Modelling the effects of pasture renewal on the carbon balance of grazed pastures

Modelling the effects of pasture renewal on the carbon balance of grazed pastures

Science of the Total Environment 715 (2020) 136917 Contents lists available at ScienceDirect Science of the Total Environment journal homepage: www...

1MB Sizes 0 Downloads 28 Views

Science of the Total Environment 715 (2020) 136917

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Modelling the effects of pasture renewal on the carbon balance of grazed pastures Lìyǐn L. Liáng a,⁎, Miko U.F. Kirschbaum a, Donna L. Giltrap a, Aaron M. Wall b, David I. Campbell b a b

Manaaki Whenua − Landcare Research, Private Bag 11052, Palmerston North 4442, New Zealand School of Science and Environmental Research Institute, The University of Waikato, Private Bag 3105, Hamilton 3240, New Zealand

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Pasture renewal causes relatively greater reduction in grazing C removal than GPP. • Pasture could gain soil C but lose productivity if renewal is too frequent. • Optimal renewal frequency depends on deterioration rate of pasture growth. • Length of fallow period affects ecosystem C balance temporarily but not annually. • Renewal timing affects seasonal ecosystem C balance but not at annual scale.

a r t i c l e

i n f o

Article history: Received 8 November 2019 Received in revised form 23 January 2020 Accepted 23 January 2020 Available online 24 January 2020 Editor: Ouyang Wei Keywords: CenW Eddy covariance Grazing NPP Pasture renewal Soil carbon

a b s t r a c t In New Zealand, pasture renewal is a routine management method for maintaining pasture productivity. However, knowledge of the renewal effects on soil organic carbon (SOC) stocks is still limited. Here we use a process-based model, CenW, to comprehensively assess the effects of pasture renewal on the carbon balance of a temperate pasture in the Waikato region of New Zealand. We investigated the effects of renewal frequency, length of fallow period, renewal timing, and the importance and quantification of age-related reductions in productivity. Our results suggest that SOC change depends on the combined effects of renewal on gross primary productivity (GPP), autotrophic and heterotrophic respiration, carbon removal by grazing and carbon allocation to roots. Pasture renewal reduces grazing removal proportionately more than GPP because newly established plants need to allocate more carbon to re-build their root system following renewal which limits foliage production. That lengthens the time before above-ground biomass has grown sufficiently to be grazed again. New plants have a lower ratio of autotrophic respiration to GPP, however, which partly compensates for the GPP loss during renewal. Our simulations suggested an average SOC loss of 0.16 tC ha−1 yr−1 if pastures were renewed every 25 years, but could gain an average of 0.3 tC ha−1 yr−1 if pastures were renewed every year. For maximizing pasture production, the optimal renewal frequency depends on the rate of pasture deterioration with more rapid deterioration rates favouring more frequent renewal. Additionally, the length of the fallow period, renewal timing, and associated environmental conditions are important factors that can affect SOC temporally, but the

⁎ Corresponding author. E-mail address: [email protected] (L.ǐL. Liáng).

https://doi.org/10.1016/j.scitotenv.2020.136917 0048-9697/© 2020 Elsevier B.V. All rights reserved.

2

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

importance of those effects diminishes at the annual or longer time scales. A major uncertainty for a full understanding of the renewal effect on SOC lies in the rate of pasture deterioration with time since previous renewal. © 2020 Elsevier B.V. All rights reserved.

1. Introduction Grazed pastures are one of the most intensively managed ecosystems in the world, producing foliage for livestock and supplying livestock products that derive from grazing animals for human consumption. In New Zealand about 20% (2.6 out of 12.1 million ha) of agricultural land is used as pasture for dairy farming (Statistics New Zealand, 2013). Grazed pasture is a key component of the national economy because of its considerable profitability. Dairy farming contributed 13.6 billion NZ dollars to New Zealand's export revenue in 2016 (NZIER, 2017). However, the growth of the dairy industry also affects the environment in many ways. The number of dairy cows in New Zealand has increased by 90% since 1990, which is one of the main drivers for the increase of greenhouse gas emissions from the agriculture sector (Ministry for the Environment, 2019). For example, since 1990, methane (CH4) emissions from enteric fermentation and nitrous oxide (N2O) emissions from dung and urine deposited by dairy cattle have increased by 129% and 112%, respectively (Ministry for the Environment, 2019). Maintaining the production of pastures while minimizing their environmental impact is important for developing sustainable agriculture. In addition to contributing to greenhouse gas emissions of enteric CH4 and N2O from soils, dairy farming could also change soil carbon stocks. In New Zealand, some soils under long-term dairy pasture on flat land have lost soil organic carbon (SOC) over the last three to four decades, ranging from 'no loss' to a loss of 2.9 tC ha−1 yr−1 across different soil types, including allophanic, gley, organic and other soils (Schipper et al., 2017). However, soils in the hill country used for dry stock grazing (beef cattle, sheep, and deer for meat production) have gained about 0.6 tC ha−1 yr−1 (Schipper et al., 2017). The changes in SOC of grazed pastures could be partly due to different management practices. There are a range of observations that implicate the management intensification under dairy farming for the SOC loss (Schipper et al., 2010) or SOC gain (Parfitt et al., 2014). However, it is not understood precisely which management mechanisms caused the apparent SOC changes. In New Zealand, pasture renewal (pasture renovation) has now become a routine practice for farmers because of its direct economic benefits (Glassey et al., 2010; Kerr et al., 2015). Pasture renewal is the killing of an existing and established pasture sward and its replacement with a new sown one. Nationally in New Zealand, the annual rate of renewal of dairy pastures is estimated to have increased from 6% in 2009 to 8% in 2013 (Kerr et al., 2015). Its purpose is to maintain high grass productivity and to halt progressive pasture deterioration that results in declining pasture productivity, which in turn affects the economic return of farming. On the one hand, pasture renewal increases pasture productivity and foliage quality, resulting in higher milk solid production and thus improved economic returns (Glassey et al., 2010; Tozer et al., 2015). On the other hand, there are concerns that pasture renewal might affect the carbon dynamics in soils and result in soil carbon loss (Ammann et al., 2013; Rutledge et al., 2015, 2017a). Pasture renewal is a disturbance to the existing pasture, and the frequency of this disturbance will inevitably affect the carbon cycle. However, it is not well understood by what mechanisms pasture renewal affects the carbon balance of grazed pastures. Conceptually, pasture renewal has a number of effects that change the carbon balance. First, there is carbon loss resulting from a lack of photosynthesis in the period between the death of an old pasture and the establishment of a new one. The amount of carbon loss depends on the length of time the paddock is fallow

(Ammann et al., 2013; Rutledge et al., 2015, 2017a). Soil carbon losses via respiration will increase with increasing length of fallow periods, especially under conditions of higher soil temperature and moisture, although the faster growth rate of a new pasture under the same conditions may also shorten the time before a pasture can regain its carbon fixation proficiency. Second, the quantity of dead roots and leaves generated during the renewal phase (e.g. spraying and/or tillage) may contribute to soil organic matter and provide a large carbon input to the pool of soil organic matter. Third, if renewal increases the carbon fixation potential of the pasture so that the newly established sward fixes more carbon than the old sward, it may provide more carbon inputs to the soil. It has long been recognized that the net primary productivity (NPP) of ecosystems, which is the difference between gross primary productivity (GPP) and autotrophic respiration (Ra), increases during the initial growing period so that more carbon can be retained by the system than at the mature stage (Drake et al., 2011; Odum, 1969; Tang et al., 2014). This results in a higher NPP to GPP ratio (NPP: GPP) or a lower Ra:GPP ratio in the initial growth stage, suggesting that more carbon could be retained by the system following pasture renewal. Fourth, if pasture renewal involves tillage, it may stimulate organic matter decomposition and result in reduced soil carbon stocks (Conant et al., 2007). This effect will be minimized if tillage is minimized, e.g. via herbicide application followed by direct drilling rather than full cultivation (Rutledge et al., 2017a). These different management factors have important implications for soil carbon stocks, plant photosynthesis, and pasture production in different ways. Understanding the effects of pasture renewal on pasture production and soil carbon stocks is therefore important for maintaining sustainable land use for dairy farming, but it is a challenge to track long-term effects experimentally. Only limited studies have explored the renewal effect on carbon stock experimentally. Carolan and Fornara (2016) selected 45 permanent grasslands with renewal history across Northern Ireland and found no significant effect of pasture renewal on SOC. Similar conclusions were reached by Fornara et al. (2020) and Gál et al. (2007), suggesting no renewal effects on soil carbon stocks. However, systematic quantification of the renewal effects under different management scenarios, like renewal frequency, length of fallow period, renewal timing, etc. on soil carbon stocks is almost impossible through experiments. In this case, model simulations provide a useful alternative to gain insights into system responses to a range of external perturbations (Kirschbaum et al., 2015). In this study, we used CenW, a model that simulates the carbon, energy, nutrient and water cycles in the ecosystem (Kirschbaum, 1999a) to investigate the effects of pasture renewal on pasture production and soil carbon stocks of a temperate grazed pasture in the Waikato region, New Zealand. We aimed to: 1) test the effects of pasture renewal frequency and age-related reductions in productivity on plant photosynthesis, grazing removal and soil carbon; 2) investigate the effects of the length of fallow period during renewal; and 3) examine the effects of renewal timing and the associated environmental effects (temperature and soil moisture). Results from this study will improve our understanding of renewal effects on carbon balances in grazed pasture and could provide useful information for determining optimal renewal frequencies that maximize pasture production while minimizing its impacts on the carbon balance.

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

3

2. Materials and methods

2.3. Modelling details

2.1. Site information and farm management

2.3.1. Description of CenW CenW (carbon, energy, nutrients and water) is a process-based model that combines the fluxes of the four components to simulate the daily carbon balance of an ecosystem (see details of CenW in Kirschbaum, 1999a). It has been applied to model the carbon balance of forests (Kirschbaum, 1999a, 1999b) and has more recently been adapted to grazed grasslands (Kirschbaum et al., 2015). The climate inputs of CenW were daily: 1) temperature (minimum, maximum and average air temperature and average soil temperature); 2) vapour pressure of the air; 3) global solar radiation; and 4) rainfall. The model also requires soil information, such as water-holding capacity, and a measure of soil texture. For a managed system like grazed pasture it is also essential to include details for any management interventions, such as grazing, fertilizer application, feed supplementation, harvesting, and pasture renewal, including seeding. In this study we integrated the 30 min EC measurements to daily time steps. The corresponding climate measurements with 30 min intervals were also integrated daily to provide inputs for CenW. The measured/gap-filled NEP and partitioned estimates of GPP and ER were used to parameterize CenW. Before conducting our scenario analyses, we parameterized CenW by comparison against eddy-covariancemeasured NEP and derived GPP and ER for 5.5 years. An optimization routine was used to adjust selected parameters to minimize the overall residual sums of squares (RSS) between model outputs and daily integrated EC-measured NEP and derived GPP and ER. The Nash–Sutcliffe coefficient of model efficiency (NSE) was used to determine the goodness of fit between model outputs and measurements (Nash and Sutcliffe, 1970). Using these optimized parameters, we conducted a series of simulations to investigate the effect of pasture renewal on the pasture carbon balance over 50 years. A climate data series was constructed by simply replicating the observed climate data between 2013 and 2018 to make up a 50-year sequence.

The study site was situated on Troughton Farm, a commercial farm that has been used for dairy farming for over 80 years. Troughton Farm is located near Waharoa in the Waikato region, in the North Island of New Zealand, at 37.78°S, 175.80°E, at an elevation of 54 m a.s.l. The climate of the site is warm temperate, with mean annual temperature and precipitation of 13.3 °C and 1250 mm, respectively, according to records collected between 1981 and 2010 (NIWA, 2015). It normally has a warm, dry period between November and March, and a cool, wet period between April and October. All precipitation is rainfall, with light frosts in winter. The dominant soil type of the experimental paddock is Te Puninga silt loam soil (Mottled Orthic Allophanic soil (Hewitt, 1998)), with carbon and nitrogen content in the topsoil (0–20 cm) of 7.2% and 0.56%, respectively. The bulk density is 918 kg m−3 with a porosity of 57% and a permanent wilting point at 25% of volumetric water content (m3 m−3 × 100). The pasture sward is dominated by perennial ryegrass (Lolium perenne) and white clover (Trifolium repens). Pasture renewal in the experimental paddock was conducted in April 2013 using direct drilling (more information about the pasture renewal can be found for the NewRye site in Rutledge et al., 2017b). Troughton Farm has 67 paddocks ranging from 2.5 to 3.5 ha. Paddocks are rotationally grazed by Jersey cows year-round at a stocking rate of 2.9 cows ha−1. The grazing frequency is higher in the spring– summer period because of rapid pasture growth under warmer conditions than in winter when pasture growth is relatively slow. On average, each paddock is grazed about 12 times per year with an average grazing duration of 21 h. For each grazing event the average grazing removal by cows is about 0.85 tDM ha−1 across the year, based on field measurements. For each year, the total amount of grazing off-take ranges from 10 to 13 tDM ha−1 yr−1. Grass is the main feed for dairy cows, but it is supplemented with about 0.8 tDM ha−1 yr−1 grass and maize silage when grass growth is insufficient to meet the animals' nutritional needs. Commercial fertilizer is applied as required.

2.2. Measurements Prior to pasture renewal, an eddy covariance (EC) system was established to monitor the net CO2 ecosystem exchange (NEE), which is equivalent (but opposite in sign) to net ecosystem production (NEP). We will use the term NEP in the following text. Pasture was renewed in April 2013 when the old pasture was killed with glyphosate herbicide before establishing a new sward using direct drilling. Continuous observations were made before, during and after the renewal events to monitor the climatic conditions and changes of energy, CO2, and water fluxes. NEP was measured at 30 min intervals by the EC system from August 2012 to February 2017. The EC set-up, data processing approaches, and associated uncertainty analyses are described in detail by Rutledge et al. (2017a). Briefly, the CO2 concentration and wind speed were measured at 20 Hz using a LI-7200 infrared gas analyser (LI-COR Inc., Lincoln, Nebraska, USA) and a CSAT3 sonic anemometer (Campbell Scientific Inc. (CSI), Logan, Utah, USA), respectively. The high-frequency data were recorded by a datalogger (CR3000, CSI) and further processed using EddyPro 6.2 (LI-COR Inc.) to calculate NEP at 30 min intervals. A seven-step data quality control procedure was applied to filter the low-quality data (see Rutledge et al., 2017a for more details). Data gaps of NEP were filled using an artificial neural network, and the gap-filled NEP was further partitioned into GPP and ecosystem respiration (ER). The associated environmental variables – including air temperature, relative humidity, soil temperature and volumetric water content, rainfall, solar radiation, and soil heat fluxes – were measured and recorded at 30 min intervals.

2.3.2. Scenarios of renewal frequency and pasture deterioration In the first set of scenarios we investigated the effect of renewal frequency on the carbon balance by running 12 scenarios with renewal intervals of 1, 2, 3, 4, 5, 7, 9, 11, 13, 15, 20 and 25 years, while keeping other factors, such as parameters and management routines, in the system identical. For each renewal event we followed the renewal method described by Rutledge et al. (2017a, 2017b), with a short fallow period of 15 days between spraying and seedling emergence, and direct drilling that disturbed the soil only slightly. The first set of simulations were run without considering pasture deterioration over time (but later simulations in this paper will include varying levels of deterioration in pasture productivity over time). All simulations with pasture renewal were compared against corresponding control scenarios with no renewal events (hereafter called nonrenewal scenario) but fertilizer and grazing events were maintained during the whole simulation period (50 years) to ensure the system started in equilibrium with no change of soil carbon over time. The simulations used two automatic routines: an automatic fertilizer routine that added fertilizers to the pasture when the nitrogen content of plant leaves fell below a given threshold (50% of nitrogen saturation), and an automatic grazing routine that simulated cattle grazing when grass production reached a given threshold (2.5 tDM ha−1). This threshold was based on the measured pasture biomass at which grazing was initiated at the experimental farm between August 2012 and February 2018. The measurements of grazing removal and pasture biomass were also used to determine an average fixed proportion of 40% to calculate the modelled grazing removal in the auto grazing routine. One of the advantages of automatic routines is that they allow dynamic adjustments to the fertilizer and grazing practices to account for weather conditions.

4

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

To investigate the combined effect of grazing frequency and pasture deterioration on carbon balance, we introduced an empirical age function adapted from Landsberg and Waring (1997) and Ryan et al. (1997) to describe pasture deterioration over time. This function takes the form of:

f age ¼

1 1 þ Rage

ð1Þ

n

2.3.3. ER partition and ecosystem carbon balance calculation We partitioned ecosystem respiration (ER) from the EC system into the autotrophic (Ra) and heterotrophic (Rh) components to investigate change in the Ra:GPP ratio of pastures after renewal. We assumed that ER equalled Rh during the renewal phase when Ra was negligible (b2.5 kgC ha−1 day−1, Fig. 2a). We selected the ER from EC measurements and the corresponding estimates of Rh from CenW over a short period after harvest and reseeding (both events happened in the same day) and performed a linear regression between observed ER and modelled Rh (Fig. 2b). Combining the modelled Rh from CenW and this derived relationship, we can determine a surrogate of EC measurements for Rh. The Ra can then be calculated as the difference between ER measurements and Rh estimates (Fig. 2c). We further compared the difference in GPP from EC and Ra estimates between the renewal year (13 April 2013 to 12 April 2014) and the corresponding averages over the following years (2014–2017, non-renewal years). We calculated the 1

0.9

0.8 Production Loss at: 5% 10% 15% 20% 25% 30%

0

NPP ¼ GPP−Ra net ecosystem production (NEP) as NEP ¼ GPP−Ra −Rh ¼ GPP−ER and net ecosystem carbon balance (NECB) as

where fage is the fraction of retained production over time, Rage is the plant age relative to the age at which productivity is halved, and n is the exponent that describes the rate of production loss (Fig. 1). Although the mechanisms behind the observed pasture deterioration are not clear, the main factors suggested have been invasion by lowquality grasses, increasing pest damage, and poor drainage, possibly related to soil compaction or pugging damage by repeated animal trampling (Parsons et al., 2011; Tozer and Edwards, 2011). In our simulations we explored the effects of six potential pasture deterioration rates characterized by production losses of 5, 10, 15, 20, 25 and 30% over 20 years after pasture renewal (Fig. 1), covering possible deterioration levels. The selected percentages of production loss corresponded to n values of 3.2, 2.4, 1.9, 1.5, 1.2 and 0.9 in Eq. (1), respectively.

0.7

net primary productivity (NPP) as

5

10 15 Plant Age [Years]

20

Fig. 1. Six different scenarios of possible pasture deterioration rate over 20 years after pasture renewal. A normalized age function was applied to describe the relative loss of production over time. Each curve is characterized by a different rate of production loss (n) over the first 20 years after pasture renewal.

NECB ¼ GPP−Ra −Rh −Rgrz where Rrzg was the carbon removal from the pasture removal by cows through grazing, including grazer respiration, methane (CH4), and carbon export as milksolids. The proportions of grazing respiration, CH4, and carbon export were set to 45.8%, 3.6%, and 15.5% of pasture removal, respectively. The rest (35.1%) of pasture removal was returned to the pasture as dung and urine (values came from Table S1 in Rutledge et al., 2017b). Other small components of NECB (Rutledge et al., 2017a), e.g., leaching, fertilizer, manure/compost, were not included here. The differences between renewed and non-renewed pasture were also calculated and denoted as ΔGPP, ΔNPP, ΔRa, ΔRh, ΔNEP, ΔNECB, etc. Here we defined a positive difference as the relative carbon gain and a negative difference as a relative carbon loss for the renewed pasture compared to the non-renewed pasture. We also assessed the difference in the carbon balance between non-renewal years and the renewal year over the renewal phase, which we considered to extend to 150 days after reseeding, by which time the difference between measured GPP and ER had become small (Fig. 2d). We also compared the Ra:GPP ratio, Rh, grazing removal (Rgrz), and root growth between renewal and non-renewal years. 2.3.4. Scenarios of lengths of fallow We further investigated the effect of different lengths of fallow period on the carbon balance of the renewed pasture. The length of the fallow period ranged from 15 to 65 days. This was combined with the scenarios that included 12 renewal frequencies and seven plant age NPP deterioration rates. Under each scenario we calculated GPP, ER, NEP, and NECB or SOC changes and compared their differences over 65 days and assessed annual carbon balances. We specifically explored the effect of the length of the fallow period on soil carbon changes with the renewal frequency of 11 years, which is a common practice in New Zealand. 2.3.5. Scenarios of renewal timing and the effects of associated environmental conditions In New Zealand, farmers commonly renew their pasture in autumn, a practice that allows the new sward to establish before winter. At our research site the pasture was renewed on 13 April 2013 and EC measurements indicated that it took about 2 months for the new pasture to return to being a carbon sink (positive NEP) (Rutledge et al., 2017a). In the last set of scenario analyses we further explored the effects of pasture renewal at different months of the year and the effects of environmental factors, e.g., soil temperature and soil moisture, on the carbon balance. We investigated the renewal effect, defined as the difference of NEP (ΔNEP) between renewal and non-renewal scenarios over the fallow period, following Rutledge et al. (2017b), by setting the renewal timing on the 13th of each month (12 months) across the year. We further grouped 12 months into four seasons: spring (September to November), summer (December to February), autumn (March to May) and winter (June to August), and explored the seasonal pattern of ΔNEP. We set the renewal frequency to 11 years with a deterioration status of 20% production loss over 20 years (n = 1.5, Fig. 1) and a fallow period of 30 days. We also explored the environmental effects on ΔNEP by conducting a regression analysis between ΔNEP and soil temperature and soil moisture, respectively.

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

2

ERCenW Rh Ra

60

1

50 40

0

5

1

80 b 1:

ER EC [kgC ha-1 day-1]

70

Ra [kgC ha -1 day-1 ]

EREC

[kgC ha -1 day-1 ]

Respiration Rate

80 a

60 y=1.00x+13.26 r 2 =0.54,p=0.0008

40

0

10

40

Days after Reseeding Rh

Ra

60 40 20 0 0

50 100 Days after Reseeding

8 d 6 [tC ha -1]

EREC

60 Rh [kgC ha

Cumulative Carbon Fluxes

[kgC ha day ]

-1 -1

Respiration Rate

80

c

5

150

-1

80 -1

day ]

GPP EC

EREC

Rh

Ra

4 2 0

0

50 100 Days after Reseeding

150

Fig. 2. Partitioning of ecosystem respiration (ER, derived from EC measurements) into autotrophic (Ra) and heterotrophic (Rh) components. a) ER from EC measurements (hollow circles), modelled ER (black line), Rh (red line) and Ra (blue line, right y-axis scale) from CenW. b) Regression relationship between ER (derived from EC measurements) and Rh estimates from CenW during a short period (12 days) after reseeding when the modelled Ra b 2.5 kgC ha−1 day−1. c) ER and partitioned Ra and Rh from the ER measurements over 150 days after reseeding using the linear relationship from panel b. d) Cumulative GPP, ER, Ra, and Rh over 150 days after reseeding, when GPP and ER are approximately equal. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

2.4. Data analysis Results from scenario simulations were further analysed using Matlab 2018a (The MathWorks Inc., Natick, MA, USA). GPP and the amount of pasture biomass eaten by grazing animals were averaged

Predictions

200 a

GPP NSE = 0.74

b

over the complete renewal cycles based on respective renewal frequencies across 50 years for each scenario. For each scenario the SOC change was calculated as the difference between the initial and final carbon stocks. Linear regression was applied to quantify the relationship between estimated ER and modelled Rh from CenW during the short

ER NSE = 0.75

50

100

150

0

100 50

0

120 GPP [kgC ha -1 day-1]

-50

50 0

NEP NSE = 0.54

100 c

100

200

0

0

50 100 Observations

d

-100 -100

0

100

EC Observations CenW Predictions

100 80 60 40 20 0 20 Mar

04 Apr

19 Apr

04 May

19 May

Fig. 3. The upper three panels (a, b, c) show the comparisons of GPP, ER, and NEP between EC observations (kgC ha−1 day−1) and CenW's simulations after parameterisation for the study site on Troughton Farm for 5.5 years. NSE is the model efficiency for daily values. The lower panel (d) demonstrates GPP dynamics before, during, and after the renewal event in 2013. The red, black, and green arrows indicate timing of herbicide application (4 April), seeds sown (13 April), and seedling emergence (19 April), respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

6

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

period after reseeding and the relationship between ΔNEP and environmental factors. A t-test was applied to test the significance of the Ra:GPP ratio between renewal and non-renewal years. For testing the NEP difference across seasons, we applied a one-way ANOVA and used the Tukey's honestly significant difference (HSD) test for the post hoc analysis.

scenario analyses, we presented a detailed assessment of the differences in carbon balances between renewal (data in 2013) and non-renewal years (data in 2014–2017). We used EC measurements and CenW outputs, including measured cumulative GPP, grazing removal, inferred Ra, Rh and the Ra:GPP ratio, and modelled root production. In the renewed pasture the cumulative GPP over 150 days was initially much smaller than in the non-renewed pasture, but the differences became progressively smaller as the new pasture developed (Fig. 4a). Consequently, cumulative Ra in the renewed pasture was relatively smaller since there were fewer and small plants during the initial grass establishment and growth. Compared to the non-renewed pasture, the renewed pasture had a GPP as much as 2.0 tC ha−1 lower during the first 50 days (Fig. 4b), although this loss was partially compensated for by the associated reduction in Ra, resulting in a maximum net reduction of −0.8 tC ha−1 in NPP. As the new pasture continued to grow while maintaining a lower Ra, the cumulative differences of NPP after 150 days turned into a small carbon gain (0.2 tC ha−1) compared to the nonrenewed pasture (Fig. 4b). This suggested that the renewed pasture was relatively more productive than the non-renewed pasture during the renewal phase. Further investigation revealed that the Ra:GPP ratio in the renewed pasture (0.47 ± 0.15) was significantly lower (p b 0.0001) than the average in the non-renewed pasture (0.56 ± 0.04). This negated the reduction in GPP, leading to a small carbon gain of NPP after 150 days. By incorporating the field measurements of grazing removal into the overall analysis, we found that the renewed pasture actually gained carbon relative to the non-renewed pasture (Fig. 4d), reaching a maximum carbon gain of 0.4 tC ha−1. While GPP was reduced through pasture renewal, that loss was negated by a corresponding reduction in Ra. An additional reduced carbon loss from grazing (largely respiration of consumed C) turned the overall effect into a net carbon gain in the renewed pasture relative to the non-renewed pasture. However, the overall carbon balance also depends on heterotrophic respiration (Rh), with increased Rh loss in the renewed pasture

3. Results and discussion 3.1. CenW performance and assessments of the carbon balances in the renewal phase

0

0

1 d

50 NPP

100 NPP +

150 R

grz

GPP +

R = a

0 -1

0.6 0.4 0.2

-2 -3

0

50 NPP + R

0.5

100 R

150

0

grz

Rgrz +

-0.5

-0.5

-0.5

-1

-1

-1

150

-1.5

0

NPP +

100 R

grz

+

R

150

h

Rh

0

50 100 Days after Reseeding

50

Root Growth Net Carbon Balance

0.5

h

0

0

0

1 f

0

-1.5

Renewed Pasture Non-Renewed Pasture

0.8

NPP

1

NPP +

[tC ha -1]

1 c

GPP Ra

2

1 e

Rgrz

0.5

3 b

Ra :GPP

GPP (Non-Renewed) Ra (Renewed) Ra (Non-Renewed)

5

-5 Differences of Carbon Balance

GPP (Renewed)

Flux Differences [tC ha -1]

10 a

[tC ha-1 ]

Cumulative GPP and Ra

Overall, we obtained good agreements between EC observations and CenW predictions for GPP, ER, and NEP, with model efficiencies of 0.74, 0.75, and 0.54, respectively (Fig. 3a–c). The overall good agreements indicated that CenW sufficiently captured the main processes of carbon cycling at this site. In particular, CenW captured the renewal event and demonstrated the GPP decline after herbicide application and the slow recovery of carbon gain after seedling emergence, as demonstrated in EC observations (Fig. 3d). During renewal, plants were sprayed with a glyphosate-based herbicide on 4 April 2013 (for more details see Rutledge et al., 2017a), but plants continued to photosynthesize unabated for a further 4 days before the herbicide took effect and photosynthesis started to decrease. It took another week before photosynthesis had fallen to near-zero and a new pasture sward was sown. Seedlings emerged 7 days after sowing, giving an effective fallow period of 15 days between spraying and seedling emergence. Photosynthetic recovery was slow after seedling emergence, and 4 weeks after seedling emergence, photosynthesis had still reached only one-quarter of pre-renewal rates (Fig. 3d), although cooler temperatures and lower incoming radiation further reduced rates as the winter season approached. To better understand the overall carbon balances during the renewal phase (i.e. 150 days after reseeding, see Fig. 2d) before conducting the

50 100 Days after Reseeding

150

-1.5

0

50 100 Days after Reseeding

150

Fig. 4. Assessments of differences in flux components between the renewal year (2013) and non-renewal years (2014 to 2017) over 150 days after April 13th of each year. a) Cumulative GPP (green solid line) from EC measurements and partitioned Ra (blue solid line) in the renewal and non-renewal years. In the non-renewed pasture, GPP (green dashed line) and Ra (blue dashed line) were averages over 150 days, starting from 14 April of each year. b) Difference in GPP (green solid line) and Ra (blue solid line) between the renewed and the averages of the non-renewed pasture. c) Difference in Ra:GPP ratio between the renewed (black dashed line) and non-renewed (grey dashed line) pasture. d) Difference in cumulative carbon balance (black solid line) between the renewed and non-renewed pasture, including both grazing removal (grey dashed line) and NPP (green dashed line). e) Difference in carbon balance from (d) (green dashed line) with the difference in heterotrophic respiration (Rh, red dashed line) added (black dashed line). f) Difference in carbon balance (grey solid line) calculated by subtracting the difference in the plant carbon gain (dashed blue line) from the carbon balance from (e). Panels d-f share same y-axis label. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

compared to the non-renewed pasture (Fig. 4e) due to the influx of dead roots to the soil carbon pools when plants were killed. This large addition of dead roots stimulated Rh from the soil and caused a large carbon loss, shifting the overall carbon balance back to a loss (−0.2 tC ha−1) (Fig. 4e). But this enhanced Rh due to changed root pools required enhanced subsequent root growth for the root pool to return to a normal root biomass. This created an effective additional carbon sink of about 0.8 tC ha−1 over the first 150 days after pasture renewal, for a final estimated carbon gain of about 0.6 tC ha−1 over the first 150 days in the renewed pasture relative to the non-renewed pasture (Fig. 4f). It should be noted that at the annual scale, pasture production or grazing removal could be higher in the renewal year relative to the non-renewal years because of the inter-annual differences in environmental conditions. Measurements in the renewal year 2013 showed the greatest pasture production over the 5-year experimental period at the same site (Rutledge et al., 2017a, 2017b), suggesting high inter-annual variations of the carbon balance. 3.2. Effect of renewal frequency on plant photosynthesis (GPP), grazing removal and SOC In this section we describe how we first tested the effects of renewal frequency without applying the pasture deterioration function and fully assessed the differences in carbon balances between the renewed and non-renewed pasture using EC measurements and CenW simulations. We then incorporated a pasture deterioration function into the scenario analyses to investigate the combined effects on carbon balances. For the following simulations, we have repeated the renewal events using the same model parameterization with the scenarios described in Section 2.3. 3.2.1. Without pasture production deterioration As expected, grazing removal (the amount of pasture eaten by cows) increased in the scenarios with lower renewal frequency (Fig. 5a). Pastures with lower renewal frequencies simply had fewer fallow periods,

7

which would have reduced photosynthetic inputs. If there is no pasture deterioration, any pasture renewal reduces GPP, with the extent of reduction decreasing with simulated increased renewal intervals (Fig. 5b). Interestingly, Fig. 5a and b shows that renewal had a proportionately greater effect on grazing removal than GPP. For example, compared to the non-renewal scenario, the grazing removal in the scenario with 1-year renewal frequency was reduced by 1.7 tC ha−1 yr−1 or 21.3%, while GPP was reduced by 1.8 tC ha−1 yr−1 or only 7.0% (Fig. 5a, b). This difference could largely be attributed to the extra carbon requirements to regrow a new root system for the emerging plants, while established plants could continue to use their existing roots. When an undisturbed pasture is grazed, its root system remains intact, and subsequently fixed carbon can be largely allocated to new foliage production, which maximizes the amount that can be grazed at the next grazing event. Allocation to roots can be limited to that needed for the maintenance of the established root system. Following renewal, however, newly established plants need to develop a new root system as well as produce above-ground foliage to enable photosynthesis. Accordingly, new pastures must initially allocate a higher proportion of fixed carbon to their developing root systems, thus limiting early foliage production. The new swards also need to grow the residual amount of biomass after grazing that remains in a non-renewed pasture. Growth of above-ground biomass is, therefore, accumulated more slowly than in an existing sward, thereby lengthening the time before grazing is possible. These simulation results are consistent with the observations made at the experimental site. The recorded interval between the last grazing before renewal (11 April 2013) and the first grazing after renewal (7 June 2013) was 57 days, while the average grazing interval for the same period (from April to June) in the years after the renewal event was 34 days. This implies there was at least one missed grazing event in the renewal year, which meant there was a smaller reduction in carbon gain than the reduction in carbon removal through grazing. In addition to the differences between GPP and grazing removal, the changing ratio between Ra and GPP (Ra:GPP) of the newly grown grass

Fig. 5. Comparison of grazing removal, GPP, and SOC under different renewal frequencies. a) Grazing removal under different renewal frequencies and the relative reduction compared to the non-renewal scenario. b) GPP and its reductions relative to the non-renewal scenario. For both a) and b), absolute values (black circles) are shown on the left axis, while reductions relative to the non-renewal scenario (blue squares) are shown on the right axis as a percentage. c) SOC dynamics under different renewal frequencies over 50 years. The red line shows the SOC dynamics without renewal events and the black arrow indicates decreasing renewal frequency. d) Average annual SOC change under different renewal frequencies. Error bars are standard errors. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

8

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

also contributes to lower carbon losses in the renewed grassland, as described in Fig. 4c. For ecosystem production, it is common to have higher NPP in the young stage relative to the mature stage, resulting from a relatively lower Ra:GPP ratio in the early stage of growth (Kira and Shidei, 1967; Odum, 1969). It is expected that renewed pastures have lower Ra than existing pastures since they have less root and leaf biomass during the renewal phase (150 days after reseeding in our case). On the one hand, renewed pastures fix less carbon (because of reduced photosynthesis during the fallow and early re-growth stage) than non-renewed pastures. On the other hand, renewed pastures lose less carbon through lower Ra and lower grazing off-take (Fig. 4b–d). To fully understand the effects of pasture renewal on the overall carbon balance of pastures, it was essential to provide an overall carbon balance to quantify the differences for all important components between the renewed and nonrenewed pastures, as we did in Fig. 4. Pasture renewal involves a disturbance of the pasture system. There are a number of factors (e.g., cultivation method, crop rotation, climate, and their interactions) that control SOC dynamics under disturbance, and that could cause SOC decreases (Conant et al., 2007; Ogle et al., 2005) or increases (López-Fando and Pardo, 2009; Ogle et al., 2005). In our study, the combined effects of pasture renewal on grazing removal, GPP, and the Ra:GPP ratio led to a net increase of SOC relative to non-renewed pasture, with the greatest increases with the most frequent renewal interval (Fig. 4c, d). In contrast, a previous study showed that the renewed pasture remained a carbon source for the following 2 years after renewal (Ammann et al., 2013). The carbon loss may be due to the long fallow period (6 months) and the enhanced disturbance of the soil by ploughing, while the fallow period in our study was only 15 days and disturbance of the soil was slight using direct drilling (Rutledge et al., 2017a). It should be noted that there was considerable inter-annual variability of SOC (ca. 2 tC ha−1) in line with changing weather conditions in the non-renewal scenario (Fig. 5c), suggesting strong effects on SOC estimates from climatic factors. We have

investigated the effect of fallow period and the environmental conditions (temperature and soil moisture) in our further scenario analyses. 3.2.2. With pasture production deterioration There was a strong interaction between renewal frequency and extent of pasture production deterioration. In general, both GPP and grazing removal decreased with very frequent renewal (Fig. 6a, b), while very long renewal periods also reduced grazing removal and GPP because the deterioration with age was too infrequently reset. This led to maximum grazing removal and GPP gains of our modelled scenarios at intermediate renewal frequencies of around 5–10 years, with optima at more frequent intervals if there was more severe pasture deterioration (Fig. 6a, b). Across the range of pasture production deterioration and renewal frequencies there could be a diverse range of SOC changes (Fig. 6c). The simulations showed SOC gains for very frequent renewal frequencies, as described in previous sections. Without age production deterioration there were then no SOC changes with infrequent renewal. If significant production deterioration was included in the simulations, it led to ongoing SOC losses at the longest renewal frequencies (Fig. 6c). However, the maximum magnitude of the SOC loss was low. For example, with a 30% loss of production and renewal every 25 years, SOC was lost at an average rate of 0.16 tC ha−1 yr−1 over a 50-year simulation period. Net SOC losses occurred when the carbon loss via production deterioration was large enough to counteract the additional carbon retained in the system due to the renewal effects on GPP and grazing removal (Fig. 6a, b). Renewal with high frequency kept the carbon inputs to the soil high and increased SOC (Fig. 6c). For example, renewal every year increased SOC at a rate of about 0.3 tC ha−1 yr−1. This SOC gain was caused by pasture renewal leading to reduced carbon losses associated with grazing, which outweighed the impact of renewal on reduced GPP. This meant that higher renewal frequency led to more carbon remaining within the system. However, the grazing removal was at its

-1

yr ]

b

GPP [tC ha

7

6

25

24

0.2

0% 5% 10% 15%

25

20% 25% 30%

0

0

5 10 15 20 Renewal Intervals [Years]

7

Loss

-0.2 0

5 10 15 20 Renewal Intervals [Years]

25

25

8 d

[tC ha-1 yr-1]

0.4 c

5 10 15 20 Renewal Intervals [Years]

Grazing Removal

0

SOC [tC ha-1 yr-1]

26

-1

-1

yr ]

-1

[tC ha

Grazing Removal

8 a

6 -0.2

Gain

0

SOC [tC ha

0.2 -1

0.4 -1

yr ]

Fig. 6. Simulations showing a) grazing removal, b) GPP, c) rate of SOC change over 50 years under different renewal frequencies and extents of pasture production deterioration, and d) the relationship between grazing removal and SOC change over seven extents of pasture deterioration. All panels share the same legend illustrated on panel c for different pasture deterioration rates.

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

GPP [tC ha-1 65 days-1]

15 f 25

2 -0.

-0.5

35 -1

45 -0.2

55

-0.4

1 3 5 9 13 20 Renewal Frequency [Years]

1 3 5 9 13 20 Renewal Frequency [Years]

-1

65

0.5

-3

65

-0.6

0.2

1

0.4

55

55

-1

1

45 0.6

15 25 35 45 55 65 Length of Fallow Period [Days]

-0.49

1.5

35

1.8

2

25

2 0.2

2.2

15 e 0.4

-0.47

10 20 30 Renewal Frequency [Years]

1

2.4

Rh [tC ha-1 65 days-1 ]

-0.45

Length of Fallow Period [Days]

d

0

0.6

Ra [tC ha-1 65 days-1 ]

2.6

-1.5

15 25 35 45 55 65 Length of Fallow Period [Days]

-2

45

-0.5

-4

-1

35

-1

-3

-1

25

-2 -3.5

RF = 1 year RF = 2 years RF = 11 years RF = 25 years

-0.5

15 c

NEP [tC ha-1 65 days-1]

-2

b

ER [tC ha-1 65 days-1]

-1

0

-1

GPP [tC ha-1 65 days-1]

0 a

-2

GPP [tC ha-1 65 days-1]

During the fallow period of the renewal event there are no carbon inputs from photosynthesis. As expected, cumulative ΔGPP decreases with increasing length of the fallow period (Fig. 7a). In the scenario with the maximum length of fallow period (65 days) and a renewal frequency of 1 year, the cumulative GPP loss (ΔGPP) relative to the nonrenewal scenario reached the maximum of ~4 tC ha−1. The difference in ΔGPP (ΔΔGPP) between the shortest (15 days) and longest (65 days) fallow period was 1.4 tC ha−1 with a 1-year renewal frequency (Fig. 7b), but this decreased greatly with decreasing renewal frequency (Fig. 7b). At a renewal frequency of 25 years the ΔΔGPP was only 0.06 tC ha−1 (Fig. 7b) because the GPP loss in a single renewal event became insignificant over the 25-year renewal cycle. Therefore, renewal frequency and the length of the fallow period together

Length of Fallow Period [Days]

3.3. Effect of the length of fallow period

determine the GPP loss during the fallow period compared to the nonrenewed pasture, showing higher loss under a longer fallow period and higher renewal frequency (Fig. 7c). GPP reductions during the fallow period further affected both Ra and Rh respiration. ΔRa decreased with increasing length of the fallow period because there was no root or leaf biomass to produce CO2, leading to a reduction in cumulative ΔRa over the maximum fallow period (65 days) of about 2.5 tC ha−1. The difference in ΔRa (ΔΔRa) between 15 and 65 days of fallow period was 0.5 tC ha−1. The difference in ΔRh (ΔΔRh) also showed a similar pattern since there were reduced carbon inputs from plant litter for soil respiration during the fallow period. As plants started to regrow, additional carbon inputs increased ΔRh. Shorter fallow periods led to increased ΔRh because more carbon supply came from the regrown grass to enhance Rh, although the differences of ΔRh among different lengths of fallow period were small, ca. 0.03 tC ha−1(Fig. 7d). The length of the fallow period and renewal frequency together determined the difference of total ecosystem respiration (ER) between renewal and the non-renewal scenarios, showing the highest ER difference under shorter fallow period and highest renewal frequency (Fig. 7e). ΔER was highest (1.9 tC ha−1 greater than without renewal) under the longest fallow period (65 days) and highest renewal frequency (1 year). Over 65 days, the renewed pasture system was a carbon source compared to the non-renewal scenario, with negative ΔNEP for all the scenarios (Fig. 7f). ΔER was the largest under the scenario with a 65-day fallow period and 1-year renewal frequency, resulting in it being the largest carbon source (−2.0 tC ha−1). In the scenario with the shortest fallow period and lowest renewal frequency, ΔNEP was close to zero (−0.02 tC ha−1). It is notable that with a lower renewal frequency (longer than every 5 years), regardless of any tested length of fallow period, the carbon loss becomes much smaller, b0.4 tC ha−1 (Fig. 7f), because the carbon losses during the fallow period in a single renewal event were progressively diluted by the increasing number of non-renewal years.

Length of Fallow Period [Days]

minimum under the most frequent renewal (Fig. 6a). Pasture production was maximized at different renewal frequencies depending on the rate of pasture deterioration, with faster pasture deterioration favouring more frequent pasture renewal. In all the deterioration scenarios, grazing removal was optimized at some intermediate renewal frequency while SOC consistently decreased with decreasing renewal frequency (Fig. 6d). An optimal renewal frequency can be found to maximize the grazing removal with minimal SOC loss (Fig. 6d). Averaging all seven deterioration rates gave an optimal renewal frequency of around 10 years, with an average grazing removal of 7.5 tC ha−1 yr−1 while maintaining a neutral SOC. It should be noted that the variation of grazing removal across different deterioration statuses under the same renewal frequency was large, reaching approximately 0.5 tC ha−1 yr−1. Renewing pasture under the optimal renewal frequency for maximizing the pasture production for grazing had only a small effect on SOC changes (Fig. 6d). Balancing the maximum grazing removal while maintaining the initial SOC could be achieved by using the optimal renewal frequency under a given deterioration level.

9

-1.5

65 1 3 5 9 13 20 Renewal Frequency [Years]

Fig. 7. The cumulative differences of GPP, ER, and NEP between renewal and non-renewal scenarios within the maximum fallow period of 65 days. a) The relationship between ΔGPP and the length of fallow period under four renewal frequencies (RF). b) The cumulative ΔGPP difference (ΔΔGPP) between shortest (15 days) and longest fallow period (65 days) during pasture renewal. c) The ΔGPP for different combinations of renewal frequencies and lengths of fallow periods. The contours are labelled with the cumulative ΔGPP between renewal and non-renewal scenarios. d) The difference in autotrophic (ΔΔRa, black circles) or heterotrophic (ΔΔRh, grey squares) respiration with a renewal frequency of 1 year compared to the non-renewal scenario across all lengths of fallow periods. e) and f) Differences of ecosystem respiration (ΔER) and net ecosystem productivity (ΔNEP) between renewal and nonrenewal scenarios under different renewal frequencies and lengths of fallow periods. The contours are labelled with the cumulative ΔER and ΔNEP within 65 days, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

10

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

Annually, the average ΔNEP was negative for all scenarios, with the largest CO2 loss in the scenario with the longest fallow period (Fig. 8a) combined with 1-year renewal frequency. ΔNEP became less negative with lower renewal frequency, until GPP reduction through pasture deterioration became the dominant factor in the CO2 balance at the annual scale, as discussed previously. Adding the grazing effects on pasture renewal, we found the ΔNECB could present a carbon sink or source relative to the control, depending on the combination of renewal frequency and length of fallow period (Fig. 8b). Maximum carbon gain was 0.24 tC ha −1 yr−1 with the shortest fallow period and 1-year renewal frequency. Carbon loss was greatest at 0.09 tC ha−1 yr−1 with the longest fallow period, a 25-year renewal frequency and the severest rate of production deterioration. The effect of the fallow period on carbon balance was prominent during the renewal period, but this effect diminished at the annual scale. Annually, the effects of renewal frequency and pasture deterioration became dominant, determining the change of soil carbon stock in the pasture system, as discussed previously. Specifically, we investigated the length of fallow period (LF) on the carbon balance by setting the renewal frequency (RF) to 11 years, similar to the 6–8% of grazed pasture that are renewed each year in New Zealand. For this practical renewal frequency, the renewed pasture could gain or lose carbon (Fig. 8d), depending on the pasture deterioration function. We noted that the magnitude of gain or loss was small (Fig. 8d), with an absolute amount that was b0.05 tC ha−1 yr−1, suggesting that the renewal effects with 11-year frequency are negligible, even for the worst deterioration function (30% production loss over 20 years). 3.4. Effect of renewal timing

NEP [tC ha

-1

-1

yr ]

NECB [tC ha

15 a

-1

-1

yr ]

b -0.4

0.2

-0.8

45

0.1

tC ha-1 yr-1

35

tC ha-1 yr-1

25

0

55

-1.2

65

-0.1

1 3 5 9 13 20 Renewal Frequency [Years] 6c

RF = 1 year, LF = 65 days

50 d

3 0 GPP R

Grz NECB

a

-3 -6 Apr

1 3 5 9 13 20 Renewal Frequency [Years]

Rh

Jul

Oct Date

Jan

Apr

SOC [kgC ha -1]

Carbon Fluxes [tC ha-1 yr-1]

Length of Fallow Period [Days]

Over the fallow period there was a clear seasonal pattern of renewal effects (the NEP difference (ΔNEP) between renewal and non-renewal scenarios), showing a difference in ΔNEP over four seasons (Fig. 9a). The average ΔNEP losses in spring and summer were 1.8 and

1.7 tC ha−1 over the fallow period, respectively, and were significantly higher than in autumn and winter, when the NEP losses were 1.1 and 0.8 tC ha−1, respectively. The differences in ΔNEP among seasons could be due to different environmental conditions. There was a strong correlation between ΔNEP and the average temperatures over the fallow period (Fig. 9b), suggesting NEP was reduced most when pasture was renewed under high temperatures. NEP losses also increased with rising soil volumetric moisture content up to 50% (Fig. 9c). A similar relationship between soil water content and ΔNEP was demonstrated by Rutledge et al. (2017b) who showed that NEP loss increased with increasing soil moisture from measurements at two paddocks. There was also a positive relationship between ΔNEP and soil moisture above 50% (R2 = 0.74, p b 0.0001, data not shown). We believe this positive relationship was an artefact, and the high ΔNEP at high soil moisture conditions was regulated by the low temperature, since there was a strong seasonal correlation between temperature and soil moisture in our study region (Liang et al., 2018), with lower temperature and higher soil moisture occurring in winter. At an annual scale, interestingly, although the ΔNEP in autumn and winter was higher than in spring and summer (Fig. 9d), there was no significant difference for the ΔNEP (p = 0.096) regardless of the renewal timing. The annual pattern suggested that the carbon losses during the fallow period could be compensated for by the regrowth of pasture, resulting in a similar NEP at the annual scale. Rutledge et al. (2015) obtained similar results following pasture renewal under severe summer drought at different sites in the Waikato region. They suggested that climatic conditions and management practices had a large impact on CO2 exchange, with a severe drought in one year and cultivation in another both causing large but short-term (about 3 months) net losses of CO2–C (1–2 tC ha−1). However, CO2 was regained later in both years so that on annual timescales, the site was a CO2 sink or CO2 neutral, supporting our conclusion based on model simulations. There were no relationships between annual ΔNEP and temperature (Fig. 9e) or soil moisture (Fig. 9f) over the fallow period, respectively.

25 No Age Function Applied Age Function Applied

0 -25 -50

20 40 60 Length of Fallow Period

Fig. 8. Annual carbon balance relative to the non-renewed pasture over different renewal frequencies (RF) and lengths of fallow period (LF). a)Annual NEP difference. b) NECB difference. c) An example of carbon balance among different carbon fluxes with the longest fallow period (65 days) and 1-year renewal frequency. d) SOC changes across different fallow periods under an 11-year renewal frequency with age function (Fig. 1, 30% loss of productivity) and without an age function applied.

NEP [tC ha-1 30 days-1]

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

0 a B

0 b

-1

A

A

-2

-3

Spring Summer Autumn Winter

0 c

R2 =0.42 , p<0.0001

B

-1

-1

-2

-2

-3

10

15

20

11

25

-3

20

Soil Temperature [°C] NEP [tC ha-1 yr-1]

3

d

3

2 1 A

A

A

A

0 -1

Spring Summer Autumn Winter

e

p=0.75

3 2

1

1

0

0

10

15

20

Soil Temperature [°C]

30

40

50

Soil Water Content [%]

2

-1

R2 =0.42, p = 0.002

25

-1

f

p =0.41

20

30

40

50

Soil Water Content [%]

Fig. 9. The seasonal and annual pattern of renewal effects (ΔNEP) and the relationship between ΔNEP and environmental controls. a) Cumulative differences of ΔNEP between renewal and non-renewal scenarios across spring, summer, autumn and winter. b) and c) Regression between the corresponding cumulative ΔNEP and mean soil temperature and soil water content over 30 days after renewal, respectively. d) Cumulative differences of annual ΔNEP between renewal and non-renewal scenarios. e) and f) The relationships between annual ΔNEP and soil temperature and soil water content, respectively.

4. Conclusions and future research needs Our results show that CenW simulations of GPP, ER, and NEP agreed well with EC measurements at Troughton Farm after parameterization. By keeping the optimal parameters in CenW, we conducted a series of simulations for renewal frequency under different scenarios of pasture production deterioration to investigate the renewal effects on the carbon balance in grazed pasture. Our simulations suggest that pastures could lose SOC if pasture is renewed too infrequently, while pastures could gain SOC but pasture production will decline if pastures are renewed too frequently. To understand this outcome requires a consideration of trends after renewal in GPP, autotrophic and heterotrophic respiration, and grazing removal. Pasture renewal had a greater effect on grazing removal than GPP. The ratio of autotrophic respiration to GPP was also lowered by renewal, and the carbon losses through heterotrophic respiration could be compensated for by the fast regrowth of root system in the renewed pasture. Together, changes in these four key carbon-flux components resulted in more carbon being retained in the pasture, thus increasing SOC if pasture renewal was very frequent. The optimal frequency depends on pasture production deterioration if we aim to maximize grazing removal with only minimal effects on soil carbon stocks. We found that with faster pasture deterioration, more frequent renewal frequency optimized grazing removal. There was no SOC loss under the optimal or more frequent pasture-renewal scenarios. On average, a renewal frequency of around 10 years is optimal, resulting in an average grazing removal of 7.5 tC ha−1 yr−1 while maintaining a similar SOC relative to the non-renewed pasture. Length of the fallow period, renewal timing, and associated environmental conditions are important additional factors that affect the carbon balance. Longer fallow periods adversely affect carbon balances, and the magnitude of carbon losses are also regulated by the renewal frequency associated with pasture deterioration. Specifically, under a renewal frequency of 11 years – a realistic and practical

management option – the effect of the tested fallow periods on the carbon balance was small, with an SOC gain or loss within 0.05 tC ha−1 yr−1. For renewal timing, we found a clear seasonal pattern for NEP loss over the fallow period, with smaller NEP in spring or summer, compared to autumn or winter. Environmental conditions played an important role in regulating NEP across seasons, showing greater losses at higher temperatures and moderate soil moisture conditions. However, at the annual scale, there were no significant differences for NEP losses across four seasons, suggesting that the carbon losses during the fallow period could be compensated for by the regrowth of pasture after renewal. One of the major uncertainties for simulating the pasture carbon balance during pasture renewal lies in the extent of pasture deterioration with age. Observations suggest that pasture production could deteriorate at different rates. This could be due to inherent physiological age effects of the pasture sward, management practices like overgrazing, or weed invasion (Tozer et al., 2010; Tozer and Edwards, 2011). Quantifying the rate of pasture deterioration requires long-term and continuous measurements of pasture productivity. Reducing the uncertainties of pasture deterioration will provide increased accuracy in quantifying the renewal effect on carbon balances. Overall, we found that occasional pasture renewal, e.g., renewed pasture under optimal frequency, leads to very limited SOC changes in the long-term. However, improved pasture renewal strategies, not only by changing the renewal frequency, renewal timing or renewal method, but also by using different foliage species, could act as a mitigation measure. A recent study by Rutledge et al. (2017b) suggested that increasing species diversity in re-established pastures could provide even more beneficial changes in carbon stocks than renewal with ryegrass-clover only. This could result from the greater root mass in the diverse pasture than the ryegrass-clover pasture (McNally et al., 2015). Incorporating understandings of the effects of plant traits on carbon dynamics into process-based models like CenW will be useful to provide a full assessment of SOC change and environmental impacts of grazed pastures.

12

L.ǐL. Liáng et al. / Science of the Total Environment 715 (2020) 136917

Declaration of competing interest The authors declare that there is no conflict of interest. Acknowledgements This research was supported by the New Zealand Agricultural Greenhouse Gas Research Centre and the Ministry of Business, Innovation and Employment Strategic Science Investment Fund. We appreciate valuable comments from Louis Schipper, Andrew McMillan and Paul Mudge and editing help from Ray Prebble. We thank the farm owners, Ben and Sarah Troughton, for providing access to the research site and for logistical assistance in collecting management information. Comments from three anonymous reviewers substantially improved the manuscript. References Ammann, C., Leifeld, J., Jocher, M., Neftel, A., Fuhrer, J., 2013. Effect of grassland renovation on the greenhouse gas budget of an intensive forage production system. Adv. Anim. Biosci. 4, 284. Carolan, R., Fornara, D.A., 2016. Soil carbon cycling and storage along a chronosequence of re-seeded grasslands: do soil carbon stocks increase with grassland age? Agric. Ecosyst. Environ. 218, 126–132. https://doi.org/10.1016/j.agee.2015.11.021. Conant, R.T., Easter, M., Paustian, K., Swan, A., Williams, S., 2007. Impacts of periodic tillage on soil C stocks: a synthesis. Soil Tillage Res. 95, 1–10. https://doi.org/10.1016/J. STILL.2006.12.006. Drake, J.E., Davis, S.C., Raetz, L.M., Delucia, E.H., 2011. Mechanisms of age-related changes in forest production: the influence of physiological and successional changes. Glob. Chang. Biol. 17, 1522–1535. https://doi.org/10.1111/j.1365-2486.2010.02342.x. Fornara, D., Olave, R., Higgins, A., 2020. Evidence of low response of soil carbon stocks to grassland intensification. Agric. Ecosyst. Environ. 287, 106705. https://doi.org/ 10.1016/j.agee.2019.106705. Gál, A., Vyn, T.J., Michéli, E., Kladivko, E.J., McFee, W.W., 2007. Soil carbon and nitrogen accumulation with long-term no-till versus moldboard plowing overestimated with tilled-zone sampling depths. Soil Tillage Res. 96, 42–51. https://doi.org/10.1016/j. still.2007.02.007. Glassey, C.B., Roach, C.G., Strahan, M.R., Mcclean, N., 2010. Dry matter yield, pasture quality and profit on two Waikato dairy farms after pasture renewal. Proc. New Zeal. Grassl. Assoc. 91–96. Hewitt, A.E., 1998. New Zealand Soil Classification. 2nd ed. Manaaki-Whenua Press, Lincoln, New Zealand. Kerr, G.A., Brown, J., Kilday, T., Stevens, D.R., 2015. A more quantitative approach to pasture renewal. Proc. New Zeal. Grassl. Assoc. 251–258. Kira, T., Shidei, T., 1967. Primary production and turnover of organic matter in different forest ecosystems of the Western Pacific. Japanese J. Ecol. 17, 70–87. Kirschbaum, M.U.F., 1999a. CenW, a forest growth model with linked carbon, energy, nutrient and water cycles. Ecol. Model. 118, 17–59. https://doi.org/10.1016/S0304-3800 (99)00020-4. Kirschbaum, M.U.F., 1999b. Modelling forest growth and carbon storage in response to increasing CO2 and temperature. Tellus Ser. B Chem. Phys. Meteorol. 51, 871–888. https://doi.org/10.3402/tellusb.v51i5.16500. Kirschbaum, M.U.F., Rutledge, S., Kuijper, I.A., Mudge, P.L., Puche, N., Wall, A.M., Roach, C.G., Schipper, L.A., Campbell, D.I., 2015. Modelling carbon and water exchange of a grazed pasture in New Zealand constrained by eddy covariance measurements. Sci. Total Environ. 512–513, 273–286. https://doi.org/10.1016/J.SCITOTENV.2015.01.045. Landsberg, J.J., Waring, R.H., 1997. A generalised model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning. For. Ecol. Manag. 95, 209–228. https://doi.org/10.1016/S0378-1127(97)00026-1.

Liang, L.L., Campbell, D.I., Wall, A.M., Schipper, L.A., 2018. Nitrous oxide fluxes determined by continuous eddy covariance measurements from intensively grazed pastures: temporal patterns and environmental controls. Agric. Ecosyst. Environ. 268, 171–180. https://doi.org/10.1016/J.AGEE.2018.09.010. López-Fando, C., Pardo, M.T., 2009. Changes in soil chemical characteristics with different tillage practices in a semi-arid environment. Soil Tillage Res. 104, 278–284. https:// doi.org/10.1016/J.STILL.2009.03.005. McNally, S.R., Laughlin, D.C., Rutledge, S., Dodd, M.B., Six, J., Schipper, L.A., 2015. Root carbon inputs under moderately diverse sward and conventional ryegrass-clover pasture: implications for soil carbon sequestration. Plant Soil 392, 289–299. https://doi. org/10.1007/s11104-015-2463-z. Ministry for the Environment, 2019. New Zealand's Greenhouse Gas Inventory 1990–2017. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I — a discussion of principles. J. Hydrol. 10, 282–290. https://doi.org/10.1016/0022-1694 (70)90255-6. NIWA, 2015. National Climate Database. National Institute of Water and Atmospheric Research. NZIER, 2017. Dairytrade's Economic Contributionto New Zealand. Odum, E., 1969. The strategy of ecosystem development. Science 164 (3877), 262–270. https://doi.org/10.1126/SCIENCE.164.3877.262. Ogle, S.M., Breidt, F.J., Paustian, K., 2005. Agricultural management impacts on soil organic carbon storage under moist and dry climatic conditions of temperate and tropical regions. Biogeochemistry 72, 87–121. https://doi.org/10.1007/s10533-004-0360-2. Parfitt, R.L., Stevenson, B.V., Ross, C., Fraser, S., 2014. Changes in pH, bicarbonateextractable-P, carbon and nitrogen in soils under pasture over 7 to 27 years. New Zeal. J. Agric. Res. 57, 216–227. https://doi.org/10.1080/00288233.2014.924536. Parsons, A.J., Edwards, G.R., Newton, P.C.D., Chapman, D.F., Caradus, J.R., Rasmussen, S., Rowarth, J.S., 2011. Past lessons and future prospects: plant breeding for yield and persistence in cool-temperate pastures. Grass Forage Sci. 66, 153–172. https://doi. org/10.1111/j.1365-2494.2011.00785.x. Rutledge, S., Mudge, P.L., Campbell, D.I., Woodward, S.L., Goodrich, J.P., Wall, A.M., Kirschbaum, M.U.F., Schipper, L.A., 2015. Carbon balance of an intensively grazed temperate dairy pasture over four years. Agric. Ecosyst. Environ. 206, 10–20. https://doi.org/10.1016/j.agee.2015.03.011. Rutledge, S., Wall, A.M., Mudge, P.L., Troughton, B., Campbell, D.I., Pronger, J., Joshi, C., Schipper, L.A., 2017a. The carbon balance of temperate grasslands part II: the impact of pasture renewal via direct drilling. Agric. Ecosyst. Environ. 239, 132–142. https:// doi.org/10.1016/J.AGEE.2017.01.013. Rutledge, S., Wall, A.M., Mudge, P.L., Troughton, B., Campbell, D.I., Pronger, J., Joshi, C., Schipper, L.A., 2017b. The carbon balance of temperate grasslands part I: the impact of increased species diversity. Agric. Ecosyst. Environ. 239, 310–323. https://doi.org/ 10.1016/j.agee.2017.01.039. Ryan, M.G., Binkley, D., Fownes, J.H., 1997. Age-related decline in forest productivity. Adv. Ecol. Res. 27, 213–262. Schipper, L.A., Parfitt, R.L., Ross, C., Baisden, W.T., Claydon, J.J., Fraser, S., 2010. Gains and losses in C and N stocks of New Zealand pasture soils depend on land use. Agric. Ecosyst. Environ. 139, 611–617. https://doi.org/10.1016/J.AGEE.2010.10.005. Schipper, L.A., Mudge, P.L., Kirschbaum, M.U.F., Hedley, C.B., Golubiewski, N.E., Smaill, S.J., Kelliher, F.M., 2017. A review of soil carbon change in New Zealand's grazed grasslands. New Zeal. J. Agric. Res. 60, 93–118. https://doi.org/10.1080/00288233.2017.1284134. Statistics New Zealand, 2013. Agricultural and Horticultural Land Use. Tozer, K.N., Edwards, G.R., 2011. What factors lead to poor pasture persistence and weed ingress ? Pasture persistence – Grassl. Res. Pract. Ser. 15, 129–138. Tozer, K.N., Cameron, C.A., Thom, E.R., 2010. Pasture persistence: farmer observations and field measurements. Pasture persistence – Grassl. Res. Pract. Ser. 15, 25–30. Tozer, K., Rennie, G., King, W., Mapp, N., Aalders, L., Bell, N., Wilson, D., Cameron, C., Greenfield, R., 2015. Pasture renewal on Bay of Plenty and Waikato dairy farms: impacts on pasture performance post-establishment. New Zeal. J. Agric. Res. 58, 241–258. https://doi.org/10.1080/00288233.2015.1015742. Tang, J., Luyssaert, S., Richardson, A.D., Kutsch, W., Janssens, I.A., 2014. Steeper declines in forest photosynthesis than respiration explain age-driven decreases in forest growth. Proc. Natl. Acad. Sci. 111, 8856–8860. https://doi.org/10.1073/PNAS.1320761111.