Arsenic behaves differently depending on what chemical form it is in. Total arsenic alone does not tell you whether a plant can remove it. This tool takes a raw water, works out which arsenic species are likely present, follows them through pre-oxidation, coagulation and separation and any polishing media, and tells you what comes out the other end.
New to this? Flick the worked example on below. It loads a geothermally-influenced river with a raised As(III) fraction, which is the situation that has caught New Zealand plants out. Turn it off at any time to clear back to an empty sheet.
Arsenite, As(III), is uncharged at drinking-water pH. It does not stick to coagulant floc, it is poorly held by adsorptive media, and it slips through low-pressure membranes. Arsenate, As(V), is an anion and is removed well by all of them. Pre-oxidation converts As(III) to As(V) and is usually the difference between a plant that complies and one that does not.
Choose the oxidant before you choose the dose. Free chlorine at the front of the plant works on arsenite, but it is the one oxidant that meets the entire raw organic carbon load before any of it has been coagulated out, and halogenated by-product formation scales with exactly that. On a New Zealand water carrying DOC, permanganate or a manganese-dioxide contact bed gives the same arsenic outcome with no halogenated by-products. Front-end chlorine solves an arsenic exceedance by creating a THM and HAA exposure, and it should be treated as a short-term measure while a non-halogenated oxidant is put in.
The panel on the right flags the by-product exposure the selected oxidant creates and the organic carbon it is being asked to sit in.
Coagulation does not precipitate arsenic. It creates a metal-hydroxide surface that arsenate adsorbs onto, and then the solids-separation stage takes that surface out of the water. Two things therefore control removal: how much sorbing surface you make, and how completely you take it away again.
The charts show how treated arsenic responds to coagulation pH and to dose, so you can see which lever is worth pulling on your water.
Adsorptive media and anion exchange are the polishing barriers. They sit after filtration and take dissolved arsenate out to very low levels. They are also the answer when coagulation alone cannot make the MAV with an acceptable margin.
Feed to this tab is the treated water leaving the coagulation and separation stage. If you are designing a stand-alone media plant, set the coagulant to None.
This is where the whole plant is assembled and judged. Pick a configuration preset to populate every stage at once, or leave it on Custom and use whatever you set on the individual tabs.
A plant can sit comfortably below the MAV for years and then fail within days without the total arsenic in the river changing at all. Only the form changed.
Arsenic removed from the water does not disappear. It concentrates into coagulation sludge or spent adsorptive media, and that stream then has to go somewhere.
Leachability screening here is indicative only. Disposal acceptance requires a laboratory leach test on the actual solids.
Everything leaves the tool from here and nowhere else.
The JSON carries a schema version so an older file still loads correctly.
Predicts the arsenic species present in a source water, tracks them through pre-oxidation, coagulation and solids separation and adsorptive polishing, and reports treated-water arsenic against the maximum acceptable value. It is a screening and diagnostic tool for operators and process engineers, intended to answer two questions: why is this plant not removing arsenic, and what would fix it.
Arsenic occurs in natural water in several forms, and they behave completely differently in a treatment plant.
Arsenic in central North Island waters is geothermal in origin. Geothermal fluid discharges arsenic mostly as arsenite, and it enters rivers and lakes where oxidation to arsenate proceeds over hours to days, and where sediments in the hydro lakes act as a store and a source. Total arsenic in these waters commonly runs at 10 to 40 µg/L against a MAV of 10 µg/L, so every plant on these sources depends on treatment, not on the raw water being clean.
In late 2024 several plants drawing on one of these rivers went above the MAV within days of each other, while total arsenic in the river stayed inside its normal long-term range. What changed was the proportion of dissolved, reduced arsenic. Plants that had always complied were suddenly presenting their coagulant with a species it could not adsorb. The emergency intervention was front-end chlorination: free chlorine ahead of the coagulant oxidises arsenite to arsenate before the coagulant sees it, and it can be put in within a day at a plant that already has chlorine on site. This tool is built around that failure mode.
Oxidising arsenite ahead of the coagulant is the fix. Doing it with free chlorine is a separate decision, and on a New Zealand water it is usually the wrong one as a permanent measure. Front-end chlorine is the only oxidant option that puts free chlorine in contact with the entire raw organic carbon load before any of it has been coagulated out. Trihalomethane and haloacetic acid formation scales with the product of free chlorine, organic carbon and contact time, and once formed those by-products are not removed by anything downstream. Many of the affected sources carry appreciable DOC, so a front-end chlorine dose held for ten or twenty minutes ahead of coagulation converts an arsenic exceedance into a DBP exposure.
Chlorine dioxide deserves a specific word, because it is often reached for as the low-THM alternative to chlorine. It is a poor arsenite oxidant, several orders of magnitude slower than free chlorine, and it carries its own by-product ceiling: roughly seventy per cent of the ClO2 applied is reduced to chlorite, with a few per cent going to chlorate. Both are inorganic anions with their own maximum acceptable values, both form immediately, and neither is removed by coagulation, filtration or adsorptive media. The chlorite MAV therefore caps the ClO2 dose at around 1 mg/L, and at that dose the arsenite conversion is negligible. Dosing harder to get the arsenic simply trades an arsenic exceedance for a chlorite one. The tool models the chlorite and chlorate formed and flags the dose ceiling.
Permanganate achieves the same arsenite conversion on the same timescale and forms no halogenated by-products. A manganese-dioxide contact bed does the same with no chemical dose at all, and takes out iron and manganese while it is there. Both carry a manganese control obligation rather than a by-product obligation, and manganese is managed at the filter, inside the plant, where an operator can see it. Ozone works but is a capital item and brings bromate into play on geothermal and coastal-influenced water. Chlorine dioxide and chloramine do not oxidise arsenite fast enough to be arsenic barriers at all.
The tool therefore ranks a permanganate or manganese-dioxide route ahead of a chlorine route wherever the two give a similar treated arsenic, and it flags every chlorine option with the by-product cost it carries. Front-end chlorine remains a legitimate emergency response while a non-halogenated oxidant is being installed, and the tool will model it, but it is not offered as the default answer.
Where measured speciation is not available, the tool estimates it from redox indicators. This is a screening heuristic, not a thermodynamic model, because the As(III)/As(V) couple is kinetically controlled and is routinely far from equilibrium in real water.
Measured, field-preserved speciation is always preferred. Arsenite oxidises in the sample bottle and thioarsenates are unstable, so unpreserved samples systematically under-report the difficult fractions.
Matrix demand is allocated sequentially to sulfide, then iron(II), then manganese, then organic carbon, using conventional stoichiometric ratios. Whatever oxidant survives that is applied to arsenite by second-order kinetics reduced to pseudo-first order.
Second-order rate constants used here place free chlorine, ozone and permanganate as fast oxidants (half-lives of seconds at practical doses), chlorine dioxide and monochloramine as effectively useless for arsenite at plant contact times, hydrogen peroxide as useful only at elevated pH, and dissolved oxygen as negligible. For free chlorine the reactive species is HOCl, so the rate is scaled by the HOCl fraction (pKa 7.54), which is why pre-chlorination slows appreciably above pH 8.
Solid manganese-dioxide media is handled as a surface reaction with a first-order dependence on empty bed contact time rather than a dissolved-oxidant dose.
Every coagulant is defined by four numbers, all of them editable in Settings so the tool can be set to the product actually delivered to the plant: the per cent active ingredient w/v, the specific gravity of the neat product, the mass fraction of metal in the active ingredient, and the pH window over which the coagulant forms floc.
The three dose bases (ppm v/v, mg/L of neat product, mg/L as metal ion) are therefore interchangeable, and all three are reported whichever one is entered. The per cent active is w/v, which is why the specific gravity appears in the conversion to a mass fraction; acids and caustics are quoted w/w and do not carry that term.
The effective pH window is a hard constraint on everything else in the model. Arsenate sorption onto metal hydroxide improves steadily as pH falls, so an isotherm on its own will always recommend a lower pH. Real coagulants do not oblige. Polyaluminium chloride is effective over a narrow band, roughly pH 6.5 to 7.2, and falls apart on either side of it. Aluminium chlorohydrate coagulates well at the high end, around pH 7 to 8. Alum has a wider window and reaches down to about pH 5.8. Ferric salts coagulate across a wide range. Outside its window a coagulant does not form separable floc, so the dosed metal is not available as a sorbent, it does not settle or filter, and it leaves as residual dissolved metal.
The tool applies the window directly. Beyond it, the fraction of dosed metal carried as separable floc tapers from one to a floor value over a settable number of pH units, the sorbent mass in the isotherm is reduced in proportion, and floc carryover is multiplied by the reciprocal. The result is a pH curve with a genuine optimum at the bottom of the window rather than a theoretical optimum somewhere the plant could never operate, and every pH recommendation the tool makes is bounded by the window of the coagulant in use.
Coagulation is modelled as adsorption onto the freshly precipitated hydrous ferric oxide or hydrous aluminium oxide, using a Freundlich isotherm with an exponent of 0.5, solved simultaneously with a mass balance. This gives a closed-form solution.
The Freundlich coefficient is adjusted for pH, for the arsenic species, for competing anions and for contact time:
Kf0 is calibrated so that an iron dose giving a Fe:As mass ratio near 20:1 at pH 7 achieves about 90 percent removal of arsenate, which is the long-standing rule of thumb. Aluminium is set at roughly half the performance of iron per unit metal mass and falls away faster as pH rises, consistent with reported jar-test behaviour. Species affinities scale arsenite, thioarsenate and organic arsenic down from arsenate.
Silica and phosphate are the dominant competitors and the silica coefficient is made pH-dependent, since silicate deprotonates and competes harder as pH climbs. Sulfate, bicarbonate, fluoride and organic carbon are included as weaker competitors.
Separation is then applied as a carryover fraction: the arsenic that is on the floc leaves with the floc, except for the small proportion of solids that reaches treated water. Low-pressure membrane filtration (MF/UF) carries over almost nothing; a clarifier and granular filter carry over a little; direct filtration carries over more.
Media capacity is expressed as an operating capacity in mg arsenic per gram of media at the stated breakthrough, corrected for pH, contact time, species and competition. Bed volumes to breakthrough follow from a mass balance on the bed:
Ion exchange is treated differently, because run length is set by competing anions rather than by arsenic. Sulfate is the controlling competitor, with nitrate and bicarbonate weaker. Bed volumes scale inversely with the total competing anion equivalents in the feed.
This tool provides engineering estimates only. It does not constitute a compliance determination.
Model coefficients. Change these to match your own jar test or column test calibration. Every value is used exactly as shown in the equations set out in the About panel.