Application of DNDC biogeochemistry model to estimate greenhouse gas emissions from Italian agricultural areas at high spatial resolution

Agriculture, Ecosystems and Environment 139 (2010) 546–556

Contents lists available at ScienceDirect

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

Application of DNDC biogeochemistry model to estimate greenhouse gas emissions from Italian agricultural areas at high spatial resolution Emanuele Lugato a,∗ , Michel Zuliani b , Giorgio Alberti b , Gemini Delle Vedove b , Beniamino Gioli a , Franco Miglietta a , Alessandro Peressotti b a b

Istituto di Biometeorologia, Consiglio Nazionale delle Ricerche, via Caproni 8, 50145 Firenze, Italy Dipartimento di Scienze Agrarie e Ambientali, Università di Udine, viale delle Scienze 208, 33100 Udine, Italy

a r t i c l e

i n f o

Article history: Received 29 May 2010 Received in revised form 10 September 2010 Accepted 24 September 2010 Available online 25 October 2010 Keywords: DNDC model Greenhouse gases emission Spatial simulations Italian agricultural areas N2 O-N

a b s t r a c t Agriculture is responsible for 13.5% of total greenhouse gas (GHG) emissions (IPCC, 2007), in particular accounting for more than 80% of global anthropogenic nitrous oxide (N2 O) emissions. However uncertainties in GHG inventories are still high as biogeochemical cycles are strongly influenced by climatic and environmental conditions and also depend on local agricultural practices. In order to improve the GHG balance assessment in cropland, higher order methods (Tier 3) are recommended such as, for example, model applications at spatial level. In the present study, a GIS-model integration was performed with the aim of improving the national GHG inventory in Italy. The de-nitrification decomposition (DNDC) model, chosen due to its ability to simulate carbon (C) and nitrogen (N) cycles, has been tested against measured data coming from eddy-covariance stations and soil flux chambers (for CO2 and N2 O fluxes) belonging to Carbo-Italy network. Despite the varying site specific parameters, the results confirmed the ability of the model to represent the real C balance in irrigated maize crop under both conventional and minimum tillage. Modelled N2 O emissions fitted the measured data well, but the corresponding emission factor from fertilizers was much lower than the IPCC default (0.008 vs 0.0125 kg N2 O-N kg−1 N, respectively). A platform of simulation was then built to run DNDC for the entire national territory, linking the model with geographical databases. To implement the model, a high spatial resolution grid (1 km × 1 km) was adopted in order to develop a tool that could be used by local administrations and easily upload information at high spatial resolution (e.g. remote sensed information). A tree management (e.g. a combination of different management and land use) was also built to simulate crops with a ‘business as usual’ (BaU) scenario and with alternative management practices (AMP) and potentially create infinite combinations in each cell simply by varying a relative land use weight. Although the total area under agriculture has not been simulated so far, this platform of simulation appears promising to improve the national GHG inventory and derive C credits from the agricultural sector. © 2010 Elsevier B.V. All rights reserved.

1. Introduction After ratification of the Kyoto protocol (UNFCCC – http://unfccc.int), great efforts are being required from the participating nations to reduce greenhouse gas (GHG) emissions in the first commitment period (2008–2012). For most Annex 1 countries, including those in Europe, agricultural soils are a major source of nitrous oxide (N2 O) (Lokupitiya and Paustian, 2006). In Italy, the overall agricultural sector is a source of GHGs emitting 6.7% of the total emissions, with an amount of 37.2 Mt of CO2 equiv. (Romano et al., 2009), of which more than half (21.6 Mt of CO2 equiv.) is emitted as N2 O from soils. Although the national

∗ Corresponding author. Tel.: +39 055 3033749; fax: +39 055 308910. E-mail address: [email protected] (E. Lugato). 0167-8809/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.agee.2010.09.015

methodology is in accordance with Tier 1 and 2 approaches proposed by the IPCC, still empirical emission factors are used to assess the emission from fertilizer (0.0125 kg N2 O-N kg−1 N from synthetic fertilizers). Disaggregated data at sub-national level, including models and inventory measurement systems required by higher order methods (i.e. Tier 3), are not available so far in Italy and comparisons with the other two approaches cannot be performed at the moment. Instead, some inventories, based on emission factors and regional regression equations derived from all available measurements, have been carried out at European level (EU 15) (Freibauer, 2003). The Italian arable soils emit between 2 and 6 kg N2 O-N ha−1 yr−1 corresponding to 23 Mt of CO2 equiv., a value slightly higher than the national estimate of 19.3 Mt of CO2 equiv. in 1995 (Romano et al., 2009). Recently, the de-nitrification decomposition (DNDC) model was linked to an economic model to estimate nitrogen and carbon losses

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

from arable soils in Europe (Leip et al., 2008). The country-scale aggregated N balance reported for Italy has a total N input of 193 kg ha−1 and N2 O emission of 3.8 kg N ha−1 , which is equivalent to an emission factor of 0.0197 kg N2 O-N kg−1 N from all N inputs (synthetic and organic fertilizers, N fixation and deposition). Despite the large soil organic carbon (SOC) pool in the agricultural soils and the recent institutionalization of the ‘National Registry for Carbon sinks’ by a Ministerial Decree on 1st April, 2008, the last Italian greenhouse gas Inventory did not report CO2 emissions from the agricultural sector. The main reason, as stated in the report, was that “it wasn’t possible to point out different sets of relative stock change factors [FLU (land use), FMG (management), FI (input factor)] for the period 1990–2007 under investigation; therefore, as no management changes can be documented, resulting change in carbon stock has been reported as zero”. Although a detailed spatial assessment of SOC stock has not been made yet in Italy, some data at field level coming from long-term experiments are available (Lugato et al., 2006, 2007; Morari et al., 2006; Triberti et al., 2008). These results generally confirm the steady state condition for SOC, somewhat justifying the exclusion of the CO2 in the agriculture GHG balance. On the other hand, the lack of a national platform that could link the land use/management factors with the SOC changes could deny the recognition of C credits through sequestration, as contemplated by the Article 3.4 of the Kyoto protocol (Smith, 2005). Some modelling work (Lugato and Berti, 2008) evidenced the positive effect of some management practices to sequester CO2 in north-east Italy, even in relation to different climate change scenarios. However, the Article 3.4 of the Kyoto protocol requires a good level of verifiability of higher order methods (Tier 3), that should also be commonly agreed (IPCC, 2007). In this context, the link between international recognized biogeochemical models and spatial databases are becoming a powerful and acceptable tools for the national GHG assessment and scenario analysis (Easter et al., 2007; Li et al., 2010; Smith et al., 2010). With the aim of improving the national GHG inventory in Italy toward a Tier 3 approach, a GIS-model integration was performed in this paper. A platform of simulation with a grid of 1 km × 1 km was built, allowing a high resolution assessment of the main GHGs (CO2 , N2 O and CH4 ). This platform was also designed to easily simulate the potential effects of land use/management changes on C sequestration with the aim to guide the future agricultural policies that could potentially generate C credits from the agricultural sector.

2. Materials and methods 2.1. Platform of simulation Many modelling systems coupling a biogeochemical model and spatially distributed data, often organized in geographic information systems (GIS), have been developed recently for conducting regional-scale soil carbon or GHG inventories (Easter et al., 2007; Smith et al., 2010). When operating at spatial level, the region is generally divided into smaller cells that are frequently aggregated in uniform soil, climate and land use units to reduce the number of combinations. Our approach, in developing the platform of simulation, was instead to adopt a grid of 1 km × 1 km covering the entire country and to associate the information necessary for the simulations to each cell. This strategy is very time consuming since we considered more than 170,000 cells, which could have been eventually aggregated to larger homogenous units considering that all the geographical databases with such a small scale are not available. However, we prefer to maintain a detailed level of spatialization in order to:

547

(1) develop a tool that could be used by local administrations in case they could provide information on smaller areas; (2) easily implement layers at higher spatial resolution about soil, climate and land use derived from remote sensing. Thus, we developed a high spatial resolution platform coupling spatial information, partly derived from a satellite approach, with the biogeochemical model DNDC. This is a widely used model, often adopted as a reference in international projects such as NitroEurope, but also used to estimate GHG emissions at regional or national scales (Giltrap et al., 2010). 2.2. Site specific model calibration The DNDC (Denitrification–Decomposition) model is a processoriented model simulating temporal changes in the levels of soil organic carbon (SOC) as well as nitrogen cycling (Li et al., 1992, 2006). It consists of two components: the first one includes soil climate, crop growth and decomposition sub-models and predicts soil environmental factors driven by ecological factors such as climate, soil, vegetation and anthropogenic activity; the second component utilizes the resulting soil environmental factors as inputs to nitrification, denitrification and fermentation sub-models. DNDC has been independently tested worldwide by a wide range of researchers during the past decades (Smith et al., 1997; Cai et al., 2003; Beheydt et al., 2007; Hastings et al., 2010). Before the spatial application, the model (version 92) was tested with measured field data coming from some experimental sites of the CarboItaly network. In particular, DNDC was run at the Beano site (46◦ 00 N 13◦ 01 E) where irrigated maize (Zea mays L.) was cultivated during the last thirty years and the soil is usually tilled (plough at 0–35 cm depth). Soil can be classified as a Chromi-Endoskeletic Cambisol (FAO, 2006) with the following characteristics in the 0–30 cm horizon: total SOC = 48.4 ± 8.5 t C ha−1 , total N = 4.2 ± 1.1 t N ha−1 , soil bulk density = 1.25 ± 0.15 g cm−3 , soil field capacity = 23% (v/v), wilting point = 12% (v/v) and pH = 7.1 ± 0.02. In December 2006 an experiment about land use change (continuous maize vs alfalfa; Alberti et al., 2010) and agricultural management (tillage vs minimum tillage; Alberti et al., in preparation) was set up. Environmental variables (i.e. air temperature and humidity, rain, soil water content, incoming radiation, etc.) and net ecosystem exchange (NEE) were measured at half-hour time step at this field using a full equipped weather station and an eddy covariance station, respectively (Alberti et al., 2010). Furthermore, six plots (3 blocks × 2 treatments), 10 m × 10 m each, were established for detailed process level studies on C dynamics, crop growth and biomass partitioning, soil respiration, belowground C deposition and crop residue decomposition. The applied treatments in these plots were conventional tillage (CT) and minimum tillage (MT) on maize. Both treatments received the same amount of N mineral fertilizer (Alberti et al., 2010): 54 kg N ha−1 at sowing and two top dressing applications (between the end of April and June) applying totally 203, 359 and 420 kg N ha−1 in 2007, 2008 and 2009 respectively. Continuous soil respiration measurements coupled with continuous soil temperature and soil moisture measurements were performed using three automated soil respiration systems with six chambers each (Delle Vedove et al., 2007; Alberti et al., 2010). Three chambers were placed in each plot: two chambers were used to estimate total soil respiration and one was used to estimate heterotrophic respiration (Rh) on a root exclusion subplot. Soil below this last chamber was isolated with a root exclusion stainlesssteel cylinder opened on both ends (32 cm diameter, 40 cm height; Alberti et al., 2010). Soil CO2 flux was measured every 2 h. Grain yield was assessed by weighing the amount of grain removed from

548

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

Fig. 1. (a) Italian administrative regions, (b) annual precipitation (mm) and (c) mean air temperature (◦ C) based on satellite Multi sensor Precipitation Estimate-MPE and Land Surface Temperature-STT satellite products.

each plot and above- and below-ground dry biomass and leaf area index (LAI) were measured three times during the growing season by destructive sampling. Measurements of N2 O were performed manually and periodically (twice per week till one month after fertilization, once per week or every two weeks thereafter) using the soil respiration chambers. Gas samples were withdrawn using a syringe at time zero and after 30 and 60 min after closure and stored in 20 ml evacuated airtight vials. N2 O concentrations in the gas samples were then determined using a gas chromatograph in the laboratory and the rate of change (dN2 O/dt) was calculated from the linear regression of N2 O concentrations within the chamber over time. The flux was then computed knowing air temperature and air pressure at the time of the measurement and total volume and basal area of the chamber. The model implementation considered only data coming from CT and MT. The standard set of non-site specific parameters in DNDC were left unchanged while site specific parameterization included specifying the field capacity and wilting point, soil texture, bulk density and total soil C content as determined at the beginning of the experiment. The model was run with a standard management from 2000 to 2006 and with the planned experimental practices in the next three years, maintaining the default pools partition (see discussion below). 2.3. Database for the regional simulation 2.3.1. Climate Meteorological forcing was obtained by means of weather stations, global reanalysis data, and satellite products based on the MeteoSat 2nd Generation (MSG) platform, for years 2006 and 2007. Incoming global radiation was derived from the DSSF (Down-welling Surface Short-wave radiation Flux) satellite product, distributed by LSA-SAF (https://landsaf.meteo.pt) at 30 min time resolution and 5 km spatial resolution. Precipitation was derived from the MPE (Multi sensor Precipitation Estimate) satellite product, distributed by Eumetsat at 15 min time resolution and 5 km spatial resolution. Wind speed and air humidity were derived from NCEP/NCAR reanalysis dataset at 6 h time resolution and 2.5◦ spatial resolution. Air temperature was derived from a network of available national weather stations and the LST (Land Surface Temperature) satellite product distributed by LSA-SAF: in this case a data assimilation procedure was developed, based on using weather stations data to derive correlations between air and surface temperature, spatializing such correlations with geostatistics, and finally applying them to the gridded LST satellite product. All products have been spatially interpolated to obtain 1 km gridded

datasets, and temporally processed to obtain daily sums (radiation, precipitation), averages (wind speed and air humidity), maxima and minima (air temperature). Annual precipitation and average air temperature for the year 2007, used for these simulations, are presented in Fig. 1. 2.3.2. Land use and management practices Statistical information about agricultural land use was obtained from the CAPRI database (http://epp.eurostat.ec.europa.eu) downscaled to clusters of 1 km2 pixels (spatial unit-TU). For each TU, we considered the following crops: maize (silage and grain), wheat (durum and soft), soybean, sunflower, which represent the main arable crops and vineyard as a wood crops widespread in all the national territory (Table 1). The management practices for the corresponding crops were obtained from a database built for this project, analysing varying sources of information: ISTAT, statistics from regional and local offices\agencies and opinion of experts. This database contained detailed information on date of planting and harvest, management of residues, tillage, average amount of fertilizer inputs (mineral and organic) and crop yields at Italian “province” spatial units (for a total of 103 province). A Visual Basic routine was developed to read the database and automatically create the management (farm) input model files. DNDC simulates the crop growth at a daily time step, using a pre-defined logistic function (S-curve), with the asymptote indicating the potential yield without growth limiting factors. This value was set using the crop yields provided by the database, considering that DNDC is able to quantify a biomass reduction in case of limiting condition during the growing season (temperature, water or nitrogen stress). Irrigated areas were taken from the raster layer ‘water management system in agricultural land’ (WM1), contained in the European Soil Database (v2.0). In the irrigated area, the model applies water whenever water deficit occurs. 2.3.3. Soil input data and SOC pools partition Soil data required by the model at spatial level were: initial SOC concentration, pH, clay content and bulk density. All these properties were extracted from the SPADE-2 database (http://eusoils.jrc.ec.europa.eu/projects/spade/spade2.html) and downscaled to each 1 km2 TU. Simulations were run twice, with a range of SOC varying ±10% respect to the average value in each TU. This range was chosen according to Hastings et al. (2010), who reported the errors associated with soil measurements that form input parameters used by DNDC and other soil models. Furthermore, the two model runs in each cell allowed to calculate the variance ( 2 ) for the DNDC outputs.

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

549

Table 1 Comparison between the utilized agricultural area (UAA) of the Italian administrative regions reported by the National Institute of Statistic (ISTAT) and the land use simulated with DNDC for the selected crops based on CAPRI database. UAA (1000 ha) ISTAT Regions Abruzzo Basilicata Calabria Campania Emilia R. Friuli V.G. Lazio Liguria Lombardia Marche Molise Piemonte Puglia Sardegna Sicilia Toscana Trentino A.A. Umbria Valle D’Aosta Veneto Total

Arable

CAPRI Woody crop

Grassland

Total

Arable

Woody crop

Total

Arable crop simulated (%)

Woody crop simulated (%)

177 341 183 298 813 164 323 8 695 389 141 500 639 381 619 508 9 221 0 562

78 50 211 152 132 24 137 16 35 34 22 92 490 78 407 171 45 41 1 107

180 151 120 112 108 39 214 26 265 74 38 448 69 613 225 127 345 77 67 152

434 542 514 563 1053 228 674 49 995 496 200 1040 1197 1072 1252 806 399 339 68 820

66 213 60 106 337 133 137 0 424 213 83 289 415 59 313 233 2 107 0 436

34 9 9 24 58 16 27 0 18 18 5 43 95 10 102 48 11 13 0 68

100 221 69 129 395 149 164 1 441 231 89 332 510 68 416 281 13 120 0 503

38 62 33 35 42 81 42 5 61 55 59 58 65 15 51 46 25 48 0 78

43 17 4 15 44 67 20 3 50 53 25 47 19 12 25 28 24 32 3 63

6969

2323

3452

12744

3626

607

4233

52

26

We used the automatic SOC pool partition of the model, without running previous equilibrium period for the following reasons:

(1) the equilibrium state of the more recalcitrant pools requires a very long period of simulation (e.g. 5000–10,000 yr) and a representative reconstruction of the past land use (Easter et al., 2007). Thus, both conditions are too complicated to make this task achievable especially with high spatial resolution adopted in the present study; (2) as the model simulates only a 1 yr rotation (i.e. a monosuccession) in the regional mode, spin-up run with a continuous crop may not guarantee a true partition of more intermediate/labile SOC pools. In fact, the amount of C input is highly variable in time depending to the amount and management of

residues in succeeding crops. Spin-up run may be effective in a continuous crop condition, which is common in some part of the country indeed, but practically impossible to be disentangled from the rotation condition in every TU.

A default SOC distribution among different pools was adopted also by Hastings et al. (2010) in their analysis of uncertainty propagation using DNDC. They showed that DNDC was an appropriate model to replicate the NEE observations for the Oensingen cropland site, with summed annual predictions accurate but initial winter emissions underestimated, probably due to the default assignment of carbon to the soil decomposition pools. In this context, new method to isolate measurable SOC fractions that could be modelled or strictly related to the ‘conceptual’ pools of the actual SOC

Fig. 2. Land use tree of the main arable crops (plus vineyards) simulated with the most common management practices (business as usual scenario-BaU) and with some alternative practices for C sequestration (AMP); i represents the specific combination (e.g. silage maize irrigated), n the TU number and w the relative area covered by the specific combination of crop/management in each TU.

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

2.3.4. Land use reconstruction and simulation Regional simulation was run using the specific interface receiving all of the inputs from the created platform. We also built a management tree (Fig. 2), where the crops considered were simulated with the “real” (BaU – business as usual) and with some alternative management practices (AMP). Each 1 km × 1 km cell, containing a specific crop, was simulated for the AMP even if not present in the usual scenario, creating a large number of resulting databases for each modelled variable. These were recombined assigning a weight (win ) to the specific crop/management combination output (Yi ) for every n TU; the weights were based on the relative presence of this crop/management combination area (hai ) with respect to  the sum of the combinations (hai ) in the each cell (win = hain / hain ). At the end the result was the following

 i

where

T2

is the total variance and

2 i Yin

DNDC

Observed

400 300 200 100

Date

Fig. 3. Simulated and observed trends of net ecosystem exchange (NEE) and N2 O-N emissions at the maize conventional tillage (CT) experiment in Beano.

4000 2000 0 Oct-06 -2000

Apr-07

Nov-07

Jun-08

Y2 in

is the variance of model

∗ ha2in

Dec-08

Jul-09

Jan-10

Aug-10

-4000 -6000 -8000 DNDC -10000

output (e.g. N2 O) for the specific crop/management combination in the n TU. We also calculated the distribution of the coefficient of variation (CV) for the CO2 , N2 O and CH4 equivalent emissions, using the n TU values:

CVn =

500

0 Jan-07 May-07 Aug-07 Dec-07 Apr-08 Aug-08 Dec-08 Apr-09 Aug-09 Dec-09

∗ ha2in

i



-150

Observed

Date

Fig. 4. Cumulated observed net ecosystem exchange (NEE) with the 20% error bars and simulated one at the maize conventional tillage (CT) experiment in Beano.

the variability of measured data (Fig. 4), built plotting the values with 20% error bars (Oren et al., 2006; Hastings et al., 2010). In MT, the model behaviour was evaluated against measures of soil CO2 flux carried out with the dynamic close chamber system (Fig. 5). The total respiration (heterotrophic + root) simulated

En

where En is the total emission of equivalent CO2 , N2 O and CH4 in the n TU. 3. Results and discussion 3.1. Maize simulation at the experimental site of Beano At the Beano site, the modelled daily NEE in CT followed closely the measured one (Fig. 3), showing strong C uptake during the crop growing season and net emissions in the rest of the year. Observed SOC change, estimated as the difference between net ecosystem production (NEP = −NEE) and grain removal and assuming that others C flows were negligible (i.e. DOC, lateral flows), indicated a general SOC depletion which was more evident in 2009. DNDC generally simulated a lower SOC turnover than the reality (Table 2), with values more close to a steady state but with a decreasing trend in the three years simulated (261, 3 and −447 kg C ha−1 in 2007, 2008 and 2009, respectively). However, considering that the uncertainty of yearly average NEE is on the order of 1000–500 kg C ha−1 yr−1 (Rannik et al., 2006; Baldocchi, 2003), the model results tended to lie within

Soil respiration (kg C ha-1)

n

Y2 in

-50 -100

100

N2O Flux (g N ha-1)

=



Observed

0

Yin ∗ win for every TU, where win was equal

to 0 whenever a specific combination was not present in the n TU. −1 Since the model outputs were expressed as a rate (e.g. kg ha ), cumulated cell values (Ycn ) were then obtained as Ycn = Yrn ∗ hain . Even if this architecture is very redundant due to the simulation of specific crop/management combination that does not occur in the usual management, it is very powerful in the scenario analysis. Once the resulting databases are created it is possible to generate a large number of alternative combinations, considering the AMP simulated and by simply varying the weights. The uncertainty in the aggregated GHG fluxes was calculated using the variance (Y2 ) of the modelled outputs (see Section 2.3.3) and the general law of error propagation: T2

DNDC

50

-200

NEE (kg C ha-1)

weighted sum Yrn =

100

NEE (kg C ha-1)

models (Zimmermann et al., 2006) could improve our prediction accuracy.

N2O Flux (g N ha-1)

550

500

DNDC

Observed

80 60 40 20 0 DNDC

Observed

400 300 200 100 0 Jan-07 May-07 Aug-07 Dec-07 Apr-08 Aug-08 Dec-08 Apr-09 Aug-09 Dec-09

Date Fig. 5. Simulated and observed trends of total soil respiration and N2 O-N emissions at the maize reduced tillage (MT) experiment in Beano.

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

551

Table 2 Observed and simulated soil organic carbon change and simulated soil N2 O emissions in the maize conventional (CT) and reduced tillage (MT) experiment at Beano site. Simulated (kg ha−1 )

Observed SOC

Yield

SOC

Yield

N2 O-N

CO2 equiv.

EF

−126 [−563 ± 427] −873[−952 ± 499] −1337[−1259 ± 328] −2336[−2774 ± 734]

4840 ± 145 4280 ± 149 4543 ± 181

261 3 −447 −183

4845 4285 4541

2.33 2.57 4.59 9.49

136 1190 3789 5115

0.009 0.006 0.010

998 ± 284 −492 ± 599 −219 ± 186 287 ± 689

4930 ± 125 3790 ± 242 4120 ± 112

818 −190 438 1066

4947 3801 4135

1.98 2.27 3.84 8.09

−2077 1761 192 −124

0.008 0.005 0.008

a

CT 2007 2008 2009 Sum MT 2007 2008 2009 Sum

In [] is also reported the SOC ± s.e. calculated for CT as net primary production – heterotrophic respiration – grain yield. Positive values of SOC mean accumulation of C, negative values mean release. EF is the ratio between the N2 O-N simulated and the N applied with fertilizations. a Observed SOC (≈net biome production) was calculated as the difference between net ecosystem production (−NEE) and grain yield in CT and as net primary production – heterotrophic respiration – grain yield in MT.

by DNDC showed higher peaks generally delayed with respect to the measured ones. This bias, as was observed in the growing season, was likely due to a shorter growing period of the simulated crop but with a higher assimilation rate as well as the paired respiration. However, the C balance was not particularly affected since autotrophic respiration has a balance equal to 0 and both measured and simulated values were very similar during the period harvestplanting. The positive effect of reduced tillage on C sequestration was noticeable in the first year, as well predicted by the model (about 1 t C ha−1 ), but it was not sustained in 2008 (Table 2). Many experiments have already demonstrated that SOC stock increases in response to reduced or no-tillage in a Mediterranean environment (Hernanz et al., 2002; Alvaro-Fuentes et al., 2009; Lopez-Bellido et al., 2010) and also in a modelling investigation (Alvaro-Fuentes et al., 2009); it is not surprising, then, that DNDC and experimental data both confirmed this general trend. However, the model showed a lower SOC turnover (i.e. less depletion in CT and higher accumulation in MT with respect to measured values) that could be also dependent on the default SOC pools partition, likely to be biased towards more recalcitrant pools. It has to be pointed out that we were not looking for the optimum calibration, that it may be easily obtainable by setting the many parameters available. However, our interest was to evaluate DNDC performances without changing run time parameters and default pool partition. The knowledge of the model strengths and weaknesses is essential in the interpretation of the results at spatial level where a specific calibration of the more than 130,000 TU simulated is impossible. The N2 O emissions were practically zero for the most part of the year in CT field, whereas many peak were measured in the period spring-early summer, after the maize N fertilization (Fig. 3). The model satisfactory followed the general dynamics observed, although it underestimated two peaks in June 2008 and showed an higher peak in 2009 due to a large single application of fertilizer of more than 300 kg N ha−1 . The model trend in MT was very similar to the CT one (Fig. 5), but the cumulated annual values were lower in the former than in the latter (8.09 and 9.49 kg N ha−1 respectively; Table 2). The effect of tillage on N2 O emissions is in fact variable and dependent on site and weather-specific conditions (Snyder et al., 2009). In a recent review, Rochette (2008), analyzing published experimental results, reported that no-till generally increased N2 O emissions in poorly-aerated soils but was neutral in soils with good and medium aeration. Moreover, N2 O emissions from long-term tillage under a continuous corn cropping system in Ohio (Ussiri et al., 2009) were 1.82, 1.96 and 0.94 kg N ha−1 yr−1 in the moldboard, chisel and no-till, respectively, evidencing the positive effect of the long-term no till practices. Since the model does not change the physical soil properties (bulk density, porosity and consequently hydraulic properties) as a direct consequence of the

mechanical tillage action, the differences in the CT and MT simulated values are mainly related to substrate availability (DOC) and electron acceptors (NO3 − ), which are the driving factors in DNDC (Li, 2007), as well as the redox potential. The simulated mouldboard plough effect, leading to a higher SOC mineralization with respect to MT (Table 2), has increased the NO3 − and DOC availability leading to higher N2 O emissions. The fundamental role of these variables was also assessed by Gomes et al. (2009) in long-term cover crops-based rotations. They concluded that managementcontrolled soil variables such as mineral NO3 − , NH4 + and DOC contents influenced more the N2 O fluxes than environmentalrelated variables such as water-filled pore space and air and soil temperature. Considering the sources of uncertainty associated with the flux measurements used to close the C balance, we judged the DNDC behaviour as quite satisfactory for C as well as for N cycle. Cumulated soil fluxes (Table 2), expressed as CO2 equivalent, highlighted the positive effect of MT resulting in a sink of CO2 even at the end of the third year. On the contrary CT was source of CO2 emitting more than 5000 kg CO2 equiv. ha−1 at the end of 2009. 3.2. Spatial crop distribution The five crops considered showed a specific geographical distribution (Fig. 6) that is mostly dependent on the north to south precipitation gradient. In the north Italy and in particular in the Po plain, higher precipitation and extended irrigation areas are favourable for maize cultivation, covering up to 60% of the territory. The maize is usually cultivated as continuous crop or in rotation with wheat (almost soft wheat) and soybean: the latter is cropped only in the Po plain. Durum wheat is particularly diffuse in the central and south Italy, where it is often cultivated in monosuccession. Sunflower is present in the southern Po plain and in central Italy, while the vineyards are found throughout the country. Comparing the land use simulated with the national statistic sources (Table 1), we covered 52% of the arable land and 26% of the woodland crops. However in some regions of the north-eastern part, like Veneto and Friuli Venezia Giulia (Fig. 6), we simulated a high percentage of the utilized agricultural area UAA due to the large presence of the maize: in other regions like Puglia and Basilicata (south-eastern), extensively covered by durum wheat, the UAA simulated was more than 60%. 3.2.1. Regional simulations All the specific crop/management combinations reported in Fig. 2 were simulated, producing a huge amount of outputs; every TU was in fact simulated with its own land use (BaU) and with alternative management practices (AMP). To illustrate this, we have

552

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

Fig. 6. Distribution and area (ha) covered by the crops considered in each TU (TU = 100 ha). Rectangular areas in the last figure indicates areas with a higher % of utilized agricultural area simulated.

reported some results for durum wheat, that is the most cultivated arable crop in Italy (Fig. 7). Yields were generally lower than 2000 kg C ha−1 (about 5200 kg ha−1 of harvested grain) especially in the south Italy, whereas they were higher especially in the north-east. In this area, the N fertilization was >140 kg N ha−1 to sustain crop productivity, decreasing toward the central and south Italy where the average application was 100–120 kg N ha−1 . The N2 O emissions were in general low (<0.5 kg N ha−1 ) and were not dependent to the spatial distribution of the N fertilization, probably because higher fertilization were offset by the higher uptake. Relatively small areas, in which also higher soil CO2 respiration was simulated, showed higher N2 O emissions since SOC decomposition increases the substrate availability (DOC and NO3 ) for the biological denitrification. Nevertheless, considering the IPCC emission factor of 0.0125 kg N2 O-N/kg N, we would have expected values ranging between 1.2 and 1.7 kg ha−1 of N2 O-N, substantially higher than those simulated. However some recent researches (Halvorson et al., 2008; Del Grosso et al., 2008; Burton et al., 2008) reported that emission factors, including the IPCC (2007) ones, over-estimated seasonal N2 O emissions, whereas they may increase when the N input rate exceeds the crop and soil uptake capacity (Grant et al., 2006; Halvorson et al., 2008). Moreover few studies have investigated the N2 O emission in Mediterranean environment, where low precipitation (Fig. 1) maintains a high redox potential unfavourable to denitrification. In fact, Meijide et al. (2009) reported N2 O emission <400 g N ha−1 yr−1 in a non-irrigated Mediterranean barley field, under organic and mineral fertilization.

Low emissions (<700 g N–N2 O ha−1 yr−1 ) were also measured in a Mediterranean vineyard with and without cover crop (Steenwerth and Belina, 2008). All the crops simulated (Fig. 8) were combined for their actual land use to assess the total GHG emissions. It has to be emphasized that, although the national balance is incomplete, there are some areas (see Section 3.2 and Fig. 6) where simulated crops almost matched the ‘real’ UAA. In the north-eastern part, the soil was in general a small sink whereas in the south-east it was almost in equilibrium, with values ranging between −500 and 500 kg C ha−1 . This variation is in the range of the inter-annual variability under constant land use and management and practically undetectable even with a higher number of soil samples per hectare (Smith, 2004; Conant and Paustian, 2002). Indeed, these results are in agreement with the net biome production (NBP) estimated by Ciais et al. (2010) using the model ORCHIDEE-STICS over European countries. They showed fluxes ranging between 0.1 and 0.3 t C ha−1 yr−1 in the north-eastern Italy and between 0.1 and −0.4 t C ha−1 yr−1 in the south-eastern part. Across the nation some areas showed a SOC decrease greater than 1 t C ha−1 , even if this trend could be strongly dependent on the land use not simulated. In particular, when a large proportion of grassland is present in the TU, the rate of SOC change could be not representative since grassland sustains higher SOC level than the only arable crop simulated. Moreover, the national SOC change value was about −0.139 t C ha−1 for the land use considered, confirming that croplands are close to an equilibrium state.

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

553

Fig. 7. Durum wheat simulations: (a) distribution and hectares covered, (b) grain yield, (c) soil respiration, (d) nitrogen fertilization and (e) nitrous oxide emissions.

The N2 O emissions were generally below 1 kg N ha−1 , with some hot spots mostly coincident with the area showing the higher SOC decline. In fact, Butterbach-Bahl et al. (2004) found SOC to be the most sensitive parameter for N2 O from agricultural soils. The cumulated emissions for the simulated land use were 1.52 Mt of CO2 equiv. (Fig. 9).

The simulations indicated the positive role of cultivated soils as a sink of CH4 (Fig. 8), with values ranging between 0 and −1.5 kg C ha−1 . Upland drained soils have generally positive redox potential (Eh), which will not provide an environment suitable for methanogenic bacteria to emit CH4 (Borken et al., 2003). Italian CH4 emissions for

Fig. 8. Simulated regional soil GHG emissions of the crops considered with the BaU scenario: (a) soil organic carbon change (SOC), (b) N2 O-N emissions and (c) CH4 –C emissions. Positive value of SOC means carbon uptake, negative value indicate a release of C.

554

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

the simulated crops were equal to −0.08 Mt of CO2 equiv. (Fig. 9). 3.2.2. Sources of uncertainty The accuracy of the spatial simulation is strongly dependent to the model accuracy and the quality of input data. Experimental sites allow to test the model performance, calibrate the crop parameters and assess the level of uncertainty (Hastings et al., 2010). The simulations with a SOC range of +/−10%, likely one of the most important input parameter in controlling the C and N2 O emissions, allowed to associate a measure of variability to the results. The CV distribution of the CO2 and CH4 equivalent emissions were similar and up to 30% for about 70% of the TUs simulated (Fig. 9), while more than 60% of them showed a CV within the 10% for the N2 O. When the variance was propagated by the quadrature method (see Section 2.3.4), the cumulated uncertainty appeared very low for all the fluxes as evidenced by the small error bars (95% confidence interval) in Fig. 9. In fact this method supposed that errors add randomly thus they have a tendency to offset each other especially with a very high number of “samples”. Moreover, in order to quantify the error attributable to the geographical input data quality, we also compared the model input

and output at the Beano experimental site and in the TU simulated at regional level, containing the site coordinates (Table 3). It is worth to point out that at the Beano, despite the experimental activity, the crop was managed by a farmer applying the common practices of the area. From the corresponding TU, we selected the maize grain irrigated and conventionally tilled to represent the local condition. As reported in Table 3, some soil properties implemented from the geographical database were very similar to those measured (pH, bulk density and clay content). Nevertheless the SOC concentration was 33% lower than the measured one, likely leading to a higher SOC accumulation and a lower N2 O-N emission. This low value was also affected by the lower N fertilization rates compared to the local practices. The amount of water applied by DNDC as irrigation (95 mm) was slightly lower than the farmer application (110 mm). However in the first case, the water was supplied with high efficiency whenever required, whereas, in the second, many other factors such as irrigation turn, wind disturbance, evaluation of the crop water stress and farmer habits make this practice more inefficient. Thus more likely, irrigation applied by the farmer could have created temporary anaerobic conditions enhancing N2 O fluxes, a situation not expected in the model since it does not apply water exceeding crop

Total VEN VAL UMB

SIC

40 20 0

PUG

100

Relative Frequency (%)

SAR

PIE MOL MAR

40 20

100

Relative Frequency (%)

LIG

EMI CAM

CO

60

0

FRI

CV (%)

80

LOM

LAZ

N2O N2O

60

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200

TOS

NO 80

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200

TRE

CV(%)

CH

80 60 40 20 0

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200

Relative Frequency (%)

100

CAL

CO 2 CO2

CV (%)

BAS

CH 4 CH4

ABR -1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1.6

2.0

2.4

GHG emissions (Mt CO2 eq.) Fig. 9. Simulated annual GHG emissions (CO2 , N2 O and CH4 ) from soil expressed as CO2 equivalent for the land use simulated and grouped by Italian administrative regions. Positive value of SOC means carbon uptake, negative value indicate a release of C. Bars indicate the 95% confidence interval. Distribution of CV across TUs was represented in the graphics inside the box.

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

555

Table 3 Comparison between inputs and outputs of the simulations performed at the Beano experimental site for 2007 and at regional level (spatial unit, TU). Input

Beano site TUa

Output

SOC

Clay

(%)

(%)

1.73 1.17

14.0 13.6

pH

7.2 7.7

BD

Fertilization

Yield

SOC

N2 O-N

(g m−3 )

(kg N ha−1 )

(kg C ha−1 )

(kg C ha−1 )

(kg N ha−1 )

1.2 1.3

257 207

4845 (4840) 3926

261 (−126) 1010

2.33 0.75

Values in brackets refer to measured data. a TU results refer to the irrigated maize conventionally tilled belonging to the cell containing the coordinates of Beano site; BD = bulk density.

covered by grain maize; however, it allowed us to highlight the potential of our platform for the simulation of alternative scenarios. In fact we could vary the surface we want to dedicate to an AMP and also locate it in the territory with a high spatial resolution.

a

0.0 -2.0 -4.0 -6.0

4. Conclusions

1109 kg/ha dSOC

GHG flux (Mt CO2 eq)

2.0

CT

MT

-8.0

CO2

CH4

N2O

Net flux

b

CT

RR

6.0 4.0 2.0

1904 kg/ha dSOC

GHG Flux (Mt CO 2 eq)

8.0

0.0 -2.0 -4.0

CO2

N2O

CH4

Net flux

Fig. 10. Annual net GHG fluxes and overall GHG balance of the grain maize with alternative management practices: (a) conventional (CT) vs reduced tillage (MT) and (b) incorporation (CT) vs removal of residues (RR).

demand. Our simulations clearly highlighted how much the result accuracy could be affected by the input data. In particular, the SPADE-2 soil database is based on map units of 1:1,000,000 scale, that is likely too large to represent the soil variability in Italy. For this reason the implementation of a national harmonised dataset, not available at the beginning of the project, will be one the first improvement expected. 3.2.3. Scenario analysis Some of the considered crops were simulated with alternative management practices (Fig. 10) to assess the potential mitigation effects or impacts of the strategies adopted. To illustrate this, we reported two cases involving grain maize crop: (1) the adoption of a minimum tillage (MT) and (2) the residue removal (RR) from the field, both compared with the BaU management based on plough and residue incorporation (CT). The conversion of all the grain maize area to minimum tillage allowed the accumulation of about 1100 kg C ha−1 of SOC on average, saving more than 4 Mt of CO2 equiv. (95% confidence interval ± 0.03 Mt of CO2 equiv.) including the lower N2 O emissions. On the contrary, the residue removal could strongly impact the SOC balance leading to a very high soil C depletion (1904 kg C ha−1 on average) and consequently increasing the GHG emissions (8.2 ± 0.04 Mt of CO2 equiv.). These simulations were based on the assumption of AMP adoption in the total area

DNDC was able to represent the C and N cycles satisfactory at the Beano site, despite changing only the site specific parameters. Modelled N2 O emissions fitted the measured data well, but the corresponding emission factor from fertilizers was much lower than the IPCC default. The reduced tillage confirmed to be a reliable practice to sequester C, as indicated by the experimental data and the modelling results. At spatial level, the CO2 fluxes were higher than the N2 O emissions (expressed as CO2 equivalent) but they varied broadly among the Italian regions. Indeed, the comparison between the experimental site and the corresponding simulated spatial unit highlighted the bias related to the input accuracy on the GHG simulated fluxes. The idea of developing a management tree and combining the results by a land use weight, allowed us to assess the effect of alternative management practices in each spatial unit, without further running the model. The simulated conversion of all the maize from conventional to minimum tillage gave a quantitative estimation of the potential role of the agricultural sector on C sequestration. In conclusion, the first results are promising but further improvements are needed to implement all the agricultural land uses, to acquire national harmonised input dataset and to test the model with measured data from other experimental sites belonging to the Carbo-Italy network. Acknowledgements This research was funded by the Italian National Research Programme “CarboItaly”, Italian project PRIN 2007, EU NitroEurope and the EU C-extreme Project. We would like also to thank Simona Castaldi for the analysis of N2 O samples and Erandi Elokupitiya for her precious help in revising the manuscript. References Alberti, G., Castaldi, S., Zavalloni, C., Delle Vedove, G., Peressotti, A., CO2 and N2 O emissions from conventional and minimum-tilled soils. In preparation. Alberti, G., Delle Vedove, G., Zuliani, M., Peressotti, A., Castaldi, S., Zerbi, G., 2010. Changes in CO2 emissions after crop conversion from continuous maize to alfalfa. Agriculture, Ecosystems & Environment 136, 139–147. Alvaro-Fuentes, J., Lopez, M.V., Arrue, J.L., Moret, D., Paustian, K., 2009. Tillage and cropping effects on soil organic carbon in Mediterranean semiarid agroecosystems: testing the century model. Agriculture Ecosystems & Environment 134, 211–217. Baldocchi, D.D., 2003. Assessing the eddy covariance technique for evaluating carbon dioxide exchange rates of ecosystems: past, present and future. Global Change Biology 9, 479–492. Beheydt, D., Boeckx, P., Sleutel, S., Li, C.S., Van Cleemput, O., 2007. Validation of DNDC for 22 long-term N2 O field emission measurements. Atmospheric Environment 41, 6196–6211. Borken, W., Xu, Y.-J., Beese, F., 2003. Conversion of hardwood forests to spruce and pine plantations strongly reduced soil methane sink in Germany. Global Change Biology 9, 956–966.

556

E. Lugato et al. / Agriculture, Ecosystems and Environment 139 (2010) 546–556

Burton, D.L., Li, X., Grant, C.A., 2008. Influence of fertilizer nitrogen source and management practice on N2 O emissions from two black chernozemic soils. Canadian Journal of Soil Science 88, 219–227. Butterbach-Bahl, K., Kesik, M., Miehle, P., Papen, H., Li, C., 2004. Quantifying the regional source strength of N-trace gases across agricultural and forest ecosystems with process based models. Plant and Soil 206, 311–329. Cai, Z.C., Sawamoto, T., Li, C.S., Kang, G.D., Boonjawat, J., Mosier, A., Wassmann, R., Tsuruta, H., 2003. Field validation of the DNDC model for greenhouse gas emissions in East Asian cropping systems. Global Biogeochemical Cycles, 17. Ciais, P., Wattenbach, M., Vuichard, N., Smith, P., Piao, S.L., Don, A., Luyssaert, S., Janssens, I.A., Bondeau, A., Dechow, R., Leip, A., Smith, P., Beer, C., Werf, G.R.V.D., Gervois, S., Oost, K.V., Tomelleri, E., Freibauer, A., Schulze, E.D., 2010. The European carbon balance. Part 2: Croplands. Global Change Biology 16, 1409–1428. Conant, R.T., Paustian, K., 2002. Spatial variability of soil organic carbon in grasslands: implications for detecting change at different scales. Environmental Pollution 116, S127–S135. Del Grosso, S.J., Halvorson, A.D., Parton, W.J., 2008. Testing DAYCENT model simulations of corn yields and nitrous oxide emissions in irrigated tillage systems in Colorado. Journal of Environmental Quality 37, 1383–1389. Delle Vedove, G., Alberti, G., Peressotti, A., Inglima, I., Zuliani, M., Zerbi, G., 2007. Automated monitoring of soil respiration: an improved automatic chamber system. Italian Journal of Agronomy 4, 377–382. Easter, M., Paustian, K., Killian, K., Williams, S., Feng, T., Al-Adamat, R., Batjes, N.H., Bernoux, M., Bhattacharyya, T., Cerri, C.C., Cerri, C.E.P., Coleman, K., Falloon, P., Feller, C., Gicheru, P., Kamoni, P., Milne, E., Pal, D.K., Powlson, D.S., Rawajfih, Z., Sessay, M., Wokabi, S., 2007. The GEFSOC soil carbon modelling system: a tool for conducting regional-scale soil carbon inventories and assessing the impacts of land use change on soil carbon. Agriculture Ecosystems & Environment 122, 13–25. FAO, 2006. World Reference Base for Soil Resources 2006. World Soil Resources Reports n. 103. Food and Agricultural Organization, Rome, p. 128. Freibauer, A., 2003. Regionalised inventory of biogenic greenhouse gas emissions from European agriculture. European Journal of Agronomy 19, 135–160. Giltrap, D.L., Li, C., Saggar, S., 2010. DNDC: A process-based model of greenhouse gas fluxes from agricultural soils. Agriculture Ecosystems & Environment 136, 292–300. Gomes, J., Bayer, C., Costa, F.D., Piccolo, M.D., Zanatta, J.A., Vieira, F.C.B., Six, J., 2009. Soil nitrous oxide emissions in long-term cover crops-based rotations under subtropical climate. Soil & Tillage Research 106, 36–44. Grant, R.F., Pattey, E., Goddard, T.W., Kryzanowski, L.M., Puurveen, H., 2006. Modeling the effects of fertilizer application rate on nitrous oxide emissions. Soil Science Society of America Journal 70, 235–248. Halvorson, A.D., Del Grosso, S.J., Reule, C.A., 2008. Nitrogen, tillage, and crop rotation effects on nitrous oxide emissions from irrigated cropping systems. Journal of Environmental Quality 37, 1337–1344. Hastings, A.F, Wattenbach, M., Eugster, W., Li, C.S., Buchmann, N., Smith, P., 2010. Uncertainty propagation in soil greenhouse gas emission models: an experiment using the DNDC model and at the Oensingen cropland site. Agriculture Ecosystems & Environment 136, 97–110. Hernanz, J.L., Lopez, R., Navarrete, L., Sanchez-Giron, V., 2002. Long-term effects of tillage systems and rotations on soil structural stability and organic carbon stratification in semiarid central Spain. Soil & Tillage Research 66, 129–141. IPCC, 2007. Summary for policymakers. Climate change 2007: the physical science basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.). Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Leip, A., Marchi, G., Koeble, R., Kempen, M., Britz, W., Li, C., 2008. Linking an economic model for European agriculture with a mechanistic model to estimate nitrogen and carbon losses from arable soils in Europe. Biogeosciences 5, 73–94. Li, C., 2007. Quantifying greenhouse gas emissions from soils: scientific basis and modeling approach. Soil Science and Plant Nutrition 53, 344–352. Li, C., Farahbakhshazad, N., Jaynes, D.B., Dinnes, D.L., Salas, W., McLaughlin, D., 2006. Modeling nitrate leaching with a biogeochemical model modified based on observations in a row-crop field in Iowa. Ecological Modelling 196, 116–130. Li, C., Frolking, S., Frolking, T.A., 1992. A model of nitrous oxide evolution from soil driven by rainfall events. I. Model structure and sensitivity. Journal of Geophysical Research 97, 9759–9776.

Li, H., Qiu, J.J., Wang, L.G., Tang, H.J., Li, C.S., Van Ranst, E., 2010. Modelling impacts of alternative farming management practices on greenhouse gas emissions from a winter wheat–maize rotation system in China. Agriculture Ecosystems & Environment 135, 24–33. Lokupitiya, E., Paustian, K., 2006. Agricultural soil GHG emissions: a review of national inventory methods. Journal of Environmental Quality 35, 1413–1427. Lopez-Bellido, R.J., Fontan, J.M., Lopez-Bellido, F.J., Lopez-Bellido, L., 2010. Carbon sequestration by tillage, rotation, and nitrogen fertilization in a Mediterranean vertisol. Agronomy Journal 102, 310–318. Lugato, E., Berti, A., 2008. Potential carbon sequestration in a cultivated soil under different climate change scenarios: a modelling approach for evaluating promising management practices in north-east Italy. Agriculture Ecosystems & amp; Environment 128, 97–103. Lugato, E., Berti, A., Giardini, L., 2006. Soil organic carbon (SOC) dynamics with and without residue incorporation in relation to different nitrogen fertilisation rates. Geoderma 135, 315–321. Lugato, E., Paustian, K., Giardini, L., 2007. Modelling soil organic carbon dynamics in two long-term experiments of north-eastern Italy. Agriculture Ecosystems & Environment 120, 423–432. Meijide, A., Garcia-Torres, L., Arce, A., Vallejo, A., 2009. Nitrogen oxide emissions affected by organic fertilization in a non-irrigated Mediterranean barley field. Agriculture Ecosystems & Environment 132, 106–115. Morari, F., Lugato, E., Berti, A., Giardini, L., 2006. Long-term effects of recommended management practices on soil carbon changes and sequestration in northeastern Italy. Soil Use and Management 22, 71–81. Oren, R., Hsieth, C.I., Stoy, P., Albertson, J., McCarthy, H.R., Harrell, P., Katul, G.G., 2006. Estimating the uncertainty in annual net ecosystem carbon exchange: spatial variation in turbulent fluxes and sampling errors in eddy-covariance measurements. Global Change Biology 12, 883–896. Rannik, U., Kolari, P., Vesala, T., Hari, P., 2006. Uncertainties in measurement and modelling of net ecosystem exchange of a forest. Agricultural and Forest Meterology 138, 244–257. Rochette, P., 2008. No-till only increases N2 O emissions in poorly-aerated soils. Soil & Tillage Research 101, 97–100. Romano, D., Arcarese, C., Bernetti, A., Caputo, A., Cóndor, R.D., Contaldi, M., De Lauretis, R., Di Cristofaro, E., Federici, S., Gagna, A., Gonella, B., Liburdi, R., Taurino, E., Vitullo, M., 2009. Italian Greenhouse Gas Inventory 1990–2007, ISPRA – Institute for Environmental Protection and Research, Rome. Smith, P., 2004. How long before a change in soil organic carbon can be detected? Global Change Biology 10, 1878–1883. Smith, P., 2005. An overview of the permanence of soil organic carbon stocks: influence of direct human-induced, indirect and natural effects. European Journal of Soil Science 56, 673–680. Smith, P., Smith, J.U., Powlson, D.S., McGill, W.B., Arah, J.R.M., Chertov, O.G., Coleman, K., Franko, U., Frolking, S., Jenkinson, D.S., Jensen, L.S., Kelly, R.H., KleinGunnewiek, H., Komarov, A.S., Li, C., Molina, J.A.E., Mueller, T., Parton, W.J., Thornley, J.H.M., Whitmore, A.P., 1997. A comparison of the performance of nine soil organic matter models using datasets from seven long-term experiments. Geoderma 81, 153–225. Smith, W.N., Grant, B.B., Desjardins, R.L., Worth, D., Li, C., Boles, S.H., Huffman, E.C., 2010. A tool to link agricultural activity data with the DNDC model to estimate GHG emission factors in Canada. Agriculture Ecosystems & Environment 136, 301–309. Snyder, C.S., Bruulsema, T.W., Jensen, T.L., Fixen, P.E., 2009. Review of greenhouse gas emissions from crop production systems and fertilizer management effects. Agriculture Ecosystems & Environment 133, 247–266. Steenwerth, K., Belina, K.M., 2008. Cover crops and cultivation: impacts on soil N dynamics and microbiological function in a Mediterranean vineyard agroecosystem. Applied Soil Ecology 40, 370–380. Triberti, L., Nastri, A., Giordani, G., Comellini, F., Baldoni, G., Toderi, G., 2008. Can mineral and organic fertilization help sequestrate carbon dioxide in cropland? European Journal of Agronomy 29, 13–20. Ussiri, D.A.N., Lal, R., Jarecki, M.K., 2009. Nitrous oxide and methane emissions from long-term tillage under a continuous corn cropping system in Ohio. Soil & Tillage Research 104, 247–255. Zimmermann, M., Leifeld, J., Schmidt, M.W.I., Smith, P., Fuhrer, J., 2006. Measured soil organic matter fractions can be related to pools in the RothC model. European Journal of Soil Science 58, 658–667.