Optimizing alfalfa productivity and persistence versus greenhouse gases fluxes in a continental arid region

Study site and alfalfa field

The study was carried out in Grassland Agricultural Trial Station of Lanzhou University (latitude 39°15′N, longitude 100°02′E), Gansu Province, China. The field (≈ 280 ha) used in this study is 1,390 m above the sea level and classified as a typical salinized meadow with temperate continental arid climate in the Northwest inland regions (Zhu et al., 1997). The mean annual precipitation is about 123 mm with ≥ 65% occurring during the growing seasons from April to October (Kobayashi et al., 2018). Irrigation is necessary and usually applied bimonthly during the growing seasons with a rate of 120 mm respectively in April, June or August in the study site. The annual mean air temperature is 7.6 °C (from −28 °C between December and February to 38° C between June and Mid-August). The soil pH value is about 8.0, and the soil at the study site is classified as Aquisalids according to USDA soil taxonomy (Zhu et al., 1997).

To optimize the alfalfa stand ages in relation to biomass, soil properties and GHG effluxes, we used a Randomized Complete Block design in 2014. A long-term established forage study with differing stand ages was used for this experiment, and data collected for this experiment took place over one year (Zhu et al., 1997; Kobayashi et al., 2018). There were three blocks (about 2.3 ha for each block) and each block was evenly divided into four subblocks, four stand age treatments (i.e., 2, 3, 5 and 7 years old, sown in late August 2012, 2011, 2009 and 2007, respectively) were randomly assigned into each subblock. For each subblock, three sampling plots (30 m width and of 100 m length) were randomly set up for forage harvest, and soil and GHG sampling.

Alfalfa biomass and soil property

To determine alfalfa productivity of different stage ages, one quadrat (1 m × 1 m) in each sampling plot was randomly selected and the hay yield was measured by cutting above-ground biomass during early blooming periods (10 June, 20 July and 01 October). To measure the under-ground root biomass (RBM), another quadrat of the same size in each sampling plot was randomly selected, RBM was collected by digging 30 cm depth after gas collection (see next section for details). The harvested materials were oven-dried at 60 ° C for 48 h, and then weighted.

To determine the soil characteristics in relation to field stage age, we also randomly selected two sampling sites in each sampling plot, and soil samples were collected at a 0–10 cm depth using the bucket auger (five cm diameter) after gas collection (see next section for details). Soil samples were naturally dried then extracted by passing through a 0.25-mm sieve. Soil organic carbon (SOC) was measured by Chromic acid REDOX titration (Nelson & Sommers, 1996). Soil total nitrogen (STN) was determined following the methods of Bremner & Mulvaney (1982). Meanwhile, two cores (8.4 cm diameter × 6 cm length) were sampled by inserting soil profile of 0–10 cm depth in each quadrat and cores were dried at 105 ° C for 48 h. The soil water content (SWC) was then estimated as: (original wet weight - soil dry weight)/soil volume.

GHG efflux

GHG effluxes from soils are more likely to occur in spring, summer and autumn than in winter (Liu, Wang & Xu, 2010), thus GHG samplings were only carried out during the growing seasons of April, June, July, August and October in 2014. Two sampling sites were randomly selected in each sampling plot on 13 April. Gases were sampled four times (i.e., 5:00, 10:00, 14:00 and 18:00) for three successively sunny days in each mid-month, after removing the above-ground plant and litter (Liu et al., 2017). The mean GHG fluxes during the three successive days were treated as the average daily fluxes for that month.

Gas was collected using a static opaque chamber (30 cm ×30 cm ×30 cm) (Liu et al., 2017). For each sampling event, four gas samples were taken within 30 min at a time interval of 10 min (i.e., 0, 10, 20 and 30 min). The chamber was also equipped with an electronic thermometer. The air temperature inside the chamber was recorded during gas sampling and applied to calculate gas flux (see below). Soil temperature (ST) was also measured by a mercury thermometer inserted five cm into the soil at the sampling site before and after gas sampling and the mean temperature of the two measurements was applied to detect its effect on GHG effluxes.

Gas concentration was measured within 24 h, i.e., CH₄ and CO₂ were simultaneously analyzed by a CH₄/CO₂ Spektrum Analyser with syringe injection (Model No. 908-0011-0001, Los Gatos Research, USA), and N₂O was analyzed by a N₂O Spektrum Analyser (Model No. 908-0015-0000, Los Gatos Research, USA). According to Liu et al. (2017), the daily GHG fluxes were estimated as: GHG_daily = (a × flux_7:00 + b ×flux_12:00 + c × flux_16:00+ d × flux_18:00), where a, b, c and d are the constant gas flux duration (i.e., a = 11 h from 20:00 to 7:00, b = 5 h from 7:00 to 12:00, c = 4 h from 12:00 to 16:00, and d = 4 h from 16:00 and 20:00). The hourly GHG fluxes were thus estimated as: GHG_hourly = GHG_daily/e, where e = 24 (number of hours per day); and the monthly GHG fluxes were then calculated as: GHG_monthly = GHG_daily ×f, where f = 30 or 31 (number of days per month between April and October 2014). The total gas flux during growing seasons was the sum of monthly fluxes (from April to October). Gas fluxes in May and September were not measured and thus gap-filled using linear interpolation of the arithmetical means of gas fluxes for the two close months (Chen et al., 2013).

The flux of GHG describes the change of gas in unit time in the sampling box. Generally, a positive value indicates gas effluxes, and a negative value suggests gas absorption. The specific formula is (Liu et al., 2017): ${\displaystyle F=\rho \frac{V}{A}\cdot \frac{P}{P_0}\cdot \frac{T_0}{T}\cdot \frac{dCt}{dt}}$ where F is the gas flux (kg/m²/h), ρ is the gas density (kg/m³) under standard conditions (ρ_CO2 = 1.965 kg/m³, ρ_CH4 = 0.715kg/m³ and ρ_N2O = 1.965 kg/m³ respectively for CO₂, CH₄ and N₂O), V is chamber volume (m³), A is the base area of the chamber (m²), P is the atmospheric pressure (kPa) of the sampling sites (approximately 85.48 kPa at 1,390 m above sea level), P₀ is atmospheric pressure under standard conditions (101.325 kPa), T₀ is the temperature under standard conditions (273.15 K), T is the temperature (K) inside the chamber, and dC_t/dt is the average rate of concentration change with time (ppm min⁻¹).

The total GHG efflux is estimated as the global warming potential (GWP) for a 100-year time horizon, CO₂-eq. One GWP of CH₄ accounts for 25CO₂-eq and one GWP of N₂O for 298 CO₂-eq (IPCC, 2006). Water use efficiency (WUE) was calculated according to Sun et al. (2018): WUE = hay yield/(irrigation + precipitation + Δ SWC_{October−April}). The precipitation and irrigation from April to October 2014 was 70 and 360 mm, respectively. GHG efflux intensity measuring the ratio of GHG effluxes per unit hay yield (GEI_hay) was also estimated according to Dyer et al. (2010): GEI_hay (kg CO₂-eq/kg hay) = GHG efflux/hay yield.

Statistical analysis

All other statistical analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). Results of a Shapiro–Wilk test (UNIVARIATE Procedure) indicated that data collected from this study were normally distributed. The difference in hay yield and WUE, and total GHG efflux (CO₂-eq) and GEI_hay between different stand ages were analyzed using least significant difference test (LSD test, GLM Procedure). The correlations of soil properties (i.e., ST, SWC, SOC, STN) and RBM to GHG effluxes were determined (CORR Procedure). The variations of soil properties and RBM and hourly GHG fluxes in response to alfalfa stand age (y, year) and seasonal progress (m, month) were analyzed using a general linear model (GLM Procedure): variation = a + b × m + c × m² + d × y + e × y² + f × m × y, where a is intercept, and b, c, d, e and f are estimated regression coefficients. The significant coefficients were only included in the final model. A stepwise multiple regression analysis was applied to determine the possible effects of soil properties and RBM on CH₄, CO₂ and N₂O fluxes (GLM Procedure) and the significant factors were only included in the final model. The proportional contributions of soil properties and RBM to CH₄, CO₂ and N₂O fluxes were then calculated as: the sum of squares for each test factor, divided by the total sum of squares then multiplied by the regression coefficient (i.e., R²) of the model.

Alfalfa biomass, GHG fluxes and soil property in relation to alfalfa stand age

Both total annual hay yield and WUE significantly increased with the stand age from 2 to 5 years then significantly decreased after which time (LSD = 0.50 and 1.15 respectively for hay yield and WUE, P < 0.0001) (Fig. 1). The first cutting respectively accounted for 56.1, 55.9, 55.2 and 59.5% of total annual forage yield respectively from the 2-, 3-, 5- and 7-year-old fields, which was significantly greater than that of the second or third cutting (P < 0.05).

The GEI_hay was significantly lower in 3- and 5-year-old fields than in 2- and 7-year-old ones (LSD = 0.12, P < 0.0001) (Fig. 2A), and a significantly higher annual GHG efflux was detected in 5-year-old fields (LSD = 1.58, P = 0.0030) (Fig. 2B).

Alfalfa stand age had no significant effect on ST and SWC (F_1,56 = 1.70 and 3.62 respectively for ST and SWC, P > 0.05) (Figs. 3A–3B). However, ST significantly increased from mid-spring (April) until summer (July) (F_1,57 = 322.89, P < 0.0001) and then significantly decreased after July (F_1,57 = 358.83, P < 0.0001) (Fig. 3A); while a reverse seasonal pattern was detected for SWC, i.e., it significantly decreased until July (F_1,57 = 322.89, P < 0.0001) then significantly increased (F_1,57 = 358.83, P < 0.0001) (Fig. 3B).

Both SOC and STN increased and peaked in 5-year-old fields (F_1,55 = 5.28 and 13.76 respectively for SOC and STN, P < 0.05) then significantly decreased after which year (F_1,55 = 5.75 and 14.08 respectively for SOC and STN, P < 0.05) (Figs. 3C–3D). However, SOC significantly increased with seasonal progress and peaked in July (F_1,55 = 13.09, P = 0.0006) after which month it significantly decreased (F_1,55 = 13.61, P = 0.0005) (Fig. 3C); but a reverse seasonal pattern was detected for STN, i.e., it significantly decreased from April to July (F_1,55 = 7.78, P = 0.0073) then significantly increased (F_1,55 = 9.35, P = 0.0034) (Fig. 3D).

Both stand age and growing season initially promoted the RBM (F_1,54 = 268.71 and 29.96 respectively for stand age and month, P < 0.0001) (Fig. 3E). But the RBM started to decline after August (F_1,54 = 269.42, P < 0.0001), and the decrease of RBM became slow (F_1,54 = 24.00, P < 0.0001) due to significant positive interaction between stand age and seasonal progress (F_1,54 = 23.79, P < 0.0001) (Fig. 3E).

The dynamics of CH₄, CO₂ and N₂O fluxes also largely depended on alfalfa stand age and season (Fig. 4). CH₄ uptake was detected in the present study and it significantly increased when alfalfa aged up to 5 years old (F_1,54 = 36.24, P < 0.0001) then significantly decreased (F_1,54 = 42.06, P < 0.0001) (Fig. 4A). While CH₄ uptake significantly decreased from April to July (F_1,54 = 149.67, P < 0.0001) and then significantly increased after July (F_1,54 = 149.05, P < 0.0001) (Fig. 4A). The CH₄uptake was generally higher (−15.4 to −25.0 µg/m²/h) in 5-year-old fields for a given month.

The seasonal and annual dynamics of CO₂ and N₂O effluxes were similar, i.e., the effluxes significantly increased when alfalfa aged to 5 years old (F_1,55 = 15.62 and 15.35 for CO₂ and N₂O respectively, (P < 0.001) and the significantly decreased F_1,55 = 13.03 and 13.32 for CO₂ and N₂O respectively, P < 0.0001); similarly the effluxes significantly increased since April (F_1,55 = 322.37 and 195.10 for CO₂ and N₂O respectively, P < 0.0001) and then significantly decreased after July (F_1,55 = 363.12 and 200.82 for CO₂ and N₂O respectively, P < 0.001) (Figs. 4B–4C). The greatest effluxes of CO₂ (i.e., 551.1 mg/m²/h) and N₂O (8.0 µg/m²/h) were estimated in 5-year-old fields in July.

GHG efflux in relation to soil property and root biomass

The CH₄ uptake significantly decreased with increasing ST and SOC but increased with increasing SWC and STN (Table 1). While CO₂ and N₂O effluxes significantly increased with increasing ST and SOC but decreased with increasing SWC and STN (Table 1). RBM had no significant effect on CH₄, whereas CO₂ and N₂O effluxes significantly increased with the increase of RBM (Table 1).

When both soil property and RBM were considered, SWC was the only factor that significantly affected CH₄ fluxes (Table 2). While three factors (i.e., SWC, ST and SOC) significantly affected CO₂ effluxes, and four factors (i.e., SWC, ST, SOC and RMB) significantly affected N₂O effluxes (Table 2). SWC accounted for ≥ 65% variation of CO₂ and N₂O effluxes. ST explained about 15% variation of CO₂ effluxes, which was 4.3 times less than did SWC but 3.5 times more than did SOC. For N₂O effluxes, ST only accounted for only <5% variation, which was 7.3 and 2.1 times less than did SWC and SOC, respectively, and RBM accounted for only <2% of variation (Table 2).