IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL.

74, 2025 5500711

An IoT System for the Measurement and the

Interpretation of the Photosynthesis-Related
CO2 Fluxes
Irene Cappelli , Member, IEEE, Lorenzo Parri , Member, IEEE, Marco Tani , Member, IEEE,
Valerio Vignoli , Member, IEEE, and Ada Fort , Member, IEEE

Abstract— The estimation of the CO2 concentration variations quantification of the farm global emissions, the CO2 uptake
due to plants photosynthesis has become a topic of outstanding by vegetation through photosynthesis and its release through
interest in view of evaluating the carbon footprint of greenhouses cellular respiration need to be estimated for full carbon budget
and farms. In this framework, distributed sensing solutions can
be an alternative to expensive and power-consuming gas analyz- and net ecosystem exchange assessment. However, while the
ers. In this article, a low-cost sensor node thought for the accurate CO2 concentration range in exhaust gases of industrial plants
monitoring of photosynthesis-related quantities (i.e., temperature, or combustion-engines is in the order of some % v/v [5],
pressure, water vapor concentration, and CO2 concentration) much lower concentrations should be detected in the case
is presented. The system is validated through measurement of environmental CO2 monitoring (i.e., in the tens [6] or
campaigns involving plants enclosed in an accumulation chamber
in the presence of artificial light with controlled intensity, hundreds [7] of ppm), hence, great attention should be paid to
photoperiod, and spectral composition. A simplified preliminary the choice of the sensing technique, the conditioning, and the
model is also proposed to interpret the obtained results from acquisition of the sensor response. Most of the CO2 sensors
the point of view of the kinetics of the gas exchanges, achieving integrated in IoT devices exploit nondispersive-infrared
the estimation of the CO2 uptake and release rates ascribable (NDIR) [8] and photoacoustic [9] sensing techniques: optical-
to the plants activity.
based sensors are usually preferred since they offer a good
Index Terms— CO2 measurement, environmental monitoring, compromise between measurement performance, cost, and
gas emissions monitoring, Internet of Things (IoT), LoRa,
power consumption, although their integration in low-power
photosynthesis, smart agriculture.
systems poses a significant challenge due to their reliance on
highly power-consuming optical sources.
I. I NTRODUCTION Among the existing photosynthesis measurement methods,
canopy chambers and eddy covariance techniques are the most
T HERE is a growing interest in the estimation of
greenhouse gas emissions and global carbon footprint of
farms in view of evaluating the impact of human activities
used for direct carbon fluxes estimation through small-scale
measurements (at the plant or the canopy level) or large-scale
on the environment [1]. In this sense, those enabling measurements (at the ecosystem size). Eddy covariance tech-
technologies involved in the Internet of Things (IoT) and the niques are based on the continuous measurement of the air flux
smart agriculture frameworks can come into play to offer above the canopy by using ultrasonic anemometers in conjunc-
large-scale distributed data collection systems oriented toward tion with gas analyzers and further postprocessing [10]. The
the low power and low-cost perspectives [2], [3] in place of temporal resolution can be quite dense, but the deployment
expensive and power-consuming gas analyzers. While low-cost costs are high and not often compatible with small-sized
systems for pervasive gas sensing are especially devoted to farms. Moreover, particular requirements must be satisfied
air quality monitoring in urban or industrial contexts [4], for accurate measurements, such as atmospheric stability and
to our knowledge, there are no contributions aimed at land homogeneity and flatness. On the other hand, canopy
monitoring the vegetation-related CO2 . Indeed, apart from the chambers allow tests on single plants or on single leaves
with less expensive instruments than micrometeorological
Received 15 August 2024; revised 5 November 2024; accepted 25 November towers. Two main approaches can be identified: steady-state
2024. Date of publication 25 December 2024; date of current version open systems [11], [12], [13], [14] and transient-state closed
13 January 2025. This work was supported by the Agritech National Research
Center through European Union Next-GenerationEU (Piano Nazionale di systems [15], [16]. In the first case, flow measurements
Ripresa e Resilienza (PNRR)–Missione 4 Componente 2, Investimento 1.4– are performed by using close chambers or cuvettes with
D.D.1032 17/06/2022, CN00000022) under Grant B63C22000640005. The inlet and outlet openings for the continuous injection of air
Associate Editor coordinating the review process was Dr. George Dan Mois.
(Corresponding author: Irene Cappelli.) fluxes with controlled CO2 and H2 O concentration and flow.
The authors are with the Department of Information Engineering and Quantitative estimates of the produced CO2 are achieved by
Mathematics, University of Siena, 53100 Siena, Italy (e-mail: cappelli@ measuring the concentration variation at the inlet and outlet
diism.unisi.it; [email protected]; [email protected]; valerio.vignoli@
unisi.it; [email protected]). of the chamber and knowing the airflow rate and the chamber
Digital Object Identifier 10.1109/TIM.2024.3522409 volume. Ideally, this measurement approach allows to carry
1557-9662 © 2024 IEEE. All rights reserved, including rights for text and data mining, and training of artificial intelligence
and similar technologies. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
5500711 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 74, 2025

out measurements on entire plants and over long time periods the model has a general formulation, the fit parameters
since the samples under test are maintained in ideal conditions are plant-specific since they embody information about the
of CO2 , water vapor, and temperature, however, complex actual exchange ability of the plant under test and they
instrumentation (i.e., flow controllers, pumping systems, gas incorporate the dependence of these rates on the test con-
cylinders, and on site power supply) is required, which ditions including temperature and abiotic stresses. Tests are
hinders the usability in prolonged open field deployments performed for a prolonged period to prove the usability of
and significantly increases material and installation costs. the designed system, but shorter tests could be sufficient to
The same approach is exploited by expensive and portable estimate the plants uptake and release rates according to the
instruments as the commercial LI-6800 by LI-COR, which proposed model. Indeed, as it is general for transient-state
performs simultaneous fluorescence and gas exchange mea- close systems, this approach is designed for short-study only
surements on a single leaf and not on the entire plant. This and the estimate of the slope of the first hours of carbon
kind of device allows for very accurate measurements but at uptake is sufficient to predict the uptake rate, considering
the expense of higher cost and not continuous but specific also that the plant starts working in conditions different
acquisitions. from realistic ones as the concentration inside the chamber
On the other hand, the close chamber systems allow to decreases.
reduce the complexity of the setup estimating the photosyn- The whole measurement approach differs from those exist-
thetic rate from the change rate of the CO2 concentration ing in the literature thanks to the simplicity of both the
with time, however, these approaches can be generally used low-cost sensing system and the measurement setup (not
for short-term studies only, to avoid alterations of the abiotic requiring power grid connection and sophisticated hardware).
condition inside the chamber due to its closure (i.e., temper- The system is therefore usable for all those tests conditions and
ature rise, water condensation, and water stress) which may setups comparable with the ones presented. For example, in the
affect the photosynthetic processes [17], [18]. Furthermore, field of industrial cultivation, highly technological greenhouses
the gas concentration monitoring in existing chamber-based are increasingly used. They are real controlled environments
systems, both prototype and commercial, is carried out where environmental conditions are monitored in terms of
with expensive and power-consuming infrared gas analyz- lighting protocols (usually colored artificial light), tempera-
ers (IRGAs), while to our knowledge, no low-cost yet accurate ture, and humidity. The deployment of the proposed system
IoT systems for the monitoring of the vegetation-related in this context is therefore rather straightforward although the
CO2 exist. chamber geometry should be revised and energy harvesting
In this framework, we propose a novel approach that systems as those presented in [20] should be foreseen to supply
exploits the low-cost sensing nodes characterized in [19] in the nodes. A different matter concerns experiments in tradi-
conjunction with a close accumulation chamber housing the tional greenhouses or open field contexts. Indeed, every time
plants under test. Although the flow measurements approach the test environment is characterized by a strong gas exchange
is generally the most used in literature, the proposed solution with the outside environment, the proposed measurements
based on a closed chamber is simpler and easier to be cannot be performed, unless adequate measurement chambers
effectively implemented. Indeed, in steady-state systems, the are used. Moreover, greater attention should be paid to the test
real-time adjustment of the concentration inside the chamber duration and to the measurement setup to limit alterations of
is quite challenging as it requires accurate control of the the ambient conditions inside the chamber, which might be
flows to guarantee a uniform and stable concentration around more critical than in controlled environments.
the plant and measurable variations of the CO2 concentration The rest of this article is organized as follows. In Section II,
between inlet and outlet flows. On the other hand, with the the materials and methods are resumed, providing a description
proposed approach, no flow control is needed and only the of the designed sensor node and the indication of the setup
flow exchange with the environment outside the chamber must and procedures used for the tests. Section III resumes the
be accounted. outcomes of the CO2 sensor calibration, as presented in [19],
Extensive measurement campaigns are then performed in and presents the experimental results obtained from the tests
the laboratory to test the instrumentation and prove its suc- with plants. Section IV further investigates these outcomes,
cess in accurately measuring photosynthesis-related quantities giving their interpretation using the simplified gray-box model
(i.e., temperature, pressure, water vapor concentration, and which describes the CO2 exchange fluxes. Finally, Section V
CO2 concentration). The plants under test are subjected to closes this article outlining the conclusions.
artificial lighting protocols at controlled intensity, spectral
composition, and photoperiod to have stable and repeatable II. M ATERIALS AND M ETHODS
lighting conditions during distinct measurement campaigns to In the following, the architecture of the IoT monitoring
better investigate the expected correlation between photosyn- system is briefly presented together with the description of
thetic activity and light radiation. A simplified “gray-box” the setup employed for the CO2 sensors calibration and for
model for the CO2 exchange, based on the photosynthesis the laboratory tests.
chemical reaction, is then proposed as a tool to interpret
the obtained results and to validate the outcomes of the A. IoT Monitoring System Architecture
measurements, thus providing a preliminary assessment of The block diagram of the developed IoT monitoring archi-
vegetation-related CO2 uptake and release rates. Even if tecture is reported in Fig. 1(a). It is composed of LoRa

Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
CAPPELLI et al.: IoT SYSTEM FOR MEASUREMENT AND INTERPRETATION OF PHOTOSYNTHESIS-RELATED CO2 FLUXES 5500711

Fig. 2. Schematic architecture of the sensor node distinguished into main

board and sensor hat.

size of 30 B, hence, with the considered transmission fre-

quency, a total amount of 1.2 MB of data was collected in the
database by the three nodes for each campaign.
Concerning the used sensor nodes, their architecture
is schematized in Fig. 2. Each node is composed of
a “main board,” hosting an STM32L4Q5 microcontroller
unit (MCU) (ST microelectronics), an RFM95x LoRa
transceiver (Semtech), an environmental sensor BME280
(Bosch) measuring temperature (1.2 ◦ C accuracy), relative
humidity (RH) (3% accuracy) and atmospheric pressure
(1.7 hPa accuracy), and a sensor hat housing up to six gas
sensors exploiting different sensing technologies and their
electronic front-ends [21]. For the characterization study pre-
Fig. 1. (a) Block diagram describing the architecture of the monitoring sented in this article, no MCU sleep was foreseen to have
application. (b) Example dashboard in case of three connected nodes.
frequent readings of the sensors, however, duty cycle policies
can be envisaged for in-field applications.
Focusing on the gas sensors, the “hat board” can host up
wireless sensor nodes, forming a star network with a central to four electrochemical four/three or two electrodes amper-
gateway, and a cloud server. The packets are forwarded from ometric gas sensors, one resistive sensor, one NDIR analog
the gateway to the server using the Semtech proprietary UDP output sensor, and can interface with up to two digital sen-
protocol while the uplink transmission from the end devices sors with serial output. For the presented study, an NDIR
to the gateway is compliant with the “class A” LoRaWAN CO2 sensor (IRC-A1, 5000 ppm full span) produced by
standard implementation and the EU868 regional parameters. Alphasense was used. The CO2 sensor exploits an optical
Cloud-based architecture is adopted to perform packet acqui- sensing principle: it embeds an IR lamp, operating with a
sition, data elaboration, and storage. In detail, the open-source chopped power source at 2 Hz, 50% duty cycle, and peak-to-
LoRaWAN Chirpstack server is used to retrieve and decrypt peak voltage of 5 V. The sensor provides two analog outputs
the packets, while a JavaScript-based Node-RED back end for the pyroelectric readings of the optical power: one from a
is employed for further processing and data storage in a reference optical channel with an optical bandwidth outside
PostgreSQL database. The pipeline ends with the realization the CO2 absorption peak, and one from an active optical
of a dashboard for real-time data visualization and down- channel with an optical bandwidth centered on a CO2 absorp-
load, developed using Grafana application. The developed tion peak (4.26 µm). This differential measurement strategy
dashboard is presented in Fig. 1(b), and the shown data are enables the rejection of the optical source drifts. An additional
exemplary for the case of three sensor nodes connected and analog output is dedicated to an NTC thermistor embedded
transmitting. in the sensor case, available for temperature compensation.
In the nine-day long measurement campaigns presented in Considerable attention was paid to the design of the sensor
the following, the boards were programmed to sample the front-end, using low-drift components, and ensuring channel
sensor outputs and transmit a radio packet every minute using symmetry for drift and noise rejection. The measurements
spreading factor 7 and coding rate 4/5. For each transmitted acquisition was optimized through periodic recalibration of
packet, the corresponding sensors acquisitions were saved in the analog-to-digital converter (ADC) of the MCU, based
the database and univocally identified through the timestamp on a time schedule and temperature monitoring, and by
and the node identification number. The readings from the sampling the signal at a high frequency (2 kHz). Finally,
sensors were coded in the physical payload of the LoRaWAN care was also devoted to data postprocessing to reduce noise,
packet as ASCII text with fixed delimiter, with a total payload by using averaging and low-pass filtering over time windows

Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
5500711 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 74, 2025

Fig. 3. Block diagram of the calibration chain used for the calibration of
the CO2 sensor.

of programmable length consisting of large number of samples

(minimum 1 s time windows), for estimating the optical signal
intensity from the two optical channels. To have reliable
readings, a power-on period is required, however, in the
considered application, the node was continuously powered.
More details about the CO2 sensor are reported in Section III Fig. 4. Measurement setup: plastic box housing a Spathiphyllum plant and
and in [19], while the schematics of the sensors front-ends are two sensing nodes while the horticulture lamp is located on top of the chamber.
reported in [21].
The whole system was continuously supplied at 5 V by
a power supply (GW Instek GPS-3303) or as an alternative The three nodes were continuously supplied at 5 V by a power
a rechargeable battery and solar energy harvesting can be supply (GW Instek GPS-3303). The second internal node was
exploited. The estimated mean current absorption of the node used both as redundancy during the measurements and to
was around 35 mA and was mainly due to the CO2 sensor preliminary investigate the formation of a CO2 concentration
functioning. Furthermore, the mean cost of the node, excluding gradient inside the chamber.
the sensors, is around 150 C. The gas exchange between the box and the room was
mitigated by further enclosing the box in a plastic bag,
B. Calibration and Characterization Setup moreover, the chosen test room was without window and not
frequented by people to minimize the alteration of the environ-
The CO2 sensors were calibrated using the automatic
mental conditions during the tests. The plants illumination was
characterization system (managed by a LabView VI) shown
guaranteed by a programmable commercial lamp specifically
in Fig. 3, where a mass flow-meters bench remotely controlled
conceived for horticulture applications and located on top of
via PC and connected to certified gas cylinders of synthetic
the measurement box.
air and CO2 at 1% was employed to generate fluxes with
The lamp was programmable and allowed to realize per-
controlled flow and composition. In particular, the air and
sonalized lighting protocols by defining the daily schedule,
the CO2 were mixed at different percentages to obtain two
by dimming the LEDs intensity (delivered power mea-
calibration concentrations (i.e., 0 ppm of CO2 and 500 ppm
sured in terms of photosynthetic photon flux density (PPFD)
of CO2 ) at a flow rate of 400 ml/min.
in µmol/m2 s) and by selecting the light spectral composition
The temperature compensation was achieved by exploiting
among four LEDs channels with spectral band in the blue
a climatic chamber (DY2000 Angelantoni) with temperature
([445, 455 nm]), red ([655, 670 nm]), far red ([720, 750 nm]),
profiles with increasing and decreasing temperature gradients
and white 3000 K. The selected lighting protocol, chosen
of 30 ◦ C (from 10 ◦ C to 40 ◦ C and vice versa) at different
from those used in the literature for indoor plant growth
variation speeds (0.11 ◦ C/min and 0.17 ◦ C/min). During these
applications, was a combination of blue, red, and far red
tests, the CO2 concentration could be assumed to be constant
components in a percentage of 12%, 76%, and 12%, respec-
and equal to about 500 ppm. The sensors were then charac-
tively, of the total PPFD amount [21]. The lamp spectral
terized and tested with the automatic characterization system
distribution, measured at 10 cm from the lamp by an Ocean
under different flows and CO2 concentration.
Insight cosine corrected spectrometer, is shown in Fig. 5,
normalized to the intensity measured by a Kipp & Zonen
C. Laboratory Tests on Plants CMP3 pyranometer.
A picture of the whole setup used during the laboratory tests Two measurement campaigns were conducted, nine-day
is reported in Fig. 4. long each, using the described measurement setup and the
A transparent PMMA box of 0.5 × 0.5 × 0.5 m3 was used lighting protocol resumed in Fig. 6, constituted of photope-
as measurement chamber to host the plants and the sensing riod of 12 hd−1 and three tested intensities (i.e., 300, 150,
systems. Two nodes, named node “Int1” and node “Int2” in the and 75 µmol/m2 s). In the following, the use of “PPFD”
following, were located inside the chamber while a third node may be preferred to “µmol/m2 s” to avoid confusion with
(node “Ext” hereafter) was placed outside the box and near to the CO2 uptake and release rates which are also expressed
it to collect the environmental data of the room as a reference. in µmol/m2 s.

Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
CAPPELLI et al.: IoT SYSTEM FOR MEASUREMENT AND INTERPRETATION OF PHOTOSYNTHESIS-RELATED CO2 FLUXES 5500711

Fig. 5. Spectral distribution of the chosen light treatment measured by the Fig. 7. Results collected during test 1: measured CO2 concentration for the
spectrometer and normalized to the intensity measured by the pyranometer. three sensor nodes (as per legend) and irradiance in blue.

Fig. 6. Graphical representation of the lighting protocol used during

tests 1 and 2.

Within the two tests, the number of plants was changed:

only a Spathiphyllum plant was tested during the first mea-
surement campaign while during the second campaign both
the Spathiphyllum and a Epipremnum aureumplants were
tested. The total illuminated leaf area in the considered test
conditions was estimated as an average value accounting
for the leaves dimension and the spatial distribution of the
light incident on them. A mean illuminated leaf area of
about 0.25 m2 was estimated for the Spathiphyllum, while
for the Epipremnum aureum, a value of about 0.1 m2 was
considered and these areas were assumed constant during the Fig. 8. Results collected during test 1. (a) Measured environmental
temperature for the three sensor nodes (as per legend). (b) Assessed water
tests duration. vapor concentration in ppm for the two nodes housed inside the measurement
chamber.
III. E XPERIMENTAL R ESULTS
in Section IV, an interpretation of the outcomes is provided by
In the following, the results obtained during the CO2 sensor means of a simplified gray-box model describing the kinetics
calibration and the laboratory tests with plants are reported. of the CO2 evolution inside the chamber. The consistency of
the model allows for showing the suitability of the proposed
A. CO2 Sensor Calibration and Characterization Results IoT system to monitor and assess the plant photosynthesis-
The characterization results showed that the achieved res- related CO2 fluxes.
olution was of about 10 ppm, and the calibration and The experimental outcomes are reported in Figs. 7 and 8 for
compensation strategies described in [19] allowed to reach an test 1 and in Figs. 9 and 10 for test 2. Figs. 7 and 9 show the
estimated accuracy of about 30 ppm in a temperature range of temporal trends of the CO2 concentration for the three sensors,
(10 ◦ C–40 ◦ C). In fact, from the sensor tests in the climatic together with the irradiance, for tests 1 and 2, respectively,
chamber, reported in [19], it emerged that the residual linear while in Figs. 8 and 10, the monitored temperature and
error caused by the temperature effect is about 4 ppm/◦ C. This water vapor concentrations are reported for the two tests. This
residual is further corrected during subsequent tests with a latter was computed from the measured RH, temperature, and
postcompensation correction. pressure according to the following equations:
PW = 1.0007 hPa + 3.46 × 10−6 × 6.1121 hPa


17.502T
B. Plants Monitoring: Tests Results × e 240.97◦ C+T (1)
In the following, the results obtained during the tests per- RH% PW
[H2 O] = (2)
formed on plants in the closed chamber are reported, whereas 100 P
Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
5500711 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 74, 2025

is a function of the applied radiation intensity, given that

the total illuminated leaves area of the plants and conse-
quently the stomata density can be considered constant during
the test duration. During the dark phase, the CO2 uptake
stops, and a CO2 increase is observed, due to the constant
release rate of the cellular respiration and to the balance
between internal and external CO2 caused by the chamber
leakages. Moreover, as expected, the addition of the second
plant increases the net CO2 absorption and its consequent
release during the respiration phase because of the increase
Fig. 9. Results collected during test 2: measured CO2 concentration for the in the number of activable photosynthetic sites in the case
three sensor nodes (as per legend) and irradiance in blue.
of test 2.
A rough assessment of the mean daily CO2 uptake entity
during the light hours can be directly derived from the mea-
sured temporal trends: for test 1, the CO2 uptake goes from
about 62 cm3 (0.12 g) at 75 PPFD to about 81 cm3 (0.16 g)
at 300 PPFD, while during test 2, it spans approximately in
the [75 cm3 (0.16 g), 118 cm3 (0.23 g)] range. The values
reported so far are averaged values of CO2 uptake computed
for the consecutive days of equal light treatment, however,
fluctuations over days within the same test can be observed
that can be explained with the slight variations of the measured
temperature and humidity which can affect the plants response
and the stomata opening and closure. The two tests lead also
to consistent estimations of the mean CO2 uptake per unit
of illuminated leaf area, with errors of about 20 mg/m2 : about
0.48 g/m2 at 75 PPFD, 0.56 g/m2 at 150 PPFD, and 0.64 g/m2
at 300 PPFD, considering a mean illuminated area of 0.25 m2
for the Spathiphyllum and of about 0.1 m2 for the Epipremnum
aureum.
The plants behavior is reflected also in the trends of [H2 O]
reported in Figs. 8(b) and 10(b), but with an opposite dynamic
because of the H2 O release through the plants stomata and due
to the transpiration process. The average external conditions
experienced during the tests were temperature of 24 ◦ C and
RH around 35% and 40% for tests 1 and 2, respectively; mean
RH around 85% was reached inside the chamber while the
mean temperature was around 24 ◦ C. In general, temperature
Fig. 10. Results collected during test 2. (a) Measured environmental gradients below 3.5 ◦ C were achieved and the environmental
temperature for the three sensor nodes (as per legend). (b) Assessed water
vapor concentration in ppm for the two nodes housed inside the measurement temperature in the chamber was always below 30 ◦ C which
chamber. is generally considered the limit temperature for heat stress in
plants [23].
To conclude, these results prove the usability of the pro-
where PW is the saturation vapor pressure derived from the posed system for the monitoring of vegetation-related CO2 ,
Buck equation [22] with T temperature in ◦ C and P absolute and they provide a straightforward and rough assessment of
pressure in hPa, while [H2 O] is the water vapor concentration the CO2 uptake entity, paving the way for further investiga-
(vol/vol) with RH the RH in percent. tion and for the design of improved measurement systems.
The CO2 concentration outside the chamber is quite stable Tests were performed for a prolonged period to prove the
and it is mainly affected by the environmental alterations usability of the designed system but in principle, shorter tests
inside the room. On the other side, the variations measured by could be sufficient to estimate the plants uptake and release
the two nodes inside the chamber are consistent and related rates.
to the plants activity and to the light/dark cycles, thus proving
the capability of the designed sensing system in measuring IV. P LANT M ONITORING : R ESULTS I NTERPRETATION
vegetation-related CO2 alterations. It can be observed that, A further step in the interpretation of the results was made
as expected, the CO2 concentration decreases during the by proposing a gray-box model based on the kinetics of the
light hours, reaching a minimum concentration dependent chemical reactions involved in the photosynthesis to interpret
on the tested light intensity with a plateau at low CO2 in the trend of the measured data and to assess uptake and release
case of the highest tested radiation. The rate of decrease rates in place of the rough uptake entity estimate.

Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
CAPPELLI et al.: IoT SYSTEM FOR MEASUREMENT AND INTERPRETATION OF PHOTOSYNTHESIS-RELATED CO2 FLUXES 5500711

Although the model has a general formulation, the fit param- is the preexponential term (considered constant for reduced
eters are plant-dependent since they describe the exchange temperature gradients), T is the absolute temperature in K ,
ability of the tested plant under specific test conditions. λi = (Ei /R) is the ratio between the activation energy
Moreover, the parameters incorporate also information about (assumed constant for reduced temperature gradients), and
the temperature and humidity dependence of the observed the gas constant R. Therefore, these rates incorporate also
phenomena and the specific response of the plant to possible information about the effect of the temperature variation on
abiotic stresses. the observed phenomena.
The stoichiometric formula of the photosynthesis is usually The equation [S] = [SO ] + [SA ] = constant holds since
summarized by the net equation in (3) (n is the stoichiometric the concentration of the total amount of photosynthetic sites
coefficient), which emphasizes the CO2 reaction with water, (S) is always the sum of the concentrations of activated and
its reduction to carbohydrate, and the production of O2 as nonactivated/regenerated sites and the illuminated leaf area
byproduct of the reaction [24] is assumed constant during consecutive days of the same
light
test. Moreover, the initial conditions in absence of light are
nCO2 + nH2 O −→ n(CH2 O) + nO2 . (3) [SA ] = 0 and [SO ] = [S]. From now on, [X ] is used to indicate
the molar concentration of species X .
For n equal to six, six moles of CO2 and six moles of The following balance equations can be written for [SA ]
H2 O are transformed into one mole of glucose (C6 H12 O6 ) and [SO ], where I is the absolute light intensity, I0 is a
and six moles of O2 , thus giving the net stoichiometry of reference light intensity, and m, p, and l are the reaction
one O2 produced for each CO2 fixed. However, this equation orders:
hides the complexity of the gross reactions taking place during
photosynthesis such as photon absorption and electrolysis dSA I dSO
= k1 [H2 O]m [SO ] −
(i.e., light-dependent reaction), carbon assimilation and glu- dt I0 dt
cose production (i.e., Calvin cycle), and decomposition of I dSO
= k1 [H2 O]m ([S] − [SA ]) − (7)
complex nutrients into simpler molecules to provide energy I0 dt
to the cells (i.e., aerobic cellular respiration). dSO p I l
= k2 [CO2 ] [SA ] + k3 [O2 ] [C6 H12 O6 ][SO ]. (8)
Over the years, different mathematical and empirical models dt I0
were proposed for estimating photosynthesis-related quanti-
Considering that the photon absorption is quite fast with
ties [25], [26] and most of them is thought for steady-state
respect to the other involved processes and that in the tested
measurement approaches and requires a wide set of parameters
conditions, all the photosynthetic sites exposed to light can
of difficult estimation [27]. The proposed approach aims to
be assumed active, [SA ] ≃ [S] and [SO ] ≃ 0 hold from
directly describe the kinetics of the chemical reaction involved
which (dSA /dt) ≃ −(dSO /dt) can be derived. Therefore, the
in the photosynthesis, assuming a first order reaction as in [28],
observed dynamic substantially depends on the dynamic of
also based on an a prior coarse analysis of the test results
the nonactive sites and the rate-limiting phenomenon is only
presented in Section III, which display a first-order behavior
due to (8).
during the lighting periods and an approximately zero-order
The total CO2 gas concentration satisfies [CO2 ]tot =
kinetics in the dark periods. By analyzing the process from
[CO2 ]0 −[CO2 ]ph + [CO2 ]resp where [CO2 ]0 is the initial con-
the point of view of the activation and the regeneration of
centration, [CO2 ]ph is the uptake due to photosynthesis, and
the photosynthetic sites, the three chemical reactions in (4)–
[CO2 ]resp is the emission due to cellular respiration. Conse-
(6) can be used to describe the plants activity. Adopting this
quently, (d[CO2 ]tot /dt) = −(d[CO2 ]ph /dt) + (d[CO2 ]resp /dt)
route, the reactions (4) and (5) can be seen as the sequence
holds where (d[CO2 ]ph /dt) and (d[CO2 ]resp /dt) are described
of the photon absorption in a nonactivated photosynthetic
as in (9) and (10)
site (SO ) and the formation of carbon-containing compounds
during the Calvin cycle with the consequent regeneration of d[CO2 ]ph I I
= 6k 2 [CO2 ] p [SA ] = k2∗ [CO2 ] p (9)
the activated site (SA ), whereas (6) indicates the multistep dt I0 I0
cellular respiration. While the first two reactions only occur d[CO2 ]resp
during the periods of light, cellular respiration is involved = 6k 3 [O2 ]l [C6 H12 O6 ][SO ]=k3∗ . (10)
dt
both in the absence and presence of light since it deals with
∗ (λ3 /T)
the vital processes of the cells. These reactions are charac- The rate k3∗ = k03 e incorporates [C6 H12 O6 ], [O2 ]l and
terized by different rates: while photon absorption is quite [SO ], which are assumed constant during the performed tests,
∗ (λ2 /T)
fast, the enzyme-mediated reactions of sites regeneration are while k2∗ = k02 e incorporates [S], under the assumption
slower of [SA ] ≃ [S]. Finally, the kinetics for the CO2 is described
with the following balance equations distinguishing between
k1
12H2 O + SO + light −→ 6O2 + SA (4) light (11) and dark (12) periods
k2
SA + 6CO2 + light −→ C6 H12 O6 + 6H2 O+SO (5) d[CO2 ]light d[CO2 ]ph d[CO2 ]resp
=− + + 8light
k3
C6 H12 O6 + 6O2 −→ 6CO2 + 6H2 O. (6) dt dt dt
p I ∗
= −k2∗ [CO2 ]light + k3−light
k1 , k2 , and k3 are the rate constants defined according to I0
the Arrhenius form as ki = k0i e−(λi /T) i = 1, 2, 3 where k0i


− a [CO2 ]light − [CO2 ]ext (11)

Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
5500711 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 74, 2025

d[CO2 ]dark d[CO2 ]resp

= + 8dark
dt dt
= k3−dark − a([CO2 ]dark − [CO2 ]ext ).
∗
(12)
The terms 8light and 8dark are added to model with rate
a the leakages of the chamber (i.e., the CO2 exchange
between the chamber and the outside environment), linearly
dependent on the [CO2 ] gradient (i.e., 1[CO2 ]) between the
internal and the external (i.e., [CO2 ]ext ) concentration. Two
rates due to the respiration process are considered for light Fig. 11. [CO2 ] acquired by node Int2 (blue solid lines) and model fitting
and dark periods and are assumed to be constant during the (black dashed lines) obtained from the model calibration during the first
considered tests. Finally, adjustable exponents m, p, and l three days of test at highest irradiance level for (a) test 1 and (b) test 2.
are considered to define the order of the reaction. In par-
ticular, p = 1 is supposed from the observed experimental
outcomes, while m does not come into play in the bal-
ance equations describing the [CO2 ] kinetics. The value of
l is not relevant, given that the information on [O2 ]l is
incorporated in k3∗ and [O2 ] is considered constant during
the tests.
At the equilibrium (i.e., (d[CO2 ]light /dt) = 0), assum-
ing no exchange between the measurement chamber and
the outside environment (i.e., a = 0), the CO2 has value
[CO2 ]light−eq = (k3∗ /k2∗ I) = (k3∗ /6k2 [S]I), which depends on
the light intensity and the total number of sites with an inverse
law.
To conclude, cellular respiration is modeled as a constant
rate process, indirectly dependent on the type and quantity
of plants under test which affect the value of [C6 H12 O6 ]
and [SO ] and consequently of k3∗ . The photo-generated CO2
satisfies a first-order kinetics and is directly dependent on
the light intensity, the CO2 concentration, and the total num-
ber of photosynthetic sites (hence on type and quantity of
plants under test). The rate of CO2 assimilation k2∗ can thus
be used to estimate the rate of the plant-dependent carbon
uptake.
A nonlinear least square fitting (lsqnonlin) routine was
implemented in MATLAB considering (11) and (12): p,
∗ ∗ ∗
k02 , λ2 , k03−light , λ3−light , k03−dark , λ3−dark , and a are the
unknown parameters of the model, [CO2 ]ext is the concen- Fig. 12. [CO2 ] acquired by node Int2 (blue solid lines) and model fitting
tration measured by node Ext, [CO2 ]light and [CO2 ]dark are (black dashed lines) obtained from the model validation during the entire test
duration for (a) test 1 and (b) test 2. With red-dotted lines the irradiance
the concentrations measured by node Int2, I0 is taken equal normalized to 150 PPFD.
to 150 PPFD, and the absolute temperature T in the rates
formulation is the temperature measured by node Int2. For
each test, the model was calibrated on the data corresponding model to describe the observed data. Moreover, p = 1 is
to the highest irradiance level (i.e., first three days of tests) derived from the fitting routine, thus validating the modeling
and then validated on the remaining days. Figs. 11 and 12 of the CO2 absorption with a first-order dynamic. The largest
report the fitting results for the [CO2 ] of node Int2 for the errors between measured and estimated values occur during
first three days and for the entire test duration, respectively. the transient phases of turning on and off of the lamp.
In these plots, unlike Figs. 7 and 9, the x-axis is reported This is due to light intensity modeling which will lead to a
in seconds instead of days since it is more suitable for the sudden variation in the fit results and a smoother trend for
performed investigation. In Fig. 12 also the trend of the the measurements. Nevertheless, the proposed model succeeds
irradiance normalized to I0 , (I/I0 ), is reported as red dotted in interpreting the measured data and describing the CO2
lines. dynamic.
A good adherence between fit and experimental data can In Fig. 13, the parameters extracted for test 1 (i.e., k2∗ ,
be highlighted with a maximum error for test 2, where the ∗ ∗
k3−light and k3−dark ) are reported as a function of the tem-
higher CO2 span is reached, of about 50 ppm during the model perature in the [0 ◦ C, 40 ◦ C] range exploiting the Arrhenius
calibration (about 5% of the maximum measured CO2 ) and formulation.
of about 75 ppm in validation (about 7.5% of the maximum The rates have an exponential dependence on tempera-
measured CO2 ), thus proving the suitability of the proposed ture and their variability in the considered temperature span

Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
CAPPELLI et al.: IoT SYSTEM FOR MEASUREMENT AND INTERPRETATION OF PHOTOSYNTHESIS-RELATED CO2 FLUXES 5500711

rate of 10 µmol/(m2 s) is reported for a grapevine exposed to

solar intensity of 2000 PPFD.

V. C ONCLUSION
In this article, the possibility of using low-cost IoT solutions
in conjunction with a close accumulation chamber for the
monitoring of the CO2 variations due to plants lifecycle
was investigated as alternative to those professional systems
involving steady-state accumulation chambers and expensive
gas analyzers. The used device was engineered to monitor
other physical quantities which come into play during pho-
tosynthesis such as temperature, pressure, and water vapor
concentration. The achieved results proved the capability of
the proposed system to accurately measure the CO2 and H2 O
alterations induced by plants photosynthesis, paving the way
for the monitoring of vegetation-related CO2 with low-cost
and low-complexity embedded solutions.
In detail, an ad hoc setup was deployed in the laboratory
Fig. 13. Temperature dependence of (a) k2∗ , (b) k3−light ,∗ and (c) k3−dark
∗
housing some plants inside a closed transparent measurement
in the [0 ◦ C, 40 ◦ C] interval, as derived from the fitting routine for test 1; chamber exposed to the light of a horticulture lamp with pro-
in (d) are the previous trends normalized to the maximum value.
grammable intensity, photoperiod, and spectral composition.
The evolution of CO2 concentration, temperature, RH, and
is of several orders of magnitude, hence, given this strong pressure was monitored in quasi-real time (i.e., one sampling
temperature dependence, we foresee very different rates for per minute) both inside the chamber and in the lab room. The
different operating temperatures. The estimated time constant performed experiments were quite simple but anyway allowed
describing the CO2 exchange with the environment outside to test the suitability of the measurement system and to prove
the chamber is a = 1.4·10−5 (1/s), hence for 1[CO2 ] ≃ that the measured variations were actually related to the plants
500 ppm, which is the maximum gradient reached during test1, response. From the measured temporal trends, a preliminary
leakages of 0.007 ppm/s are experienced, corresponding to assessment of mean daily CO2 uptake was found, spanning in
0.16 µmol/(m2 s). The temperature increase boosts the reaction the [62 cm3 (0.12 g), 81 cm3 (0.16 g)] range for test 1 and
while the temperature decrease slows it down. Moreover, in the [75 cm3 (0.16 g), 118 cm3 (0.23 g)] range for test 2.
the rate of [CO2 ] variation due to cellular respiration in the These results also led to the estimation of the mean CO2
presence of light is by far lower than the one experienced uptake per unit of illuminated leaf area in the [0.48 g/m2 ,
during the dark periods. 0.64 g/m2 ] range. Tests were performed for a prolonged period
Note that through the estimation of k2∗ , an assessment of the to prove the usability of the designed system, but in principle,
plant behavior in environmental condition can be obtained by shorter tests could be sufficient to estimate the rates of interest,
using the following equation: thus limiting the possible plant stress or working conditions
different from the realistic ones.
d[CO2 ]light p I
= −k ∗2 [CO2 ]out . (13) A model describing the dynamic of CO2 fluxes was also
dt I0 proposed as a possible tool to reasonably interpret the obtained
Considering an atmospheric CO2 concentration [CO2 ]out = results. It was derived from the photosynthesis stoichiometric
420 ppm [29], T = 20 ◦ C, and I = 150 PPFD, a rate formula and analyzing the chemical mechanisms from the
of CO2 uptake equal to about 0.026 ppm/s can be found, point of view of activation and regeneration of the photo-
corresponding to an uptake rate per unit leaves area of synthetic sites. The model was initially calibrated on the
0.259 mg/(m2 s) or to 0.576 µmol/(m2 s). Using (13), under first three days of tests and then validated on the remaining
intense radiation I = 2000 PPFD, an uptake of CO2 data using a nonlinear fitting routine and achieving maximum
of about 0.348 ppm/s can be predicted, corresponding to error of about 50 and 75 ppm in calibration and valida-
3.453 mg/(m2 s) or 7.68 µmol/(m2 s) for unit leaf area. Con- tion, respectively, for Test 2. The capability of the model
cerning the rate of CO2 release because of cellular respiration, to interpret the measured data was proven and preliminary
a value of about 0.008 ppm/s can be derived, corresponding estimate of CO2 absorption and release rates were provided
to a respiration rate per unit leaves area of 0.008 mg/(m2 s) achieving an uptake rate for unit leaf area of ∼8 µmol/(m2 s)
or to 0.182 µmol/(m2 s). Thus, a positive balance in the CO2 and a release rate per unit leaf area of 0.182 µmol/(m2 s)
exchange fluxes between uptake and respiration is proved. under the conditions [CO2 ]out = 420 ppm, T = 20 ◦ C,
Despite an uncertainty in the order of 20% in the estimation and I = 2000 PPFD.
of the illuminated leaf area can be present, the results are Future activities may concern the refinement of the model,
consistent with those existing in the literature. For example for example foreseeing a thermostatic system, to evaluate
in [15], a CO2 uptake in the [0.140 mg/(m2 s), 0.695 mg/(m2 s)] its accuracy in case of temperatures different from those
range is found for an olive tree exposed to solar radiation tested. This possibility was not accounted during the presented
from 0 to 800 W/m2 , while in [13], a maximum photosynthesis measurement campaigns since the preliminary goal of the
Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
5500711 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 74, 2025

study was the evaluation of the system usability for the chosen [13] J. P. Smith, E. J. Edwards, A. R. Walker, J. C. Gouot, C. Barril,
monitoring task and the validation of the model in the labora- and B. P. Holzapfel, “A whole canopy gas exchange system for
the targeted manipulation of grapevine source-sink relations using
tory conditions, where critical temperature variations were not sub-ambient CO2 ,” BMC Plant Biol., vol. 19, no. 1, pp. 1–15,
expected. However, in view of possible future applications in Dec. 2019.
noncontrolled outdoor contexts, the validation of the system [14] S. Burkart, R. Manderscheid, and H.-J. Weigel, “Design and perfor-
mance of a portable gas exchange chamber system for CO2 - and
for temperatures different from the ambient one can be of H2 O-flux measurements in crop canopies,” Environ. Experim. Botany,
great interest, together with other advancements to make the vol. 61, no. 1, pp. 25–34, Sep. 2007.
system suitable for field deployments. Furthermore, tests with [15] O. Pérez-Priego, L. Testi, F. Orgaz, and F. J. Villalobos, “A large closed
canopy chamber for measuring CO2 and water vapour exchange of
different plant species are foreseen for the future to investigate whole trees,” Environ. Experim. Botany, vol. 68, no. 2, pp. 131–138,
model adaptability, although a similar dynamic is expected Apr. 2010.
for plants tested in similar tests conditions, with different [16] P. Steduto, Ö. Çetinkökü, R. Albrizio, and R. Kanber, “Automated
closed-system canopy-chamber for continuous field-crop monitoring of
model parameters ki∗ depending on the specific plant under CO2 and H2 O fluxes,” Agricult. Forest Meteorol., vol. 111, no. 3,
test. pp. 171–186, May 2002.
[17] M. E. Dusenge, A. G. Duarte, and D. A. Way, “Plant carbon metabolism
ACKNOWLEDGMENT and climate change: Elevated CO2 and temperature impacts on photo-
synthesis, photorespiration and respiration,” New Phytologist, vol. 221,
This study was carried out within the Agritech National no. 1, pp. 32–49, Jan. 2019.
[18] E. Driesen, W. Van den Ende, M. De Proft, and W. Saeys, “Influence
Research Center and received funding from the Euro- of environmental factors light, CO2 , temperature, and relative humidity
pean Union Next-GenerationEU (PIANO NAZIONALE DI on stomatal opening and development: A review,” Agronomy, vol. 10,
RIPRESA E RESILIENZA (PNRR)–MISSIONE 4 COM- no. 12, p. 1975, Dec. 2020.
[19] I. Cappelli, L. Parri, M. Tani, V. Vignoli, and A. Fort, “An IoT
PONENTE 2, INVESTIMENTO 1.4–D.D.1032 17/06/2022, monitoring system for the accurate measurement of CO2 concentration
CN00000022). This article reflects only the authors’ views due to plants photosynthesis,” in Proc. IEEE Int. Instrum. Meas. Technol.
and opinions, neither the European Union nor the European Conf. (I2MTC), May 2024, pp. 1–6.
Commission can be considered responsible for them. [20] M. Bruzzi, I. Cappelli, A. Fort, A. Pozzebon, M. Tani, and V. Vignoli,
“Polycrystalline silicon photovoltaic harvesting for indoor IoT systems
under red- far red artificial light,” in Proc. IEEE Sensors Appl. Symp.
R EFERENCES (SAS), Aug. 2021, pp. 1–6.
[21] I. Cappelli et al., “Autonomous IoT monitoring matching spectral
[1] K. O. Yoro and M. O. Daramola, “CO2 emission sources, greenhouse artificial light manipulation for horticulture,” Sensors, vol. 22, no. 11,
gases, and the global warming effect,” in Advances in Carbon Capture. p. 4046, May 2022.
Sawston, U.K.: Woodhead, 2020. [22] A. L. Buck, “New equations for computing vapor pressure and enhance-
[2] V. K. Quy et al., “IoT-enabled smart agriculture: Architecture, ment factor,” J. Appl. Meteorol., vol. 20, no. 12, pp. 1527–1532,
applications, and challenges,” Appl. Sci., vol. 12, no. 7, p. 3396, Dec. 1981.
Mar. 2022. [23] T. B. dos Santos, A. F. Ribas, S. G. H. de Souza, I. G. F. Budzinski,
[3] D. Shadrin, A. Menshchikov, A. Somov, G. Bornemann, J. Hauslage, and and D. S. Domingues, “Physiological responses to drought, salinity, and
M. Fedorov, “Enabling precision agriculture through embedded sensing heat stress in plants: A review,” Stresses, vol. 2, no. 1, pp. 113–135,
with artificial intelligence,” IEEE Trans. Instrum. Meas., vol. 69, no. 7, Feb. 2022.
pp. 4103–4113, Jul. 2020. [24] M. R. Badger, “Photosynthetic oxygen exchange,” Annu. Rev. Plant
[4] S. Ali, T. Glass, B. Parr, J. Potgieter, and F. Alam, “Low cost sensor with Physiol., vol. 36, no. 1, pp. 27–53, 1985.
IoT LoRaWAN connectivity and machine learning-based calibration for [25] L. D. C. García-Rodríguez et al., “Mathematical modeling to estimate
air pollution monitoring,” IEEE Trans. Instrum. Meas., vol. 70, pp. 1–11, photosynthesis: A state of the art,” Appl. Sci., vol. 12, no. 11, p. 5537,
2020. May 2022.
[5] R. Pavri and G. D. Moore, “Gas turbine emissions and control,” [26] E. M. Sukhova, V. A. Vodeneev, and V. S. Sukhov, “Mathematical mod-
Atlanta, GE Energy Services, vol. 1, pp. 1–20, Nov. 2001. Accessed: eling of photosynthesis and analysis of plant productivity,” Biochemistry
Nov. 20, 2023. [Online]. Available: https://www.eplptx.com/wp- (Moscow), Suppl. Ser. A, Membrane Cell Biol., vol. 15, no. 1, pp. 52–72,
content/uploads/2020/05/Gas-Turbine-Emissions-and-Control.pdf Jan. 2021.
[6] J. Wang, A. S. Denning, L. Lu, I. T. Baker, K. D. Corbin, and K. J. Davis, [27] Q. Wang, J. Chun, D. Fleisher, V. Reddy, D. Timlin, and J. Resop,
“Observations and simulations of synoptic, regional, and local variations “Parameter estimation of the Farquhar–Von Caemmerer–Berry bio-
in atmospheric CO2 ,” J. Geophys. Res., Atmos., vol. 112, no. 4, pp. 1–19, chemical model from photosynthetic carbon dioxide response curves,”
Feb. 2007. Sustainability, vol. 9, no. 7, p. 1288, Jul. 2017.
[7] K. Nomura et al., “Long-term estimation of the canopy photosyn- [28] F. C. Rubio, F. G. Camacho, J. M. F. Sevilla, Y. Chisti, and E. M. Grima,
thesis of a leafy vegetable based on greenhouse climate conditions “A mechanistic model of photosynthesis in microalgae,” Biotechnol.
and nadir photographs,” Scientia Horticulturae, vol. 289, Nov. 2021, Bioeng., vol. 81, no. 4, pp. 459–473, Feb. 2003.
Art. no. 110433. [29] P. Friedlingstein et al., “Global carbon budget,” Earth Syst. Sci. Data
[8] M. Vafaei, A. Amini, and A. Siadatan, “Breakthrough in CO2 mea- Discuss., vol. 15, no. 12, pp. 5301–5369, Nov. 2023.
surement with a chamberless NDIR optical gas sensor,” IEEE Trans.
Instrum. Meas., vol. 69, no. 5, pp. 2258–2268, May 2020.
[9] S. Wang, M. Hua, B. Wang, Y. He, P. Zhu, and Y. Li, “Design
and implementation of a miniaturized photoacoustic spectroscopy CO2 Irene Cappelli (Member, IEEE) received the M.Sc.
sensor,” IEEE Trans. Instrum. Meas., vol. 73, pp. 1–9, 2024. degree in electronics and communication engineer-
[10] M. Mauder, T. Foken, M. Aubinet, and A. Ibrom, “Eddy-covariance ing and the Ph.D. degree in information engineering
measurements,” in Springer Handbook of Atmospheric Measurements. and science from the University of Siena, Siena,
Cham, Switzerland: Springer, 2021. Italy, in 2019 and 2023, respectively.
[11] N. Salvatori, A. Giorgio, O. Müller, U. Rascher, and A. Peressotti, She is currently a Research Fellow with the
“A low-cost automated growth chamber system for continuous measure- Department of Information Engineering and Math-
ments of gas exchange at canopy scale in dynamic conditions,” Plant ematics, University of Siena. Her current research
Methods, vol. 17, no. 1, pp. 1–11, Dec. 2021. interests include the design of low-power measure-
[12] K. Nomura et al., “Long-term and continuous measurement of canopy ment systems and energy-autonomous the Internet
photosynthesis and growth of spinach,” Environ. Control Biol., vol. 58, of Things (IoT) systems for gas and environmental
no. 2, pp. 21–29, 2020. monitoring.

Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.
CAPPELLI et al.: IoT SYSTEM FOR MEASUREMENT AND INTERPRETATION OF PHOTOSYNTHESIS-RELATED CO2 FLUXES 5500711

Lorenzo Parri (Member, IEEE) received the M.Sc. Valerio Vignoli (Member, IEEE) received the
degree in electronics and communications engineer- Laurea degree in electronic engineering and
ing and the Ph.D. degree in information engineering the Ph.D. degree in nondestructive testing from
and science from the University of Siena, Siena, the University of Florence, Florence, Italy, in 1989
Italy, in 2017 and 2021, respectively. and 1994, respectively.
He is currently a Research Fellow with the He is currently a Full Professor of electronics
Department of Information Engineering and Math- with the Department of Information Engineering
ematics, University of Siena. His current research and Mathematical Sciences, University of Siena,
interests include the design of gas measurement Siena, Italy. His research interests include the design
systems, electronic systems for industrial application of data acquisition and processing systems based
and condition monitoring, the design of embedded on chemical sensors and the design of analog and
microcontroller-based devices, and the development of sensor networks. mixed-signal electronic circuits.

Ada Fort (Member, IEEE) received the Laurea

degree in electronic engineering and the Ph.D.
degree in nondestructive testing from the University
Marco Tani (Member, IEEE) received the bache-
of Florence, Florence, Italy, in 1989 and 1992,
lor’s degree in telecommunication engineering from
respectively.
the University of Siena, Siena, Italy, in 2008.
She is currently a Full Professor at the Depart-
He is currently a Research Associate with the
ment of Information Engineering and Mathematical
Electronics and Electronics Measurement Research
Sciences, University of Siena, Siena, Italy. Recently,
Group, University of Siena. His current research
she has been involved in the study of random number
interests include advanced reduced instruction
generators based on chaotic maps. Her research
set computer (RISC) machines [advanced RISC
interests include the development of measurement
machines (ARM)]-based embedded systems and dis-
systems based on chemical and ultrasonic sensors and the development of
tributed sensors and networks.
automatic fault diagnosis systems.

Authorized licensed use limited to: UNIVERSIDAD DEL QUINDIO. Downloaded on May 06,2025 at 03:32:34 UTC from IEEE Xplore. Restrictions apply.