Which Shrimp-Pond Water-Quality Sensors Are Worth Installing in India, and Which to Skip
The verdict first. On a brackish-water shrimp pond in India, install temperature and dissolved oxygen first, add pH and salinity second, and stop there until the farm can show that it acts on every alert those four produce. Treat ammonia, nitrite, turbidity and ORP probes as optional extras that most farms are better served by replacing with a weekly test routine and a handful of cheap calculations. The exception is a large, well-staffed operation that already has the calibration discipline to keep an ion-selective probe honest.
That is an uncomfortable verdict for anyone selling a seven-sensor "complete water-quality station," and it comes from one question applied to every probe: what can the farmer do within an hour of this number changing? Oxygen has an answer (start the aerators). Temperature has an answer (change the aeration, top up water, adjust feeding). pH and salinity have answers that arrive a little slower. Ammonia, on a pond at 15 to 25 parts per thousand salinity, usually has an instrument that cannot measure the thing it is named after, and an answer that depends on two other sensors anyway.
This post ranks the candidates against that question, shows the arithmetic that separates the useful ones from the decorative ones, and ends with a recommended sensor set, an alarm-rule engine you can run, and a calibration schedule a farm crew will actually follow. It follows our earlier post on dissolved-oxygen monitoring, which covers the single most important parameter in depth. Here the question is everything else.
A sensor earns its place on a pond by what a farmer can do within an hour of its reading changing. A number that only explains yesterday belongs in a lab report, not on a bund.
What the Guidance Asks Farmers to Watch
Before ranking sensors, it helps to know what the people who write extension guidance for Indian shrimp farmers tell them to measure. The Central Institute of Brackishwater Aquaculture (CIBA) lists the parameters that should be monitored routinely during a culture period as temperature, pH, salinity, dissolved oxygen and transparency. Its guidance sets pH at 7.5 to 8.5, with a daily swing of no more than 0.5 units; a daily salinity variation of no more than 5 parts per thousand; transparency, measured with a Secchi disc, between 25 and 35 centimetres; total ammonia nitrogen no higher than about 1 mg/L and nitrite nitrogen no higher than about 0.5 mg/L; and notes that any detectable hydrogen sulphide is undesirable.
Other published sources agree on the shape but differ on the numbers, which is worth knowing before anyone writes a threshold into firmware. Temperature guidance for the whiteleg shrimp (Litopenaeus vannamei) clusters around 28 to 32 degrees Celsius, with 25 to 30 for black tiger shrimp. Salinity guidance for vannamei is quoted variously as 10 to 25, 15 to 25, or 15 to 30 parts per thousand, with the species tolerating a far wider range. Ammonia guidance ranges from a TAN ceiling of 1 mg/L to 2 mg/L, and toxic unionised ammonia limits from 0.01 to 0.1 mg/L depending on the source and the life stage. Whatever number a farm adopts, it should be written down once, with its source, and reviewed with the farm's own aquaculture adviser.
Parameter | Typical target for vannamei (sources vary) | Daily swing to worry about | What moves it |
Temperature | 28 to 32 C | Surface-to-bottom gap; sudden drops after rain | Weather, depth, aeration, water exchange |
Dissolved oxygen | Above about 4 to 5 mg/L, ideally higher | Pre-dawn minimum, not the daily average | Bloom density, biomass, feeding, aeration |
pH | 7.5 to 8.5 | More than 0.5 units in a day | Photosynthesis, alkalinity, rain, liming |
Salinity | 10 to 25 ppt (wide tolerance) | More than 5 ppt in a day | Rain, evaporation, exchange, source water |
Total ammonia nitrogen | Roughly 1 mg/L or less (sources vary) | Rising trend after bloom crashes | Feed load, mortality, bloom dieback |
Nitrite nitrogen | Roughly 0.5 to 1 mg/L or less | Rising trend late in the crop | Nitrification balance |
Transparency (Secchi) | 25 to 35 cm | Sudden clearing or darkening | Plankton density, suspended mud |
Hydrogen sulphide | Not detectable | Any detection | Anaerobic sediment |
Read the right-hand columns, not the targets. Most of the damage in a pond comes from swings and trends, and a good sensor programme is a swing-and-trend detector first and a threshold alarm second.
The Four Tests Every Probe Has to Pass
Apply the same four questions to each candidate, in order, and drop the probe at the first failure.
Does it measure the thing it is named after, in this water? Brackish ponds are not the matrix most probes were calibrated for.
What drifts, fouls or consumes itself, and how fast? In warm, nutrient-rich water, "maintenance-free" means "maintained by someone you do not employ."
Is the instrument's error smaller than the decision range? A probe with an error band wider than the interesting range cannot support a decision, however good its dashboard looks.
Can a farmer act on it within an hour, and does he know how? An alert that nobody can act on teaches the crew to ignore all alerts.
Sensor | Measures what it claims? | Drift and fouling | Error vs decision range | Actionable within an hour? | Verdict |
Temperature (thermistor or RTD) | Yes | Very low | Excellent | Yes | Install first, two depths |
Dissolved oxygen (optical) | Yes, with correct salinity setting | Low, with cleaning | Good | Yes, the most valuable | Install first (see our DO post) |
pH (glass electrode) | Yes, if the junction is healthy | Moderate; junction clogs, reference ages | Good for swings, weaker in absolute terms | Yes, slowly (aeration, water exchange, liming plan) | Install second, with a calibration routine |
Salinity (conductivity cell) | Yes | Low with cleaning | Good | Yes (after rain, before water exchange) | Install second |
Ammonia (ISE) | Poorly in brackish water | High | Often wider than the whole decision range | Only with pH and temperature | Skip; test weekly and calculate |
Nitrite, nitrate, H2S (online) | Not practical | n/a | n/a | n/a | Skip; test kit |
Turbidity (optical) | Yes, but it cannot tell mud from bloom | Moderate; lens fouling | Fine for trends | Partly | Optional |
ORP | Reads something, interpretation is shaky | Moderate | Hard to define | Rarely | Skip for automation |
The rest of the post defends each row.
Temperature: The Cheapest Sensor and the One Most Often Done Badly
A thermistor or platinum RTD in a stainless sheath is the least troublesome sensor you can hang in a pond. It does not foul in a way that matters, it does not drift appreciably, and the accuracy of a modest part is better than the pond needs. Whatever else a farm installs, it should install this.
The mistakes are placement mistakes. A single sensor just under the surface reads the warmest water in the pond on a sunny afternoon and tells the farmer almost nothing about the bottom, where the shrimp spend their time and where the sediment respires. A pair of sensors, one at about 30 centimetres and one near the bottom, gives something useful that no single sensor can: the stratification gap. A pond whose surface is two or three degrees warmer than its bottom in the afternoon is a pond that is stratifying, and a pond that stratifies mixes violently at night or after rain, which is when oxygen collapses.
Temperature also corrects other measurements. Dissolved-oxygen saturation, the ammonia equilibrium discussed below, and conductivity-to-salinity conversion all need the water temperature, so a dependable temperature channel makes every other probe more trustworthy. Many multi-parameter probes include one. Check that the one on your pH probe is not the only one, because a probe in the sun reads its own body temperature.
Install temperature at two depths. The gap between them predicts the night-time mixing that crashes oxygen, and a single near-surface probe cannot see it.
pH: Valuable, Provided Someone Owns the Calibration
pH is the second parameter where the guidance is explicit about swing: no more than 0.5 units in a day. In a pond with a phytoplankton bloom, pH rises through the afternoon as photosynthesis strips carbon dioxide out of the water and falls overnight as respiration puts it back. A pond that swings from 7.6 before dawn to 8.9 at four in the afternoon is not unusual and not harmless, and a farmer visiting once a day at a fixed hour will see only one end of the swing.
A glass pH electrode is a sound sensor with three well-known failure modes in pond water. The reference junction clogs with biofilm and fine particles, which slows the response and shifts the offset. The reference electrolyte depletes over weeks to months, and the electrode's slope declines with age. And the glass bulb coats, which slows response. A vendor guide for aquaculture probes puts routine calibration checks for pH and ion-selective sensors at every 4 to 8 weeks depending on fouling, and a station maker lists monthly two- or three-point calibration against pH 4.01, 7.00 and 10.01 buffers with junction replacement and reference rehydration. Those intervals fit the experience of crews who actually run these systems: a pH probe left alone for a quarter reads confidently and wrongly.
Two practices keep pH useful at low cost. Calibrate against fresh buffers, not against the handheld, because two uncalibrated instruments agreeing proves nothing. And cross-check a probe against a handheld meter or a colorimetric test weekly, logging the difference, because a growing gap tells you the probe is ageing before it fails.
Why pH Is Also the Ammonia Sensor You Already Own
The strongest argument for pH is that it turns a cheap, weekly ammonia test into a real-time toxicity estimate. Total ammonia nitrogen (TAN) is a mixture of ionised ammonium, which is comparatively harmless, and unionised ammonia, which is the form that damages gills. The proportion in the toxic form depends on pH and temperature, steeply. A standard equilibrium relation (Emerson and colleagues, 1975) gives the unionised fraction, and the calculation takes a few lines:
# nh3_fraction.py -- unionised ammonia from TAN, pH and temperature
def nh3_fraction(temp_c: float, ph: float) -> float:
T = temp_c + 273.15
pka = 0.09018 + 2729.92 / T # Emerson et al. (1975), freshwater form
return 1.0 / (1.0 + 10 ** (pka - ph))
def nh3_n(tan_mg_l: float, temp_c: float, ph: float) -> float:
return tan_mg_l * nh3_fraction(temp_c, ph)
for t, p in [(20, 8.0), (28, 7.5), (28, 8.0), (30, 8.5), (30, 9.0), (32, 8.8)]:
print(f"{t} C pH {p}: {nh3_fraction(t, p)*100:4.1f}% unionised; "
f"TAN 1.0 mg/L -> NH3-N {nh3_n(1.0, t, p):.3f} mg/L")
Running it gives the numbers in the table below.
Water temperature | pH | Fraction of TAN as unionised NH3 | NH3-N at TAN 1.0 mg/L |
20 C | 8.0 | 3.8 % | 0.038 mg/L |
28 C | 7.5 | 2.2 % | 0.022 mg/L |
28 C | 8.0 | 6.5 % | 0.065 mg/L |
30 C | 8.5 | 20.2 % | 0.202 mg/L |
30 C | 9.0 | 44.5 % | 0.445 mg/L |
32 C | 8.8 | 36.7 % | 0.367 mg/L |
The same one milligram per litre of TAN is a non-event at pH 7.5 before dawn and several times a typical toxicity guideline at pH 8.8 in a hot afternoon. A farm that measures TAN weekly with a test kit and pH continuously can compute the toxic fraction every five minutes at zero marginal cost, and can raise a warning on the afternoon pH peak when the TAN figure from last Tuesday is still "within limits." That one calculation does more than most ammonia probes, which is the segue into the next section. (The relation above is the freshwater form; salinity shifts the equilibrium slightly, so treat the output as an estimate and use the farm adviser's preferred correction if one exists.)
Weekly TAN by test kit plus continuous pH and temperature gives you a live unionised-ammonia estimate for free. The estimate is more useful than a direct ammonia reading in most brackish ponds.
Salinity: A Robust Sensor That Earns Its Keep Twice
Conductivity is a simple, robust measurement, and a conductivity cell converts to salinity with a temperature-compensated formula. It fouls less destructively than electrochemical probes, though the electrode surfaces still need cleaning. In a coastal Indian pond its value is practical: salinity falls after rain and rises with evaporation in dry weather, and the CIBA guidance asks for no more than 5 parts per thousand of variation in a day. A heavy shower can drop pond salinity sharply and at once, and an alert at that moment tells the farm to hold off on water exchange, check pH and alkalinity, and watch the animals.
The second use matters more than people expect. Salinity is an input to the dissolved-oxygen calculation, because the same percent saturation corresponds to fewer milligrams per litre in saltier water. Our dissolved-oxygen post worked through the error: a sensor left at the freshwater default overstated oxygen by about 18 percent in a 30 ppt pond. A live salinity channel can feed that correction automatically, which removes a manual setting that farms forget to update after the monsoon. A salinity probe therefore protects the single most important measurement on the pond, in addition to its own alarm.
Ammonia: The Seductive Sensor That Misleads Most
An online ammonia sensor sounds like the obvious luxury: ammonia kills shrimp, so measure it continuously. In freshwater and low-salinity systems it is a reasonable idea. In brackish water it is a trap, for reasons that sit in the physics of the sensor.
Most online ammonia probes are ammonium ion-selective electrodes (ISEs). The membrane responds to ions of a similar size and charge to ammonium, which means it also responds to potassium and sodium. The effect is small in freshwater and large in seawater. A manufacturer's guide to ion-selective sensors for water-quality sondes puts it bluntly: sodium ions in sea water can produce a false ammonium reading of up to about 12 mg/L N with no ammonium present, and for this reason ammonium and nitrate sensors do not work in sea water. A 2025 laboratory paper on ammonium electrodes makes the same point in engineering terms: practical ISEs are susceptible to interference from pH, temperature and potassium in complex water, and the proposed remedy is a multi-sensor compensation scheme, which means more probes and more calibration, not fewer.
The specifications of the instruments themselves tell the rest of the story. One manufacturer's ammonium ISE lists an accuracy of plus or minus 10 percent of reading or 2 mg/L, whichever is greater. Another station maker lists plus or minus 10 percent of reading or 1 mg/L. Put those numbers next to the decision range: a farm adopting the CIBA ceiling is trying to distinguish 0.5 mg/L from 1 mg/L TAN. An instrument with an error band of 1 to 2 mg/L cannot tell those apart. Drift compounds it. ISEs are described in several sources as drifting from calibration in unstable ways because of membrane ageing, biofouling and junction poisoning, with recommended calibration checks ranging from monthly to every few hours depending on the application and the claimed accuracy, plus membrane replacement every six months or so.
An honest pond-side ammonia sensor would need: a membrane that tolerates 10 to 30 ppt salt, a potassium and sodium compensation scheme validated for brackish water, temperature and pH inputs for the equilibrium, an antifouling system, and a calibration routine tighter than most farms will keep. Some vendors sell systems with these features, with claimed performance a farm can test against a lab method on its own water. That is the right way to evaluate one: a month of side-by-side comparison against a colorimetric test on the same pond, not the datasheet.
Until a farm has done that, the better design is the one in the previous section: weekly TAN by a reliable test method, continuous pH and temperature, a computed unionised fraction, and an alarm on that calculated value.
In brackish water, a probe named "ammonia" is often an ammonium-and-sodium probe with a confident dashboard. Verify it against a lab or kit method on your own pond before you let it drive a decision.
Turbidity, ORP, Nitrite and Hydrogen Sulphide
Turbidity. An optical turbidity sensor reports scattered light in NTU. Shrimp farmers use transparency, a Secchi disc reading of 25 to 35 centimetres, as a proxy for plankton density, and the two measure different things. A sensor cannot tell a green bloom from suspended mud, and both reduce transparency. As a trend channel (an abrupt clearing is the classic sign of a bloom crash, which is followed by an ammonia and low-oxygen risk) it has value, and as a number to put a threshold on, it does not translate cleanly to Secchi depth. It is optional. The Secchi disc itself is a hand-held instrument that costs almost nothing, and a farm crew that logs it twice a day gets most of the benefit.
ORP. Oxidation-reduction potential sensors read a voltage that responds to the redox state of the water, and in a pond that state is a messy mixture of dissolved oxygen, organics and sulphides. The number is hard to interpret in the water column and more useful in the sediment, where a strongly negative value flags anaerobic conditions. It does not map to an action a farmer takes within the hour. Skip it for automation.
Nitrite, nitrate and hydrogen sulphide. No inexpensive, robust online sensor serves these in a brackish pond. Nitrite tends to matter late in the crop when nitrification is unbalanced, and hydrogen sulphide is a sediment problem found by smell and by test strip. A weekly test kit routine, with a daily one in high-density ponds in the late crop, covers them. Where the budget exists, put it into more test frequency and better lab verification, not into probes that cannot be maintained.
A Recommended Set, by Farm Type
The right set depends on staffing and scale more than on technology. Three tiers capture most farms.
Tier | Who it fits | Install | Manual routine | Why this and no more |
Starter | Small farm, owner on site, a few ponds | Optical DO and temperature at two depths on the worst ponds | Secchi twice daily, pH and TAN kit each morning and weekly, salinity refractometer weekly | Covers the oxygen crash and stratification; everything else is a cheap manual check |
Standard | Mid-size farm, a supervisor and a night crew | Starter set plus pH and conductivity on each pond | Weekly TAN, nitrite and alkalinity kit; weekly handheld cross-checks of every probe | Adds the two swings the guidance names; computes unionised ammonia from live pH |
Extended | Large farm with a technician and a lab routine | Standard set plus a validated ammonia probe on selected ponds, turbidity as a trend channel | Month-long side-by-side validation before any ammonia alarm is trusted; daily Secchi | Only worth it with a technician who can keep ISEs honest |
Costs follow the same ladder, and farms in Andhra Pradesh and Tamil Nadu will have their own quotes. The point is not the price of the kit. It is the cost of the visits to keep the kit right, which is the cost most proposals leave out.
Alarm Rules That Use the Sensors You Actually Installed
A rule engine for these four sensors, plus a weekly TAN entry, is short enough to read in one sitting. It checks swings and trends as well as absolute values, and it computes the unionised ammonia estimate. Thresholds are the farm adviser's to set. The ones below are placeholders drawn from the guidance above.
# pond_rules.py -- swing-and-trend alarms for temperature, pH, salinity and ammonia
from dataclasses import dataclass
from typing import Optional
@dataclass
class Reading:
t: float # unix seconds
temp_surface_c: float
temp_bottom_c: float
ph: float
salinity_ppt: float
def nh3_fraction(temp_c, ph):
T = temp_c + 273.15
return 1.0 / (1.0 + 10 ** (0.09018 + 2729.92 / T - ph))
def evaluate(history: list[Reading], tan_mg_l: Optional[float],
ph_swing_limit=0.5, sal_swing_limit=5.0, nh3_limit=0.1,
strat_limit_c=2.5, temp_high_c=32.0) -> list[str]:
alerts = []
now = history[-1]
day = [r for r in history if r.t >= now.t - 86400]
if len(day) >= 12: # need a few hours of data
if max(r.ph for r in day) - min(r.ph for r in day) > ph_swing_limit:
alerts.append("PH_SWING: daily range above %.1f" % ph_swing_limit)
if max(r.salinity_ppt for r in day) - min(r.salinity_ppt for r in day) > sal_swing_limit:
alerts.append("SALINITY_SWING: above %.0f ppt in 24 h (rain or evaporation?)" % sal_swing_limit)
if now.temp_surface_c - now.temp_bottom_c > strat_limit_c:
alerts.append("STRATIFICATION: surface %.1f C above bottom; expect night mixing" %
(now.temp_surface_c - now.temp_bottom_c))
if now.temp_surface_c > temp_high_c:
alerts.append("TEMP_HIGH: surface %.1f C" % now.temp_surface_c)
if tan_mg_l is not None:
nh3 = tan_mg_l * nh3_fraction(now.temp_surface_c, now.ph)
if nh3 > nh3_limit:
alerts.append("NH3_ESTIMATE: %.3f mg/L N at pH %.2f (TAN %.2f from last test)" %
(nh3, now.ph, tan_mg_l))
return alerts
Three design choices deserve a sentence each. The pH alarm looks at the range over 24 hours, because the guidance concerns daily variation and a single reading cannot show it. The ammonia estimate uses the most recent manual TAN and live pH and temperature, so the afternoon pH peak triggers a warning that a weekly test could not. And every alert string says why, so the person who receives it knows what to check first.
What to Send, and How Often
Sensor choice settles half the data budget. The other half is the telemetry design, and the same restraint applies. A pond node that reports every reading every minute fills the radio channel and drains the battery to produce data nobody reads. Temperature and conductivity change slowly. pH and dissolved oxygen swing through the day and need more attention in the hours around dusk and dawn. A sensible schedule reports each channel at the rate its dynamics justify, and sends an immediate event message when a rule fires.
// 16-byte periodic uplink for a four-sensor pond node (LoRa or similar low-rate link)
typedef struct __attribute__((packed)) {
uint32_t seq; // monotonic; gaps reveal lost uplinks
int16_t temp_top_x100; // 0.01 C
int16_t temp_bot_x100;
uint16_t do_pct_x10; // percent saturation x10 (convert to mg/L at the gateway)
uint16_t ph_x1000; // pH x 1000
uint16_t cond_uS_cm; // conductivity, microsiemens per cm (salinity computed at gateway)
uint8_t flags; // bit0 probe fault, bit1 cleaning due, bit2 low battery, bit3 alarm active
uint8_t batt_dv; // battery in 0.1 V steps
} pond_uplink_t; // 16 bytes packed
Send percent saturation and conductivity, and compute milligrams per litre and salinity at the gateway, where the salinity and temperature compensation can use the farm's own settings. Daytime reporting every 15 minutes and every 5 minutes from sunset to two hours after sunrise keeps the pre-dawn window well sampled without flooding the link. A "cleaning due" flag, raised by a timer in the node, turns the weekly brush-down from something a person must remember into something the system asks for.
A Calibration and Cleaning Schedule a Crew Will Follow
Maintenance schedules fail when they are written for a technician who does not exist. This one is written for a farm crew with a bucket, a bottle of buffer and a soft brush, and every item has a failure sign that someone can see without an instrument.
Item | Interval | Action | Failure sign to look for |
All probes: surface cleaning | Weekly | Wipe with a soft brush and clean water; clear the copper guard or antifouling mesh | Slimy film, algae on the sensing face |
Handheld cross-check | Weekly | Read the same pond at the same depth with a trusted handheld; log the gap | Gap growing week on week |
Optical DO | Monthly | Water-saturated-air check; replace the sensing cap on schedule | Slow response after aerator starts |
pH probe | Every 4 to 8 weeks | Two-point calibration with fresh buffers; check the junction; top up reference fill if refillable | Slow, drifting, or sluggish after rinse |
Conductivity cell | Every 4 to 8 weeks | Check against a standard solution; clean electrodes | Reads high after rain or fouling |
Temperature probes | Quarterly | Ice-bath check at 0 C and compare the pair | Two probes disagreeing in still water |
Any ISE (if installed) | Per vendor, often monthly or more | Multi-point calibration against standards near operating temperature; membrane replacement | Sudden jumps, or readings that do not follow handheld kit tests |
Secchi disc | Twice daily | Log depth and time of day | Missing entries |
Treat the weekly cross-check as the master control. If the gap between a probe and the handheld is stable and small, the probe is fine. If it grows, the probe is the problem, whatever its own diagnostics say.
Seven Sensors, Three That Mattered
A representative composite, genericised and not a named client, shows how this plays out. A 12-pond brackish-water farm bought a multi-parameter station for each pond: temperature, DO, pH, conductivity, ammonia (ISE), turbidity and ORP. Seven probes times twelve ponds is 84 sensing points, each with its own cleaning and calibration schedule, plus telemetry for all of them.
The first month looked impressive on the dashboard, and the first two months looked like a fault-finding exercise. The ammonia channel read 3 to 6 mg/L N on days when the kit test said 0.5, and flat-lined after the first heavy rain. The ORP channel produced a number the supervisor could not link to anything. The turbidity sensors fouled within ten days on the ponds with the densest blooms. The pH probes drifted by the second month, which nobody noticed until a handheld comparison during an unrelated visit. Of the original alert rules, about half were ignored by the night crew because nobody knew what to do about them.
The farm then did something unfashionable: it dropped sensors. It kept temperature (two depths), optical DO, pH and conductivity on every pond, a daily Secchi reading and morning kit tests logged on the phone, and it added the calculated unionised-ammonia rule from this post. The ammonia, ORP and turbidity probes were retired or kept on two ponds as a month-long validation exercise. Maintenance dropped to a weekly cleaning round and a monthly calibration day. The number of alerts per pond per week fell sharply, and the proportion that led to an action rose. The figures behind this account are illustrative of the pattern we see on pilots and are not a published result; the pattern is what matters: fewer sensors, each one maintained and acted on, beat more sensors, none of which anyone trusted.
When to Skip Sensors Entirely
A farm with two ponds and an owner on site. A disciplined test routine with a pre-dawn oxygen check may beat a system that needs its own maintenance. Do the arithmetic for each farm honestly.
A farm with no reliable power or signal at the pond. Sensors with nobody to read them are decoration.
A farm that has not yet fixed the basics. If aerators are undersized or the feed management is poor, probes will report the problem more elegantly and fix nothing.
A farm that will not clean the probes. Dirty probes are worse than none, because they give confident wrong numbers. If the crew will not brush them weekly, do not buy them.
A case where the instrument cannot answer the question. Do not buy an ammonia probe to tell 0.5 from 1 mg/L if its error is 1 to 2 mg/L.
Starting on Monday
Write down the farm's thresholds with their sources. One page: target ranges, swing limits, and who decided them.
Install temperature at two depths and optical dissolved oxygen on the ponds that have crashed before.
Add pH and conductivity on each pond, with salinity feeding the dissolved-oxygen conversion.
Set up the weekly cross-check: a handheld meter, a log, and a person whose job it is.
Compute unionised ammonia from weekly TAN, live pH and temperature, and alarm on the result.
Run any ammonia probe as a month-long validation against kit or lab tests on your own water before it drives an alarm.
Write the response to every alert on one card: what to check first, who to call, what to switch on.
Review monthly: which alerts led to actions, which were ignored, and which sensors were never trusted. Retire the ones nobody uses.
The Hour That Decides It
Every probe on this list would pass a bench test, and most of them would pass a first week in the pond. The ones to install are the ones that still deserve trust in month six, when the weather has turned, the biofilm has grown and the person who installed them has moved on to the next farm. A farmer standing at the bund at three in the morning needs a short list of numbers he believes and knows how to act on. Give him four sensors he trusts before you give him seven he doubts.
EurthTech is an engineering-first product design and embedded systems company with a development center in Hyderabad. To discuss a project, write to connect@eurthtech.com or see https://www.eurthtech.com/case-studies.
Comments