Case study

UPM and UC3M: hydrology measured sensor by sensor

Research groups at Universidad Politecnica de Madrid and Universidad Carlos III de Madrid built their own dense NB-IoT network, so soil moisture and water levels are recorded through every storm, in near real time, on battery power alone.

Dense 50 x 50 m sensor grid
Storm events in near real time
Battery-only field deployment
Peer-reviewed and open access

Client: UPM and UC3M
Sector: Hydrology research
Location: Madrid, Spain
Connectivity: NB-IoT
Hardware: ESP32 NB-IoT nodes
Deployment: Thinger.io Cloud

Who they are

Hydrologists and computer scientists

The project joins hydrologists from the Universidad Politecnica de Madrid, who study what a catchment does during and right after a storm, with sensor network and data fusion specialists from the Universidad Carlos III de Madrid.

Their starting point: hydrology needs high-quality field data, and traditional networks are sparse, poorly connected and short on battery. So the team designed, calibrated and deployed its own low-cost sensors.

The project

Two universities, one network

Hydrology, UPM: CEIGRAM, the Complex Systems Group and the Department of Agroforestry Engineering
Computing, UC3M: Department of Computer Science and the Applied Artificial Intelligence Group
Field site: Parque Agrourbano de Valdebebas, Madrid
Catchments: 6.1 and 8.6 ha of crops and grasslands, loamy sand, about 6% slope
Published: Environmental Modelling & Software 182 (2024)

The network

What the network measures

A soil moisture node every 50 metres, one level gauge at each catchment outlet, and every device reporting over NB-IoT on a duty cycle the team tunes remotely. The sheet below redraws the measurements the researchers work with, as described in their published study.

Hydrological network measurements · As the researchers see themSHEET HY-01 · REV A · STYLIZED

SMS

Soil moisture, twin probes

FIG. 01

Each node carries two independent probes buried at 10 cm and read through an RC circuit, so every point yields a calibrated pair of records whose divergence the study itself tracks.

2 PROBES · RC CIRCUIT · 12-BIT ADC · 10-READING AVERAGE

WLV

Water level at the outlet

FIG. 02

One ultrasonic gauge per catchment, mounted over the downstream end point, closes the water balance by watching the stage of the channel.

A02YYUW · 40 kHz · 25-30 CM TYPICAL RANGE

GRD

A field, not a point

FIG. 03

Sensors on a regular 50 x 50 m grid turn point readings into a moisture map of the whole catchment through Voronoi and inverse distance interpolation.

50 x 50 M GRID · VORONOI · INVERSE SQUARED DISTANCE

NBT

A duty cycle worth days

FIG. 04

Every node wakes, averages its readings, transmits and hibernates for 15 minutes. If the server cannot be reached within five minutes, it forces itself back to sleep.

NB-IOT · QUECTEL BC660K · 15 MIN CYCLE · FORCED HIBERNATION

PWR

Battery life, engineered

FIG. 05

Batteries are the budget of the whole network. Tuning the operation remotely took the benchmark sensor from 4.31 to 11.24 days per charge.

3.7 V 3400 mAh · 87.7 TO 59.7 mV/DAY · CUTOFF NEAR 2700 mV

SYN

One storm, every sensor

FIG. 06

Recording times are synchronized across the network from the cloud, so a single storm is captured simultaneously at every point.

REMOTE SLEEP TIME · SYNCHRONIZED SAMPLING · EVENT-BASED

Results

What the study demonstrated

The network ran through four full battery cycles in the field and came back with the answers the team was after. No sensor was lost inside its hermetic PVC capsule, and the calibration held well enough to map soil moisture across whole catchments.

The biggest gain came from the cloud: remote tuning of measurement, sleep and retry behaviour more than doubled battery life, and forced hibernation cut the cost of a bad connection.

Findings

Remote tuning, measured in the field

Battery life per charge
Before4.31 days
Tuned11.24 days
Discharge rate
Before87.7 mV/d
Tuned59.7 mV/d
Cost of a bad connection
Before54 mV
Hibernate6 mV

Findings as published by the authors in Environmental Modelling & Software 182 (2024).

Under the hood

The cloud layer is Thinger.io

The devices are built on an IoT development toolkit manufactured by Thinger.io, an ESP32 board with a Quectel BC660K NB-IoT modem, and all of them report to the Thinger.io cloud.

Auto provisioning enrols each board on first connection, a product definition gives the fleet a shared sleep time the researchers change remotely, and one data bucket stores every reading with its battery voltage. Dashboards and the REST API feed the team's calibration curves, maps and models.

Soil moisture nodes on a 50 m grid and ultrasonic outlet gauges report over NB-IoT to the Thinger.io platform, where auto provisioning, a shared data bucket, dashboards and a remotely tuned sleep property serve the UPM and UC3M research teams.

References

Sanz, E., Trincado, J., Martínez, J., Payno, J., Morante, O., Almeida-Ñaulay, A.F., Berlanga, A., Molina, J.M., Zubelzu, S., Patricio, M.A. (2024). Cloud-based system for monitoring event-based hydrological processes based on dense sensor network and NB-IoT connectivity. Environmental Modelling & Software, 182, 106186. doi:10.1016/j.envsoft.2024.106186 (open access, CC BY).
Bustamante, A., Patricio, M.A., Molina, J.M. (2019). Thinger.io: an open source platform for deploying data fusion applications in IoT environments. Sensors, 19(5), 1044.

Keep reading

Put your water network on a platform

This monitoring system runs on the same open platform any team can start with. If you want to monitor soil moisture, water levels or any hydrological process at scale, we can help you get there.

Talk to our team Explore the Smart Water solution