Design of a Portable Analyzer to Determine the Net Exchange of CO2 in Rice Field Ecosystems

Global warming is influenced by an increase in greenhouse gas (GHG) concentration in the atmosphere. Consequently, Net Ecosystem Exchange (NEE) is the main factor that influences the exchange of carbon (C) between the atmosphere and the soil. As a result, agricultural ecosystems are a potential carbon dioxide (CO2) sink, particularly rice paddies (Oryza sativa). Therefore, a static chamber with a portable CO2 analyzer was designed and implemented for three rice plots to monitor CO2 emissions. Furthermore, a weather station was installed to record meteorological variables. The vegetative, reproductive, and maturation phases of the crop lasted 95, 35, and 42 days post-sowing (DPS), respectively. In total, the crop lasted 172 DPS. Diurnal NEE had the highest CO2 absorption capacity at 10:00 a.m. for the tillering stage (82 and 89 DPS), floral primordium (102 DPS), panicle initiation (111 DPS), and flowering (126 DPS). On the other hand, the maximum CO2 emission at 82, 111, and 126 DPS occurred at 6:00 p.m. At 89 and 102 DPS, it occurred at 4:00 and 6:00 a.m., respectively. NEE in the vegetative stage was −25 μmolCO2 m2 s−1, and in the reproductive stage, it was −35 μmolCO2 m2 s−1, indicating the highest absorption capacity of the plots. The seasonal dynamics of NEE were mainly controlled by the air temperature inside the chamber (Tc) (R = −0.69), the relative humidity inside the chamber (RHc) (R = −0.66), and net radiation (Rn) (R = −0.75). These results are similar to previous studies obtained via chromatographic analysis and eddy covariance (EC), which suggests that the portable analyzer could be an alternative for CO2 monitoring.


Introduction
The increase in the concentration of carbon dioxide (CO 2 ) in the atmosphere is one of the main factors responsible for global warming.Currently, CO 2 levels are at 419 ppm; this represents 150% of the values in the 18th century [1].This increase is mainly due to anthropogenic activities such as intensive agriculture and changes in land use, among others [2].Rice (Oryza sativa) cultivation extends from tropical to temperate regions [3].It is the second most important staple food in the world, with an annual production of 740 Mt [4].It covers 114 countries and an area of 153 Mha in total, or 11% of the world's arable land [5].In 2021, Peru produced 3.5 Mt of rice in an area of 417,000 ha [6].Currently, 90% of rice production is obtained through flood irrigation [7], making it a significant source of methane (CH 4 ).Furthermore, nitrous oxide (N 2 O) is mainly generated by nitrification and denitrification processes, which are closely related to soil moisture [8].Both gases represent approximately 30% and 11%, respectively, of global agricultural emissions [9].
Sensors 2024, 24, 402 2 of 17 Net Ecosystem Exchange (NEE) is one of the main processes that influence CO 2 concentration in the atmosphere.Agricultural ecosystems, particularly rice paddies, play a crucial role in carbon absorption.Therefore, it is important to understand their function in carbon (C) flux [10].For example, Chatterjee et al. [11] monitored lowland paddy fields for one year (dry and wet seasons) using eddy covariance (EC) to evaluate variations in NEE and find a suitable model for the better partitioning of NEE with respect to its components, such as gross primary production (GPP) and ecosystem respiration (Reco).Kumar et al. [12] calculated NEE in rice and wheat systems in the northwest Indo-Gangetic plains.This was the first estimation in a rice-wheat spring sequence using EC.Neogi et al. [13] investigated the characterization of CO 2 fluxes in tropical lowland rice paddy ecosystems using EC to better understand the environmental impact in terms of C budget in submerged soil.
The land-atmosphere exchange of matter and energy is recorded using EC [14], widely used given its solid theoretical basis.However, it is expensive, difficult to manipulate [15], and susceptible to information gaps [16].On the other hand, static chambers are used to complement the deficiencies of EC.Nevertheless, they require long monitoring periods [17].Additionally, the cost of chromatographic analysis for collected gases is high.In this regard, infrared sensors represent an opportunity to solve these challenges.They utilize the nondispersive infrared (NDIR) principle to measure the concentration of CO 2 instantly [18].In addition, they are easy to acquire, manipulate, and program.An automatic estimation and sampling method based on sensors that can replace the conventional methods mentioned and simultaneously increase the efficiency in estimating greenhouse gas (GHG) fluxes is necessary [19].
In this research, a CO 2 analyzer was designed together with a static chamber to monitor diurnal and nocturnal NEE in rice fields.The objective was to establish a novel, efficient, and dependable method of making resource management decisions for sustainable agricultural practices in Peru.

Site Description
This research was carried out in the "Experimental Irrigation Area" (AER) on the campus of the National Agrarian University La Molina (UNALM), La Molina District, Lima Province, Lima Region (12 • 04 ′ 41 ′′ S, 76 • 56 ′ 45 ′′ W, altitude: 246 m) (Figure 1).During the study, the maximum, minimum, and average temperatures were 32.3, 15.6, and 23.24 • C, respectively.The maximum precipitation was 2.6 mm with an average relative humidity of 77%.The meteorological data were recorded using the automatic station VANTAGE Pro2 Davis, Hayward, CA, USA, located at the AER (Figure 2).In addition, the physicochemical characteristics of the soil are detailed in Table 1.

Design of Portable Analyzer for CO 2 Monitoring
A portable analyzer was designed for CO 2 monitoring (Figure 3a).Its components are as follows: (a) MHZ19B CO 2 sensor from Winsen Electronics; its detection range is 0 to 5000 ± 50 ppm.It operates at optimal T a and RH conditions of 0 to 50  , respectively.(b) DHT22 T a and RH sensor from Aosong Electronics.Its measurement range for T a is −40 to 80 ± 0.5 • C, and for RH, it is 0 to 100 ± 2%.(c) Real-time clock (RTC) module "DS3231" from MMJ Smart Electronics.(d) microSD memory module from Deeoee Electronics.(e) 16 × 2 LED display from Yuxian Electronics.The Arduino DUE board (g) and "Arduino IDE", both from Arduino CC, were selected as the microcontroller unit and coding system, respectively.The components were soldered onto a multipurpose board (f) to ensure connection with the ARDUINO board.Then, the system was placed in a plastic box measuring 150 × 110 × 80 mm 3 .The device was powered by a PHILLIPS 4000 (mAh) portable battery with 5 V of output.The operational analyzer is shown in Figure 3b.
A portable analyzer was designed for CO2 monitoring (Figure 3a).Its c are as follows: (a) MHZ19B CO2 sensor from Winsen Electronics; its detectio to 5000 ± 50 ppm.It operates at optimal Ta and RH conditions of 0 to 50 °C an respectively.(b) DHT22 Ta and RH sensor from Aosong Electronics.Its m range for Ta is −40 to 80 ± 0.5 °C, and for RH, it is 0 to 100 ± 2%.(c) Real-time module "DS3231" from MMJ Smart Electronics.(d) microSD memory modul oee Electronics.(e) 16 × 2 LED display from Yuxian Electronics.The Arduino (g) and "Arduino IDE", both from Arduino CC, were selected as the microcon and coding system, respectively.The components were soldered onto a m board (f) to ensure connection with the ARDUINO board.Then, the system w a plastic box measuring 150 × 110 × 80 mm .The device was powered by a PH (mAh) portable battery with 5 V of output.The operational analyzer is show 3b.

Static Transparent Chamber Design
The monitoring system consisted of a static chamber and a portable C (Figure 4a).The chamber is made of transparent 2 mm thick transparent acr dimensions are 1 × 0.5 × 0.5 m .The gas-mixing system consisted of a portabl and 2 fans (f), both connected through a Universal Serial Bus (USB) port (g). the metal base, with dimensions of 0.5 × 0.5 × 0.15 m , has a 2 mm thick slo installed 6 cm below the soil surface before transplanting permanently.In a analyzer is attached to one of the side faces using a support (h).The finishe shown in Figure 4b.

Static Transparent Chamber Design
The monitoring system consisted of a static chamber and a portable CO 2 analyzer (Figure 4a).The chamber is made of transparent 2 mm thick transparent acrylic, whose dimensions are 1 × 0.5 × 0.5 m 3 .The gas-mixing system consisted of a portable battery (e) and 2 fans (f), both connected through a Universal Serial Bus (USB) port (g).In addition, the metal base, with dimensions of 0.5 × 0.5 × 0.15 m 3 , has a 2 mm thick slot.This was installed 6 cm below the soil surface before transplanting permanently.In addition, the analyzer is attached to one of the side faces using a support (h).The finished device is shown in Figure 4b.

Field Management
Three ponds of 3 × 4 × 0.6 m were installed and lined with geomembrane (Figure 5a,b).The seedbed was prepared on 11 November 2022 and transplanted 35 days postsowing (DPS).The distribution was five rice seedlings per hill, spaced 20 cm × 20 cm each.The vegetative, reproductive, and maturation phases lasted 95, 35, and 42 DPS, respectively.In total, the crop lasted 172 DPS (Figure 5c).The water regime maintained soil moisture between saturation and a 5 cm depth.Irrigation water came from the Rimac River and was stored in a 25 m tank.Its physicochemical characteristics are described in Table 2.The NPK fertilization dose was 230-60-90.In total, 100% of P and K and 50% of N were applied during transplanting.The remaining N was distributed during tillering, floral primordium, and flowering (Figure 5d).The nitrogen sources were urea, diammonium phosphate (DAP), and "Basacote plus 3M".

Field Management
Three ponds of 3 × 4 × 0.6 m 3 were installed and lined with geomembrane (Figure 5a,b).The seedbed was prepared on 11 November 2022 and transplanted 35 days post-sowing (DPS).The distribution was five rice seedlings per hill, spaced 20 cm × 20 cm each.The vegetative, reproductive, and maturation phases lasted 95, 35, and 42 DPS, respectively.In total, the crop lasted 172 DPS (Figure 5c).The water regime maintained soil moisture between saturation and a 5 cm depth.Irrigation water came from the Rimac River and was stored in a 25 m 3 tank.Its physicochemical characteristics are described in Table 2.The NPK fertilization dose was 230-60-90.In total, 100% of P and K and 50% of N were applied during transplanting.The remaining N was distributed during tillering, floral primordium, and flowering (Figure 5d).The nitrogen sources were urea, diammonium phosphate (DAP), and "Basacote plus 3M".

Field Management
Three ponds of 3 × 4 × 0.6 m were installed and lined with geomembrane (Figure 5a,b).The seedbed was prepared on 11 November 2022 and transplanted 35 days postsowing (DPS).The distribution was five rice seedlings per hill, spaced 20 cm × 20 cm each.The vegetative, reproductive, and maturation phases lasted 95, 35, and 42 DPS, respectively.In total, the crop lasted 172 DPS (Figure 5c).The water regime maintained soil moisture between saturation and a 5 cm depth.Irrigation water came from the Rimac River and was stored in a 25 m tank.Its physicochemical characteristics are described in Table 2.The NPK fertilization dose was 230-60-90.In total, 100% of P and K and 50% of N were applied during transplanting.The remaining N was distributed during tillering, floral primordium, and flowering (Figure 5d).The nitrogen sources were urea, diammonium phosphate (DAP), and "Basacote plus 3M".

Sensor Calibration
The MHZ19B sensor was calibrated with an automatic reference of 400 ppm by the manufacturer [20].The DHT22 sensor was calibrated by relating its readings to the hourly data obtained by the automatic weather station for 24 h (Figure 6).Ta is air temperature, Ts is the air temperature reading by the sensor, RH is relative humidity, and RHs is the relative humidity reading by the sensor.

Monitoring and Data Collection
Diurnal CO2 monitoring in rice plots begins at tillering, the stage of maximum leaf growth.There were 5 days that lasted 24 h each and were carried out simultaneously in the three plots.The preparatory phase began with the attachment of the static camera to the metal base.Then, a water seal was made on the coupling to prevent gas leakage.The analyzer was then placed and turned on in the chamber so that the CO2, Ta, and RH readings stabilized for 30 min.Monitoring per se began with closing the chamber and turning on the fans during the first 30 min of each hour.The opposite action was carried out during the remaining 30 min.

Data Processing
Emission fluxes were calculated based on CO2 concentration changes (ppm min ).Firstly, linear regression analysis was performed on 30 data [21,22].Secondly, the CO2 emission flux (μmol m d ) was calculated with Equations ( 1)- (3).T a is air temperature, T s is the air temperature reading by the sensor, RH is relative humidity, and RHs is the relative humidity reading by the sensor.

Monitoring and Data Collection
Diurnal CO 2 monitoring in rice plots begins at tillering, the stage of maximum leaf growth.There were 5 days that lasted 24 h each and were carried out simultaneously in the three plots.The preparatory phase began with the attachment of the static camera to the metal base.Then, a water seal was made on the coupling to prevent gas leakage.The analyzer was then placed and turned on in the chamber so that the CO 2 , Ta, and RH readings stabilized for 30 min.Monitoring per se began with closing the chamber and turning on the fans during the first 30 min of each hour.The opposite action was carried out during the remaining 30 min.
where K is the accumulation factor of the chamber (mol min ppm −1 m −2 d −1 ); S is the rate of change in CO 2 concentration (ppm min −1 ); P is barometric pressure (mbar); R is the ideal gas constant, 0.0831451 (bar L K −1 mol −1 ); T c is the temperature inside the chamber (K); V is the net volume of the chamber (m 3 ); and A is the net area of the chamber entrance (m 2 ).
where NEE is the net CO 2 flux of the rice ecosystem µmolCO2 m −2 s −1 , PPFD is the photosynthetic photon flux density µmolphotons m −2 s −1 , P max is the maximum photosynthetic rate, K m is an adjustment constant, and Reco is the respiration rate of the rice ecosystems µmolCO2 m −2 s −1 .For this, PPFD, P max , and K m data from Yang et al. [21] were used.The daily NEE for each phenological stage is the average of the fluxes from three analyzers.To verify the normality of the data, the Anderson-Darling test was used, which turned out to be non-parametric.Spearman correlation (R) was performed between the environmental variables, NEE, and Reco.Additionally, the Mann-Whitney U test was applied to assess significant differences between the results, at a significance level of 5%.

NEE Response to Environmental Factors
The results of the Anderson-Darling test verified the non-normality of the data, except for Ts.Then, the correlations between the environmental factors, Reco, and NEE were

Diurnal Variation in NEE
The NEE values during the study are represented in Figure 7.They are positive a night and negative during the day.This behavior is consistent with the results obtained by Bhattacharyya et al. [24], McMillan et al. [25], and Zhang et al. [26].During dayligh hours, the ecosystem functioned as a carbon dioxide (CO2) sink, with higher levels of ab sorption through photosynthesis compared with emissions through respiratory processes However, during the nighttime, the ecosystem acted as a source of CO2, primarily because of Reco [27,28].In the absence of sunlight, NEE is, on average, 58 times lower than the results of Yang et al. [21] and Bhattacharyya et al. [29].This decrease can be attributed to the higher levels of RHc during the same period (Figure 9d).As a result, the sensor did not perform at its optimal level.In contrast to portable analyzer technology, the EC meth odology used in the aforementioned studies employed open-path NDIR gas analyzers such as the LI-7200, LI-7500, and EC-150.These are specifically designed to measure fluxes in CO2, water vapor, and energy below the canopy.Therefore, their prices are excessively higher compare with a portable analyzer.In this study, the "MHZ19B" NDIR sensor wa used, which differs in application, precision, and price.However, if optimal operating conditions are guaranteed, the sensor has a high potential for accuracy and practicality.

Diurnal Variation in NEE
The NEE values during the study are represented in Figure 7.They are positive at night and negative during the day.This behavior is consistent with the results obtained by Bhattacharyya et al. [24], McMillan et al. [25], and Zhang et al. [26].During daylight hours, the ecosystem functioned as a carbon dioxide (CO 2 ) sink, with higher levels of absorption through photosynthesis compared with emissions through respiratory processes.However, during the nighttime, the ecosystem acted as a source of CO 2 , primarily because of Reco [27,28].In the absence of sunlight, NEE is, on average, 58 times lower than the results of Yang et al. [21] and Bhattacharyya et al. [29].This decrease can be attributed to the higher levels of RHc during the same period (Figure 9d).As a result, the sensor did not perform at its optimal level.In contrast to portable analyzer technology, the EC methodology used in the aforementioned studies employed open-path NDIR gas analyzers such as the LI-7200, LI-7500, and EC-150.These are specifically designed to measure fluxes in CO 2 , water vapor, and energy below the canopy.Therefore, their prices are excessively higher compare with a portable analyzer.In this study, the "MHZ19B" NDIR sensor was used, which differs in application, precision, and price.However, if optimal operating conditions are guaranteed, the sensor has a high potential for accuracy and practicality.On the other hand, total NEE during the vegetative phase (−25.2 μmolCO m s ) was 1.4 times lower than the minimum during the reproductive phase (−35 μmolCO m s ).Similarly, the maximum diurnal NEE during the reproductive stage (−13 μmolCO m s ) was 1.4 times higher than during the vegetative stage (−9.4 μmolCO m s ).This is consistent with Yang et al. [21], who determined that the maximum absorption during the vegetative and maturation phase was approximately 1.5 times lower than in the reproductive phase.In the vegetative phase, CO2 assimilation is limited because the plant is in the growth stage (Figure 10a,b).In the reproductive stage, complete development is observed, leading to maximum absorption.In the maturation stage, senescent leaves fall and add organic matter to the soil.Additionally, the plots are drained in preparation for the harvest phase.These two processes gradually increase CO2 emissions until the crop is harvested (Figure 10c,d).This behavior is similar to the results of Chen et al. [30].On the other hand, total NEE during the vegetative phase (−25.2 µmolCO2 m 2 s −1 ) was 1.4 times lower than the minimum during the reproductive phase (−35 µmolCO2 m 2 s −1 ).Similarly, the maximum diurnal NEE during the reproductive stage (−13 µmolCO2 m 2 s −1 ) was 1.4 times higher than during the vegetative stage (−9.4 µmolCO2 m 2 s −1 ).This is consistent with Yang et al. [21], who determined that the maximum absorption during the vegetative and maturation phase was approximately 1.5 times lower than in the reproductive phase.In the vegetative phase, CO 2 assimilation is limited because the plant is in the growth stage (Figure 10a,b).In the reproductive stage, complete development is observed, leading to maximum absorption.In the maturation stage, senescent leaves fall and add organic matter to the soil.Additionally, the plots are drained in preparation for the harvest phase.These two processes gradually increase CO 2 emissions until the crop is harvested (Figure 10c,d).This behavior is similar to the results of Chen et al. [30].On the other hand, total NEE during the vegetative phase (−25.2 μmolCO m s ) was 1.4 times lower than the minimum during the reproductive phase (−35 μmolCO m s ).Similarly, the maximum diurnal NEE during the reproductive stage (−13 μmolCO m s ) was 1.4 times higher than during the vegetative stage (−9.4 μmolCO m s ).This is consistent with Yang et al. [21], who determined that the maximum absorption during the vegetative and maturation phase was approximately 1.5 times lower than in the reproductive phase.In the vegetative phase, CO2 assimilation is limited because the plant is in the growth stage (Figure 10a,b).In the reproductive stage, complete development is observed, leading to maximum absorption.In the maturation stage, senescent leaves fall and add organic matter to the soil.Additionally, the plots are drained in preparation for the harvest phase.These two processes gradually increase CO2 emissions until the crop is harvested (Figure 10c,d).This behavior is similar to the results of Chen et al. [30].ern Oscillation (ENSO) is also influential.Chatterjee et al. [11] recorded average annual maximum and minimum temperatures of 39.2 and 22.5 °C.Kumar et al. [12] recorded Ta and Ts values between 31.8 to 38.2 °C and 27.7 to 28.9 °C, respectively.Neogi et al. [13] recorded a progressive increase in temperature as the vegetative cycle of rice continued.From the vegetative stage to harvest, the average temperatures were 23.6 to 33.5 °C, respectively.However, there were variations in the irrigation techniques used.This study had a maximum water depth of 5 cm during the entire study period.Chatterjee et al. [11] used a higher water regime in three units.Kumar et al. [12] irrigated their crops only when the moisture content fell below the saturation level.In turn, the irrigation regime of Neogi et al. [13] resulted in a sheet of 7-10 cm.According to Yang et al. [22], NEE is sensitive to field management strategies, with water management being one of the most important factors.In addition, soil CO2 emissions decrease when flooded with water, as this reduces the diffusivity of the upper layer of soil [34].These anoxic conditions decrease soil biological activity, as mentioned by Bao et al. [27] and Liu et al. [31].The result obtained from the Mann-Whitney U test shows that the NEE values did not present a significant difference between the analyzed studies (p > 0.05).Therefore, the CO2 analyzer generally performed optimally throughout the 24 h of monitoring and throughout the entire study period.However, more research is needed to consider it a reference method (Table 5).Regarding pp during the rice season, a total of 13.4 mm was recorded, despite the influence of the cyclone "Yaku," which coincided with 111 and 126 DPS (Figure 2d).In comparison with Chatterjee et al. [11] and Neogi et al. [13], whose average annual pp was 1500 mm.75 and 80% was happened between June and September.Kumar et al. [12], whose studies were carried out in Delhi, recorded 1198 mm during the kharif season for rice cultivation when most of the rains occur from July to September because of the southwest monsoon.
Regarding Tc, it ranged from 21.82 to 33.9 • C on average.On the other hand, Ts fluctuated between 24.95 and 25.51 • C on average.This study was carried out in the months of February to May during the summer season when the cold phase of the El Niño Southern Oscillation (ENSO) is also influential.Chatterjee et al. [11] recorded average annual maximum and minimum temperatures of 39.2 and 22.5 • C. Kumar et al. [12] recorded Ta and Ts values between 31.8 to 38.2 • C and 27.7 to 28.9 • C, respectively.Neogi et al. [13] recorded a progressive increase in temperature as the vegetative cycle of rice continued.From the vegetative stage to harvest, the average temperatures were 23.6 to 33.5 • C, respectively.
However, there were variations in the irrigation techniques used.This study had a maximum water depth of 5 cm during the entire study period.Chatterjee et al. [11] used a higher water regime in three units.Kumar et al. [12] irrigated their crops only when the moisture content fell below the saturation level.In turn, the irrigation regime of Neogi et al. [13] resulted in a sheet of 7-10 cm.According to Yang et al. [22], NEE is sensitive to field management strategies, with water management being one of the most important factors.In addition, soil CO 2 emissions decrease when flooded with water, as this reduces the diffusivity of the upper layer of soil [34].These anoxic conditions decrease soil biological activity, as mentioned by Bao et al. [27] and Liu et al. [31].The result obtained from the Mann-Whitney U test shows that the NEE values did not present a significant difference between the analyzed studies (p > 0.05).Therefore, the CO 2 analyzer generally performed optimally throughout the 24 h of monitoring and throughout the entire study period.However, more research is needed to consider it a reference method (Table 5).The MHZ19B sensor has a response time of less than 60 s, so the analyzer was programmed with a response time of one minute to perform a better analysis.On the other hand, open-path analyzers such as LI-7500A and LI-7550 have selectable response times of 0.1, 0.05, and 0.0025 s [35].In addition, a regression analysis was performed between NEE fluxes calculated from data collected with the portable analyzer and NEE fluxes calculated using EC by Chatterjee et al. [11], Kumar et al. [12], and Neogi et al. [13].The determination coefficients have values of 0.661, 0.7873, and 0.5943, respectively (Figure 12).These values can be improved if an additional calibration method is taken into consideration and by improving the Tc and RHc conditions.Furthermore, the sensitivity of the sensors is also an important factor to consider since the MHZ19B has a sensitivity of ±50 ppm compared with LI-7500A, with ±0.11 ppm [35].
Sensors 2024, 24, x FOR PEER REVIEW 13 of 17 The MHZ19B sensor has a response time of less than 60 s, so the analyzer was programmed with a response time of one minute to perform a better analysis.On the other hand, open-path analyzers such as LI-7500A and LI-7550 have selectable response times of 0.1, 0.05, and 0.0025 s [35].In addition, a regression analysis was performed between NEE fluxes calculated from data collected with the portable analyzer and NEE fluxes calculated using EC by Chatterjee et al. [11], Kumar et al. [12], and Neogi et al. [13] The determination coefficients have values of 0.661, 0.7873, and 0.5943, respectively (Figure 12).These values can be improved if an additional calibration method is taken into consideration and by improving the Tc and RHc conditions.Furthermore, the sensitivity of the sensors is also an important factor to consider since the MHZ19B has a sensitivity of ±50 ppm compared with LI-7500A, with ±0.11 ppm [35].

Conclusions
A static chamber with a portable CO 2 analyzer was designed and implemented.It is an economical, simple, and effective alternative to traditional NEE calculation methods.It

Figure 1 .
Figure 1.(a,b) Location of the study area in Lima, Peru.(c) AER map.

Figure 1 .
Figure 1.(a,b) Location of the study area in Lima, Peru.(c) AER map.

Figure 3 .
Figure 3. (a) Scheme of the portable analyzer for CO2 monitoring; (b) analyzer finished

Figure 3 .
Figure 3. (a) Scheme of the portable analyzer for CO 2 monitoring; (b) analyzer finished.

Figure 5 .
Figure 5. (a,b) Rice plot's side and top view, respectively (c) Calendar of monitoring and fertilization days (in DPS); (d) stages of rice paddy growth.

Figure 5 .
Figure 5. (a,b) Rice plot's side and top view, respectively (c) Calendar of monitoring and fertilization days (in DPS); (d) stages of rice paddy growth.

Figure 5 .
Figure 5. (a,b) Rice plot's side and top view, respectively (c) Calendar of monitoring and fertilization days (in DPS); (d) stages of rice paddy growth.

Figure 6 .
Figure 6.Correlation graph between the DHT22 sensor and the automatic station.(a) Ta; (b) RH.

Figure 6 .
Figure 6.Correlation graph between the DHT22 sensor and the automatic station.(a) T a ; (b) RH.

Figure 7 .
Figure 7.Diurnal NEE for different growth stages of rice fields.

Figure 7 .
Figure 7.Diurnal NEE for different growth stages of rice fields.

Figure 8 .
Figure 8. Spearman correlation heat map (R) between NEE, Reco, and environmental variables Light and dark colors indicate positive and negative correlations, respectively."*" indicates a sig nificant correlation at the 0.05 level (p < 0.05).

Figure 8 .
Figure 8. Spearman correlation heat map (R) between NEE, Reco, and environmental variables.Light and dark colors indicate positive and negative correlations, respectively."*" indicates a significant correlation at the 0.05 level (p < 0.05).

Table 1 .
Physicochemical characteristics of the soil in the study area.

Table 2 .
Physicochemical characteristics of water.

Table 2 .
Physicochemical characteristics of water.

Table 3 .
Diurnal NEE (μmolCO m s ) for different growth stages in rice fields.

Table A2 .
Comparison of the diurnal variation in NEE (µmolCO2 m 2 s −1 ) in floral primordium.

Table A3 .
Comparison of the diurnal variation in NEE (µmolCO2 m 2 s −1 ) in spindle state.

Table A4 .
Comparison of the diurnal variation in NEE (µmolCO2 m 2 s −1 ) in bloom.