Jump to Section
.png)
Team Behind TemPro
TemPro: A Low-Cost Temperature Profiling Probe
Looking beneath the streambed
A stream is not limited to the water visible at the surface. Water continuously enters, moves through, and returns from the sediments beneath the channel. This region, known as the hyporheic zone, connects surface water and groundwater and influences stream temperature, nutrient transport, contaminant movement, and ecological functioning.
Despite its importance, hyporheic exchange remains difficult to monitor.
Why Do We Need to Understand the Hyporheic Zone?
HOW IMPORTANT
IS THE HYPORHEIC ZONE?
...CAN WE AFFORD TO IGNORE IT?
Beneath the visible stream lies the hyporheic zone—a dynamic region where stream water enters the sediments, mixes with groundwater, and may later return to the channel. Although hidden, it regulates water temperature, provides habitat, increases water residence time, and supports the transformation of nutrients and contaminants. It can promote processes such as nitrate removal, but under unsuitable conditions it may also store and later release pollutants. Ignoring this zone would therefore provide an incomplete understanding of river flow, ecology, and water quality.
WHY IS HYPORHEIC EXCHANGE DIFFICULT
TO
QUANTIFY,
AND WHY
WILL IT
MATTER IN
THE
FUTURE?
Exchange occurs below the streambed, varies over very short distances, and changes with streamflow, groundwater levels, sediment type, and geomorphic features such as boulders, riffles, meanders, and log jams. A measurement at one point may not represent the surrounding reach, while installing instruments can disturb the sediment and alter natural flow paths. Understanding these exchanges will become increasingly important as climate change, droughts, floods, groundwater extraction, nutrient loading, and emerging contaminants place greater pressure on rivers and their natural ability to regulate temperature and water quality.
WHAT
METHODS
ARE
AVAILABLE
TO
QUANTIFY HYPORHEIC EXCHANGE?
Common approaches include piezometers and hydraulic-gradient measurements, seepage meters, salt or dye tracer experiments, environmental and isotope tracers, pore-water chemistry, temperature monitoring, geophysical surveys, and numerical modelling. Each method measures a different component or spatial scale: hydraulic methods indicate flow potential, seepage meters measure local exchange, tracer tests describe reach-scale storage, chemistry reveals transformation, and temperature enables continuous flux estimation. Because no single method provides a complete picture, there is a need for affordable, continuous, multi-depth monitoring systems that can be deployed at several locations—which motivated development of TemPro
So, if there are already so many ways to study hyporheic exchange, why do we need another approach?
The answer lies in the trade-offs. Some methods provide very detailed measurements but only at a few locations. Others cover larger areas but offer only a snapshot in time. Some require repeated field visits, laboratory analysis, or relatively expensive equipment. The figure below brings these trade-offs together by comparing these existing methods in terms of their temporal capability, spatial capability, and relative resource demand, while showing where temperature-based monitoring fits within the broader toolbox for quantifying hyporheic exchange.
Clearly, no single method performs best in every situation. Temperature profiling, however, occupies a particularly useful space: it can capture changes at minute-scale intervals, continue monitoring for weeks or months, and be deployed at multiple depths and locations. In other words, it offers a valuable combination of high temporal resolution, long-term monitoring capability, useful spatial coverage, and relatively lower resource demand.
These advantages made temperature a natural choice for monitoring hyporheic exchange. The next question was simple: can this kind of monitoring be made more affordable, practical, and easier to deploy across multiple locations?
That question is where TemPro begins.
The process occurs beneath the sediment surface, varies over short distances, and often requires several sensors to capture its spatial and temporal behaviour.
Commercial temperature-monitoring systems provide reliable observations, but the cost of deploying several instruments can limit their use in spatially distributed studies.
TemPro was developed to address this challenge.
TemPro is a low-cost, field-deployable temperature profiling system designed to record streambed temperature continuously at multiple depths. The measurements can be used to examine groundwater–surface-water interactions and estimate vertical water flux through streambed sediments.
Explore TemPro
TemPro combines an above-the-stream control and logging unit with a subsurface probe carrying temperature sensors at selected depths.
The interactive explorer below allows you to move through the instrument step by step.
Through the interactive display, visitors can examine:
-
The complete TemPro instrument
-
The control and logging enclosure
-
The battery-management and logging board
-
The wiring and sensor arrangement
-
The instrumented probe body
-
The sensor interface
-
The waterproof temperature sensor
-
The perforated protective casing
Each highlighted point opens a closer view and a short description of the selected component. Begin with the complete system and select either the control unit or the probe assembly.
Select the highlighted point to look inside the instrument.
How Does TemPro Actually Work?
At first glance, TemPro may look like a pipe connected to an electronics box. Inside, however, several small components work together to continuously watch what is happening beneath the streambed.
Think of it as a small field station with four basic jobs:
Sense the temperature → Know the time → Save the measurement → Go back to sleep.
🧠
The Brain — ESP32
At the heart of TemPro is an ESP32 microcontroller. It acts as the brain of the system. At every programmed measurement interval, it wakes up, collects temperature readings from the sensors, checks the time, organizes the information, and saves it to a microSD card. Once the job is finished, it returns to deep sleep to conserve battery power. So for most of its time in the field, TemPro is actually sleeping.
🕒
🌡
Keeping Track of Time — DS3231
Temperature becomes much more useful when we know exactly when each measurement was recorded. A small DS3231 Real-Time Clock (RTC) keeps track of the date and time independently using its own battery. This ensures every temperature measurement receives an accurate timestamp, even when the main controller is asleep. The recorded data can therefore be arranged as a continuous temperature time series for later analysis.
Measuring Temperature at Several Depths
TemPro uses DS18B20 digital temperature sensors. Each sensor has its own unique digital address. This means several sensors can communicate with the controller using the same data wire, rather than requiring a separate communication line for every sensor. This becomes particularly useful when the temperature needs to be monitored at several depths below the streambed. The sensors provide temperature measurements with a resolution of 0.0625°C and a manufacturer-specified accuracy of ±0.5°C. They are factory calibrated, so their digital output can be read directly as temperature. A small 4.7 kΩ pull-up resistor helps maintain stable communication between the sensors and the controller.
From Electronics To A Field Probe
Of course, measuring temperature beneath a flowing mountain stream is not simply a matter of placing sensors in water. The sensors also have to survive sediment movement, remain at their intended depths, and stay in contact with the surrounding streambed environment.
🛡 Protecting Each Sensor
Each temperature sensor is placed inside a small polypropylene casing measuring approximately 7 cm × 2 cm. The casing contains 6 mm openings, allowing water to move freely around the sensor. A nylon mesh is wrapped around it to help prevent fine sediment from accumulating around
the sensor while still allowing water contact. The casing therefore performs two jobs: it protects the sensor and helps keep it positioned at the required depth.
📏 Building the Profiling Probe
The main probe is made from a 1.5 m long, ¾-inch uPVC pipe.
Sensor casings can be attached at different positions along the probe, depending on the depths that need to be monitored. The sensors are fixed using waterproof epoxy-based adhesive and cable ties, while their wires run upward to the electronics housed inside the logger box. This creates a vertical temperature profile of the streambed rather than relying on a single temperature measurement.
💧 More Than Just Temperature
The lower end of the probe contains 3 mm perforations.
This part can function as a small piezometer, allowing the probe to support water-level observations and water sampling in addition to temperature monitoring.
So the same field installation can provide more than one window into what is happening beneath the stream.
How Does It Store The Data?
Every time TemPro wakes up, the controller combines the sensor measurements with their timestamps and stores them on a microSD card. The data can later be retrieved during field visits without removing the temperature sensors from the streambed. Programming and updating TemPro are performed using the Arduino IDE (https://www.arduino.cc/en/software), making the electronic system programmable and adaptable to different monitoring requirements.
How Does TemPro Save Power?
Long-term environmental monitoring is always a battle against battery consumption. TemPro deals with this by spending most of its time in deep sleep mode.
During deep sleep, the complete system with sensors consumed approximately:
4.3 mA
During data recording, consumption briefly increased to approximately:
54 mA
But this higher demand occurs only for a very short period during each logging event.
For the tested operating cycle, the resulting average current consumption was approximately:
4.44 mA
In simple terms, TemPro wakes only when it has something to measure, records the data, and then goes back to sleep.
And What Does It Cost?
One of the main motivations behind TemPro was to make multi-depth temperature monitoring more accessible.
The prototype development cost was approximately:
₹7,848 (~US$89.6)
More importantly, once the main logger is built, expanding the probe with another temperature sensor costs only about:
₹300 (~US$3.49) per additional sensor
This includes the sensor itself together with the associated wiring and interfacing.
That means increasing the number of monitored depths does not require purchasing an entirely new logging system.
THE IDEA IN ONE LINE
TemPro combines inexpensive digital temperature sensors, low-power electronics, data logging, and a field-ready profiling probe to continuously observe how temperature changes beneath the streambed.
And those temperature changes can ultimately be used to investigate how water is moving between the stream and the sediments beneath it.
From The Lab To The Stream: Does TemPro Really Work?
Building TemPro was only the first step. The more important question was whether a low-cost system built around DS18B20 temperature sensors could provide measurements reliable enough for real hydrological applications.
We therefore tested TemPro in two very different settings.
First, the sensors were evaluated under controlled laboratory conditions to check their accuracy across the range of temperatures expected in the field (following some of the works like). Then, TemPro was installed directly beneath the bed of a mountain stream and compared with commercial HOBO MX TidbiT 400 temperature loggers (fllowing suggestion from a peionerring work :).
In simple terms, we asked two questions:
Are the sensors themselves accurate?
and
Do they continue to perform well when buried beneath a real streambed?
From controlled testing to real-world validation: TemPro was evaluated both in the laboratory and under natural streambed conditions.
FIRST, HOW ACCURATE ARE THE SENSORS?
TemPro uses DS18B20 digital temperature sensors. These sensors are inexpensive and easy to integrate into a monitoring network, but affordability alone is not enough. For hydrological applications, we also need to know how closely their measurements represent the actual temperature.
Five sensors were therefore tested in a controlled circulating water bath at seven temperatures:
5, 10, 15, 20, 25, 30, and 35°C.
This range covers the temperatures likely to occur at the field site.
A traceable reference thermometer was used to determine how far each sensor reading differed from the reference temperature.
Before calibration, all five sensors already remained within the manufacturer's stated accuracy of ±0.5°C. The observed deviations were within approximately ±0.26°C.
Each sensor was then calibrated individually. Following calibration, the residual differences were reduced to approximately −0.075 to +0.05°C, or roughly within ±0.1°C.
Figure shows the variation of residuals of sensors with respect reference sensor Pre- and Post-calibration
Already within specification → improved further through individual calibration
TAKEAWAY
The laboratory tests confirmed that the inexpensive sensors were already operating within their specified accuracy range. Individual calibration could improve their agreement with the reference thermometer even further.
BUT THE LAB IS NOT A STREAM
Laboratory calibration tells us whether the sensor is accurate.
Field validation tells us whether the complete TemPro system remains reliable where it is actually meant to work.
THE FIELD TEST: RINGALI GAD
TemPro was evaluated in Ringali Gad, a second-order mountainous headwater stream in the Western Himalaya. The monitoring site was located along the upstream section of a meander bend in the Mussoorie hills of Uttarakhand, India. The stream provided a realistic and challenging environment for testing the probe.
Stream discharge: 0.01–0.22 m³/s
Stream temperature: 2.9–32.5°C
Streambed: Predominantly cobbles and gravel
Study area: Ringali Gad, Mountain Headwater stream, India
HOW DO YOU INSTALL A PROBE IN A COBBLE–GRAVEL STREAMBED?
Driving a relatively lightweight uPVC probe directly into coarse sediment could damage the probe and its sensors.
So TemPro is not hammered directly into the streambed.
Instead, a temporary galvanised-iron, or GI, pipe acts as a protective installation casing.
The installation method combines ideas from established streambed installation approaches and was adapted to suit TemPro.
The procedure for moving the GI pipe through the streambed may vary depending on the need. In our case, since it was a forested headwater catchment, we tried to manually move the pipe through the sediment.
The steel casing takes the force of installation. TemPro itself is protected inside it.
IF TEMPRO WAS ALREADY CALIBRATED IN THE LAB, WHY TEST IT AGAIN IN THE FIELD?
Laboratory calibration gave us confidence that the DS18B20 sensors were measuring temperature accurately. Under controlled conditions, all sensors were already within the manufacturer-stated accuracy range, and calibration allowed their accuracy to be improved further when required.
But a controlled water bath is very different from a streambed.
Once TemPro is installed in the field, the sensors must operate while surrounded by natural sediment, exposed to flowing water, changing temperatures and installation-related disturbance. Sensor contact with the sediment, the probe assembly, wiring, data logging and the surrounding environmental conditions can all influence how the complete monitoring system performs.
So the purpose of field validation was not simply to calibrate the sensors again. It was to answer a different question:
Can TemPro reproduce the measurements of an established commercial temperature logger under real streambed conditions?
To answer that, we compared TemPro directly with HOBO MX TidbiT 400 temperature loggers.
The HOBO provides a useful benchmark because it is a commercially available, high-accuracy waterproof temperature logger widely suited to environmental monitoring.
So rather than testing TemPro in isolation, both systems were asked to measure the same streambed, at the same depths, during the same period.
These commercial sensors have an accuracy of approximately ±0.25°C, a resolution of 0.01°C, and an IP68 waterproof casing, making them suitable as a reference for field comparison.
TemPro and HOBO sensors were deployed at corresponding depths beneath the sediment–water interface:
7.5 cm 17.5 cm 27.5 cm
TemPro sensors were mounted along the profiling probe, while the HOBO sensors were mounted on a mild-steel flat bar at corresponding depths.
The idea was straightforward:
When the streambed temperature changes, does TemPro see essentially the same change as HOBO?
TemPro began recording on 23 March.
The HOBO loggers were installed later because of transport and site-access constraints. Rather than immediately comparing the instruments after installation, the streambed was given time to settle.
After approximately four days of stabilization, the comparison period used for validation was:
13–21 APRIL
This helped ensure that the comparison represented normal streambed conditions rather than temporary disturbances caused by installing the sensors.
BEFORE LOOKING AT THE COMPARISON... WHAT WERE THE SENSORS ACTUALLY SEEING?
Both instruments were tracking the natural daily heating and cooling of the streambed.
Temperature rises and falls over a roughly daily cycle, producing the characteristic diurnal temperature signal.
Closer to the streambed surface, these fluctuations are stronger. With increasing depth, the signal becomes progressively damped and delayed as heat moves through the sediment.
So, what TemPro is tracking beneath the streambed? Natural daily temperature oscillations change with depth, providing the thermal signals used to investigate subsurface water movement.
Now For The Important Part: Did TemPro Follow HOBO?
YES — VERY CLOSELY.
During the 13–21 April comparison period, TemPro and HOBO recorded remarkably similar temperature patterns.
TemPro closely reproduced the temporal variability recorded by the HOBO sensors at all three depths. The two datasets showed strong agreement in the timing of diurnal peaks and troughs, as well as in the overall magnitude of temperature fluctuations. The close overlap between the temperature time series indicates that TemPro captured both short-term thermal dynamics and depth-dependent temperature variations under natural streambed conditions.
TemPro versus HOBO under the same streambed conditions. The closely overlapping temperature curves show that TemPro reproduced both the timing and magnitude of temperature variations recorded by the commercial reference loggers.
The curves tell an encouraging story—but instrument validation cannot rest on visual similarity alone. The next step is to put that agreement to a quantitative test and ask how strongly the data support TemPro’s performance.
So, let’s now turn to the statistics and ask how closely TemPro agreed with the co-located HOBO sensors when both systems were measuring temperature at the same depths under the same streambed conditions.
R² = 0.98–0.99
Across all three depths
BUT FOLLOWING THE SAME PATTERN IS ONLY PART OF THE TEST
Two instruments could rise and fall together while still giving slightly different actual temperatures.
So we also asked:
How far apart were the actual readings?
The answer was encouraging.
The mean absolute difference between TemPro and HOBO was only:
0.11–0.15°C
depending on sensor depth.
Even when larger deviations were given greater statistical weight, RMSE remained:
0.14–0.17°C
That means TemPro was not simply reproducing the shape of the temperature signal.
It was also producing very similar temperature values.
Where did they DIFFER?
No two instruments measure identically at every instant.
To see exactly when TemPro and HOBO differed, we calculated the difference between them throughout the comparison period.
Pale/white → almost no difference
Blue → TemPro slightly lower
Red → TemPro slightly higher
Most of the heatmap remains pale, corresponding to differences of approximately:
±0.2°C
Occasional deviations approached around:
±0.4°C
particularly during peak daytime temperatures.
Importantly, these remained within the expected ±0.5°C accuracy specification of the DS18B20 sensors used in TemPro.
So the occasional deviations were not unexpected sensor failures; they remained within the operating accuracy expected for the device.
And, R² showed that TemPro and HOBO followed very similar temperature patterns, while MAE and RMSE showed that the differences between their measurements remained small.
All these definitely vouch for the performance of TemPro
BUT CORRELATION AND ERROR ALONE DO NOT TELL THE WHOLE STORY
Does TemPro reproduce the overall magnitude and variability of the HOBO record?
and
Are the two instruments genuinely consistent and in agreement—not simply correlated?
To find out, we also checked NSE and ICC
Now, coming back to the question: Does TemPro reproduce the overall magnitude and variability of the HOBO record? NSE values ranged between 0.94 - 0.96
