Why Cities Are Installing Cheaper Sensors—and Learning Not to Trust Them Blindly
A city may know the temperature at its airport to a tenth of a degree and still have little idea how hot the pavement becomes outside a primary school.
Traditional monitoring networks were designed to produce reliable, comparable measurements over long periods. Their instruments are carefully calibrated, installed according to strict standards and maintained by specialists. This makes the data suitable for weather forecasting, legal compliance and scientific research. It also makes dense coverage expensive.
A handful of professional stations can describe conditions across a region. They cannot capture every overheated courtyard, polluted intersection or flood-prone underpass within a city.
Municipalities are responding by installing much cheaper sensors. Compact devices can now measure temperature, humidity, particulate matter, noise, soil moisture, traffic flows and water levels at a fraction of the cost of reference-grade equipment. Hundreds of sensors can be distributed across neighbourhoods that previously had no direct monitoring at all.
The result is an attractive proposition for city governments: more data, collected more frequently and at a far finer geographical scale. It is also producing a less comfortable discovery. Cheap sensors do not simply provide a lower-cost version of conventional measurements. They produce a different class of evidence, with different weaknesses and more demanding requirements for interpretation.
The most useful urban sensor network may therefore be the one that admits how uncertain its measurements are.
The official station was never designed to describe every street
Professional monitoring networks remain indispensable. Switzerland’s national air-pollution system, for example, measures major pollutants at a limited number of carefully selected sites representing different conditions, including urban streets, residential areas and rural locations. MeteoSwiss operates a much larger network of automatic weather stations equipped with high-precision instruments and formal quality controls.
These systems create the reference against which environmental change can be measured. Their consistency matters because governments use the results to assess legal limits, track long-term trends and compare one period with another.
The same standards restrict where stations can be placed.
A meteorological station intended to produce internationally comparable temperature readings generally needs an open, carefully specified environment. It should not sit beside a dark façade, under an air-conditioning outlet or inside a narrow street canyon. Yet these are precisely the places where people experience urban heat.
MeteoSwiss does not operate conventional weather stations in many dense urban locations because such sites fail to meet international measurement standards. Municipal, cantonal and university networks must therefore fill the gap when planners want to understand conditions at neighbourhood level.
This creates two legitimate but different forms of observation. A reference station tells a city what the weather is under controlled measurement conditions. A street-level sensor can reveal how that weather is transformed by buildings, traffic, shade and vegetation.
The cheaper device does not replace the official station. It allows the city to ask a different question.
Cities need patterns more often than perfect measurements
An urban planner deciding where to plant trees may not need a temperature reading accurate to the final decimal place. The more useful information may be that one street remains consistently hotter than another during the evening, or that a housing estate cools much more slowly than a nearby park.
Low-cost networks are particularly effective at exposing these spatial differences.
A city can place sensors beside schools, care homes, tram stops and public squares. Temperature measurements taken through the day can show which areas accumulate heat and where night-time cooling fails. Air-quality sensors can identify recurring pollution patterns close to busy roads. Water-level devices can detect rapid changes in drainage channels or underpasses before a conventional inspection would take place.
One expensive instrument provides confidence in a single location. A network of inexpensive instruments provides coverage.
That trade-off is often reasonable because urban decisions depend on relative patterns. A municipality may want to rank streets according to heat exposure, identify sites deserving a closer investigation or compare conditions before and after a redevelopment. The network can operate as an early-warning and diagnostic system rather than as a final legal authority.
The danger begins when the city forgets which kind of evidence it has collected.
A coloured dashboard can make provisional data appear definitive. Once measurements are displayed as a map, differences between neighbourhoods acquire an air of precision even where they may partly reflect sensor placement, equipment bias or missing data.
More points on a map do not automatically produce more knowledge.
A sensor measures its surroundings as much as the city
Low-cost devices are highly sensitive to where and how they are installed.
A temperature sensor fixed to a sunlit metal pole may record the heat absorbed by the mounting surface rather than the surrounding air. Another placed beneath dense vegetation may remain cooler than the pavement only a few metres away. A device near an open window, ventilation system or parked vehicle can produce readings shaped by a highly local source.
Height also matters. Conditions close to the ground differ from those at the level where official stations usually measure. This may be useful when the objective is to understand the exposure of a child, pedestrian or building façade. It becomes misleading when the reading is compared directly with a conventional meteorological station.
Air-quality sensors introduce further complications. Humidity can affect readings from optical particulate-matter devices because water changes the apparent size of airborne particles. Some gas sensors react to substances other than the pollutant they are supposed to measure. Components age, become contaminated and gradually drift away from their original calibration.
Two apparently identical devices can therefore produce different values in the same place.
This does not make them useless. It means that procurement is only the beginning of the work. A city that buys 500 sensors has also acquired 500 installation decisions, maintenance obligations and potential sources of measurement error.
The cheaper the hardware becomes, the easier it is to underestimate the cost of making its data credible.
Calibration cannot be completed once and forgotten
A sensor normally leaves the factory with a general calibration. Urban environments rarely resemble laboratory conditions, and the relationship between the sensor signal and the real-world quantity can change after installation.
The most reliable approach is co-location. Low-cost devices are placed temporarily beside a reference-grade station so their readings can be compared. Statistical correction models are then developed to account for systematic over- or underestimation and for the effects of humidity, temperature or other variables.
A correction created at one location may not work equally well elsewhere. A model trained during winter may perform poorly during a hot summer. A device calibrated beside a traffic-heavy street may respond differently in a residential area with another pollution mixture.
Sensors also drift. A network that performs well during its first month may become less accurate after a year of weather exposure, dirt and component ageing.
Calibration must therefore be treated as a continuing process. Some devices need to return periodically to a reference site. Others can be checked against neighbouring sensors or evaluated through algorithms that detect implausible changes. Networks can assign confidence scores based on stability, agreement with surrounding devices and recent calibration history.
The sophistication of the correction system may eventually exceed that of the sensor itself.
Research in Zurich has explored dynamic calibration methods that evaluate each particulate-matter sensor according to accuracy, stability, responsiveness and agreement with the surrounding network. This reflects a broader shift in urban monitoring. Cities are no longer asking only what a sensor measured. They are also asking how much confidence that particular measurement deserves at that moment.
A single number is being replaced by a number accompanied by a judgment about reliability.
A dense network can repeat the same mistake hundreds of times
Large sensor networks appear resilient because they contain many devices. Quantity, however, does not correct a shared defect.
A city may purchase one model of sensor, install it according to one specification and process all measurements through one algorithm. If that sensor responds incorrectly to high humidity, the problem can affect the entire network simultaneously. Hundreds of readings may agree with one another because all of them are wrong in the same way.
This is the difference between precision and accuracy. A network can produce highly consistent data without measuring the underlying condition correctly.
The same problem appears in placement. Sensors are often installed where cities already have access to electricity, connectivity or municipal property. Lampposts, public buildings and transport infrastructure become convenient hosts. The resulting network may have impressive coverage while systematically missing private courtyards, informal gathering places or neighbourhoods with weaker municipal infrastructure.
Data density can therefore conceal social and geographical gaps.
A map based on citizen-owned weather stations presents another form of bias. Households that buy connected devices are not distributed evenly across the city. Some neighbourhoods may contribute dozens of measurements while others remain almost invisible. Devices may also be positioned according to household convenience rather than a scientific sampling plan.
Cities need to understand not only what the network contains, but whom and where it fails to represent.
The intended decision should determine the required accuracy
Not every municipal decision requires the same standard of evidence.
A low-cost temperature network may be sufficient to identify neighbourhoods for a heat-mitigation programme. It may be unsuitable for determining whether a contractor met a legally binding environmental threshold. A noise sensor can alert an authority to recurring disturbances without providing the certified evidence needed for enforcement proceedings.
The distinction should be made before deployment.
Cities frequently begin with the technology: a supplier offers a platform, a funding programme supports smart infrastructure or a department wants to pilot a new sensor. Only later do officials decide how the measurements might be used. This reverses the appropriate sequence.
The first question should concern the decision. Is the network intended to identify patterns, trigger inspections, inform public warnings, evaluate investments or enforce regulation? Each purpose has different requirements for precision, availability, calibration and documentation.
Exploratory monitoring can tolerate greater uncertainty. The network may be designed to locate anomalies worth investigating with professional equipment. Operational systems require stronger guarantees because an error can trigger a costly intervention or fail to detect a genuine danger.
Legal and health-related decisions demand the highest standard. A city should not allow an inexpensive sensor to acquire regulatory authority merely because its dashboard looks convincing.
The relevant comparison is not between cheap and expensive devices. It is between the quality of the evidence and the consequence of getting the decision wrong.
Public dashboards can overstate what the city knows
Publishing sensor data can improve transparency. Residents can see heat conditions, air pollution or noise levels in their own neighbourhood rather than rely on one city-wide average. Community groups can identify local problems and challenge planning decisions with more specific evidence.
The same transparency can create false certainty.
A real-time map rarely shows all the uncertainty behind the colour assigned to a street. It may not explain when a device was last calibrated, whether readings are provisional or how humidity affected the measurement. Users see a red zone and reasonably assume that the difference from an orange zone is meaningful.
Cities need to design dashboards with the same care they apply to the sensing infrastructure. Measurements should be labelled according to their status. Confidence ranges, missing data and suspected device failures should remain visible rather than being smoothed away. A preliminary low-cost reading should not be presented in the same visual language as a verified official measurement.
This is not merely a technical issue. A city can damage public confidence when residents are encouraged to act on data that officials later dismiss as insufficiently reliable.
The reverse is equally problematic. Authorities should not invite citizens to participate in monitoring and then disregard uncomfortable results because the equipment is inexpensive. Low-cost data may not prove a regulatory breach, but it can provide credible grounds for a more rigorous investigation.
Transparency requires the city to explain what the network can establish and where its knowledge ends.
Maintenance is the part procurement plans often omit
The initial economics of low-cost sensors are compelling. A municipality can deploy many more devices within the budget required for one professional station.
The comparison becomes less favourable when the full operating system is included.
Devices need power, connectivity, weather protection and secure mounting. Batteries must be replaced. Communications fail. Firmware requires updates. Sensors disappear, become damaged or record implausible values. Metadata about location, height, surrounding surfaces and maintenance history must remain accurate.
Data infrastructure creates additional costs. Measurements need to be transmitted, stored, cleaned and documented. Algorithms must identify gaps and anomalies. Someone must decide whether a sudden change represents an environmental event or a broken device.
Networks also produce organisational dependencies. A supplier may control the platform, calibration model and data format. Replacing the hardware can become difficult if historical data cannot be transferred or correction methods are proprietary. A cheap sensor can therefore introduce an expensive form of vendor lock-in.
Municipalities should evaluate the cost of a reliable measurement over several years, not the purchase price of the device. A network that gradually deteriorates without anyone noticing is worse than a smaller system whose limitations are understood.
The real asset is not the sensor. It is the maintained chain connecting physical measurement to a defensible decision.
Switzerland is beginning to standardise the new urban networks
Swiss cities provide a useful view of the transition because the country combines established national monitoring systems with growing municipal demand for neighbourhood-level climate information.
Urban heat is particularly difficult to capture through conventional stations. Built surfaces store energy during the day and release it slowly at night, producing large differences between dense neighbourhoods and surrounding areas. Street canyons and places with little vegetation may become substantially hotter than nearby open sites.
Municipal and academic networks have therefore expanded urban temperature monitoring. Affordable weather devices and citizen stations can contribute enough observations to produce more detailed heat maps than the official network alone.
The national response is not to treat these measurements as equivalent to professional stations. A current Swiss initiative is developing best-practice guidance for low-cost urban climate networks, covering installation, sensor characteristics, metadata, quality control and data formats. The emphasis on common standards reveals where the sector has reached: cities have demonstrated that cheap networks can be useful, but comparability can no longer be assumed.
Standardisation is necessary because each municipality otherwise risks building a unique system whose results cannot be compared with another city or even with its own future network. A sensor replacement, software update or change in mounting practice can break continuity in the data series.
Common guidance will not eliminate uncertainty. It can make the uncertainty identifiable, documented and manageable.
Cheap sensors are most valuable when paired with expensive ones
The strongest urban monitoring system is not a choice between dense low-cost networks and sparse professional stations. It combines them.
Reference instruments provide accuracy, continuity and an anchor for calibration. Low-cost sensors provide spatial resolution and the ability to observe environments that formal standards deliberately avoid. Mobile measurements, satellite data and statistical models can add further layers.
Each component corrects a weakness in another.
A cheap network can indicate where a city should place a temporary reference instrument. The reference station can determine whether the pattern is genuine and improve the calibration of surrounding devices. Models can then extend the observations into areas where no sensor is installed.
This layered approach is more credible than expecting one technology to perform every function. It also allows cities to allocate budgets according to risk. Critical locations may receive professional instruments, while broader screening is handled by inexpensive devices.
Sensor networks should be judged by the decisions they improve rather than by the number of devices installed. A city does not become more intelligent because thousands of objects are connected to the internet. It becomes better informed when those objects reveal something useful that the existing system could not see.
Trust should be engineered rather than assumed
Low-cost sensors have opened parts of the city that were previously difficult to measure. They can reveal how environmental conditions vary from one block to the next, show where public investments are underperforming and give residents access to information that once belonged almost exclusively to specialist agencies.
Their weakness is not that they are cheap. It is that their apparent simplicity disguises a complicated process of calibration, maintenance and interpretation.
A city that understands those limits can use inexpensive sensors confidently. It can distinguish a pattern from a verified measurement, pair provisional data with professional investigation and publish results without overstating their certainty.
A city that ignores the limits may produce a sophisticated map of measurement errors.
The mature smart-city approach is therefore neither enthusiasm nor distrust. It is conditional trust: confidence supported by reference stations, metadata, calibration histories, uncertainty estimates and a clear definition of what each reading is permitted to decide.
Cheap sensors allow cities to observe more. Learning when not to believe them may prove just as valuable.
