Introduction

Achieving climate-resilient agriculture to counteract the threat of global warming and make a positive impact on the sustainable development goal of “Zero Hunger” is a prime focus of the global scientific community. Monitoring of carbon budget and subsequently the sink and sources of CO2 from terrestrial ecosystems plays a dominant role in mitigating climate change due to rising atmospheric CO2. The quantification of the carbon budget of any ecosystem demands long-term monitoring of the biosphere–atmosphere exchange of scalars and energy in combination with governing eco-physiological variables (Baldocchi et al. 2001). Continuous and long-term monitoring of CO2 fluxes from diverse ecosystems has thus become critical in developing a mitigation strategy to reduce carbon emission and enhance carbon sequestration.

Substantial usage of the eddy covariance technique worldwide since the 1980s (e.g., AsiaFlux, EUROFLUX, AmeriFlux networks) proved its capability to measure and analyze the carbon sequestration potential in the terrestrial ecosystem. The phenomenal quality of this technique is to measure CO2 and H2O fluxes on varying spatial (ranging from a single crop field to an extensive forest landscape) and temporal scales (varying from half-hourly to yearly). Although a majority of previous studies were mainly focused over grassland and forest ecosystems (Joon Kim et al. 1992; Mielnick and Dugas 2000; Ansley et al. 2002; Kaul et al. 2009; Hernandez-Ramirez et al. 2011), the study on CO2 exchange over agricultural ecosystem is gradually expanding. Scientists have executed a detailed analysis of intra- and inter-annual variations of CO2 fluxes (Murayama et al. 2003; Aubinet et al. 2009; Glenn et al. 2010). Some have dealt with data processing (Soegaard et al. 2003; Saito et al. 2009) while others focused on the response of environmental variables on CO2 fluxes in different crop sites (Suyker et al. 2004; Moureaux et al. 2008; Béziat et al. 2009; Ceschia et al. 2010). Besides this wide network to measure CO2 and H2O fluxes on varying spatial and temporal scales, poor representation of the Indian subcontinent for eddy covariance (EC) towers may lead to large uncertainty in quantifying the South Asian Carbon budget (Patra et al. 2013). The EC technique is a unique micrometeorological method with high sampling rate and short response time. It provides the continuous flux measurements over large area with detailed information on the short-term fluxes variation.

The establishment of EC flux-towers aimed at quantification of carbon exchanges over varied ecosystems in India was initiated with the inception of the National Carbon Project under the Geosphere-Biosphere Programme of Indian Space Research Organisation (Dadhwal et al., 2010; Jha et al. 2013; Patel et al. 2011; Watham et al. 2014). Later ecosystem flux research gained momentum to cover vast Indian landmass with large heterogeneity in land covers, climate regimes, and physiography (Chatterjee et al. 2018; Deb Burman et al. 2019). Previous studies in India were mainly focused over forest ecosystems (Jha et al. 2013; Watham et al. 2014; Sarma et al. 2018) and few over cropland preferably dominant cereal-based cropping system (Patel et al. 2011; Bhattacharyya et al. 2013; Deb Burman et al. 2020). These efforts seem meager to represent the vast landmass of cropland with diverse cropping systems in India. Furthermore, to the best of our knowledge, there is no carbon flux report addressed over the sugarcane (C4)-wheat (C3)-based cropping system. This cropping system having a combination of cash (sugarcane) and cereal (wheat) crop offers financial and food security of rural livelihood. Sugarcane possesses the capability of being most advantageous among agronomic crops owing to its higher productivity and return on energy investment (Goldemberg et al. 2008; Waclawovsky et al. 2010), further catalyzing its propensity towards higher greenhouse gas reductions in comparison to other agronomic biofuels (de Vries et al. 2010; Davis et al. 2012).

India earmarked its topmost position in the sugarcane and wheat production, accounting for 22.82% and 12.05%, respectively, in the total world’s production. These crops are dominantly manifested in a paddy-autumn sugarcane-ratoon-wheat–based system in the sub-tropical region of India. Further, the difference in photosynthetic pathways (C3 and C4) of these crops enhances the scope of comparative analysis of carbon sequestration potential of the sugarcane-wheat cropping system. Over the last two decades, substantial work being carried out across the world to elucidate variation in carbon and water fluxes over agroecosystems (Pearcy and Ehleringer 1984); (Baldocchi 1994; Li et al. 2006; Aubinet et al. 2009; Béziat et al. 2009). Owing to the non-intrusive nature of the EC technique, it has been widely utilized for carbon accounting of croplands comprising both C3 and C4 crops, for example, carbon balance over a C3 (wheat)-C4 (maize) cropping system (Yongqiang et al. 2002; Lei et al. 2011), C3 (wheat) crop (Patel et al. 2011; Wang et al. 2013; Saha et al. 2015; Chi et al. 2017), and C4 (sugarcane) crop (Teixeira et al. 2013; Cabral et al. 2013; Anderson et al. 2015). Keeping all the facts in mind, the comparative study over C4 (sugarcane) and C3 (wheat) crop rotation mainly practiced in the fertile region of Indo-Gangetic plains of India (Pathak et al. 2014) is the prime need in the context of the research program on climate change. The major objectives of this study are two-fold: (1) to quantify CO2 fluxes and seasonal variation over the sugarcane-wheat cropping system with characteristic behavior of C4 and C3 photosynthetic pathways, (2) to assess the environmental control on net ecosystem exchange (NEE) of sugarcane and wheat crop in the sub-tropical climate of Upper Indo-Gangetic plains.

Material and methods

Site description

The study site was an extensive agricultural land located in southern direction, about 28 km far from Saharanpur district of Uttar Pradesh. This area can be characterized under a dry sub-humid, megathermal climate. The mean annual temperature is 23.4 °C, with a maximum (43.9 °C) and minimum (3.7 °C) temperature in June and January, respectively. The mean annual precipitation is 1204 mm, a large portion of which falls from July to September. The distinctive cropping system in the Saharanpur region consists of sugarcane, wheat, rice, and maize.

In the present study, we dealt with a ratooned sugarcane-wheat-sesbania cropping system, having a sugarcane crop from February 2014 to October 2014, wheat crop in winter from November 2014 to April 2015, and sesbania (Sesbania rostrata) from end April 2015 to June 2015. The soil in the region is coarse sandy loam, with mean bulk density of 1.55 g cm−3.

Flux and micrometeorological measurement

Instruments for measuring eddy covariance and micro-climatic variables were installed on a tripod mast of measurement tower located at the study site (29° 52′ 19.139″ N and 077° 34′ 01.621″ E). CO2 fluxes between the land surface and the atmosphere were measured from January 2014 to June 2015 using EC system installed at a height of 3.5 m above the ground surface. The EC system consists of the open-path gas analyzer (Li-7500, IRGA, Li-Cor., Lincoln, NE, USA) along with a 3D sonic anemometer (CSAT-3, Campbell Scientific). Measured data were sampled at a frequency of 10 Hz using a fast response data logger (Model CR1000, Campbell Scientific, USA). The vertical CO2 flux density was obtained as the covariance between the CO2 mixing ratio (c′) and vertical fluctuations (w′).

Apart from the flux measurements, the ancillary meteorological measurements during the whole study period were also recorded. The net radiation was measured using the net radiometer (CNR1, Kipp and Zonen) installed at 1.5 m above the ground surface. It is a four-component device with a pair of pyranometer and pyrgeometer, used to measure shortwave and longwave radiation, respectively. The ambient temperature and relative humidity was recorded using the temperature-humidity sensor (HMP-45, Vaisala, Inc., Finland). Soil temperature probes were inserted at the depth of 0.075 m and 0.15 m to measure soil temperature. Soil water content (SWC) was measured using CS616 soil moisture sensors placed at 0.05, 0.10, and 0.15 m depth. The soil heat flux component was measured using a HFP01 soil heat flux plate (Campbell Scientific Inc.) placed at the depth of 0.075 m.

The rainfall data was taken from the nearby meteorological station located at 29° 58′ N latitude and 77° 33′ E longitude. We also computed the water availability index (WAI) to examine the water availability in the study region (Wang and Li Xin 2020). It is the ratio of actual evapotranspiration and potential evapotranspiration (PET) and ranges from 0 to 1 representing well-watered to water-stressed conditions. The AET and PET were derived from EC measurements and by the Penman-Monteith equation (Allen et al. 1998), respectively.

Field LAI measurements

The ACCUPAR Model LP-80 (Decagon devices, Pullman, WA, USA), a battery-operated linear photosynthetically active radiation (PAR) ceptometer, was used to calculate leaf area index (LAI) non-destructively in real time in the field. The AccuPAR calculates the LAI based on the above- (Pa) and below-canopy PAR (Pb) measurements along with the parameters related to the position of the sun and the canopy architecture using transmission and scattering equation (Supplementary material) of Norman and Jarvis, 1975.

We measured Pa and Pb values above and below the crop canopy, respectively, to obtain LAI. Pa measurements were obtained by placing LP-80 above the canopy in an unshaded region. Pb was obtained as an average of the 5 individual readings taken in different probe direction parallel or perpendicular to the crop rows. Similarly, the measurements were taken randomly from the 3 points in the field and then averaged to obtain the LAI for that particular time.

Data analysis and processing

Raw data obtained from EC flux tower was further processed in the datalogger starter software (PC200W) to split the obtained dataset into half-hourly intervals. Subsequently, data were processed offline using Eddy Pro 6.2.0 Software (LI-COR, Lincoln, NE, USA). Obtained half-hourly EC data was rejected on the following criteria: (1) incomplete half-hourly dataset; (2) night-time dataset under stable conditions corresponding to u*< 0.1 ms−1 (Figure S1). Approximately 16.44% and 28.32% data was lost due to instrument malfunction, power failure, and the above specified criteria.

Gap-filling

EC system can rarely produce good quality data for whole year. Several reasons exist for the gaps, and it is required to fill the gaps for annual estimates (Anthoni et al. 2004). Gap-filling was done using the following techniques:

  1. 1

    For small gaps (< 2 h), missing data was linearly interpolated (Aubinet et al. 1999).

  2. 2

    For the gaps over 2 h, night-time and day-time fluxes were filled separately by establishing relationship between meteorological parameters, provided the corresponding meteorological parameters remain complete.

    Day-time gaps were filled by expressing a rectangular hyperbolic function (Michaelis-Menten equation) between the gross primary productivity (GPP) and PAR on a monthly basis (Falge et al. 2002).

$$ {F}_{\mathrm{GPP},\mathrm{day}}=\frac{\alpha \ast \mathrm{PAR}\ast \mathrm{Pmax}}{\alpha \ast \mathrm{PAR}+\mathrm{Pmax}} $$
(1)

where FGPP,day is day-time GPP, α represents initial slope of light use efficiency curve (μ mol μ mol−1 photons), and Pmax is the maximum photosynthetic rate at light saturation (μ mol CO2 m−2 s−1). The model (Eq. 1) was fitted using Origin 9.0 software (Micro Cal Software Inc.).

Night-time missing data were filled by using a simple exponential relationship between net ecosystem exchange and air temperature under unstable condition (u* > 0.1 ms−1)

$$ {F}_{\mathrm{NEE},\mathrm{night}}=a\ {\exp}^{\left(b\ast \mathrm{Ta}\right)} $$
(2)

where FNEE,night is net ecosystem exchange (NEE) during night-time, a and b are empirical constants, and Ta is the air temperature. Ecosystem respiration during day-time (Rd) was extrapolated from Eq. (2). The negative NEE values imply net CO2 uptake by the vegetation while positive NEE values indicate net CO2 release into the atmosphere by the vegetation.

Q10 was calculated according to the Van’t Hoff equation:

$$ {Q}_{10}={e}^{10\mathrm{b}} $$
(3)

where Q10 is the temperature sensitivity of FNEE,night or night-time respiration (Rn) and b is the empirical constant which reflects the temperature sensitivity of Rn

Net ecosystem respiration (Rn) during night-time can be considered as a source of entire CO2 flux due to absence of photosynthetic activity (Vote et al. 2015). Subsequently, net ecosystem respiration (Reco) was the sum of Rn (net ecosystem respiration during night-time) and Rd (net ecosystem respiration during day-time).

$$ {R}_{\mathrm{eco}}=\mathrm{Rn}+\mathrm{Rd} $$
(4)

Subsequently, NEE was partitioned into GPP and Reco using (Eq. 5)

$$ \mathrm{GPP}=-\mathrm{NEE}+{R}_{\mathrm{eco}} $$
(5)

where GPP denotes gross primary productivity (μ mol CO2 m−2 s−1), NEE is net ecosystem exchange (μ mol CO2 m−2 s−1), and Reco is net ecosystem respiration (μ mol CO2 m−2 s−1).

In our study site, incomplete flux dataset was available for the months of January, February, March, November, and December of 2014. To represent the flux dataset on a yearly basis, we gap-filled the aforementioned months based on the available meteorological parameters. We applied the Michaelis-Menten equation and exponential temperature equation to fill day-time and night-time data, respectively.

Path analysis

Dependence of NEE on five primarily affecting environmental variables {PAR, Ta, vapor pressure deficit (VPD) and SWC, WAI} was evaluated using path analysis with EC data collected over sugarcane and wheat growing season. Path analysis is an extension of multiple regression which is a useful analysis tool when correlative information of relationship among variables is known in prior (Li 1981). Further, it acts as an appropriate tool, when the independent nature of predictor variables is not certain and is expected to depend on one another. Path analysis is effectively used in many ecological studies (Bassow and Bazzaz 1998; Xu and Qi 2001; Huxman et al. 2003; Zhuang et al. 2004; Saito et al. 2009).

In the present study, path analysis was done to partition the correlation coefficient into direct and indirect effect on variables. The direct effect is the standardized partial regression coefficient of the dependent variable y and independent variable i, which indicates the direct impact of i on y. The indirect effect indicates how the independent variables influence each other, which in turn affects the dependent variable.

$$ {r}_{\mathrm{yi}}={P}_{1\mathrm{y}}{r}_{1\mathrm{i}}+{P}_{2\mathrm{y}}{r}_{2\mathrm{i}}\dots \dots +{P}_{\mathrm{ny}}{r}_{\mathrm{ni}}\ \left(i=\mathrm{1,2,3}\dots n\right) $$
(6)

where ryi represent the correlation coefficient between dependent variable y and independent variable i, Pny is the direct effect of the independent variable n on dependent variable y, rni is the correlation between independent variables (i and n), and Pny × rni is used to express the indirect path coefficient of i through n on y.

Satellite-based GPP products

In the present study, we utilized the available GPP products from January 2014 to June 2015. The GPP products include the GPP from Moderate Resolution Imaging Spectroradiometer (MODIS) dataset and Vegetation Photosynthesis Model (VPM) (Supplementary material 2.2) The MODIS GPP (GPPMOD) obtained from the MODIS version 6 (MOD17A2H.006) is a cumulative 8-day composite with a 500-m spatial resolution. The GPPMOD values were extracted on an 8-day basis from the study site using the Google Earth Engine platform (GEE).

GPP from the VPM model (GPPVPM) was also utilized to compare with the observed GPP from the EC tower site (GPPEC). The GPPVPM product is available at 8-day temporal resolution with a moderate spatial resolution of 500 m over the entire globe from the years 2000 to 2016. The GPPVPM product was downloaded from the US Department of Agriculture (https://data.nal.usda.gov/dataset/global-moderate-resolution-dataset-gross-primary-production-vegetation-2000%E2%80%932016). The obtained GPP products (GPPVPM and GPPMOD) available at an 8-day interval were then compared with the 8-day GPPEC.

Results and discussion

Variation in meteorological variables

The variation in meteorological variables over the 2014–2015 growing season is shown in Fig. 1. The observed seasonal trend was almost similar for VPD and PAR. The average air temperature was 26.49 °C and 22.30 °C for 2014 and 2015 growing season, respectively. Daily average SWC (measured at 7.5 cm) varied significantly ranging from 0.25 to 0.47 m3 m−3 and 0.23 to 0.51 m3 m−3 for the years 2014 and 2015, respectively. The mean day-time VPD values range from 1.02 to 38.55 hPa and 1.64 to 48.60 hPa in the years 2014 and 2015, respectively.

Fig. 1
figure 1

Variation of meteorological parameters at Saharanpur flux site (SFS) in 2014 (left) and 2015 (right). (a) Sum of photosynthetically active radiation (MJm−2 day−1), (b) daily mean air temperature (°C), (c) mean day-time vapor pressure deficit (h Pa), (d) soil water content (m3 m−3) at a depth of 7.5 cm, (e) rainfall (mm), and (f) leaf area index. Bar indicates ±SD. Dashed line in middle of the graph represents bifurcation between the years 2014 and 2015

Full size image

The seasonal average of PAR during sugarcane (Feb 2014–October 2014), wheat (Dec 2014–April 2015), and sesbania (May 2015–June 2015) growing season was 9.97, 5.74, and 7.88 M J m−2 day−1, respectively. The average seasonal air temperature was 30.03 °C, 19.54 °C, and 30.11 °C for sugarcane, wheat, and sesbania growing season, respectively. Volumetric soil water content (SWC) was higher during the sugarcane growth period as compared to that in wheat season, reaching its maximum in July (0.41 m3m−3). SWC in the wheat growing season varies from 0.37 m3m−3 (March) to 0.32 m3m−3 (February). The daily rainfall (RF) shows clear seasonal patterns, and was mainly distributed in the growing period. Considering the study period from January 2014 to June 2015, the annual rainfall shows large variation, ranging from 618.6 mm (Jan 2014 to Dec 2014) to 224.2 mm (Jan 2015–June 2015). RF is highest during June–August and reduced during the winter season.

The meteorological parameter assessed during sugarcane-wheat-sesbania growing season showed great variability which further provided an excellent opportunity to quantify the impact of climate variability on CO2 exchange.

Diurnal dynamics of carbon dioxide fluxes

The mean of 30-min NEE has been represented through NEE fingerprint plot (Fig. 2A). The diel NEE variation revealed that the CO2 absorption dominates the carbon release after around 8:00 hr. The blue colored portion in the fingerplot depict CO2 uptake which peaks around 10:30 to 14:00 hrs. The 30-min NEE growing season ranged from −49.8 to 39.3 μ mol m−2 s−1 over the sugarcane growing season. NEE ranged from −22.63 to 20.75 μ mol m−2 s−1 during the wheat growing season. During the sesbania growing season, the NEE ranged from −13.41 to 16.40 μ mol m−2 s−1, which later increased up to 31.81 μ mol m−2 s−1, after its incorporation into the soil.

Fig. 2
figure 2

(A) NEE fingerprint plot representing half-hourly NEE (μmol m−2 s−1) during whole study period. The thick dashed line represents the start of the (a) sugarcane growing season, (b) fallow period, (c) wheat growing season, and (d) sesbania growing season. (B) Diurnal variation of net ecosystem CO2 exchange (NEE) in different months of growing seasons of crops (a) sugarcane and (b) wheat

Full size image

Figure 2B shows the monthly diurnal variation in NEE during the sugarcane and wheat growing seasons. It depicted that the ecosystem acted as a strong CO2 sink during the active vegetative period and a CO2 source during the senescence phase. Diurnal variation in winter months (December and January) was not quite remarkable, owing to bare soil, low temperature, and low PAR. During the main growing season of sugarcane and wheat, there was notable diurnal variation in NEE. A distinctive U-shaped curve was observed with one absorption peak (at mid-day near noon) and two emission peaks (6:00 and 19:00 hrs). The uptake of CO2 in the form of negative NEE has distinct diel variation and matches with the intensity of radiation controlled by movement and position of sun in the sky. The NEE became negative with sunrise and attained peak negative value at noon hours, then after CO2 uptake tended to decline until evening hours and NEE became positive during night hours. Later in the day, the upward peak of the curve was the result of either downward ramping of the photosynthesis, stems from a daily increase in temperature, which coincidently increases respiratory losses from soil and vegetation (Larcher, 1995) or reduction in leaf gas exchange via stomata (Goulden et al. 2004). This trend was in line with the previous findings in a range of ecosystems (Grassi et al. 2009; Wagle et al. 2015).

The diurnal peak was higher in July and March (Fig. 3) in sugarcane and wheat growing season, respectively, on account of the direct relationship of NEE with canopy closure and higher PAR values. The maximal peaks of diurnal CO2 uptake in sugarcane and wheat crop were −2.21 mg CO2 m−2 s−1 and −1.49 mg CO2 m−2 s−1, which were observed in the months of July and March, respectively. Both appeared during a maximum vegetative phase which was similar to the seasonal variation in LAI. The minimal NEE values in December 2014 and January 2015 reached up to 0.077 mg CO2 m−2 s−1and 0.017 mg CO2 m−2 s−1, depicting the weak CO2 absorption strength after sugarcane harvesting. Similarly, NEE values in June 2015 turned positive (0.14 mg CO2 m−2 s−1), depicting the CO2 source status of the study site. Higher respiration due to higher temperature and microbial activity outstripped the CO2 uptake during that period. The daily difference of NEE in a crop field was determined by the extent of CO2 source and sink (Jun et al., 2006). The daily difference of monthly averaged NEE was higher over the study site during the active growing period (April–August (sugarcane) and February–April (wheat)). In the sugarcane crop, it ranged from 2.04 to 3.61 mg CO2 m−2 s−1, while in wheat crop the range varied from 2.82 to 2.95 mg CO2 m−2 s−1. The higher difference in NEE during active growing season attributed to higher photosynthetic potential during the day and respiration at night.

Fig. 3
figure 3

Dynamics of net ecosystem CO2 exchange (NEE), gross primary productivity (GPP), and ecosystem respiration (Reco) in sugarcane-wheat-sesbania system during the 2014 and 2015 growing season. The first section displays the carbon fluxes (GPP, NEE, and Reco) in sugarcane growing season while 2nd, 3rd, and 4th sections display fallow, wheat, and sesbania growing season, respectively

Full size image

Seasonal variation in CO2 fluxes

In Fig. 3, the daily averaged seasonal variations of GPP, NEE, and Reco are shown. Over the entire course of the study, based on positive and negative values of NEE, the ecosystem can be divided into four sections: two stages of CO2 uptake in sugarcane and wheat growing seasons (negative NEE), one stage of CO2 emission after sugarcane harvesting,, and another after wheat harvesting (positive NEE). Ecosystem exhibited as both carbon source and sink, depending upon the crop type and crop phenology, aligned with two major physiological activities (photosynthesis and respiration). Reco was greater than GPP in early and late growing season due to low photosynthetic activity and higher respiration rates; NEE was positive, and the ecosystem was a carbon source. GPP increases with enhancement in growth stages in the sugarcane-wheat system, eventually exceeding Reco due to which ecosystem acted as a carbon sink. NEE had negative periods during 32 DOY (−2.696±1.212 g C m−2 day−1)–270 DOY (−1.198±0.795 g C m−2 day−1) and 60 DOY (−3.286±1.848 g C m−2 day−1)–120 DOY (−1.513±0.771 g C m−2 day−1) displaying sugarcane and wheat crop canopy, respectively. Cumulative nocturnal average of CO2 efflux during the month of October (274 DOY–282 DOY) was 37.5% higher than the CO2 uptake, shifting NEE values clearly on the positive side tending to act as a carbon source.

Also, in December, the ecosystem maintains its carbon source characteristics as the surface was respiring due to the low vegetative growth of the wheat crop. Wheat active growing season started from the end of January (25 DOY) to mid-April (114 DOY), under which NEE values were constantly over positive side up to February depicting early growth stage in wheat crop. Crop experiences maximum CO2 absorption after it attains canopy closure, which lasted for a shorter duration. The anthesis stage of wheat crop also possesses a limiting factor in carbon uptake rates, as the respiration exceeds distinctively due to the involvement of high respiration costs in order to produce reproductive organs (Baldocchi 1994; Rochette et al. 1995). After harvesting (114 DOY), ecosystem respiration again exceeded, owing to the soil warmth and new input of fresh decomposable material, shifting the ecosystem into a carbon source. Maximal CO2 uptake approached −8.94 g C m−2 day−1 (195 DOY) over sugarcane stand (C4) and −6.79 g C m−2 day−1 (78 DOY) over the wheat crop (C3). Differential CO2 uptake rates in both crops were in good agreement with the results of Falge et al. (2002) and Jun et al. (2006), who reported that the maximal CO2 uptake rates in C4 crops are higher than in C3 crop, owing to their photosynthetic pathway. Net assimilation fluxes were observed between −9 and −13 g C m−2 day−1 for the wheat crop (Anthoni et al. 2004; Béziat et al. 2009). Similar values were observed for sugar beet (Moureaux et al. 2008), rapeseed (Béziat et al. 2009), and soybean (Hollinger et al. 2005).

Just after wheat harvesting, sesbania (green manure) seeds were sown on 115 DOY, which after about 30 days, further incorporated into the soil (144 DOY), thereby increasing respiration at a higher rate. Daily average CO2 absorption during green manuring phase reaches its maximum at −0.23 g C m−2 day−1 (141 DOY), which after higher respiration rate (due to its incorporation into the soil) continuously decreases reaching even lower than daily averaged nocturnal CO2 efflux thus acting as a carbon source during that phase. Reco was greater than GPP even during the sesbania growing season, which lowered down the carbon sequestration capacity of the ecosystem. This phenomenon could be due to the application of artificial irrigation on 132 DOY (Fig. 1) which induced favorable conditions for higher soil microbial activity thus reducing the CO2 sequestration capacity of the ecosystem.

Cumulative carbon fluxes

The daily cumulative carbon fluxes were quantified during different crop growing seasons and illustrated in Fig. 4. The NEE, GPP, and Reco were −898.24, 3696.87, and 2822.69 in the year 2014 (Fig. 4a). The values of cumulative carbon fluxes were −923.04, 3316.65, and 2433.18 g C m−2 during sugarcane growing season. The cumulative carbon fluxes were −123.32 (NEE), 1024.25 (GPP), and 895.83 (Reco) g C m−2 in the year 2015 (Fig. 4b), representing wheat-sesbania (active growth phase (Ga) + Soil incorporation (Is)) growing season. When wheat and sesbania were analyzed separately, the NEE, GPP, and Reco values over the wheat growing season were −192.30, 621.47, and 488.34 g C m−2. Similarly, during Ga (DOY 114-DOY 144) in sesbania, cumulative NEE, GPP, and Reco reached up to 7.19, 167.97, and 175.17 g C m−2, respectively. The shorter growth phase in sesbania inclined the ecosystem more towards carbon source after wheat harvesting. Moreover, after Is, a considerable increase in carbon emission from the soil was found and the cumulative fluxes during that phase (DOY 144–DOY 170) reached up to 60.49 (NEE), 347 (GPP), and 407.50 (Reco) g C m−2. The sudden and effective shift of the ecosystem towards a carbon source after Is may be ascertained to higher soil respiration which was further influenced by elevated temperature and addition of the labile substrate (Dash et al. 2014). Soil respiration activates carbon emission from the soil which is more strongly driven by the activities of soil microbes and plant roots. Furthermore, addition of complex organic substrates (e.g., sesbania root) deployed more C emission from the soil, ascribed to the activation of a higher number of the microbial community present in the soil (Beare et al. 1990; Hoyle and Murphy 2007).

Fig. 4
figure 4

Cumulative carbon fluxes in (a) sugarcane (2014), (b) wheat–sesbania (2014–2015) growing season. The dashed light gray lines in section (a) represent the start and end of the sugarcane growing season. The vertical thick lines are the harvesting period in wheat and sesbania. The inset depicts cumulative NEE (g C m−2) highlighting the variation from carbon sink to carbon source after wheat harvesting and sesbania incorporation into the soil

Full size image

The seasonal flux values are comparable to numerous published carbon flux values for C4 and C3 crops on a global basis, although published values are from studies conducted in areas with different local climates, soil characteristics, and management practices. Compared to recent studies in different sites (Table 1), the NEE of ratooned sugarcane in Saharanpur found less than all sites except over the ratooned sugarcane in the Sau Paulo State, Brazil (Cabral et al. 2013; Xin et al. 2020) and the first cycle of sugarcane grown in the Louisiana region of the USA (Xin et al. 2020). This could be explained by two points: (1) our study was focused on ratooned sugarcane (2nd cycle), during which lesser carbon sequestration was observed as compared to the previous growth cycle (Cabral et al. 2013); (2) shorter growth cycle of sugarcane (9 months) in India in comparison to the other sugarcane growing regions in the world (12 months).

Table 1 Comparative analysis of carbon fluxes of sugarcane and wheat growing season at different study sites
Full size table

The cumulative carbon flux value over wheat crop (C3) was higher than the results reported by Li et al. (2006), while lower than the values reported by Wang et al. (2013) and Chen et al. (2015) over wheat crop. Cumulative carbon fluxes were also lower than the values reported by Bhattacharyya et al. (2013) over rice crop in Cuttack, Orissa. Nevertheless, the maximum values of NEP and GPP were similar to the values reported by Patel et al. (2011); Bhattacharyya et al. (2013); and Wang et al. (2013) over C3 cropping system.

Comparative analysis of C3 and C4 crops

The magnitude of CO2 flux is often related to the amount of growing plant tissue, represented by indices such as above-ground biomass and LAI (Frank and Dugas 2001; Flanagan and Johnson 2005). CO2 fluxes were found temporally disproportionate with the leaf area index in sugarcane growing season specifically at the senescence stage, similarly as observed in corn by Baldocchi (1994). This may be due to temporal changes in canopy architecture which causes canopy assimilation rates to diverge as leaf area increases (Baldocchi 1994). Keeping the LAI and GPP values only for active phases in both growing seasons determines a strong linear relationship (r = 0.92). Such a linear relationship between GPP and LAI was also related in other studies (Li et al. 2005; Vitale et al. 2016). LAI imparts a major impact on the CO2 exchange rate of a plant canopy primarily through its control over absorption and availability of PAR for photosynthesis (Hodges and Kanemasu 1977; Hipps et al. 1983). To distinguish the impact of biological and environmental factors on canopy photosynthesis of C3 and C4 crops, we normalized GPP by LAI (Vitale et al. 2016). Further, the GPP/LAI ratio was expressed as an individual function of PAR and air temperature. The relationship between GPP/LAI and air temperature depicted the optimum temperature (To) for photosynthesis in C3 and C4 crops. C3 crop exhibits a To in the range of 12–20 °C (Fig. 5b), while C4 crop exhibits a greater photosynthetic rate at higher To (25–30 °C) than C3 crops. Moreover, the photosynthetic rate in C4 crop reveals narrower temperature range in comparison to C3 (Fig. 5a), as the photosynthetic rate in C4 plunged at low temperature (Yamori et al. 2014).

Fig. 5
figure 5

(a) The relationship between GPP and air temperature in sugarcane (C4) and wheat growing seasons, and (b) and (c) depict response of GPP per unit of LAI (GPP/LAI) to air temperature (°C) and photosynthetic active radiation (μ mol m−2 s−1), respectively

Full size image

Similarly, GPP per unit LAI relationship with PAR reveals a steady increase in CO2 fixation in C4 crop with an increase in light intensity in comparison to the C3 crop (Fig. 5c). This pattern elucidated more efficient use of solar energy in the case of the C4 photosynthetic pathway (Wang et al. 2012). Further to comprehend the CO2 fixation a bit more, we selected two points of the active growth sections in both growing seasons (June (sugarcane growing season), March (wheat growing season)). CO2 fixation in C4 crops was higher than in the C3 crop at approximately the same light intensity (C4 (250.26 μ mol m−2 s−1), C3 (241.61 μ mol m−2 s−1)).

Response of CO2 fluxes to environmental parameters

Daily NEE evaluation using path analysis

Path analysis was used to examine the impact of major environmental (predictor) variables on regulating NEE at daily time step as mentioned under methodology for both sugarcane (n = 173) and wheat (n = 120) growing seasons (Fig. 6). The direct (i.e., path coefficient/DE), indirect (i.e., coupling effect/IE), and total effect (TE) of environmental variables on NEE are shown in Table S1. All the environmental variables were significantly correlated with NEE. In both growing seasons, PAR was the highly correlated variable with NEE (Table S2). PAR has the greatest DE, followed by VPD and Ta in sugarcane and wheat growing season, respectively. The direct path coefficient of PAR on NEE was −0.66 (sugarcane) and −0.74 (wheat), showing the comparatively much stronger impact of PAR on NEE during the wheat growing period.

Fig. 6
figure 6

Path diagrams illustrating effect of photosynthetic active radiation (PAR), air temperature (Ta), vapor pressure deficit (VPD), soil water content (SWC), and water availability index (WAI) on daily net ecosystem exchange (NEE) during (a) sugarcane growing season (n = 173) and (b) wheat growing season (n = 120). Analysis was based on daily average of all variables considered. Black arrows represent positive paths; gray arrows represent negative paths. Bold numbers indicate standard path coefficient. Arrow width indicates the significant correlation at p<0.05 between the environmental variables

Full size image

The IE depict how the environmental variables influenced NEE through other variables under consideration. The path analysis showed that VPD (−0.410) and SWC (0.205) had stronger IE on NEE during sugarcane growing season, while Ta (−0.539) and VPD (−0.304) had stronger IE on NEE during wheat growing season. The TE of SWC (−0.205) and WAI (−0.229) on NEE was higher in sugarcane growing season, in comparison to the wheat growing season. This indicates higher dependency of sugarcane on water for growth and development (Cabral et al. 2013).

Response of day-time NEE to PAR

The relationship between day-time NEE and incident PAR was established under different VPD sections (Fig. 7). Net CO2 uptake increases with the rise in VPD up to 20 h Pa which further started diverting in a decreasing trend with a further rise in VPD. We found R2 between NEE and incident PAR for different VPD ranges over the whole sugarcane and wheat growing season. R2 values were higher in VPD values ranging from 0 to 20 h Pa (0.61 (sugarcane) and 0.40 (wheat)) (95% confidence interval, p<0.01) in comparison to VPD ranging from 20 to 40 h Pa (0.41 (sugarcane) and 0.25 (wheat)) (95% confidence interval, p<0.01). The net CO2 uptake reached up to −2.0 mg CO2 m−2 s−1 and −1.45 mg CO2 m−2 s−1 when VPD ranges from 0 to 20 hPa in sugarcane and wheat growing season, respectively. On the contrary, the net CO2 uptake decreases and reached up to −1.6 mg CO2 m−2 s−1 and −1.0 mg CO2 m−2 s−1 when VPD ranges from 20 to 40 hPa in sugarcane and wheat growing season, respectively.

Fig. 7
figure 7

Relationship between NEE and PAR in (a) sugarcane growing season and (b) wheat growing season at 0<VPD<20 hPa and 20 hPa< VPD<40 hPa. Dark gray colored closed up triangles represent the net CO2 uptake at VPD range from 0 to 20 hPa, while light gray colored open circles represent the net CO2 uptake at VPD range from 20 to 40 hPa

Full size image

The decrease in net CO2 uptake beyond 20 hPa was may be due to the effect of higher VPD on stomatal conductance. Decrease in stomatal conductance further influences carboxylation rate as vegetation absorbs CO2 from the atmosphere through stomata.

Response of night-time NEE (Rn) to air temperature

Respiration comprises of both autotrophic and heterotrophic respiration, which are predominantly controlled by temperature and soil moisture (Flanagan and Johnson 2005; Nakano et al. 2008). Therefore, we plotted the Rn against the air temperature in both sugarcane (Fig. 8a) and wheat growing seasons (Fig. 8b), under varying soil water conditions measured at a depth of 7.5 cm (SWC). As expected, Rn during sugarcane growing season was significantly larger than during wheat growing season which reflects greater plant respiration of sugarcane owing to its greater biomass (Nakano et al. 2008)

Fig. 8
figure 8

Response of night-time respiration (Reco) to air temperature (Ta) over (a) sugarcane growing season and (b) wheat growing season under water sufficient conditions (open black circle), water deficit condition (open black square), and water excess condition (closed grayish triangle). Date was bin averaged at 1 °C

Full size image

In our study, Rn was exponentially related to the corresponding temperature. An exponential relationship between night-time respiration and air temperature has been previously described (Suyker et al. 2004; Wang et al. 2013; Vote et al. 2015; Chen et al. 2015). Optimum soil moisture (0.33 ≤ SWC < 0.42 m3 m−3 (sugarcane) and 0.31 ≤ SWC < 0.37 m3 m−3 (wheat)) leads to higher correlation between NEE and air temperature. Further, the correlation values were arranged under different soil moisture conditions as water sufficient condition> water deficit condition> water excess conditions ranging from 0.64 to 0.87 (95% confidence interval, p<0.0001) and 0.38 to 0.75 (95% confidence interval, p<0.0001) over sugarcane and wheat growing season, respectively.

Q10 was calculated from Eq. 3, which varied from 1.12 to 2.18 (2.03–2.18 (sugarcane), 1.12–1.54 (wheat)) influenced by environmental temperature, substrate availability, and soil moisture (Davidson et al., 2006; Nakano et al. 2008; Jingxue et al., 2019). Contrary to the previous studies reported (Nakano et al. 2008; Jingxue et al., 2019), Q10 was higher under water sufficient condition in both growing seasons. Lower Q10 values under water deficit and water excess conditions could be the impact of suppressed respiration activity under both the conditions.

Comparison of GPP product against observed GPP

The available satellite GPP products (GPPMOD and GPPVPM) were compared against the GPPEC (Fig. 9). The GPPMOD (R2 = 0.02, RMSE = 8.74 g C m−2 day−1) has showed very low performance in comparison to the GPPVPM (R2 = 0.63, RMSE = 4.63 g C m−2 day−1). Various studies also depicted the poor results of GPPMOD over the agricultural sites (Zhang et al. 2012; Wang et al. 2017). The poor performance of GPPMOD can be attributed to its two major assumptions: (1) constant LUE value for individual biome type; (2) APAR (PAR × fPAR) is the energy absorbed by the entire canopy (Zhang et al. 2017). The GPPMOD is primarily based on the leaf quantity (amount of the leaf area). Though recent studies have depicted that not only leaf quantity but also leaf quality (photosynthetic rate of the individual leaf) is also a major determinant of the photosynthetic capacity (Wu et al. 2016). The leaf quality is based on the leaf chlorophyll content and it is not only dependent upon the environment and nutrient availability but also regulated by the phenology. Further, the phenological variations regulate the seasonal GPP variations. Unlike GPPMOD, GPPVPM take leaf quality into consideration in the form of fraction of PAR absorbed by chlorophyll (fPARchl), which is computed from the EVI. It can better capture the seasonal variation in the photosynthetic capacity and can further improve the seasonal representation of the GPP (Zhang et al. 2009, 2014).

Fig. 9
figure 9

Relationship of (a) satellite-based GPP products (GPPVPM and GPPMOD) and (b) MODIS-based EVI against GPPEC. The half-filled diamonds represent the relationship between GPPEC and GPPVPM while the dark gray colored stars represent the relationship between GPPEC and GPPMOD. Solid black diamond, solid green up triangles, and open diamond represent the EVI vs GPP during sugarcane growing season, wheat growing season, and whole study period, respectively

Full size image

The 8-day variation in the GPPVPM and GPPEC from January 2014 to June 15 is depicted through Fig. 9a and Fig. S2. The result showed that the GPPVPM has moderately captured the seasonal variation in GPPEC over sugarcane-wheat-based system. Considering GPPEC as the reference, the GPPVPM and GPPMOD underestimated the GPP by 36% and 66%, respectively. Similar results have been observed by Xin et al. (2020) over the sugarcane crop in Louisiana, USA. We also assessed the biophysical variation of EVI in terms of GPP dynamics (Fig. 9b). The relationship between EVI and GPPEC (R2 = 0.49) was also established to address the moderate results from the comparative analysis of GPPEC and GPPVPM. The EVI showed the seasonal phenological variations during the whole growing season, but performed moderately to capture the peak growing period (Fig. 10B). This could be due to the shorter footprint (250–300 m) of the EC site and heterogeneity of the land surface. The GPP products are available at 500 m resolution, which takes into account a large footprint area around EC site with higher heterogeneity which ultimately lead to the consideration of the mixed pixels.

To the best of our knowledge, the comparative study of satellite-based GPP products and GPPEC was never done in India particularly over the sugarcane (C4)-wheat (C3)-based system, though various comparative studies have been well evaluated over the croplands in India which only include the crops with the C3 photosynthetic pathway (Patel et al. 2011). The continuous evaluation of the GPPEC and satellite-based GPP over different ecosystems and over the vegetation with different photosynthetic pathways is imperative to accurately upscale the GPP at regional or global scale. Owing to the continuous and systematic measurements over different vegetation and ecosystems, satellite remote sensing plays an important role for global GPP quantification. Further, the GPP estimation at regional, continent, or at global scale can improve our understanding of the feedbacks between the atmosphere and the terrestrial biosphere in context of the climate change and further can facilitate climate-policy making (Canadell et al. 2007; Xiao et al. 2008).

Conclusions

In order to understand the dynamics of CO2 fluxes (NEE, Reco, and GPP) over C4-C3 crops in sub-tropical dry sub-humid climate, the seasonal variation of carbon fluxes and the controlling environmental variables over sugarcane-wheat-sesbania-based system was analyzed from 2014 to 2015 using the EC measurements and the ancillary environmental variables.

Results indicated that sugarcane being a C4 (high biomass crop) engaged in more carbon sequestration in comparison to wheat. The diurnal dynamics of NEE over the ecosystem depicted the carbon sink capability in sugarcane from May to August whereas wheat exhibits a peak of carbon sequestration by the end of March in the year 2015. Just after wheat harvest, the ecosystem becomes a carbon source up to sesbania sowing, which further acted as a carbon source owing to the increase in respiration activity due to the incorporation of sesbania into the soil (144 DOY). The cumulative NEE was −923.04 g C m−2 and −192.30 g C m−2 for sugarcane and wheat growing season, respectively. CO2 fluxes were jointly affected by multiple environmental parameters. Path analysis indicated that PAR was the most important predictor variable affecting daily NEE over both growing seasons. The variation of GPPEC with the satellite products was also evaluated. This is the first case study in India to evaluate the flux dynamics over C4-C3 cropping system. Furthermore, the present study evaluates the broad aspects ranging from micrometeorology to remote sensing. The results are also important for accurate modelling of agroecosystem for this area and further planning for the effects of climate change.