«International 17 Workshop th Nitrogen The was jointly organised by Teagasc and AFBI Printed by Print Depot Suggested citation Authors, 2012. Title ...»
All melons were harvested in September. N2O and CO2 samples were taken twice a week during the whole experimental period (c. 2 months), using static chambers (3.1 l). Gas samples were analyzed by gas chromatography using an electron capture detector for N2O and a flame ionization detector with methanizer for CO2. Soil moisture and mineral N (NH4+ and NO3-) were measured using the methodology described in Sanchez-Martin et al. (2008). Soil samples were taken coinciding with gas sampling days. Differences between treatments in the cumulative emissions were analysed using analysis of variance (ANOVA, P 0.05).
3. Results & Discussion Application of nitrate fertilizer reduced net N2O emissions by c. 74% compared to the urea fertilized plots. Taking into account that urea fertilization significantly increased the soil NH4+ content, our results suggest that nitrification was a major source of these higher emissions.
Irrigation management had no significant effect on N2O emissions, however, the low frequency treatment reduced net CO2 emissions by c. 35%. This effect may be partially attributed to decreased ecosystem respiration due to the lower temporal water availability of this treatment (Meijide et al., 2010).
4. Conclusion Based in our results, replacing NH4+ fertilizer (urea) with NO3- fertilizer could be an option to mitigate N2O emissions in the Mediterranean climate. Additionally, lowering the irrigation frequency may decrease CO2 emissions.
References Meijide, A., Cárdenas, L., Sánchez-Martín, L. and Vallejo, A. 2010. Carbon dioxide and methane fluxes from a barley field amended with organic fertilizers under Mediterranean climatic conditions. Plant and Soil 328, 353-367.
Sanchez-Martin, L., Arce, A., Benito, A., Garcia-Torres, L. and Vallejo, A., 2008. Inﬂuence of drip and furrow irrigation systems on nitrogen oxide emissions from a horticultural crop. Soil Biology and Biochemistry 40, 1698-1706.
Nitrogen Workshop 2012
Synergetic leaching model based on pathway and pressure factors Kuzmanovski, V.a, Džeroski, S.a, Schulte, R.P.O.b, Debeljak, M.a a Department of Knowledge Technologies, Jozef Stefan Institute, Jamova 39, 1000 Ljubljana, Slovenia b Teagasc, Johnstown Castle, Environmental Research Centre, Co Wexford, Rep. of Ireland
1. Background & Objectives The implementation of the Nitrates Action Programme in 2006 has led to a reduction of agricultural nutrient pressures on surface waters and groundwater in Ireland. This has coincided with an apparent halt – or even tentative reversal – of the historic deterioration of water quality over time.
However, the current implementation of the Water Framework Directive poses more stringent challenges to minimising nutrient loss from agriculture, particularly in areas where “high status” water quality must be preserved. Maintenance of “high status” water quality requires customised mitigation strategies that account for local geo-climatic conditions and farming systems. Such strategies require an in-depth quantitative insight into the interactions between nutrient pressures and transport mechanisms. In this study we use data mining methodology to find the synergetic patterns of the interactions between pressure (soil mineralization, drought and grass growing season) and pathway (soil drainage, net rainfall and rainfall intensity) attributes which impact leaching of nitrogen and phosphorus to water.
2. Materials & Methods Analysis were done on a dataset that contains data for the following pressure factors: phosphorus and nitrogen field’s input, duration of grass growing season, soil moisture deficit (which represents the soil water dynamics) and pathway factors: soil drainage, net rainfall and the number of intense drainage events (Schulte et al., 2006). The dataset contains 352 records, each representing one grid cell of 10 x 10 km (out of 798 grid cells into which Ireland has been divided). To find interactions among attributes that might have synergetic effects on nutrients transfer to water, we used decision trees as a flexible and robust data mining method, ideally suited for the analysis of complex ecological data. They can deal with nonlinear relationships, high-order interactions and missing data, and in the same time are simple to understand and give easily interpretable results (De’ath and Fabricius, 2000). Decision trees predict the value of one or several dependent variables (targets) from a set of independent variables (attributes). We modelled the concentration of nitrogen and phosphorus in water individually, as well as simultaneously, where we predicted the concentrations of both nutrients at once. Models for predicting the concentration of individual nutrients were learned with model trees in the Weka data mining suite (Witten & Frank 2005) while the simultaneously predictions of both nutrients were made with predictive clustering trees (Blockeel et al., 1998) implemented in the Clus data mining system (Blockeel and Struyf, 2002).
3. Results & Discussion Soil moisture deficit (SMDmax) has been selected as the most important attribute because it appears at the top most position of all induced decision tree models. The model with the best validation performance was the model for predicting nitrogen concentration (Fig. 1, Table 1) (correlation coefficient of 10-fold cross validation is 0.799), while the worst performance was observed for the phosphorous model (correlation coefficient of 10-fold cross validation is 0.475).
Table 1. Nitrogen concentration for different sections of the nitrogen model (Figure 1) LM1 LM2 LM3 LM4 LM5 LM6 LM7 LM8 0.
7733 1.0522 1.5704 2.7285 6.0889 8.6308 11.4505 14.3184 The model for predicting both, nitrogen and phosphorus at the same time, showed good prediction performance with the correlation coefficient of 10-fold cross validation of 0.874 for nitrogen and
0.669 for phosphorus.
Figure 1. Regression tree for predicting the nitrogen concentration in water.
The structure of the nitrogen model (Figure 1) shows that strongly correlated attributes for estimating nitrogen concentration in water are soil moisture deficit (“SMDMax”), field’s input of nitrogen (“sum_N_ha_farmed”), drainage factor (“drainage_factor”) and grass growing season (“growth_season”). Hence, when the soil moisture deficit is less than 45.259 mm (wet soil), low field’s input of nitrogen (less than 174.489 kg ha-1) and low drainage factor (poorly drained field) lead to minimal concentration of nitrogen (LM1: 0.7733 mg L-1) which is smaller than in case of moderate or well-drained soil (LM2: 1.0522 mg L-1). The theory says that smaller soil moisture deficit “allow” field’s input of nitrogen to control the concentration in water (Schulte et al., 2006), on the other hand, when soil is well-drained then water has open paths to go deeper, collecting the accumulated nitrogen in the soil, otherwise the water will stay in the soil and it will stop nitrogen leaching under the root zone. The Table 1 shows the predicted concentrations of nitrogen in water for different combinations of the attributes included in the model. Note that each path in the decision tree can be interpreted as a rule that explains the phenomena leading to the correct conclusion. It should be noted that the performance measures of the phosphorus model shows that the used attributes cannot predict its concentration in water and some additional attributes should be included into database.
4. Conclusion Our approach has identified synergetic interactions among the attributes describing the pressure of nutrients loss and transport pathways. Discovered patterns can be used for formulating further mitigation strategies which can make significant contributions to the reduction of water pollution from agriculture. The proposed methodology could be applied also to less aggregated data which would enable identification of areas with different risks of nutrients losses (Debeljak et al., 2010).
References Blockeel H., Raedt L. De, and Ramon J. 1998. Top-down induction of clustering trees, Proceedings of the 15th International Conference on Machine Learning (Shavlik, J., ed.), pp. 55-63, Blockeel, H. and Struyf, J. 2002. Efficient algorithms for decision tree cross-validation. Journal of Machine Learning Research, 3:621-650.
Witten, I.H. and Frank, E. 2005. Data Mining: Practical Machine Learning Tools and Techniques. Morgan Kaufmann, San Francisco, CA, USA, second edition.
De’ath G., Fabricius K.E., 2000. Classification and regression trees: A powerful yet simple technique for ecological data analysis. Ecology 81, 3178-3192.
Debeljak, M., Kocev, D., Towers, W., Jones, M., Griffiths, B. and Hallett P 2009. Potential of multi-objective models for risk-based mapping of the resilience characteristics of soils: demonstration at a national level. Soil Use & Man, 25:66-77.
Schulte R.P.O., Richards K., Daly K., Kurz I., McDonald E.J. and Holden N.M. 2006. Agriculture, meteorology and water quality in Ireland: A regional evaluation of pressures and pathways of nutrient loss to water. 17, 117-133.
Nitrogen Workshop 2012
Terrestrial carbon and nitrogen losses and indirect greenhouse gas emissions via groundwater Jahangir, M.M.R.a, b, Johnston, P.b, Khalil, M.I.c, Hennessy, D.d, Humphreys, J.d, Fenton, O.a, Richards, K.G.a a Teagasc Environment Research Centre, Johnstown Castle, Co. Wexford, Rep. of Ireland b Department of Civil, Structural & Environ. Engineering, The University of Dublin, Trinity College, Dublin 2, Ireland c Environmental Protection Agency, Johnstown Castle Estate, Co.Wexford, Ireland d Teagasc, Animal and Grassland Research and Innovation Centre, Moorepark, Fermoy, Co. Cork, Ireland
1. Background & objectives Reactive nitrogen (N) emissions to waters greatly contribute to groundwater pollution, freshwater and coastal eutrophication, algal bloom and hypoxia. While most research has focused on direct emissions, indirect N2O emissions via groundwater is a significant, but very poorly understood component of the global N2O budget (Clough et al., 2007). Groundwater is also an important vector of indirect emissions of CO2 and CH4 (Minamikawa et al., 2010) with significant discharges to surface waters and effects on aquatic biogeochemistry. The dynamics of dissolved C and N in groundwater is a key ‘‘missing piece’’ in our understanding of global C and N balances. This research aimed to (i) measure the amount of dissolved C and N losses from terrestrial ecosystems to the aquatic ecosystems via groundwater, and (ii) estimate the contribution of indirect emissions of GHG to the atmosphere.
2. Materials & Methods The investigation were carried out at two low permeability (L: L1, Johnstown Castle; L2, Solohead) and two high permeability (H: H1, Oak Park; H2, Dairygold) sites in Ireland. Among the sites, L1, L2 and H2 were grassland and H1 was arable. Groundwater sampling was carried out monthly between Feb, 2009 and Jan, 2011 for hydrochemistry and dissolved gases. For dissolved N2O, CO2 and CH4, samples were degassed using He headspace extraction technique and analysed by gas chromatography. The samples for N2 were analysed in a high precision membrane inlet mass spectrometer (MIMS). Prior to groundwater sampling watertable (WT) depth was measured using an electronic dip meter. A water balance was used to calculate the effective rainfall (ER).
3. Results & Discussion Total N input was 300, 213, 150 and 297 kg N ha-1, respectively at L1, L2, H1 and H2 sites. Among the grassland sites, the number of livestock units (LU) was lower at L2 (2.0 LU) than L1 (2.2 LU) and H2 (2.2 LU). Rainfall was well above average (130-140%) in 2009 and below average (87in 2010 across sites. The period of ER at the L sites was longer than the H sites (Table 1). The annual WT fluctuation ranges were 1.9, 0.7, 3.5 and 5.3 m below ground level.
Mean N2O-N conc. over the two years differed significantly between sites (p0.001) (Figure 1).
Dissolved CO2 conc. was significantly higher at grassland than arable sites. Mean CH4 conc. was higher at the L sites than the H sites. Mean dissolved N (DN=NO3--N+NO2--N+N2-N+N2ON+NH4++dissolved organic N) loads in groundwater over the two years accounted for 12, 8, 38, and 27% of the surface N input. The major fraction of DN was NO3-N (81-92%) at H sites and N2 (46at L sites. Loads of dissolved C (dissolved organic C (DOC)+CO2+CH4) discharged ranged from 78-344 kg ha-1 at L and 30-217 kg C ha-1 at H sites.
0.40 0.30 0.20 0.10
Figure 1 (a) N2O, (b) CO2 and (c) CH4 conc. in groundwater at four experimental sites (mean ± SE, n=24)
4. Conclusions Estimation of losses of dissolved carbon and nitrogen via groundwater is important to reduce the uncertainties in the terrestrial C and N balances. Quantifying dissolved N2O, CO2 and CH4 in groundwater beneath an agricultural system is of huge importance for global GHG budgets.
References Clough T.J., Addy K., Kellogg D.Q., Nowicki B.L., Gold, A.J. and Groffman P.M. 2007. Dynamics of nitrous oxide in groundwater at the aquatic–terrestrial interface, Global Change Biology 13, 1528–1537.
Minamikawa K., Nishimura S., Sawamoto T., Nakajima Y. and Yagi K. 2010. Annual emissions of dissolved CO2, CH4, and N2O in the subsurafce drainage from three cropping systems, Global Change Biology 16, 796-809.
Nitrogen Workshop 2012
The complexity of the recharge processes and their effect on seasonal variations of nitrate concentration in shallow groundwater and streams: observations and modeling L. Ruiz a,b, M. Rouxe1 a,b, Gascuel-Odoux a,b, L. Aquilina c, A. Aubert a,b, M. Faucheux a,b, P.
Merot a,b, J. Molenat a,b,e, M. Sebilo d a INRA, UMR1069, Sol Agro et hydrosystème, F-35000 Rennes.
b Agrocampus Ouest, UMR1069, Sol Agro et hydrosystème, F-35000 Rennes.
c Université de Rennes 1, CNRS, UMR 6118, Géosciences Rennes,F-35000 Rennes.
d UPMC Univ Paris 06, UMR Bioemco, 4 place Jussieu, 75252 Paris Cedex 05, France e INRA, UMR, LISAH, F-34000 Montpellier