Daily and seasonal patterns of CO 2 fluxes and evapotranspiration in maize-grass intercropping no consórcio milho-pastagem

Studies that investigate the relationships between CO 2 fluxes and evapotranspiration (ET) are important for predicting how agricultural ecosystems will respond to climate changes. However, none was made on the maize-grass intercropping system in Brazil. The aim of this study was to determine the ET and CO 2 fluxes in a signal grass pasture intercropped with maize, in São João, Pernambuco, Brazil, in a drought year. Furthermore, the soil water storage (SWS) and leaf area index (LAI) were determined. The latent heat flux was the main consumer of the available energy and the daily and seasonal ET and CO 2 variations were mainly controlled by rainfall, through the changes in soil water content and consequently in SWS. The agroecosystem acted as an atmospheric carbon source, during drier periods and lower LAI, and as an atmospheric carbon sink, during wetter periods and higher LAI values. In a dry year, the intercropping sequestered 2.9 t C ha -1 , which was equivalent to 8.0 kg C ha -1 d -1 . This study showed strong seasonal fluctuations in maize-grass intercropping CO 2 fluxes, due to seasonality of rainfall, and that this agroecosystem is vulnerable to low SWS, with significant reduction in CO 2 uptake during these periods.


Introduction
Long-term flux measurements of different crop species are necessary to improve our understanding of management and climate effects on carbon flux variability as well as cropland potential in terrestrial carbon sequestration (Béziat et al., 2009). Quantifying carbon dioxide (CO 2 ) fluxes in terrestrial ecosystems is critical for better understanding of global carbon cycling and observed changes in climate (Hernandez-Ramirez et al., 2011).
In Brazil, these studies focus mainly on rain forest and grass (Randow et al., 2004;Meirelles et al., 2011) and none measured these fluxes in maize-grass intercropping system using the eddy covariance method in semiarid lands of Brazil. Thus, it is not clear how much carbon could be captured by the maize-grass intercropping system in semiarid regions of Brazil.
Here, a micrometeorological experiment over maize-grass intercropping system was conducted in the northeastern region of Brazil from January to December 2013 to improve the current understanding of CO 2 , water and energy exchanges over this ecosystem. Therefore, the main objectives of the present work are: (1) to quantify the seasonal and diurnal variations in the CO 2 , energy and water fluxes over the maizegrass intercropping and (2) to understand the biotic and abiotic factors controlling these fluxes in this agroecosystem in a drought year.

Material and Methods
The water, energy and CO 2 fluxes were measured in an area cultivated with signal grass (Brachiaria decumbens Stapf) intercropped with maize (Zea mays L.). The area of study was in São João, Pernambuco, Brazil (8º 52' S; 36º 22' W; 705 m).The climate is hot and humid, with average annual rainfall of 782 mm, and the wettest trimester consists of the months of May, June and July and the soil is classified as Regolithic Neosol (Gondim et al., 2015).
In 2012, one of the worst droughts recorded in the last 50 years occurred in this region. Consequently, the grassland that was planted in 2000 was destroyed. To recover these grasslands for grazing, the producer planted maize intercropped with signal grass on June12, 2013. Maize was sown with a spacing of 1.0 m between rows, with seven seeds per meter. Between the rows of maize, two rows of signal grass were sown. Maize emergence occurred on June 27, 2013, and harvested on September 14, 2013. Maize was planted with signal grass to recover the pasture in the system known as "Santa Fé". This system is one of the most frequently used cultivation systems for recovering pastures in this region.
This study was conducted throughout the year of 2013, considering three sub-periods: the no cropped period (from January 1 to June 27); the intercropping period (from June 28 to September 14) and the grass alone period (from September 15 to December 31).
Soil water content was measured at depths of 0.10, 0.20, 0.30, 0.40 and 0.50 m using a water content reflectometer (model CS 615, Campbell Scientific Inc., Logan, UT, USA) and then the soil water storage (SWS) was calculated for the layer of 0-0.50 m, according to Lima et al. (2013).
The leaf area index (LAI) was determined seven times for maize and eight times for grass, at approximately fifteen-day intervals from June to December 2013 in five 1 m 2 plots. The LAI was obtained using an automatic leaf area meter (model LI-300, LI-COR Inc., Lincoln, NE, USA).
Meteorological and eddy covariance measurements were made at a height of 1.5 m above the maize/grass layer by a tower installed in the center of an area (420 x 400 m) that had a flat topography. The height of the sensors increased as the maize/grassland grew taller over the study period, such that the distance between the sensors and the crops did not change.
The tower was equipped with sensors to measure wind speed and direction (model 034B Windset, Campbell Scientific Inc., Logan, UT, USA), relative humidity and air temperature (model HMP45C, Vaisala, Campbell Scientific Inc., Logan, UT, USA) and rainfall (model TE 525WS-L, Texas Electronics, Dallas, TX, USA).The solar global radiation (Rs) was measured with a pyranometer (model LI-200X, LI-COR Inc., Lincoln, NE, USA) and net radiation (Rn) was measured with a net radiometer (model NR-Lite, Kipp & Zonen, Campbell Scientific Inc., Logan, UT, USA). Soil heat flux (G) was measured at depth of 0.02 m by using one soil heat flux plate (model HFT3, REBS, Seattle, WA, USA). The measurements from all of the sensors were recorded by a data logger (model CR1000, Campbell Scientific Inc., Logan, UT, USA) every 60 s. The mean/sum data were logged every 1800 s. The tower also was equipped with a measurement system of eddy covariance, which consisted of a three-dimensional sonic anemometer (model CSAT3, Campbell Scientific Inc., Logan, UT, USA) and a fast response CO 2 /H 2 O infrared gas analyzer anemometer (model EC-150, Campbell Scientific Inc., Logan, UT, USA). These data were recorded at a rate of 10.0 Hz.

Results and Discussion
The daily average of air temperature and vapor pressure deficit and daily total of rainfall and soil water storage are shown in Figure 1. The maximum air temperature was 26.3 °C, minimum air temperature was 17.4 °C and an average air temperature was 21.4 °C. Little variation in air temperature was observed during the study period ( Figure 1A).
The vapor pressure deficit (VPD) was associated with the air temperature and rainfall. During periods of lower temperature and higher rainfall, the VPD was lower, and vice versa. Krishnan et al. (2012) examined the seasonal and interannual variability of the energy balance of the two grasses in Arizona (a semiarid region) and found that the VPD varied from 1.47 to 1.98 kPa, with the highest values occurring during the periods with higher temperatures.
A total rainfall of 416.8 mm ( Figure 1B) occurred during the study period, which corresponded to 65% of the normal climatologic precipitation in this region during the studied months, showing that 2013 also was a dry year. The maximum rainfall of 28.0 mm occurred on April 28, 2013. The month of July had the highest amount of precipitation (76.2 mm), followed by April (58.5 mm) and June (54.7 mm). The soil water storage (SWS) was higher in June and July 2013 and lower between January and March and from September to December 2013. This result corresponded with the rainfall ( Figure 1B).
All of the above-mentioned factors (air temperature, VPD, rainfall and SWS) followed a seasonal cycle. In wettest months (April, May, June, July and August) with lower air temperatures, the VPD was lower and the SWS was higher due to the more frequent rainfall events. The opposite trend occurred from January to March and from September to December. These seasonal meteorological and soil variations have been observed by many authors (Aires et al., 2008;Meirelles et al., 2011), who have studied water and energy fluxes on pastures across many regions around the world.
Regarding LAI, the highest value (2.59 m² m -2 ) was obtained for maize 48 days after planting (July 30, 2013). For the period of intercropping system, the highest LAI occurred on September 13, 2013 (0.63 m² m -2 ). For grass alone, the highest LAI occurred December 12, 2013 and was 0.68 m² m -2 (Table  1). Gondim et al. (2015) studied LAI in grass in the same region and found similar results.
Seasonal variations of energy balance components, i.e., net radiation (Rn), latent heat (LE), sensible heat (H) and soil heat (G) fluxes are shown in Figure 2. The data in Figure 2 represent daily averaged values.
The Rn showed greater variation and lower values in July due to the higher rainfall and greater cloud cover. The highest Rn values occurred from January to March and from September to December. The daily average Rn value was 137 W m -2 . Aires et al. (2008) observed the same pattern of seasonality in Rn when measuring energy fluxes in a pasture comprised of C3/C4 plants in southeastern Portugal. Regarding G, the daily values varied from -11.6 to 9.7 W m -2 (-1.0 to 0.84 MJ m -2 d -1 ) with an average value of 0.18 W m -2 (0.02 MJ m -2 d -1 ). Gondim et al. (2015) determined the energy fluxes of pasture (Brachiaria decumbens L.) and observed a mean G value of 0.61 MJ m -2 d -1 .
The variations of LE and H can be due to the variations of rainfall and SWS ( Figure 1B) because LE was greater than H during periods of greater water availability (higher rainfall and SWS). The opposite trend occurred at the start (January to March) and end (September to December) of the experiment when the amount of rainfall and SWS decreased. A.  Higher values of LE correspond with more energy being directed to the process of soil evaporation and plant transpiration (ET). In contrast, higher H correspond to the use of more energy for heating air. Therefore, from January to March and from September to December, most of the available energy in the maize-grass intercropping system was used to heat the air (H). This result can be explained by the decreased rainfall and SWS at the beginning of the dry season ( Figure  1B). Krishnan et al. (2012) also observed higher H values in grasslands during the dry season.

Soil water storage (mm)
Actual (ET) and reference (ETo) evapotranspiration in maize-grass intercropping during the period from January to December 2013 are represented in Figure 3.
The ETo varied from 1.6 to 7.0 mm d -1 , with a mean of 4.0 mm d -1 . The maximum ETo (above 3.5 mm d -1 ) values occurred from January to March and from September to December, most likely due to the higher air temperature, VPD ( Figure 1A) and Rn ( Figure 2). However, in the late autumn and winter, due to the low atmospheric demand, the minimum values of ETo (1.5 to 3.0 mm d -1 ) occurred. These results are according to Gondim et al. (2015), who measured the ETo values for same region of this research.
The ET varied from 0.1 to 4.0 mm d -1 with an average value of 1.2 mm d -1 . The maximum ET values (3.0 to 4.0 mm d -1 ) occurred during the periods of higher SWS (June, July and August). Meirelles et al. (2011) measured the ET of Brachiaria brizantha using the same methodology of this research (eddy covariance). These authors found that the ET varied from 1.6 to 4.3 mm d -1 , with an average value of 2.6 mm d -1 . In addition, these authors found that the maximum ET values occurred during the periods of higher soil moisture. Krishnan et al. (2012) studied two grassland areas in a semiarid climate over four years and observed ET values of 2.8 to 3.6 mm d -1 .
The seasonal variation of CO 2 fluxes and SWS are shown in Figure 4.
In the period from January 1 to June 12 2013, when the area was not cropped as well as in the period when maize emerged (from June 13 to 27, 2013), the major part of the CO 2 fluxes was positive, with an average of 0.37 µmol m -2 s -1 . Thus, the area served as a carbon source, emitting 0.69 t C ha -1 in a period of 179 days (January 1 to June 27, 2013).
At the beginning of the intercropping, when the SWS was low, maize was in its early growth stage with a low LAI, and the signal grass had not yet emerged (Table 1), the CO 2 fluxes were positive, reaching values of 0.58 µmol m -2 s -1 .However, when maize had a greater LAI, the signal grass had emerged and the SWS was greater than 40 mm, the CO 2 fluxes were lower. The period from July 26 to September 14, 2013, had the lowest values of CO 2 (negative values), average of -3.08 µmol m -2 s -1 , indicating that the intercropping system acted as a carbon sink. In this same period, 1.63 t C ha -1 was sequestered, which was equivalent to 31.9 kg C ha -1 d -1 . After the harvest of maize (from September15 to December 31, 2013), period of signal grass alone, the values of CO 2 showed a tendency of decrease, but still remained negative, with average of -1.33 µmol m -2 s -1 , which corresponds to 13.8 kg C ha -1 d -1 .
During the experimental period, the average value of CO 2 fluxes was -0.78 µmol m -2 s -1 , which corresponded to a carbon sequestration of 2.9 t C ha -1 and 8.0 kg C ha -1 d -1 . Several studies (Randow et al., 2004;Wolf et al., 2011;Zhang et al., 2014) have shown that the CO 2 flux values in pastures tend to be negative (i.e., pastures usually act as a sink for CO 2 ).
The results obtained in this study agree with those obtained by Randow et al. (2004), who measured CO 2 fluxes in pastures and forests in the Amazon region and found an average day time CO 2 flux of -11.2 µmol m -2 s -1 in the pastures during the wet season and -7.6 µmol m -2 s -1 during the dry season. On the other hand, Wolf et al. (2011), who measured the CO 2 fluxes in pastures in Panama, found values ranging from 5.6 (dry season) to -24.5 (wet season) µmol m -2 s -1 . Despite the great difference of climate between Amazon and Northeast regions, the data of our study showed results similar to those of Randow et al. (2004) and Wolf et al. (2011), once that CO 2 fluxes were also higher in the wet season compared with the dry season.
This behavior of CO 2 fluxes can be explained by the higher SWS ( Figures 1B and 4) and LAI for the signal grass and maize (Table 1), which probably promoted higher water absorption, increasing the LE and ET (Figures 2 and 3), due to increased photosynthesis. Consequently, the increased photosynthesis resulted in greater carbon absorption from the atmosphere (Figure 4). Several authors have observed results similar to those our research. Merbold et al. (2009) compared ecosystem fluxes across a range of vegetation types and climate zones of Africa and found that maximum carbon assimilation rates were highly correlated with mean annual rainfall. On the other hand, Otieno et al. (2010) observed a strong correlation between the CO 2 fluxes and soil water content and concluded that the savanna grassland is vulnerable to low soil moisture, with significant reduction in CO 2 uptake during drought, as observed in this study.
The variation between the crops regarding their roles as a carbon source or sink will depend on several factors related to the weather (higher or lower rainfall, solar radiation, air temperature, etc.), the type of plant (C3 or C4 metabolism), pasture condition, soil water storage, among others. For example, Zhang et al. (2014) showed, in a study of 11 years in temperate grasslands in China, that carbon sequestration occurred, but the ability of the grassland to sequester carbon depended on the type of the pasture and the environmental conditions. In addition, an extreme climate event, such as drought, can significantly reduce carbon sequestration, as observed in periods of lower rainfall and soil water storage ( Figures 1B and 4) in this research.

Conclusions
1. The daily and seasonal variations of ET and CO 2 fluxes were mainly controlled by rainfall, through the changes in soil water content and consequently in soil water storage.
2. There were strong seasonal fluctuations in CO 2 fluxes of maize-grass intercropping, with significant reduction in CO 2 uptake during periods of low rainfall and soil water storage.
3. In a drought year, the intercropping sequestered 2.9 t C ha -1 , which was equivalent to 8.0 kg C ha -1 d -1 .