DEV Community

Eman Tanveer
Eman Tanveer

Posted on

Building Better Monitoring Systems: From Environmental Sensors to Useful Data

Environmental monitoring has a data problem.

Modern sensors enable continuous data capture, remote monitoring facilitates data collection at global scales, and large-platform storage makes room for petabytes of information.

But raw data capture is only the beginning.

The more interesting engineering challenge for forest or environmental monitoring is to transform spurious values into information, usable by people.

A useful monitoring system requires the establishment and management of a pipeline of activity:

Sensors → Data collection → Validation → Storage → Analysis → Visualization → Decision-making

In this post, you'll find some practical considerations about the aforementioned topics.

1. Starting With Questions, Not Sensors

A classic pitfall in the design of an environmental monitoring system is to begin with your assets.

For example, it might seem logical for an organization to ask themselves:

"What type of sensor are we going to use?"

A better approach is to identify what you're trying to investigate.

For example:

"What environmental mechanism we want to measure?"

Suppose I'm attempting to measure vegetation stresses.

I'd need to consider if things like the soil's moisture, temperature, precipitation levels, humidity, observations about plants, and so on are pertinent in this case.

Essentially, I need to align the measurements that I want to make with the mechanisms of interest.

It helps to avoid the danger of a data project simply having the accumulation of some disparate data points from unrelated or peripheral sensor captures.

2. Dealing With Real-World Messiness

Rarely will the incoming data from an environmental sensor be a clean time series.

In the real world, we can find:

Missing data

Duplicates

Sensor calibration issues

Communication problems

Outliers

Unexpected and unrealistic values

Effects of sensor maintenance

Seasonality

Irregular intervals in samples

A monitoring pipelines should account for some of these factors.

For example, simple validation can flag some of the more clearcut abnormalities:


if value < minimum_reasonable_value:

flag("possible error")

if value > maximum_reasonable_value:

flag("possible error")

Enter fullscreen mode Exit fullscreen mode

Not necessarily reject the value. Unusual data can provide indications of legitimate events in real-world systems.

A more realistic approach is often to first identify them for later review and investigation.

3. Looking Beyond Single Data Points

At its core, monitoring an environmental phenomenon is a time-series analysis.

Generally speaking, singular data points don't tell us much about what's going on in an environmental system, in isolation.

Let's see an example.

Suppose I'm seeing the following values for a measurement of the soil's moisture:

Day 1: 38%

Day 2: 35%

Day 3: 31%

Day 4: 27%

Day 5: 24%

Looking at the values on a day-to-day basis provides some useful information, compared to having no data at all.

If this is part of a time series that is constantly decreasing, the values might provide more insight, depending on external circumstances.

Was there a period without rain? Was the time abnormally hot? Is this site known to be dry during this time of year?

This is why a monitoring system is likely to benefit from storing the timestamps and context around a measurement, rather than simply the value.

A potential structure for a time series record might resemble the following:


{

"timestamp": "2026-09-24T10:00:00Z",

"location": "site-01",

"sensor": "soil-moisture-03",

"value": 24.7,

"unit": "percent",

"quality": "valid"

}

Enter fullscreen mode Exit fullscreen mode

This might sound trivial but being able to carry relevant contextual, timestamp information allows for downstream analysis that would otherwise have been impossible.

4. Sensor Quality Control Considered

Quality control measures should be taken at different stages of the monitoring stack.

Device level

Data level

System level

Battery status

Connectivity

Calibration

Hardware issues

Sensor configuration

Missing values

Duplicates

Unrealistic values

Irregular changes

Units

Data ingestion problems

Database outages

API errors

Communication issues

Processing problems

The idea here is that having clearly defined quality control checks at different layers within your monitoring stack can greatly reduce complexity and troubleshooting time.

If you know that some problems are likely to be device-specific, some database-specific, you can isolate the source of the problem and troubleshoot accordingly.

5. Location: Part of the Data

One key aspect of an environmental monitoring system is that location is a critical enabler.

Sensors can be separated by a few meters can have vastly different measurements due to differences in the local environment around them, such as soil density, vegetation, elevation, slope, exposure, drainage capabilities, and so on.

As a result, the system metadata should carry location information alongside the sensor readings.

A monitoring system can benefit massively from holding or collecting information like:

Sensor ID

Latitude

Longitude

Height/Elevation

Location (Site)

Type of measurement

Installation time

Measurement interval

Calibration information

The ability to provide this kind of meta information becomes significantly more valuable when monitoring systems are designed for multiple sites (rather than a single point).

Once a system can take into account the location in which these sensors exist, it's much easier to integrate data against maps, satellite imaging, weather services, and so on.

6. Know The Difference Between Data Collection And Monitoring

Even if your network of sensors generates thousands of data points everyday, it doesn't immediately give you much monitoring power beyond the initial installation.

Monitoring is an interpretative process.

A more valuable system architecture might resemble something as follows:

Raw sensor data → Quality control → Aggregation → Trend detection → Visualization → Human interpretation.

The algorithms applied at different stages will depend on the use-case, but the principle is simple: raw data isn't immediately valuable monitoring information.

7. Longevity Considerations

Environmental monitoring systems are generally expected to last for months or years.

A consideration for this makes the difference between a useful long-term monitoring site and a short-term prototype.

For example:

A sensor network might work absolutely beautifully in your test environment, only to encounter network issues a month later when it's on an autonomous expedition or in another field of activity.

You want your environmental monitoring system to be reliable, resilient, and robust for an extended period.

The design principles necessary for these qualities are sometimes at odds with more ambitious features, and simple maintenance-robust designs can often provide more long term value than more experimental system designs.

Essentially: it's not very helpful if a technically sophisticated monitoring system cannot last beyond the 12-month mark in the field because of its complexities.

8. Visualizing Useful Information

There isn't much point in displaying a hundred charts if no one can derive value from them.

Some simple questions that the data visualization might try to answer include:

Is there an environmental condition changing?

Are there any anomalies in the patterns coming from certain monitoring sites?

Has a certain measurement fallen within or beyond thresholds?

Are environmental factors shifting together?

Is there suspicious data showing up?

Having an appropriate chart or visualization for the job is far more important than trying to display all possible information.

The objective might be as simple as having a time series chart for time series data.

Once this baseline, it's possible to look to enrich or complement it with more complex displays to further support interpretation needs.

9. Field Observations + Technology = Powerful Insights

One of the biggest challenges in designing environmental monitoring systems is the ability to understand what is driving particular patterns.

A monitoring system will collect measurement values, but the same system cannot necessarily explain what events or conditions may cause them.

For example, an abnormal reading might be due to something like:

A real worldly occurrence.

Sensor movement.

Faulty sensor/equipment.

Local weather.

Seasonal time.

Changes to the local flora.

A good system must be able to support the integration and alignment of automation observations with human observations.

This enables users or managers to make an assessment of how real-world events or conditions may impact the data they're seeing.

This is particularly important for the identification of anomalies.

10. Systems Designed For the End-User

Ultimately, the aim of a monitoring system is to make it easier to support some type of decision with respect to a real-world situation.

It can be, for example, something like:

A decision made by forestry managers

Scientists

Land managers

Researchers

Conservation authorities

Various consulting entities

Depending on whom your system aims to support, it can influence what you build.

It may influence sensor choices, database designs, dashboard structures, and so on.

If you're part of an organization looking for the best technology to explore the world of environmental and forestry monitoring, you might like to learn more about [our] solutions at https://enviroforest.com/.

Conclusion

Building an environmental monitoring system isn't just about connecting sensors to a database.

The true engineering challenge of building one is connecting physical reality with trustworthy information.

That's why it's important to think about:

Measurement → Data quality → Contextual information associated with the measurement → Further processing and transformation steps → Visualizing and interpreting information to get insights.

For developers, environmental forest monitoring represents a fascinating intersection of IoT, time series processing, geospatial information, data science, database management, and environmental expertise.

And perhaps the largest piece of advice for environmental and forest monitoring systems is also simple:

Don't build systems that gather insights.

Build systems that help people achieve their purpose.

Top comments (0)