The carbon balance of temperate grasslands part I: The impact of increased species diversity

Agriculture, Ecosystems and Environment 239 (2017) 310–323

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment journal homepage: www.elsevier.com/locate/agee

Research paper

The carbon balance of temperate grasslands part I: The impact of increased species diversity S. Rutledgea,b , A.M. Walla , P.L. Mudgec , B. Troughtond, D.I. Campbella , J. Prongera , C. Joshie , L.A. Schippera,* a

School of Science and Environmental Research Institute, University of Waikato, Private Bag 3105, Hamilton 3240, New Zealand National Institute for Public Health and the Environment, Centre for Environmental Quality, PO Box 1, 3720 BA Bilthoven, The Netherlands Landcare Research, Private Bag 3127, Hamilton, New Zealand d 103A Troughton Road, Waharoa, New Zealand e Department of Mathematics and Statistics, University of Waikato, Private Bag 3105, Hamilton 3240, New Zealand b c

A R T I C L E I N F O

Article history: Received 31 August 2016 Received in revised form 26 January 2017 Accepted 29 January 2017 Available online 6 February 2017 Keywords: Grassland Pasture Management Pasture renewal Eddy covariance Carbon balance

A B S T R A C T

Appropriate management of grasslands may aid the sequestration of carbon (C) in soil organic matter, thereby increasing soil quality and offsetting other CO2 emissions to the atmosphere. In ungrazed grassland trials, higher species diversity has been found to enhance soil C sequestration due to increases in plant production and C inputs to the soil via roots. Little is known about the impact of increased species diversity on the change in soil C stocks in intensively grazed pastoral systems common in temperate regions. Here, we report the CO2 and C balances of three blocks of an intensively managed dairy farm in temperate New Zealand. Two blocks underwent pasture renewal (PR): one block was renewed back to ryegrass-clover (NewRye), while the second was sown in a mix of higher species diversity including ryegrass, clover, timothy, prairie grass, lucerne, chicory and plantain (NewMix). A third block served as an undisturbed ryegrass-clover control (Control). We hypothesised that the block renewed to a more diverse pasture would have higher pasture production and C sink strength than either the unmodified pasture or the pasture renewed to ryegrass-clover. Net ecosystem production (NEP) was measured using eddy covariance, and other inputs and outputs of C (e.g. C in pasture removed by grazers and C deposited in dung) were calculated. NEP and the net ecosystem carbon balance (NECB) were determined for four years: one year before PR, and three years after. In the year before PR, we measured important differences in annual NEP (and thus NECB) which suggested unanticipated inherent site differences between blocks which affected C cycling: the NEP of the NewRye block was 116 to 160 g C m
2 y
1 higher than that of the other blocks. These differences between blocks were accounted for when considering differences in NECBs as a result of the treatments in years following renewal. During the three years following PR, dry matter production was similar for the NewMix and NewRye blocks (15,033 and 14,708 kg DM ha
1 y
1 respectively) and greater than the Control block (13,116 kg DM ha
1 y
1). For 11 out of the 12 block-years the soil-pasture systems lost C as indicated by negative NECBs. Taking into consideration the pretreatment differences between blocks, the NewMix pasture had a higher NECB (by 254 g C m
2 over 3 years) than the NewRye pasture, indicating smaller C losses at the NewMix block. In contrast, there was no difference in NECB between NewMix and Control blocks. When PR is undertaken to increase pasture performance, renewal to a more diverse sward may decrease C losses compared to renewal to a ryegrassclover sward. In contrast, there was no evidence that PR to a higher diversity sward decreased C losses compared to an unmodified ryegrass-clover sward. Many C balance studies based on eddy covariance techniques have not accounted for pre-existing differences between treatment blocks and we have demonstrated that this may be critically important for drawing conclusions about the true effect of imposed treatments on C balances. © 2017 Elsevier B.V. All rights reserved.

* Corresponding author. Current address: School of Science University of Waikato, Private Bag 3105, Hamilton 3240, New Zealand. E-mail address: [email protected] (L.A. Schipper). http://dx.doi.org/10.1016/j.agee.2017.01.039 0167-8809/© 2017 Elsevier B.V. All rights reserved.

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

1. Introduction Soils form the largest terrestrial store of carbon (C) globally (Batjes, 1996; Janzen, 2004) and are therefore an important part of the global carbon cycle. Small changes in soil C content over large areas could substantially intensify or mitigate current increases in atmospheric CO2 (Smith, 2008; Kell, 2012; Paustian et al., 2016). Moreover, soil C is important for soil health and biomass production because soil organic matter improves soil water holding capacity, nutrient cycling and soil structure (Milne et al., 2015). There is increased interest in identifying management practices that increase the photosynthetic input of C into soil or reduce the rate of loss of C to increase C sequestration in agricultural soils (Cole et al., 1997; Caldeira et al., 2004; Freibauer et al., 2004; Smith et al., 2008; Lal et al., 2011; Paustian et al., 2016). Grazing lands are important stores of soil C due to their large spatial extent globally (25% of the Earth's land area, Cole et al., 1997; Asner et al., 2004; Steinfeld et al., 2006) and their high C content compared to soils under different land uses (Conant et al., 2001; Tate et al., 2005). Despite this, there is evidence that many grassland soils are not C saturated (Beare et al., 2014), and have potential to sequester more atmospheric C provided suitable management practices are put in place. A range of management practices have been proposed that may enhance soil C sequestration in grasslands, including: increasing primary productivity by alleviating nutrient or water deficiencies through fertilisation and irrigation (e.g. Conant et al., 2001; Finn et al., 2016; Smith et al., 2016), optimising grazing management (Allard et al., 2007; Klumpp et al., 2011), introducing earthworms (Schon et al., 2015), increasing the duration of grass leys or converting leys to permanent grasslands (Soussana et al., 2010), and finally, increasing plant species diversity and/or adding deeprooting species – a topic which has received considerable research interest in the past decade (Fornara and Tilman, 2008; Steinbeiss et al., 2008; Kell, 2012; McNally et al., 2015). Interest in using increased plant diversity to enhance soil C sequestration has largely arisen from results of two large (ungrazed) grassland experiments established on former arable land, where rates of C accumulation were significantly higher in plots with greater plant species diversity (Tilman et al., 2006; Fornara and Tilman, 2008; Steinbeiss et al., 2008; Lange et al., 2015). There are a number of mechanisms by which increased plant diversity could lead to greater soil C, but an increase in plant production (usually measured above ground) and increased C inputs to the soil via roots are the general mechanisms. The higher rates of C sequestration observed by Fornara and Tilman (2008) in more diverse grassland plots was largely attributed to higher C inputs, due to “the presence of highly complementary functional groups” (legumes and C4 grasses). They found a highly significant positive relationship between C accumulation and root biomass, suggesting that roots were important for soil C storage, which is consistent with a number of other studies (Rasse et al., 2005; Lu et al., 2011; Fornara and Tilman, 2012; Fornara et al., 2013). Increased C inputs via roots, which stimulated microbial activity and soil C stabilisation, were also concluded to be key mechanisms for the increase in C sequestration with increasing plant species diversity observed by Lange et al. (2015), Steinbeiss et al. (2008) and Cong et al. (2014). Greater rooting depth may also play a role in the observed increase in C sequestration in more diverse grasslands. Higher diversity increases the chances of deeper rooted species being present, and further, at the same site as Fornara and Tilman (2008), Mueller et al. (2013) determined that interactions between species in more diverse plots meant that root depth distributions were twice as deep as expected compared to monocultures, due to phenotypic plasticity. Deeper roots could increase soil C by

311

allowing plants access to more water and nutrients, thus enhancing production and C inputs to the soil. Probably of more importance, deeper roots would be a conduit for C to be incorporated into deeper soil horizons where C concentrations are lower and storage capacity higher (McNally et al., 2015). Observed increases in plant productivity in ungrazed grasslands with higher diversity has led to interest in testing this approach in intensively grazed pastoral systems, such as those common in temperate regions such as New Zealand (NZ) (Pembleton et al., 2015; Mason et al., 2016). Currently, simple binary mixes comprised of perennial ryegrass (Lolium perenne L.) and white clover (Trifolium repens L.) typically dominate NZ pastures grazed by dairy cattle. These pastures are easy to manage and are rotationally grazed year round, but have relatively shallow roots (Crush et al., 2005), which can lead to reduced production during dry conditions. Research in NZ has shown annual pasture production of more diverse swards to be higher, or the same as traditional ryegrass-clover pastures, with production during drier summer/autumn conditions consistently higher (RuzJerez et al., 1991; Nobilly et al., 2013; Woodward et al., 2013; Mason et al., 2016), likely due to increased root biomass – particularly lower in the soil profile (McNally et al., 2015), allowing greater access to soil water. While initial interest in more diverse pastures in NZ was largely driven by a quest to increase pasture production and/or quality, and thus milk and meat production (Totty et al., 2013; Pembleton et al., 2015), potential environmental benefits such as reduced nitrogen leaching losses and nitrous oxide emissions (Totty et al., 2013; Beukes et al., 2014), improved water use efficiency and soil C storage (McNally et al., 2015) are now gaining increased research attention. In summary, studies in ungrazed grasslands have shown increased soil C sequestration in plots with greater diversity of plant species. These increases in soil C have largely been attributed to enhanced productivity leading to increased C inputs to the soil through more and/or deeper roots. In an intensively grazed New Zealand grassland, McNally et al. (2015) clearly demonstrated that more diverse pastures had higher root biomass (particularly at depth) than traditional ryegrass-clover pastures. Annual production from more diverse pastures in New Zealand has been reported to be higher, or the same as traditional ryegrass-clover pastures, with production during drier/warmer conditions being consistently higher in more diverse swards. Paddock scale measurements of CO2 fluxes in New Zealand have revealed that predominantly ryegrass-clover pastures often switch from being net sinks of CO2 in spring to net sources in summer-autumn if pastures become moisture stressed (Campbell et al., 2015; Rutledge et al., 2015; Hunt et al., 2016). Based on this combination of information, we hypothesised that a recently sown more diverse pasture would have higher C sink strength relative to either unmodified or resown ryegrass-clover pastures. We tested this hypothesis by measuring the net ecosystem carbon balance (NECB) of a grazed New Zealand grassland renewed to a more diverse sward and comparing the NECB with that of a grassland renewed with a conventional ryegrass-clover pasture and with that of an existing untouched conventional ryegrassclover pasture. Measurements of CO2 fluxes were made using the eddy covariance technique for one year before and three years after pastures were renewed. Non CO2-C fluxes such as supplementary feed imports and biomass removed by grazing cattle were also quantified. An important consideration when evaluating the potential benefit of diverse swards is that at least short-term losses of C would be expected during the pasture renewal process due to killing the existing rye-clover pasture (Rutledge et al., 2014). This potential loss may offset any benefit from increased C storage from diverse pastures and needs to be factored into conclusions about

312

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

net changes in soil C stocks. This pasture renewal effect occurs over relatively short timeframes and there are important nuances associated with soil moisture content, site preparation and fallow period (Rutledge et al., 2014). The pasture renewal effect at our present study site is more fully explored in a companion paper (Rutledge et al., 2017). 2. Methods 2.1. Site description The study was conducted on Troughton Farm, a commercial dairy farm (37.77 S, 175.8 E, 54 m elevation) in the Waikato region, North Island, New Zealand. The farm has been used for dairying for at least 80 years. Thirty year (1980–2010) mean annual temperature and precipitation at a climate station 13 km to the south-west of the farm were 13.3  C and 1249 mm respectively. Although frosts occur in clear and calm conditions, no snow is received in winter. The study took place on three blocks on the farm which were at most 1215 m apart (Fig. 1), chosen because they were flat and the soils similar. Soils on the three experimental blocks of paddocks were a complex of types formed in rhyolitic and andesitic volcanic ash on rhyolitic alluvium (McLeod, 1992). Soils were silt loams ranging from well to poorly drained and had relatively high soil C contents between 7.6 and 9.3% in the Ap horizon (McLeod, 1992). The Te Puninga soil, classified as a Mottled Orthic Allophanic soil by Hewitt (1998), was the dominant soil type on all three experimental blocks as determined by spatial soil sampling. This soil was a moderately gleyed imperfectly drained soil with a soil C content of 9.3% (McLeod, 1992) and a permanent wilting point of 0.25 m3 m
3 in the topsoil (LandcareResearch, 2015). 2.2. Farm management The 207 ha farm was grazed by approximately 660 Jersey cows, equating to an average stocking density of 3.2 cows ha
1. The majority of paddocks were between 2.5 and 3.5 ha area, and they

were rotationally grazed throughout the year. During times of the year with lower pasture production (i.e. winter and under dry conditions in late summer), additional (supplementary) feed was provided to the animals, mostly in the form of grass or maize silage and hay. Initially the animals received the supplementary feed predominantly in the paddock, however from June 2015 onwards most of the feed was received on a dedicated feed pad rather than in the paddock. Although a proportion of paddocks on the farm were harvested for silage during the spring and early summer months, this never took place on paddocks that were part of this experiment. Paddocks were typically fertilised with commercial fertilisers twice yearly, while from winter 2012 onwards, regular application of duck, chicken and goat manures also took place (1–2 applications per year). Combined nutrient applications averaged 58 kg N ha
1 y
1, 13 kg P ha
1 y
1 and 38 kg K ha
1 y
1 along with small amounts of trace elements. The commercial fertilisers supplied 49% of the N, 5% of P and 26% of K with the remainder being derived from duck, chicken and goat manures which were largely composed of animal bedding (wood chips) and excrement. Farm management as described above is considered fairly typical for the region. Day-to-day management decisions, for example when to graze the paddocks, were largely outside the control of the research team. While the grazing regime was similar between the three blocks with regards to number of grazings and stocking rates (Fig. S1), it was not practicably feasible to graze all paddocks in the three blocks at the same time, resulting in asynchronous grazing of the experimental paddocks. 2.3. Experimental design Before pasture renewal (PR), swards on all experimental blocks were predominantly made up of ryegrass and white clover. Two blocks underwent pasture renewal: one was resown back into ryegrass-clover (hereafter referred to as ‘NewRye’; for future reference the measurement site has received the code NZ-Tr3 within the Fluxnet network; D. Papale, pers. comm. 2016) and the other into a more diverse sward consisting of grasses, legumes and herbs (hereafter referred to as ‘NewMix’; Fluxnet code NZ-Tr2). The third block served as a control and did not receive any treatment (hereafter referred to as ‘Control’; Fluxnet code NZ-Tr1). This experimental setup with three blocks was chosen to allow separation of the ‘pasture renewal effect’ and the ‘sward effect’ on C dynamics. The impact of pasture renewal has been explored in detail in the companion paper (Rutledge et al., 2017), while the focus of the current paper is on the ‘sward effect’. In the current paper, we compare C fluxes collected before pasture renewal to fluxes for the period after pasture renewal of the two treatment blocks. We refer to four 12-month periods as ‘years’: The before-PR ‘year’ (Year 1) ran from 1 April 2012 to 1 April 2013. The after-PR ‘years’ ran from June to June for 2013–2014, 2014–2015 and 2015–1016 for Years 2, 3 and 4, respectively. This way, the two months when C fluxes were expected to be most impacted by pasture renewal (1 April–31 May 2013) are disregarded in the current analysis, but analysed in detail in Rutledge et al. (2017). 2.4. Pasture renewal

Fig. 1. Map of the part of Troughton Farm with the three experimental blocks where NECB measurements were made. The inset shows the location of Troughton Farm on the North Island of New Zealand.

The NewRye pasture was made up of the conventional combination of perennial ryegrass (Lolium perenne; Agricom ‘One50-AR370 ; seeding rate 19 kg seeds ha
1) and white clover (Trifolium repens; PGG Wrightson seeds ‘Kopu II’ and Agricom ‘Tribute’; both 3 kg ha
1). The NewMix pasture contained, in addition to the same ryegrass (10 kg ha
1) and two white clovers (both at 1.5 kg ha
1), grasses timothy (Phleum pratense; 1.0 kg ha
1), cocksfoot (Dactylis glomerata; Agricom ‘Kara’; 2 kg ha
1) and

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

prairie grass (Bromus wildenowii; Agricom ‘Atom’; 5 kg ha
1), lucerne (or Alfalfa; Medicago sativa; Agricom ‘Torlesse’; 8 kg ha
1), chicory (Cichorium intybus; Agricom ‘Choice’; 2 kg ha
1) and plantain (Plantago lanceolata; Agricom ‘Tonic’; 1.5 kg ha
1). This mix of seeds was chosen because a trial in the same region using a very similar mix of species found good pasture productivity, increased drought resilience and more even distribution of dry matter production throughout the year compared to the conventional rye grass-clover pasture (Woodward et al., 2013). Seasonal root sampling of that sward in 2012–2013 confirmed higher root mass and deeper roots in the moderately diverse sward compared to a ryegrass-clover control sward (McNally et al., 2015). 2.5. Pasture production and botanical composition Aboveground biomass production and botanical composition of the three blocks were monitored after pasture renewal. In every block, 10 cages (85  55 cm) were placed at random locations to exclude cattle, ensuring not to place them close to gates, fence lines or water troughs. At a frequency similar to the grazing rotation at the time of year, a 45  45 cm quadrant of pasture within the cages was clipped to 4 cm height to simulate grazing. Samples were dried at 95  C for 24 h and weighed to determine dry biomass. Cages were moved to a different randomly allocated location after every sampling. Every three months, botanical composition was also measured by subsampling the cut biomass and sorting approximately 400 pieces per sample into the different species. 2.6. Carbon balance measurements The NECBs (expressed as g C m
2 period
1) for each of the three experimental blocks on the farm were calculated as (adapted from Chapin et al., 2006): NECB = NEP + Fsupp + Fmanure + Ffertiliser + Fexcreta return
Fpasture removed
Fsupp removed
Fleach

(1)

Where NEP is the net ecosystem production (equal in size to net ecosystem exchange (NEE) but of opposite sign, here excluding grazer respiration), where positive values of NEP indicate the ecosystem is a sink for CO2. Fsupp is the C imported into the experimental blocks as supplemental cattle feed (grass silage, maize silage, hay and straw), Fmanure is the C imported as animal (chicken, duck and goat) manure/bedding used as fertiliser, Ffertiliser is the C imported in manufactured fertiliser (custom blends which included urea or lime), Fexcreta returned is the C deposited on the experimental blocks in cattle dung and urine. Fpasture removed is the total amount of C ingested by cows from pasture, Fsupp removed is the C ingested from supplementary feed (Fsupp), and Fleach is the C lost as dissolved organic C via leaching as reported for the site by Sparling et al. (2016). Several terms of the NECB were left out because they did not apply to our system, or were deemed negligible (e.g. no C was imported onto the three experimental blocks as effluent from the dairy shed, no pasture was removed via mechanical harvesting, and loss of C through erosion would have been negligible because of the flat topography and lack of surface runoff channels). Here we assume that changes in vegetation biomass would have been negligible on an annual time scale which means that a non-zero NECB would imply changes of C stored as soil C: NECB  Dsoil C/Dt. Details of CO2 flux measurements (used to derive NEP) and measurements and calculations of non-CO2-C fluxes are provided in the two sections below. It is important to note that in Eq. (1), the Fpasture removed and Fsupp removed terms mean cows are treated in the same way as ‘mechanical’ harvesters would be in a ‘cut and carry’

313

situation, except they return a portion of ingested C via dung and urine (after Hunt et al., 2016). 2.6.1. CO2 fluxes 2.6.1.1. Data acquisition and flux processing. NEE of the pasture/soil system was measured using eddy covariance (EC) technique. As explained below, respiration from grazing cattle was deliberately excluded from NEE measurements following Felber et al. (2016) and Hunt et al. (2016). Measurements started in October 2011 and November 2011 on NewMix and Control blocks respectively, well before the start of Year 1 (the before-PR ‘year’). In contrast, measurements on the NewRye block started in August 2012, midway through Year 1. The EC systems were the same for the three blocks, and included a CSAT3 sonic anemometer (Campbell Scientific Inc., Logan, UT, USA) and a LI-7200 infrared gas analyser (LI-COR Inc., Lincoln, NE, USA) mounted at 1.55 m height. The relatively low installation height was chosen to ensure at least 80% of the flux originated from within the experimental blocks (Rutledge et al., 2017). Half-hourly fluxes were calculated from 20 Hz data using EddyPro version 5.2.0 (EddyPro1 Version 5 [Computer software], 2015). Raw data were screened for spikes, values below or above absolute limits, drop-outs, and insufficient amplitude resolution (LI-COR Inc., 2015). Fluxes were calculated using the “block average as detrend” method and using mixing ratios of H2O and CO2 rather than concentrations (Burba et al., 2012). Double-axis coordinate rotation was used, and time lags between time series of H2O and CO2 concentration and vertical wind speed were determined using the EddyPro automated time lag optimisation routine. This routine first determined the time lag for each half hour by covariance maximisation. It then determined the median time lag for H2O and CO2 separately for all relative humidity classes (0–10%, 10–20%, etc.). This RH-dependent median time lag was then applied as the default time lag for half hours when no clear covariance maximum could be found. A fully analytical correction was used for correcting of high-pass filtering effects in the low frequency range (Moncrieff et al., 1997). The method outlined by Fratini et al. (2012) was used to correct for low-pass filtering effects in the high frequency range. The software also tested for steady state conditions and developed turbulence, which resulted in a three-level quality (QC) control flag similar to that of Mauder and Foken (2004) with flags 0–2 (with 0 being best quality). 2.6.1.2. Data quality control, gap filling and partitioning. Halfhourly values for NEE were discarded when: i) a gas analyser reported warnings or the sonic anemometer reported errors during more than 0.5% of the half hour, ii) unrealistically high or low CO2 flux values were calculated, iii) fluxes were collected under low turbulence conditions as indicated by large standard deviation of the CO2 concentration or low values of the standard deviation of the vertical wind speed (Acevedo et al., 2009) with a threshold of sw of 0.11 m s
1 found suitable at our sites, iv) tests for stationarity and developed turbulence as performed in EddyPro (Mauder and Foken, 2004) resulted in a combined quality control flag value of 2, v) more than 5% of the flux originated from paddocks with grazing cattle as estimated using the Kormann and Meixner (2001) footprint model, or when vi) the source area of the flux was behind the tower. As a final quality check, CO2 fluxes were identified as ‘soft spikes’ when the modelled flux value calculated by the gapfilling method described below deviated strongly from measured flux values, with the threshold residual determined after careful inspection of the data. Gaps in the NEE time series were filled using an artificial neural network approach (ANN; Papale and Valentini, 2003) using Matlab

314

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

code created by Cove Sturtevant (Biometeorology Lab, University of California Berkeley), whose methods were inspired by instruction from Dario Papale (University of Tuscia). Separate ANN’s were run for daytime (PPFD > 10 mmol m
2 s
1) and night-time data. The feed-forward ANN consisted of one hidden layer with 4 hidden nodes. Sigmoid transfer functions were applied to the weighted sums from the hidden and output layers and training was performed using the Levenberg–Marquardt algorithm. Both daytime and night-time NEE datasets were split randomly into training, validation and independent testing sets which made up 70, 15 and 15% of the entire datasets, respectively. Multiple runs with different initial random seeds for biases and weights were used to ensure the global minimum of the difference between the observations and ANN predictions was found, rather than local minima. The mean squared error and r2 were used to test the networks’ performance. Both ANNs were trained and fitted 50 times and the medians of the predicted values was used to fill gaps. The ANNs were run on blocks of data of two months. Input variables for the daytime ANN were PPFD, vapour pressure deficit, air and soil (40 mm) temperatures, volumetric moisture content at 100 mm and phytomass index (see below). Input variables for the night-time ANN consisted of air temperature, soil temperatures at 40 and 100 mm, volumetric soil moisture content at 100 mm and phytomass index. The phytomass index (PI) is an empirical factor (Lohila et al., 2004) which accounts for changes in photosynthetic capacity of the sward primarily caused by periodical removal of much of the aboveground biomass by grazing cattle. PI is a dimensionless index (between 0 and 1) calculated on a daily time step as the difference between daily averaged night-time NEE and daily averaged daytime NEE during non-light-limiting conditions (= DNEE), standardized by the maximum DNEE as found generally just before grazing, such that PI = DNEE/DNEEmax. The phytomass index, first used by Aurela et al. (2001) and Lohila et al. (2004) to predict primary production, was found to be very suitable to aid gapfilling in an rotationally grazed pasture in New Zealand (Campbell et al., 2015), when measures of aboveground biomass were unavailable. The potential circularity of using PI (derived from NEE) to predict missing NEE values was not of concern, because the ANN’s sole purpose was to gapfill the NEE time series. Time series of NEE of the NewRye block were not gapfilled using the approach described above for the period of Year 1 when measurements were not yet available (between 1 Apr 2012 and the start of measurements on 17 August 2012). Instead, cumulative NEP for the NewRye block over that period was assumed to equal the average of the cumulative NEPs of the two other blocks. To partition NEE into photosynthesis (gross primary production or GPP) and ecosystem respiration (ER), we first ran the night-time ANN on daytime data to estimate daytime ER. Daytime GPP was then estimated by subtracting modelled daytime ER from gapfilled NEE. Night-time GPP was assumed to be zero. This paper will report on net ecosystem production (NEP) which was assumed to equal –NEE. Annual and monthly sums were calculated by summing the gap-filled half-hourly data. We followed the sign convention where both GPP and ER have positive values. NEP is positive when the ecosystem is fixing carbon (i.e. acts as a sink) and negative when C is emitted to the atmosphere (i.e. acts as a source). Note that NEP and ER in this paper do not include respiration from grazing cattle as we deliberately discarded data collected during half hours when cattle were in the experimental block (Hunt et al., 2016). 2.6.2. Non-NEP components of the carbon balance For all non-NEP components of the NECB (e.g. Fsupp, Fmanure, etc.), the total C flux was the sum of C in all export or import events to the experimental block for the period of interest (e.g. a single

event, or an annual total) converted to a per area basis. The following descriptions outline how total C export or import was calculated for each event. 2.6.2.1. Carbon exports. In addition to respiration from plants, soil and dung (captured by the EC system), the other major C export pathways occur following grazing of pasture and supplementary feed by cattle. Fpasture removed was the total amount of C ingested by cattle from pasture on the experimental blocks. The pasture C ingested by cattle for each grazing event was calculated using: Fpasture removed ¼ Pasture-C  Utilisationpasture

ð2Þ

where; Pasture-C was the amount of C in pasture available for grazing on the experimental blocks (calculations are described in more detail below); and Utilisationpasture was the proportion of available pasture above 4 cm height predicted to be ingested by cows (0.85 based on results from an experimental farmlet trial in the Waikato with similar management to the current study (Macdonald et al., 2008)). Pasture-C at the time of each grazing was calculated from pasture growth rates and the C content of bulk pasture (45.2%). Pasture growth rates for the period following pasture renewal were determined from cuts to 4 cm (typical dairy cattle grazing height) made at similar intervals to the grazing rotation. Since measurements of pasture growth were not made prior to pasture renewal, pasture growth rates for this period were modelled using the McCall pasture growth model (McCall and Bishop-Hurley, 2003) specifically calibrated for the site using pasture growth rate data measured on the Control site. There were only very minor seasonal differences in C content, and therefore the annual C content of the bulk pasture was calculated for each of the swards by multiplying the annual botanical composition (derived from four seasonal samplings per year) by the annual C content of each species (calculated from four seasonal samplings in one year). The C contents for each species matched very closely with values measured on the same species (and cultivars) in a similar pasture trial in the Waikato a year earlier (Mason et al., 2016). This similarity provides confidence that pasture carbon contents are consistent and well constrained. Fsupp removed was the total amount of C ingested by cattle from supplementary feed imported and fed on the experimental blocks for each grazing event using: Fsupp removed ¼ Fsupp  Utilisationsupp

ð3Þ

where; Fsupp was the amount of C in supplementary feed imported into the experimental blocks (described in more detail in Section 2.6.2.2); and Utilisationsupp was the proportion of supplementary feed fed on paddocks ingested by cows (0.8; DairyNZ, 2012). Fleach was the amount of dissolved organic carbon (DOC) leached below 0.6 m. DOC leaching was calculated from measurements of DOC concentrations (from 100 suction cup lysimeters) on two of the experimental blocks (Sparling et al., 2016), and modelled drainage volumes using the Woodward model (Woodward et al., 2001). The drainage model used measured evaporation data from the eddy flux towers (Pronger et al., 2016). 2.6.2.2. Carbon imports. Fsupp was the C imported as supplementary feed (grass silage, maize silage, hay and straw) and fed to cows while grazing in the experimental blocks. The mass of C imported for each supplementary feeding event was calculated using: Fsupp ¼ Suppmass  Supp dry matter content  SuppC Content

ð4Þ

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

Supplementary feed mass (Suppmass) was quantified by weighing loads of bulk feed (grass and maize silage), or from an average bale weight for each batch of bales (grass silage, hay and straw). A number of subsamples of each feed type and ‘batch’ were dried to determine dry matter (DM) content (Suppdry matter content) and a separate sub-sample analysed for C content (SuppC Content). DM content tended to be quite variable, but C content was very consistent across the different feed types and sampling times. Fmanure was C imported in manure (used as fertiliser) brought in from nearby chicken, duck and goat farms and consisted of a mixture of animal excrement and animal bedding (wood chips, hay and straw). The amount of C in the manure applied to the experimental paddocks for each application was calculated from the weight of manure applied (Manuremass; from tractor-mounted scales), and the C (ManureC Content) and DM content (Manuredry matter content) of the manure using: Fmanure ¼ Manuremass  Manuredry matter content  ManureC Content

ð5Þ

As for supplemental feed imports, subsamples of each manure type and ‘batch’ were dried to determine dry matter content and then analysed for C content. Ffertiliser is the C imported in manufactured fertiliser, which was generally custom mixes, often including urea, and occasionally lime. Where proprietary fertilisers with known compositions were used, the C content of the fertiliser was calculated from the composition of the individual components (e.g. urea). For fertilisers with unknown compositions, the C content of the fertiliser was analysed (on a wet weight basis). In both cases C import was calculated by multiplying the carbon content by the mass applied. Fexcreta return is the C in cattle dung (Dung-C) and urine (Urine-C) deposited on the experimental blocks. Dung-C was the larger of the two components, and was calculated for each grazing event from the mean of dung excreted on the day of grazing and the previous day (proportioned according to the time spent grazing on the experimental paddocks), using: Dung-C ¼

Total dung
C  Time on paddock 2

ð6Þ

where: Total dung
C ¼ Current Day

Si¼Previous Day Pasture-Ci  Utilisationpasture  ð1
digestibilityÞ Current Day

þ Si¼Previous Day Supp-Ci  Utilisationsupp  ð1
digestibilityÞ

ð7Þ

and: “Time on paddock” was the time cattle spent grazing the experimental paddock(s) in days; Pasture-C is the amount of C in pasture available to be grazed (including paddocks outside experimental blocks); Supp-C was the amount of C in supplements fed to the cows (including supplements fed on a feed pad and paddocks outside the experimental blocks); Utilisationpasture was the proportion of available pasture ingested by cows as outlined previously; Utilisationsupp is the proportion of supplementary feed ingested by cows (0.8 for supplementary feed fed on paddock, DairyNZ (2012) and 0.95 for supplementary feed fed on a feed pad (B. Troughton pers. comm. 2016)); 1–digestibility gives the proportion of ingested organic matter excreted in dung. Digestibility of pasture can vary between seasons, and the digestibility of different supplementary feeds can also vary. Therefore, samples of pasture and supplementary feed fed out on the experimental blocks were taken periodically and analysed for metabolisable energy (ME), from which digestibility was calculated by rearranging the standard equation: ME = 0.16*Digestibility (Hill Laboratories, n.d.) to: Digestibility = ME/0.16. Measured ME values

315

closely matched those reported by DairyNZ for New Zealand pastures (DairyNZ, 2012) and therefore DairyNZ values were used to fill in gaps. Digestibility between swards was considered to be the same, based on three years of measurements by Woodward et al. (2013) in the Waikato, where metabolisable energy (ME) was 11.7 MJ ME kg
1 DM for standard pastures and 11.8 MJ ME kg
1 DM for diverse pastures. The average C content of ingested feed in this study was 45.2%, and the average C content of dung samples from multiple cows/seasons grazing similar pastures to the current study was 46.6% (Rutledge et al., 2015). Therefore, we assumed that the digestibility of carbon was the same as digestibility of organic matter. Based on calculations in Hunt et al. (2016), data from personal communication with Garry Waghorn (Senior Scientist DairyNZ; 24 May 2016), and energy partitioning values for high quality pasture in Waghorn et al. (2007), we assumed that C deposited on the experimental blocks in urine (Urine-C) was 14% of the C deposited in dung (Dung-C). Fexcreta return was the sum of Dung-C and Urine-C for the period of interest. 2.7. Supporting meteorological and soil measurements Ancillary measurements at each EC site included relative humidity and air temperature measured at 1.5 m (fully aspirated HMP 155, Vaisala, Helsinki, Finland), shortwave incoming radiation at 1 m (NR01, Hukseflux Thermal Sensors, Delft, Netherlands), and rainfall 450 mm above ground level (tipping bucket raingauge, TB5, Hydrological Services). Soil temperature was measured using a four junction averaging thermocouple (TCAV, CSI) with two probes at 20 mm and two at 60 mm (referred to as 40 mm here forth), and using thermistors (107 probes; CSI) at 50, 100 and 200 mm. Soil moisture was measured at 50 and 100 mm using water content reflectometers (CS616, CSI). Meteorological and soil parameters were measured every 1 s and averaged every 30 min. 2.8. Uncertainties associated with the NECB calculation 2.8.1. Uncertainty estimates for CO2 fluxes In this study, the total uncertainty in annual sums of NEP was estimated by quantifying the contribution from two sources of random uncertainty (Goulden et al., 1996). We applied the method described by Dragoni et al. (2007) to determine both the random measurement uncertainty (Hollinger and Richardson, 2005) and the uncertainty introduced by the gapfilling procedure. The choice of threshold value for sw for rejecting measurements during poorly developed turbulence which is considered the largest source of systematic uncertainty (Morgenstern et al., 2004; Wohlfahrt et al., 2008), was assumed to be equal or very similar between farm blocks (Ammann et al., 2007, 2009). Because the main purpose of this study was to compare the CO2 and C balance between parallel blocks on the farm, these systematic uncertainties, assumed similar between blocks, were not considered. The reported uncertainties for NEP and NECB therefore only allow for comparisons between sites (differentially), rather than for comparison to absolute values. 2.8.2. Uncertainty estimates for non-CO2-C fluxes The calculation of the size of the non-NEP components of the NECB utilised measurements and estimates of factors like wet weight of supplemental feed, dry matter and C contents, time cows spent on the paddocks, pasture utilisation and digestibility, etc. For determination of the uncertainties relating to the non-NEP components of the NECB, we estimated a probability distribution for all these factors needed to calculate the non-NEP components. To do this we used an approach that was inspired by Bayesian elicitation (O’Hagan et al., 2006). A mean and variance was assigned to all factors required for the calculations, and unless

316

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

measurements provided strong evidence to the contrary, distributions were assumed to be Normal (or Gaussian). The method used to assign a mean and variance differed depending on available data. Where four or more data points were available for any factor, the mean and variance were calculated from the measured data. If there were between one and three measured data points for a factor the mean was used, but the variance was determined by utilising the maximum calculated variance of other factors of the same type. In situations where no measured data were available, the assigned mean was the calculated mean of all other factors of the same type, while the variance was either the variance calculated between all other factors of the same type (where available), or estimated from ‘expert opinion’. The approach used to assign a mean and variance to the Fpasture removed component is described in some more detail in the Supplementary materials, because this component has the largest uncertainty of all non-NEP NECB components, and is dealt with a little differently from the other components. By assigning a mean and variance to all factors, probability distributions were obtained for all factors in the calculation. A Monte Caro simulation (1000 runs), whereby we randomly drew from the probability distributions of all factors needed to calculate these non-NEP components, resulted in an estimate of a probability distribution of the NECB. We report the 95% probability interval, obtained by using the 2.5th and 97.5th percentiles of this distribution. Values were assumed different when 95% probability intervals did not overlap. 3. Results 3.1. Weather and pasture 3.1.1. Weather Mean annual temperature did not vary much over the four-year study period, with values between 13.6 and 13.8  C (Table 1). Annual rainfall ranged from 1105 mm in Year 1 to 1212 mm in Year 3. Rainfall over the months January–March (mid-summer and early-autumn) was distinctly lower in Years 1, 2 and 3 compared to Year 4 (Table 1). This low rainfall was also reflected in the volumetric soil moisture contents (VMC), with the 14-day running means of VMC (averaged across the three blocks) below the permanent wilting point (PWP) of 0.25 m3 m
3 for 48, 31 and 53 consecutive days, for Years 1, 2, and 3 respectively (Fig. 2a). During Year 4, VMC was below PWP for only 9 consecutive days. 3.1.2. Pasture production and species composition Pasture production showed a strong seasonal pattern with highest pasture production in spring (September-November), and lowest pasture production in winter (June–August; Fig. 2f). Average annual pasture production over the three years after PR was similar for the NewMix and NewRye blocks (15,033 and 14,708 kg DM ha
1 y
1 respectively), but lower for the Control

Table 1 Mean annual rainfall and temperature for the four measurement years (averaged across the three experimental blocks). Also shown is the sum of rainfall over the mid-summer and early autumn period (January–March), and the number of consecutive days when 14-day running mean volumetric soil moisture content (VMC) at 100 mm depth was below the permanent wilting point (PWP) of the dominant soil type in the experimental blocks.

Annual rainfall (mm) Mean annual temperature ( C) January–March rainfall (mm) Number of days when VMC < PWP

Year 1

Year 2

Year 3

Year 4

1105 13.6 105 48

1123 13.8 112 31

1212 13.6 130 53

1193 13.8 340 9

block (13,116 kg DM ha
1 y
1). There were no large consistent seasonal differences in production between the three experimental blocks (Fig. 2f). Seasonal sampling confirmed that the pasture species composition at the Control and NewRye blocks were very similar and dominated by ryegrass (Fig. 3). Botanical composition measurements in the NewMix block after PR showed higher species diversity reflecting the mix of species sown. Together ryegrass, chicory, and plantain were the three most dominant species in the sward. Lucerne did not establish at high densities, even though the seeding rate of 8 kg ha
1 should have allowed that. There were seasonal changes in pasture composition in the more diverse sward: in summer, herbs plantain and chicory made up a larger proportion of the sward (up to 65%), while ryegrass was more dominant in winter (data not shown). Over the three years following establishment, the ryegrass component at the NewMix block became more dominant with the proportion of ryegrass increasing from 19% in Year 2 to 46% in Year 4 (Fig. 3). 3.2. CO2 fluxes and annual net ecosystem carbon balances 3.2.1. CO2 fluxes In general, GPP and ER showed a similar seasonal pattern to pasture production, with maximum values in spring/summer and lowest values in autumn/winter (Fig. 2). Of particular note were the rapid reductions in GPP and ER as the soil dried out from midsummer during the first three years (Fig. 2). Large net losses of CO2 (negative NEP) during the pasture renewal period for the NewRye and NewMix blocks were clearly evident (see shaded area in Fig. 2), and this is explored in detail in Rutledge et al. (2017). On an annual basis, NEP was positive (indicating a net uptake of CO2) for all years (Table 2) – but note that cattle respiration was not included in NEP. In the year before PR, NEP differed between the three blocks, with highest value in the NewRye block, followed by the Control and then the NewMix blocks (Table 2, Fig. S2 column A). A similar pattern occurred for cumulative NEP for three years after PR, with NEP of the NewRye block higher than the other two blocks (Fig. S2 column B). 3.2.2. Annual net ecosystem carbon balances Most eddy covariance based NECB studies assume that treatment blocks had similar NEP and NECBs prior to imposing treatments and that any observed differences between blocks were due to ‘treatment’ effects. In contrast, we had about a year of measurements prior to pasture renewal that demonstrated some important differences between blocks that needed to be accounted for when considering differences in NECBs between treatments. Consequently, we present both NECBs ‘as measured’ in the three years after pasture renewal and also with a correction for initial differences between blocks. This second approach assumes that the differences between blocks as observed in the initial year (before PR) would have continued and stayed constant in the following years if PR would not have taken place; the consequences of this assumption are discussed further below. While annual values for NEP were always positive (indicating net CO2 uptake), annual NECBs were always negative (indicating a net C loss), with the exception of the NewRye block in the year before PR (Table 2, Fig. 4 column A). There appeared to be a trend of decreasing NECB (increase in loss of C) with time for all three blocks. The similar trend between blocks indicates this was not related to treatment. It is not clear what would cause such a trend, and ongoing measurements at the same blocks will determine if the trend continues. The patterns of annual NECBs between blocks was similar to that for NEP. In the year prior to PR, the NECB of the NewRye block (+45 g C m
2 y
1) was higher than that of the Control and NewMix

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

317

Fig. 2. (a) Monthly normal and measured rainfall, and volumetric soil moisture content at 0.10 m depth. The dashed line at 0.25 m3 m
3 refers to the permanent wilting point; (b) monthly normal air temperature, and air and soil temperature measured at 0.05 m depth; (c) net ecosystem exchange (NEP); (d) gross primary productivity (GPP); (e) ecosystem respiration (ER); (f) daily pasture growth rate averaged seasonally. The black arrows in panels c, d and e shows when NewMix and NewRye blocks were sprayed with herbicide on 4 April 2013 to kill existing pasture as part of pasture renewal. The greyed out area depicts the two-month period directly following pasture renewal on NewMix and NewRye blocks, which was disregarded in the current paper. Normal rainfall and air temperature were measured between 1981 and 2010 at a nearby weather station. Farm rainfall, soil moisture contents and temperatures were very similar between blocks and averages for the three blocks are shown here. All lines except for those in panel f are 15-day running means.

blocks which were similar (–71 and
115 g C m
2 y
1 respectively, Table 2 and Fig. 4 column A). Differences between blocks were also evident after pasture renewal, with a cumulative total NECB for the three years after pasture renewal of
282 g C m
2 over three years for the NewRye block, which – although negative – was again higher than the Control (–429 g C m
2 over three years) and the NewMix (
507 g C m
2 over three years) blocks (Fig. 4 column B, Table 2). When expressed on an annual basis, this corresponds to

average annual NECBs of
143,
169 and
94 g C m
2 y
1 for the Control, NewMix and NewRye blocks, respectively. The higher NECB in the NewRye block was largely due to higher NEP. In the three years after PR, cumulative NEP of the NewRye block was 205 g C m
2 over three years higher than the NewMix block and 176 g C m
2 over three years higher than the Control block. In contrast, the differences between non-CO2-C imports (manure, supplementary feed, excreta, fertiliser) and exports

318

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

`

Fig. 3. Average annual pasture species composition on the three experimental blocks for the three years after pasture renewal on the NewMix and NewRye blocks.

(pasture/supplementary feed removed, leaching) between the NewRye and NewMix blocks, and NewRye and Control blocks were only 20 g C m
2 over three years and 29 g C m
2 over three years, respectively (calculated from values in Table 2). The differences in NEP (and NECB) between blocks before PR suggests that despite careful selection of the locations of blocks and EC systems on the farm (we matched soil complexes as closely as possible, and for the year prior to pasture renewal, the management of these blocks was virtually identical), there were some inherent underlying differences between blocks which affected C fluxes. To account for these potential block-specific factors, we also expressed the cumulative NECBs for the period after pasture renewal taking into account the NECB of the first year as outlined below. After PR, the difference between the NECB for each year and the NECB of Year 1 was calculated. These values were then summed for the three years after PR, such that 4

Change in NECB af terPR ¼ Sn¼2 ðNECByear n
NECByear 1 Þ in units of g C m
2 per three years

ð8Þ

Taking into account the ‘baseline NECB’ from year 1, the Control and NewMix blocks both lost similar amounts of C (216 and 162 g C m
2 respectively) over the three years after PR. In comparison, the NewRye block lost 416 g C m
2 over three years after accounting for its pre-treatment baseline. This was 200– 254 g C m
2 more than the other two blocks, and probability intervals between the NECBs of the NewRye block and the other two blocks did not overlap (Fig. 4 column C, Table 2). In summary, after taking into account the difference in baseline (year 1) NECBs between blocks, the Control and NewMix swards had lost less C compared to the NewRye sward (Fig. 4 column C). 4. Discussion 4.1. Differences between swards Our experimental design allowed us to take two approaches for determining whether the NECB of a recently sown diverse sward (NewMix) was more positive (more C stored or less C lost) than conventional ryegrass-clover swards via: 1) comparison to a renewed ryegrass-clover pasture (NewRye), and 2) comparison to an unmodified ryegrass-clover pasture (Control). For the first comparison, when pre-existing block differences were taken into account (Eq. (8)) and although it was still a net C source, the NewMix pasture had a higher NECB (by 254 g C m
2 over three years) than the NewRye pasture (Fig. 4 column C,

Table 2). This greater C retention in the NewMix than the NewRye pasture was in agreement with trends reported in previous studies in ungrazed grassland experiments that found increases in soil C stocks with increasing plant species diversity (Tilman et al., 2006; Fornara and Tilman, 2008; Steinbeiss et al., 2008; Cong et al., 2014; Lange et al., 2015). While experiments in ungrazed grasslands have shown soil C stocks can increase with increasing plant diversity, there is much less information on the role of increased plant diversity on soil C in intensively grazed pasture systems. In one of the few diversity experiments in grazed grasslands, Skinner and Dell (2016) found no significant difference in the change in C stocks under two species and five species pastures for the full soil profile to 1 m, although both soil C stocks and root biomass were higher in the five species pasture between 0.1 and 0.3 m depth. Aboveground production in the five species mixture was 31% higher than the two species pasture. In contrast, Skinner et al. (2006) reported C losses from the top 0.05 m under an 11 species pasture but no change under two or three species mixes. This apparent loss of soil C occurred despite higher aboveground production in the 11 species pasture. In studies that found increasing soil C stocks with increasing plant species diversity, gains have been attributed to greater plant production and subsequent C inputs to the soil. In our study there was no difference in above ground pasture production between the NewMix and the NewRye pastures (Fig. 2f), and therefore the higher NECB of the NewMix block compared to the NewRye block was likely due to proportionally greater allocation of C belowground compared to what was occurring in the baseline year. This suggestion of greater belowground allocation was consistent with the significantly higher root biomass and calculated root C inputs reported by McNally et al. (2015) for a recently sown diverse sward compared to a recently sown standard ryegrass-clover sward in the Waikato. Based on the hypothesis of greater root biomass and C inputs in more diverse swards, we had also expected the NewMix pasture to retain more C than the unmodified Control ryegrassclover sward (second comparison). However, the NECBs of the Control and NewMix pastures were almost exactly the same, irrespective of how NECBs were calculated (Fig. 4) and despite higher pasture production in the NewMix block (Fig. 2f). We speculate that this lack of difference between NewMix and Control was associated with the pasture renewal process. Even when excluding the two months directly following PR from the analyses, when C cycling was most impacted by PR (April–May 2013; Rutledge et al., 2017), the NewRye block lost 200 g C m
2 relative to the Control block when pre-treatment block differences were taken into account (Fig. 4 column C). This loss may have been due to effects associated with pasture renewal such as ongoing increased respiration from decomposition of dead roots of the old sward. In contrast to relative C losses at NewRye block compared to the Control block, the NewMix pasture maintained C in the years following PR compared to the unmodified ryegrass-clover pasture. Ongoing decomposition of roots following renewal likely occurred in the same measure at the NewMix block as it would have at the NewRye block, and to maintain the same C balance as the Control block, the NewMix sward would have needed greater C inputs and storage in the soil (Fig. 4, column C) than the NewRye sward. Pasture renewal is common on New Zealand dairy farms (e.g. on average occuring every 10–15 years; Pasture Renewal Charitable Trust, 2013; Kerr et al., 2015) to rejuvenate pasture production, as we also observed (Fig. 2f). Our results suggest that renewing conventional ryegrass-clover pastures to more diverse mixes may potentially lead to greater soil C retention compared to resowing ryegrass-clover pastures, with no discernible difference in pasture production.

Table 2 Components of the carbon balance for the before-pasture renewal year (Year 1; April 2012–March 2013), and for three years after pasture renewal (1 June 2013–31 May 2016; Years 2–4), in g C m
2. The Control block did not undergo pasture renewal. NEP excludes grazer respiration. Values in the square brackets represent the asymmetric 95% probability intervals. Note that probability intervals on NEP only include random uncertainties, because systematic uncertainties were assumed similar between blocks. These uncertainties for NEP and NECB can therefore be used to compare NEP and NECB between sites, but not for comparisons to absolute values. Year 1 (before PR) (g C m
2 y
1) Control a

New Mix b

Year 2 (g C m
2 y
1) New Rye *,c

Control a

Year 3 (g C m
2 y
1)

New Mix b

New Rye c

Control a

226 [+14/
16] 52a [4]

167 [+15/
14] 89b [9]

272 [+17/
15] 71c [+4/
5]

Manure

53a [+10/
8]

80ab [20]

Fertiliser Pasture removed

0
464a [+26/
28]
11a [2]

0
462a [+30/
37]
32b [4]

86b [+17/
16] 0
448a [+24/
27]
14a [2]

119a [+15/
13] 1
583a [+30/
34]
41a [+4/
3]

111a [+21/
24] 1
640a [+33/
38]
72b [7]

137a [+23/
22] 1 -643a [+35/
39] -57c [4]

Dung returned

121a [+3/
11]

128ab [+7/
8]

139a [+2/
13]

163a [+7/
9]

187b [10]

Urine returned

17a [+1/
2]
2
71a [+24/
32]

18a [2]
2
115a [+37/
39]

19a [+1/
2]
2 45*,b [+21/
36]

22a [2]
2 -43a [+30/
36]

26a [2]
2
134b [+40/
42]

201 [15]

Supplement feed imported

Supplement feed removed

Leaching NECB

a

New Rye a

Control a

63 [+14/
15] 29a [+3/
4]

45 [15]

33a [2]

64 [+14/
15] 58b [5]

30a [1]

122b [6]

11
460a [+26/
28] -27a [2]

11
483ab [36]
46b [4]

144c [+16/
14] 11 -537b [+30/
35]
23a [3]

179ab [+8/
10]

135a [+8/
7]

136a [+9/
8]

25a [2]
2
17a [+37/
45]

19a [2]
2
194a [26]

19a [2]
2
123b [+34/
31]

Year 2–4 relative to Year 1y (g C m
2 over three years)

Year 2–4 (g C m
2 over three years)

New Mix b

New Rye c

Control

New Mix

a

a

46a [5]

78 [+15/
16] 42a [8]

178 [+16/
14] 55a [8]

337 [+26/
28] 131a [+8/
7]

30a [1]

26b [1]

0c

5
502a [29]
37a [4]

5
611b [+35/
42]
34a [+6/
7]

4
521a [+28/
30]
44a [6]

181a [+15/
13] 17
1545a [+70/
73]
105a [6]

148a [+9/
8]

194a [+9/
11]

216b [12]

186a [10]

20a [2]
2
147ab [+31/
35]

27a [2]
2
192a [+24/
27]

30a [3]
2
251b [+30/
34]

26 [2]
2
118c [+25/
26]

492a [+19/
22] 68a [6]
5
429a [+57/
66]

308 [26] 189b [+14/
12] 259b [+23/
22] 17
1735b [+80/
86]
151b [+10/
11] 538b [+25/
22] 74a [+6/
7]
5
507a [+70/
71]

New Rye b

513 [+26/
23] 154c [+11/
10] 281b [+24/
26] 17
1701b [+77/
82]
124c [+8/
9] 513ab [+22/
23] 71a [6]
6
282b [+71/
76

Control a

New Mix b

New Rye


265 [+51 /
52]


37 [+52/
53]


230a [+49/
48]


216a [+96/
81]


162a [+113/
112]


416b [+115/
77]

abc: Values in rows within the same year with different lower case letters were considered to be different (based on the 95% probability intervals not overlapping). * Data for the year before PR was incomplete for the NewRye block NEP between Apr 2012 and the start of measurements on 17 August 2012 was assumed to equal the average of the NEP of the two other blocks. y Using Eq. (8) or similar.

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

248 [+13/
14] 18a [3]

66 [15]

14a [2]

115 [+15/
14] 40b [5]

NEP

New Mix

Year 4 (g C m
2 y
1)

319

320

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

A

Control

0.3

B

C

a)

d)

g)

b)

e)

h)

c)

f)

i)

0.2 0.1 0

NewMix

0.3 0.2 0.1 0

NewRye

0.3 0.2 0.1 0 -200

-100

0

100

-600

-400

-200

-600

-400

-200

0

NECB

NECB

NECB relative to Year 1

(gC m -2 y -1 )

(gC m -2 over three years)

(gC m -2 over three years)

200

Fig. 4. The net ecosystem carbon balance (NECB) for each experimental block prior to pasture renewal (April 2012 to April 2013; column A) and the cumulative NECB for three years after pasture renewal (1 June 2013 to 31 May 2016, column B). Column C shows cumulative NECB for the three years after pasture renewal relative to the year before PR (Eq. (8)). In each panel, the thick black lines depicts the calculated best estimate of the NECB, with the bars representing the frequency distribution of the 1000 runs of the Monte Carlo simulations used to estimate the uncertainty on the NECB. Dotted lines show the 95% probability intervals. NECBs were assumed different between years and sites when the probability intervals did not overlap.

4.2. Strengths and limitations One major strength of the NECB approach used in this study is that it can provide information on changes in soil C stocks in pastures more quickly than repeated soil sampling (e.g. Schipper et al., 2014). For example, from our calculated probability bounds, we were able to detect a difference in NECB between treatments of about 140 g C m
2 over three years. In comparison, using a soil sampling approach, Schipper et al. (2014) was only able to just detect a significant difference of 780 g m
2 when resampling 25 Gley soils (to 0.30 m) 28 years (on average) after initial sampling. However, there are a number of other issues that require careful consideration when using the NECB method to test alternative management approaches to increase C storage. These strengths and limitations are discussed below. 4.2.1. Importance of pre-treatment measurements Perhaps the greatest challenge in a study with a before-afterimpact-control design as presented here, is accounting for between-block variability prior to applying treatments. Our interpretation of the potential benefits of the NewMix sward in terms of retaining more C in the soil than the NewRye pasture relies on the assumption that the differences in NECB between blocks observed in Year 1 (before treatments were imposed) continued to occur in the three subsequent years. While there are progressively more studies that compare NECBs of paired (grassland) sites to study the impact of management on the C balance, very few studies include pre-treatment measurements

(e.g. Allard et al., 2007; Ammann et al., 2007; Veenendaal et al., 2007; Skinner, 2008; Hirata et al., 2013; Matsuura et al., 2014; Hunt et al., 2016), or they have very short run-in periods (e.g. Zenone et al., 2011, 4 months). Our findings strongly suggest that a lack of pre-treatment measurements may lead to incorrect conclusions about the size and direction of C balance changes. Had we not accounted for block differences prior to the establishment of treatments, we would have concluded that the resown ryegrassclover pasture (NewRye) led to an greater C retention relative to the new mixed sward pasture (NewMix) and the unmodified Control ryegrass-clover pasture (Fig. 4 column B). It should be noted that even with the applied correction for pre-treatment site differences, potentially considerable unaccounted uncertainty remains because of the assumption that the difference in NECBs observed in the single initial year of this study provided a fully representative underlying non-treatment trajectory for the following years. Large pre-treatment differences in NEP and NECB were not anticipated because we matched treatment blocks with regards to soil types, site history, and position in the landscape as closely as possible when selecting sites. Indeed, Pronger et al. (2016) demonstrated that annual evaporation rates for the three sites were all within 3% of one another. We consider it unlikely that the pre-PR differences in NEP and NECB were a result of systematic differences in grazing management between blocks. We confirmed that the number of grazings per year were very similar between blocks (Fig. S1). While the specific timing and intensity of grazing may have differed between blocks, a Monte Carlo simulation to

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

quantify the impact of variation in grazing on the annual NEP at a farm in the same region, and under similar grazing management as Troughton Farm, showed that grazing variation likely resulted in, at most, 30 g C m
2 y
1 differences in annual NEP (Campbell et al., 2015). This variation is smaller than the observed differences in NEP in Year 1 between the NewRye block and the other two blocks, which were between 47 and 133 g C m
2 y
1. Therefore, we consider that the observed pre-treatment differences in NEP were not artefacts of differences in management, but instead caused by inherent site differences. The NewRye block was located furthest away from the other two blocks which were essentially side-byside (Fig. 1) and it was possible the position of the NewRye block in the landscape was sufficiently different to cause undetected differences in drainage and soil moisture dynamics, which affected C cycling. If the pre-treatment differences in NEP and NECB we observed at Troughton Farm are normal inherent within-farm variability, caution must be exercised that any reported treatment effects are not simply systematic site differences. Large inter-annual variability in NECBs in managed temperate pastures, observed in this study and by others (for example by Jones et al. (2016), Soussana et al. (2007) and Ammann et al. (2007, 2009)), indicate that, ideally, a multi-year pre-treatment period should be allowed for when setting up experiments. However, EC measurements and full quantification of non-CO2-C fluxes (e.g. imported feed) are expensive, and resources often do not allow for replication of treatments or studies with multiple-year run-in periods to account for inherent site differences. Spatial variability may be less of an issue when treatments are imposed on large areas with homogenous past management histories (e.g. large cropping farms) and soils, but may be more of a challenge for intensive rotationally grazed pastoral systems (like the current study) where the unit of management (paddock/field) is generally about the same size as the EC flux footprint. More studies with longer pre-treatment sideby-side comparison periods would be needed to assess whether the within-farm variability observed at Troughton Farm is commonplace. Reducing any (supposedly modest) between block pre- and post-treatment variation in NEP and NECB associated with differences in grazing timing may be possible by synchronising grazing between blocks. However, such alignment will generally not be practical within rotationally grazed systems, because the number of cattle required to graze the flux footprints of multiple EC towers at the same time is prohibitive (e.g. in the current study, 25 ha at a winter stocking rate of 300 cows ha
1 d
1 would require 7500 cows). Further, growth rates of different pasture swards (the ‘treatment’) can vary, and therefore optimum management of each sward will require asynchronous grazing. It may be possible to partially align grazings during some periods of the year (e.g. graze one paddock from each block in a rotation), and this could be explored in future studies. 4.2.2. Potential importance of choice of pasture species While we have some evidence that conversion to a more diverse sward maintained more C in comparison to conversion back to a ryegrass-clover mix, the amount of extra C retained was not as large as we had initially anticipated. Indeed, setting aside the dry matter production gains, there was no C storage advantage of the NewMix sward in comparison to the unaltered Control sward. While increasing plant species diversity may increase production, root biomass and soil C stocks (e.g. Tilman et al., 1996; Mueller et al., 2013; Cong et al., 2014), it has been shown that plant species identity or specific functional traits (e.g. N fixing ability or tap roots) can be at least as important as greater diversity for increasing productivity and soil C stocks (Hooper and Vitousek, 1997; Steinbeiss et al., 2008). Lucerne (or alfalfa, Medicago sativa)

321

in particular is known for deep rooting coupled to good above ground productivity during dry conditions. Pure lucerne crops can be strong CO2 sinks on an annual basis (Gilmanov et al., 2014), and Skinner (2008) found that a lucerne based pasture had higher NECB (by 65 g C m
2 y
1) than a grass based pasture over four years. While lucerne was sown at our NewMix block, it did not establish as well as expected, seemingly outcompeted by grasses in the sward, and never made up more than 2.3% of above ground biomass (Fig. 3). Further, the previously observed increase in late summer-autumn production of a diverse sward in the Waikato (Woodward et al., 2013), did not consistently occur at our NewMix site (Fig. 2f), and all blocks remained net sources of CO2 during these drier periods (Fig. 2c). By comparison, in the Woodward et al. (2013) trial, lucerne contributed 15% of biomass production on average and up to 85% during a dry summer. At this same trial site, McNally et al. (2015) found significantly higher root biomass, and calculated higher C inputs for the diverse sward than the standard ryegrass-clover sward. Had lucerne abundance been higher in the diverse sward of the current study, we speculate that C retention would have been greater. Optimal species mixes to meet both production and environmental (e.g. C sequestration) goals are likely to vary between soil, climate and management regimes. The current experimental design allowed for testing of only one diverse pasture mixture. We would argue that we have not necessarily identified the best species mix for optimising dry matter production and C storage for the existing climate and soil conditions at Troughton Farm. 5. Conclusions Our findings suggest that if the need for pasture renewal (PR) to increase pasture production or quality arises, renewal to a more diverse sward may reduce C loss compared to PR to a ryegrassclover sward. Pasture production did not differ between the two renewed swards (which both had higher production than the unmodified ryegrass-clover sward), and therefore farmers may be open to sowing more diverse swards when renewing pastures. However, there was no evidence that PR to a more diverse sward increased C retention compared to an unmodified ryegrass-clover sward. The identification of the potential benefit of the more diverse sward in terms of retaining more C in the soil than the resown ryegrass-clover pasture was dependent on measurements of NECBs prior to treatments being imposed, so baselines for each experimental block could be determined. Very few carbon balance studies have made such pre-treatment measurements. Because our results indicate that such pre-treatment measurements can have a major influence on the conclusions drawn, they should be made a focus for future studies. In the current study, resources allowed for the testing of only one pasture of increased species diversity. It is possible (or even likely), that the species composition we trialled was not optimal for the site conditions. Identification of the optimal plant species combination (or individual species) for both pasture production and soil C storage, for specific combinations of soil, climate and management regime remains an important research priority. Acknowledgements This research was funded by the New Zealand Agricultural Greenhouse Gas Research Centre. Research program leader David Whitehead is thanked for his support. We received guidance and advice on pasture renewal from Sharon Woodward, Errol Thom and Deanne Waugh from DairyNZ, Katherine Tozer from AgResearch and Ben Trotter from Agricom. Garry Waghorn from DairyNZ provided advice on partitioning of carbon ingested by cows between respiration, dung and urine. Chris Morcom, Janine

322

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323

Ryburn, Dean Sandwell provided support in the field and laboratory. We thank Paul Jamieson from MilkTestNZ for collecting milk samples for us. We appreciate the helpful discussions with our colleague John Hunt about the methodology used in this manuscript. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.agee.2017.01.039. References Acevedo, O.C., Moraes, O.L.L., Degrazia, G.A., Fitzjarrald, D.R., Manzi, A.O., Campos, J. G., 2009. Is friction velocity the most appropriate scale for correcting nocturnal carbon dioxide fluxes? Agric. Forest Meteorol. 149, 1–10. Allard, V., Soussana, J.F., Falcimagne, R., Berbigier, P., Bonnefond, J.M., Ceschia, E., D'Hour, P., Henault, C., Laville, P., Martin, C., Pinares-Patino, C., 2007. The role of grazing management for the net biome productivity and greenhouse gas budget (CO2, N2O and CH4) of semi-natural grassland. Agric. Ecosyst. Environ. 121, 47– 58. Ammann, C., Flechard, C.R., Leifeld, J., Neftel, A., Fuhrer, J., 2007. The carbon budget of newly established temperate grassland depends on management intensity. Agric. Ecosyst. Environ. 121, 5–20. Ammann, C., Spirig, C., Leifeld, J., Neftel, A., 2009. Assessment of the nitrogen and carbon budget of two managed temperate grassland fields. Agric. Ecosyst. Environ. 133, 150–162. Asner, G.P., Elmore, A.J., Olander, L.P., Martin, R.E., Harris, A.T., 2004. Grazing systems, ecosystem responses, and global change. Annu. Rev. Environ. Resour. 29, 261–299. Aurela, M., Tuovinen, J.-P., Laurila, T., 2001. Net CO2 exchange of a subarctic mountain birch ecosystem. Theor. Appl. Climatol. 70, 135–148. Batjes, N.H., 1996. Total carbon and nitrogen in the soils of the world. Eur. J. Soil Sci. 47, 151–163. Beare, M.H., McNeill, S.J., Curtin, D., Parfitt, R.L., Jones, H.S., Dodd, M.B., Sharp, J., 2014. Estimating the organic carbon stabilisation capacity and saturation deficit of soils: a New Zealand case study. Biogeochemistry 120, 71–87. Beukes, P.C., Gregorini, P., Romera, A.J., Woodward, S.L., Khaembah, E.N., Chapman, D.F., Nobilly, F., Bryant, R.H., Edwards, G.R., Clark, D.A., 2014. The potential of diverse pastures to reduce nitrogen leaching on New Zealand dairy farms. Anim. Prod. Sci. 54, 1971–1979. Burba, G., Schmidt, A., Scott, R.L., Nakai, T., Kathilankal, J., Fratini, G., Hanson, C., Law, B., McDermitt, D.K., Eckles, R., Furtaw, M., Velgersdyk, M., 2012. Calculating CO2 and H2O eddy covariance fluxes from an enclosed gas analyzer using an instantaneous mixing ratio. Global Change Biol. 18, 385–399. Caldeira, K., Morgan, M.G., Dennis Baldocchi Brewer, P.G., Chen, C.-T.A., Gert-Jan Nabuurs Nakicenovic, N., Robertson, G.P., 2004. A portfolio of carbon management options. In: Field, C.B., Raupach, M.R. (Eds.), The Global Carbon Cycle: Integrating Humans, Climate and the Natural World. Island Press, Washington, D.C. Campbell, D.I., Wall, A.M., Nieveen, J.P., Schipper, L.A., 2015. Variations in CO2 exchange for dairy farms with year-round rotational grazing on drained peatlands. Agric. Ecosyst. Environ. 202, 68–78. Chapin, F., Woodwell, G., Randerson, J., Rastetter, E., Lovett, G., Baldocchi, D., Clark, D., Harmon, M., Schimel, D., Valentini, R., Wirth, C., Aber, J., Cole, J., Goulden, M., Harden, J., Heimann, M., Howarth, R., Matson, P., McGuire, A., Melillo, J., Mooney, H., Neff, J., Houghton, R., Pace, M., Ryan, M., Running, S., Sala, O., Schlesinger, W., Schulze, E.D., 2006. Reconciling carbon-cycle concepts, terminology, and methods. Ecosystems 9, 1041–1050. Cole, C.V., Duxbury, J., Freney, J., Heinemeyer, O., Minami, K., Mosier, A., Paustian, K., Rosenberg, N., Sampson, N., Sauerbeck, D., Zhao, Q., 1997. Global estimates of potential mitigation of greenhouse gas emissions by agriculture. Nutr. Cycl. Agroecosyst. 49, 221–228. Conant, R.T., Paustian, K., Elliott, E.T., 2001. Grassland management and conversion into grassland: effects on soil carbon. Ecol. Appl. 11, 343–355. Cong, W.-F., van Ruijven, J., Mommer, L., De Deyn, G.B., Berendse, F., Hoffland, E., 2014. Plant species richness promotes soil carbon and nitrogen stocks in grasslands without legumes. J. Ecol. 102, 1163–1170. Crush, J.R., Waller, J.E., Care, D.A., 2005. Root distribution and nitrate interception in eleven temperate forage grasses. Grass Forage Sci. 60, 385–392. DairyNZ, 2012. Facts and Figures for New Zealand Dairy Farmers. DairyNZ. Dragoni, D., Schmid, H., Grimmond, C.B., Loescher, H.W., 2007. Uncertainty of annual net ecosystem productivity estimated using eddy covariance flux measurements. J. Geophys. Res. 112. EddyPro1 Version 5 [Computer software], 2015. Infrastructure for Measurements of the European Carbon Cycle Consortium. NE.LI –COR, Inc, Lincoln. Felber, R., Bretscher, D., Münger, A., Neftel, A., Ammann, C., 2016. Determination of the carbon budget of a pasture: effect of system boundaries and flux uncertainties. Biogeosciences 13, 2959–2969. Finn, D., Page, K., Catton, K., Kienzle, M., Robertson, F., Armstrong, R., Dalal, R., 2016. Ecological stoichiometry controls the transformation and retention of plant-

derived organic matter to humus in response to nitrogen fertilisation. Soil Biol. Biochem. 99, 117–127. Fornara, D.A., Tilman, D., 2008. Plant functional composition influences rates of soil carbon and nitrogen accumulation. J. Ecol. 96, 314–322. Fornara, D.A., Tilman, D., 2012. Soil carbon sequestration in prairie grasslands increased by chronic nitrogen addition. Ecology 93, 2030–2036. Fornara, D.A., Banin, L., Crawley, M.J., 2013. Multi-nutrient vs. nitrogen-only effects on carbon sequestration in grassland soils. Glob. Change Biol. 19, 3848–3857. Fratini, G., Ibrom, A., Arriga, N., Burba, G., Papale, D., 2012. Relative humidity effects on water vapour fluxes measured with closed-path eddy-covariance systems with short sampling lines. Agric. Forest Meteorol. 165, 53–63. Freibauer, A., Rounsevell, M.D.A., Smith, P., Verhagen, J., 2004. Carbon sequestration in the agricultural soils of Europe. Geoderma 122, 1–23. Gilmanov, T.G., Baker, J.M., Bernacchi, C.J., Billesbach, D.P., Burba, G.G., Castro, S., Chen, J., Eugster, W., Fischer, M.L., Gamon, J.A., Gebremedhin, M.T., Glenn, A.J., Griffis, T.J., Hatfield, J.L., Heuer, M.W., Howard, D.M., Leclerc, M.Y., Loescher, H. W., Marloie, O., Meyers, T.P., Olioso, A., Phillips, R.L., Prueger, J.H., Skinner, R.H., Suyker, A.E., Tenuta, M., Wylie, B.K., 2014. Productivity and carbon dioxide exchange of leguminous crops: estimates from flux tower measurements. Agron. J. 106, 545–559. Goulden, M.L., Munger, J.W., Fan, S.M., Daube, B.C., Wofsy, S.C., 1996. Measurements of carbon sequestration by long-term eddy covariance: methods and a critical evaluation of accuracy. Global Change Biol. 2, 169–182. Hewitt, A.E., 1998. New Zealand Soil Classification, 2nd edition Manaaki-Whenua Press, Lincoln, New Zealand. Hill Laboratories, n.d. Metabolisable Energy Testing at Hill Laboratories. Accessed 7 June 2016 from: http://hill-laboratories.com/file/fileid/45374. Hirata, R., Miyata, A., Mano, M., Shimizu, M., Arita, T., Kouda, Y., Matsuura, S., Niimi, M., Saigusa, T., Mori, A., Hojito, M., Kawamura, O., Hatano, R., 2013. Carbon dioxide exchange at four intensively managed grassland sites across different climate zones of Japan and the influence of manure application on ecosystem carbon and greenhouse gas budgets. Agric. Forest Meteorol. 177, 57–68. Hollinger, D.Y., Richardson, A.D., 2005. Uncertainty in eddy covariance measurements and its application to physiological models. Tree Physiol. 25, 873–885. Hooper, D.U., Vitousek, P.M., 1997. The effects of plant composition and diversity on ecosystem processes. Science 277, 1302–1305. Hunt, J.E., Laubach, J., Barthel, M., Fraser, A., Phillips, R.L., 2016. Carbon budgets for an irrigated intensively grazed dairy pasture and an unirrigated winter-grazed pasture. Biogeosciences 13, 2927–2944. Janzen, H.H., 2004. Carbon cycling in earth systems – a soil science perspective. Agric. Ecosyst. Environ. 104, 399–417. Jones, S.K., Helfter, C., Anderson, M., Coyle, M., Campbell, C., Famulari, D., Di Marco, C., van Dijk, N., Topp, C.F.E., Kiese, R., Kindler, R., Siemens, J., Schrumpf, M., Kaiser, K., Nemitz, E., Levy, P., Rees, R.M., Sutton, M.A., Skiba, U.M., 2016. The nitrogen, carbon and greenhouse gas budget of a grazed: cut and fertilised temperate grassland. Biogeosci. Discuss. 2016, 1–55. Kell, D.B., 2012. Large-scale sequestration of atmospheric carbon via plant roots in natural and agricultural ecosystems: why and how. Philos. T. Roy. Soc. B. 367, 1589–1597. Kerr, G., Brown, J., Kilday, T., Steven, D., 2015. A more quantitative approach to pasture renewal. Journal of New Zealand Grasslands 77, 251–258. Klumpp, K., Tallec, T., Guix, N., Soussana, J.-F., 2011. Long-term impacts of agricultural practices and climatic variability on carbon storage in a permanent pasture. Glob. Change Biol. 17, 3534–3545. Kormann, R., Meixner, F.X., 2001. An analytical footprint model for non-neutral stratification. Boundary Layer Meteorol. 99, 207–224. LI-COR Inc, 2015. EddyPro1 5 Help and User's Guide. LI-COR, Inc., Lincoln, NE. https://www.licor.com/env/help/EddyPro5/index.htm. Lal, R., Delgado, J.A., Groffman, P.M., Millar, N., Dell, C., Rotz, A., 2011. Management to mitigate and adapt to climate change. Journal of soil and water conservation 66, 276–285. LandcareResearch, 2015. Smap Soil Report – Te Puninga silt loam. Landcare Research, pp. 3. http://smap.landcareresearch.co.nz/home. Lange, M., Eisenhauer, N., Sierra, C.A., Bessler, H., Engels, C., Griffiths, R.I., MelladoVazquez, P.G., Malik, A.A., Roy, J., Scheu, S., Steinbeiss, S., Thomson, B.C., Trumbore, S.E., Gleixner, G., 2015. Plant diversity increases soil microbial activity and soil carbon storage. Nat. Commun. 6, 6707. Lohila, A., Aurela, M., Tuovinen, J.-P., Laurila, T., 2004. Annual CO2 exchange of a peat field growing spring barley or perennial forage grass. J. Geophys. Res. 109, D18116 (18110.11029/12004JD004715). Lu, M., Zhou, X.H., Luo, Y.Q., Yang, Y.H., Fang, C.M., Chen, J.K., Li, B., 2011. Minor stimulation of soil carbon storage by nitrogen addition: a meta-analysis. Agric. Ecosyst. Environ. 140, 234–244. Macdonald, K.A., Penno, J.W., Lancaster, J.A.S., Roche, J.R., 2008. Effect of stocking rate on pasture production, milk production, and reproduction of dairy cows in pasture-based systems. J. Dairy Sci. 91, 2151–2163. Mason, N.W.H., Orwin, K., Lambie, S., Woodward, S.L., McCready, T., Mudge, P.L., 2016. Leaf economics spectrum–productivity relationships in intensively grazed pastures depend on dominant species identity. Ecol. Evol. 6, 3079–3091. Matsuura, S., Miyata, A., Mano, M., Hojito, M., Mori, A., Kano, S., Sasaki, H., Kohyama, K., Hatano, R., 2014. Seasonal carbon dynamics and the effects of manure application on carbon budget of a managed grassland in a temperate, humid region in Japan. Grassland Sci. 60, 76–91.

S. Rutledge et al. / Agriculture, Ecosystems and Environment 239 (2017) 310–323 Mauder, M., Foken, T., 2004. Documentation and Instruction Manual of the Eddy Covariance Software Package TK2. Arbeitsergebn, Univ Bayreuth, Abt Mikrometeorol, pp. 42 (ISSN 16148916). McCall, D.G., Bishop-Hurley, G.J., 2003. A pasture growth model for use in a wholefarm dairy production model. Agric. Syst. 76, 1183–1205. McLeod, M., 1992. Soils of Part Northern Matamata County, North Island, New Zealand.DSIR Land Resources Scientific Report, ISSN1170-5965. N.Z. Department of Scientific and Industrial Research, Lower Hutt, pp. 96. 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 ryegrassclover pasture: implications for soil carbon sequestration. Plant Soil 392, 289– 299. Milne, E., Banwart, S.A., Noellemeyer, E., Abson, D.J., Ballabio, C., Bampa, F., Bationo, A., Batjes, N.H., Bernoux, M., Bhattacharyya, T., Black, H., Buschiazzo, D.E., Cai, Z., Cerri, C.E., Cheng, K., Compagnone, C., Conant, R., Coutinho, H.L.C., de Brogniez, D., Balieiro, F.d.C., Duffy, C., Feller, C., Fidalgo, E.C.C., da Silva, C.F., Funk, R., Gaudig, G., Gicheru, P.T., Goldhaber, M., Gottschalk, P., Goulet, F., Goverse, T., Grathwohl, P., Joosten, H., Kamoni, P.T., Kihara, J., Krawczynski, R., La Scala Jr, N., Lemanceau, P., Li, L., Li, Z., Lugato, E., Maron, P.-A., Martius, C., Melillo, J., Montanarella, L., Nikolaidis, N., Nziguheba, G., Pan, G., Pascual, U., Paustian, K., Piñeiro, G., Powlson, D., Quiroga, A., Richter, D., Sigwalt, A., Six, J., Smith, J., Smith, P., Stocking, M., Tanneberger, F., Termansen, M., van Noordwijk, M., van Wesemael, B., Vargas, R., Victoria, R.L., Waswa, B., Werner, D., Wichmann, S., Wichtmann, W., Zhang, X., Zhao, Y., Zheng, J., Zheng, J., 2015. Soil carbon, multiple benefits. Environ. Dev. 13, 33–38. Moncrieff, J.B., Massheder, J.M., de Bruin, H., Elbers, J., Friborg, T., Heusinkveld, B., Kabat, P., Scott, S., Soegaard, H., Verhoef, A., 1997. A system to measure surface fluxes of momentum sensible heat, water vapour and carbon dioxide. J. Hydrol. 188-189, 589–611. Morgenstern, K., Black, T.A., Humphreys, E.R., Griffis, T.J., Drewitt, G.B., Cai, T.B., Nesic, Z., Spittlehouse, D.L., Livingstone, N.J., 2004. Sensitivity and uncertainty of the carbon balance of a Pacific Northwest Douglas-fir forest during an El Nino La Nina cycle. Agric. Forest Meteorol. 123, 201–219. Mueller, K.E., Tilman, D., Fornara, D.A., Hobbie, S.E., 2013. Root depth distribution and the diversity-productivity relationship in a long-term grassland experiment. Ecology 94, 787–793. Nobilly, F., Bryant, R.H., McKenzie, B.A., Edwards, G.R., 2013. Productivity of rotationally grazed simple and diverse pasture mixtures under irrigation in Canterbury. Proc. New Zeal. Grassland Assoc. 75, 165–172. O’Hagan, A., Buck, C.E., Daneshkhah, A., Eiser, J.R., Garthwaite, P.H., Jenkinson, D.J., Oakley, J.E., Rakow, T., 2006. Uncertain Judgements: Eliciting Experts' Probabilities. John Wiley and Sons, Hoboken, New Jersey. Papale, D., Valentini, R., 2003. A new assessment of European forests carbon exchanges by eddy fluxes and artificial neural network spatialization. Global Change Biol. 9, 525–535. Pasture Renewal Charitable Trust, 2013. Results from Statistics NZ Agricultural Census for 2011–12 Reported by the Pasture Renewal Charitable Trust. . http:// www.pasturerenewal.org.nz/article/138.html. Paustian, K., Lehmann, J., Ogle, S., Reay, D., Robertson, G.P., Smith, P., 2016. Climatesmart soils. Nature 532, 49–57. Pembleton, K.G., Tozer, K.N., Edwards, G.R., Jacobs, J.L., Turner, L.R., 2015. Simple versus diverse pastures: opportunities and challenges in dairy systems. Anim. Prod. Sci. 55, 893–901. Pronger, J., Campbell, D.I., Clearwater, M.J., Rutledge, S., Wall, A.M., Schipper, L.A., 2016. Low spatial and inter-annual variability in evaporation from a year round intensively grazed temperate pasture system. Agric. Ecosyst. Environ. 232, 46–58. Rasse, D.P., Rumpel, C., Dignac, M.F., 2005. Is soil carbon mostly root carbon? Mechanisms for a specific stabilisation. Plant Soil 269, 341–356. Rutledge, S., Mudge, P.L., Wallace, D.F., Campbell, D.I., Woodward, S.L., Wall, A.M., Schipper, L.A., 2014. CO2 emissions following cultivation of a temperate permanent pasture. Agric. Ecosyst. Environ. 184, 21–33. 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. Rutledge, S., Wall, A.M., Mudge, P.L., Troughton, B., Campbell, D.I., Pronger, J., Joshi, C., Schipper, L.A., 2017. The carbon balance of a temperate grassland part II: The impact of pasture renewal via direct drilling. Agric. Ecosyst. Environ. 239, 132– 142. RuzJerez, B.E., Ball, P.R., White, R.E., Gregg, P.E.H., 1991. Comparison of an herbal ley with a ryegrass-white clover pasture and pure ryegrass sward receiving fertiliser nitrogen. Proc. New Zeal. Grassland Assoc. 53, 225–230. Schipper, L.A., Parfitt, R.L., Fraser, S., Littler, R.A., Baisden, W.T., Ross, C., 2014. Soil order and grazing management effects on changes in soil C and N in New Zealand pastures. Agric. Ecosyst. Environ. 184, 67–75.

323

Schon, N.L., Mackay, A.D., Gray, R.A., Dodd, M.B., van Koten, C., 2015. Quantifying dung carbon incorporation by earthworms in pasture soils. Eur. J. Soil Sci. 66, 348–358. Skinner, R.H., Dell, C.J., 2016. Yield and soil carbon sequestration in grazed pastures sown with two or five forage species. Crop Sci. 56, 2035–2044. Skinner, R.H., Sanderson, M.A., Tracy, B.F., Dell, C.J., 2006. Above- and belowground productivity and soil carbon dynamics of pasture mixtures. Agron. J. 98, 320– 326. Skinner, R.H., 2008. High biomass removal limits carbon sequestration potential of mature temperate pastures. J. Environ. Qual. 37, 1319–1326. Smith, P., Martino, D., Cai, Z., Gwary, D., Janzen, H., Kumar, P., McCarl, B., Ogle, S., O'Mara, F., Rice, C., Scholes, B., Sirotenko, O., Howden, M., McAllister, T., Pan, G., Romanenkov, V., Schneider, U., Towprayoon, S., Wattenbach, M., Smith, J., 2008. Greenhouse gas mitigation in agriculture. Phil. Trans. R. Soc. B : Biol. Sci. 363, 789–813. Smith, P., House, J.I., Bustamante, M., Sobocka, J., Harper, R., Pan, G., West, P.C., Clark, J.M., Adhya, T., Rumpel, C., Paustian, K., Kuikman, P., Cotrufo, M.F., Elliott, J.A., McDowell, R., Griffiths, R.I., Asakawa, S., Bondeau, A., Jain, A.K., Meersmans, J., Pugh, T.A.M., 2016. Global change pressures on soils from land use and management. Glob. Change Biol. 22, 1008–1028. Smith, P., 2008. Land use change and soil organic carbon dynamics. Nutr. Cycl. Agroecosyst. 81, 169–178. Soussana, J.F., Allard, V., Pilegaard, K., Ambus, P., Amman, C., Campbell, C., Ceschia, E., Clifton-Brown, J., Czobel, S., Domingues, R., Flechard, C., Fuhrer, J., Hensen, A., Horvath, L., Jones, M., Kasper, G., Martin, C., Nagy, Z., Neftel, A., Raschi, A., Baronti, S., Rees, R.M., Skiba, U., Stefani, P., Manca, G., Sutton, M., Tuba, Z., Valentini, R., 2007. Full accounting of the greenhouse gas (CO2 N2O, CH4) budget of nine European grassland sites. Agric. Ecosyst. Environ. 121, 121–134. Soussana, J.F., Tallec, T., Blanfort, V., 2010. Mitigating the greenhouse gas balance of ruminant production systems through carbon sequestration in grasslands. Animal 4, 334–350. Sparling, G.P., Chibnall, E.J., Pronger, J., Rutledge, S., Wall, A.M., Campbell, D.I., Schipper, L.A., 2016. Estimates of annual leaching losses of dissolved organic carbon from pastures on Allophanic Soils grazed by dairy cattle Waikato, New Zealand. New Zeal. J. Agr. Res. 59, 32–49. Steinbeiss, S., Bessler, H., Engels, C., Temperton, V.M., Buchmann, N., Roscher, C., Kreutziger, Y., Baade, J., Habekost, M., Gleixner, G., 2008. Plant diversity positively affects short-term soil carbon storage in experimental grasslands. Glob. Change Biol. 14, 2937–2949. Steinfeld, H., Gerber, P., Wassenaar, T., Castel, V., Rosales, M., de Haan, C., 2006. Livestock’s Long Shadow: Environmental Issues and Options. Food and Agriculture Organization of the United Nations, Rome, pp. 298. Tate, K., Wilde, R.H., Giltrap, D.J., Baisden, W.T., Saggar, S., Trustrum, N.A., Scott, N.A., Barton, J.P., 2005. Soil organic carbon stocks and flows in New Zealand System development, measurement and modelling. Can. J. Soil Sci. 85, 481–489. Tilman, D., Wedin, D., Knops, J., 1996. Productivity and sustainability influenced by biodiversity in grassland ecosystems. Nature 379, 718–720. Tilman, D., Hill, J., Lehman, C., 2006. Carbon-negative biofuels from low-input highdiversity grassland biomass. Science 314, 1598–1600. Totty, V.K., Greenwood, S.L., Bryant, R.H., Edwards, G.R., 2013. Nitrogen partitioning and milk production of dairy cows grazing simple and diverse pastures. J. Dairy Sci. 96, 141–149. Veenendaal, E.M., Kolle, O., Leffelaar, P.A., Schrier-Uijl, A.P., Van Huissteden, J., Van Walsem, J., Möller, F., Berendse, F., 2007. CO2 exchange and carbon balance in two grassland sites on eutrophic drained peat soils. Biogeosciences 4, 1027– 1040. Waghorn, G.C., Burke, J.L., Kolver, E.S., 2007. Principles of feeding value. In: Rattray, P. V., Brookes, I.M., Nicol, A.M. (Eds.), Pasture and Supplements for Grazing Animals. New Zealand Society of Animal Production, Hamilton, New Zealand, pp. 35–59. Wohlfahrt, G., Hammerle, A., Haslwanter, A., Bahn, M., Tappeiner, U., Cernusca, A., 2008. Seasonal and inter-annual variability of the net ecosystem CO2 exchange of a temperate mountain grassland: effects of weather and management. J. Geophys. Res. 113 doi:http://dx.doi.org/10.1029/2007JD009286. Woodward, S.J.R., Barker, D.J., Zyskowski, R.F., 2001. A practical model for predicting soil water deficit in New Zealand pastures. New Zeal. J. Agr. Res. 44, 91–109. Woodward, S.L., Waugh, C.D., Roach, C.G., Fynn, D., Phillips, J., 2013. Are diverse species mixtures better pastures for dairy farming? Proc. New Zeal. Grassland Assoc. 75, 79–84. Zenone, T., Chen, J., Deal, M., Wilske, B., Xu, J., Bhardwaj, A., Hamilton, S., Philip, R.G., 2011. CO2 fluxes of transitional bioenergy crops: effect of land conversion during the first year of cultivation. Global Change Biol.- Bioenergy 3, 401–412.