheap; analysis is not. Collecting the full event sequence and analysing only part of it, once the continuous record shows which samples matter, routinely saves a large share of event analysis cost with no loss of information. It only works if preservation and hold times are respected, so it applies to metals, nutrients and organics — not microbiology.

The only place this tool exports from. Choose the sections you want, then produce a PDF, a CSV of the determinand schedule, or a JSON session file.

• PDF — opens the browser print dialogue. Choose Save as PDF as the destination.
• CSV — the determinand schedule with frequencies, sample counts and costs. Written with a byte order mark so that micro symbols, degree signs and cubic metre units survive opening in a spreadsheet.
• JSON — the complete session, including settings and any overrides you have made. Load it back to resume.