Methods Directory UMM :Data Elmu:jurnal:E:European Journal of Agronomy:Vol13.Issue2-3 July2000:

The model that was chosen is LEACHN Hutson and Wagenet, 1992. The main soil parameters hydraulic conductivity, retention curve and me- teorological data are considered as stochastic vari- ables. The scaling factor approach Miller and Miller, 1956 has been used to insert the stochastic variability of soil hydrological parameters into the model. This approach allows one to include the hydrological variability in a model through only one parameter, the scaling factor. The scaling theory has been extensively used since long Sim- mons et al., 1979; Russo and Bresler, 1980; Rao et al., 1983; Warrick 1990: several works used the scaling factor for modelling Peck et al., 1977; Warrick and Amoozegar-Fard, 1979; Hopmans and Stricker, 1989 or to insert stochastic compo- nents into deterministic models Boulier and Vau- clin, 1986; Vachaud et al., 1988; Braud et al., 1995. The objective of this paper was to calculate and to discuss the predicted probability to exceed a given level of NO 3 leaching loss for several crop- ping systems.

2. Methods

The research was organized in the following phases: choice of simulation model, creation of a specific software for repeated model utilization, definition of the variability associated with soil hydrological parameters and to climatic data, choice of cropping systems to be studied and, finally, selection of the most representative output values. 2 . 1 . The model An analysis of nitrate leaching from cropping systems was carried out using the LEACHN model Hutson and Wagenet, 1992. This model has already been extensively applied to different environments Ramos and Carbonell, 1991; Lotse et al., 1992; Jemison et al., 1994; Acutis et al., 1997a,b; Ducco, 1997. It solves the Richards’ equation to predict soil water dynamics and simu- lates the nitrogen cycle according to Johnsson et al. 1987. Three pools of organic matter are defined: humus, litter and manure, each character- ized by a CN ratio and by a mineralization rate, which depends on soil temperature and soil water contents. The humification process is simulated, allowing for re-organization of mineral nitrogen and CO 2 losses. Crop residues straw and roots are added to the litter pool. The ammonia pool is subjected to nitrification, volatilization and leach- ing and the nitrate pool is subjected to denitrifica- tion and leaching. Mineral nitrogen movements are simulated using the convection-diffusion equa- tion. The soil temperature is simulated using the Fourier’s law, with the upper boundary condition set to the air temperature; the thermal conductiv- ity is expressed as a function of soil water content, following De Vries 1963. The model was analyzed for parameter sensitiv- ity, calibrated and validated on experimental re- sults obtained on a sandy-loam and a loamy soil in the Western Po valley Northern Italy in open field conditions using ceramic cups Ducco, 1997; Zavattaro, 1998 and in lysimeters Acutis et al., 1997a,b. Some examples of calibration and vali- dation of this model, for the water and the nitro- gen sections, are reported in Fig. 1. Ducco 1997 and Acutis et al. 1997a,b re- ported that the most sensitive inputs to predict nitrogen leaching of a particular cropping system using this model are: a the hydrological charac- teristics of the soil in particular the retention curve; and b the climatic variables: weekly evapotranspiration, air temperature and rainfall. These inputs, which are highly variable, were therefore chosen to be stochastically studied. The nitrogen transformation rates were not changed in the stochastic dataset because the sen- sitivity of the simulated N leaching to changes in the humus mineralization and nitrification rates was low. In fact, as Fig. 1 shows, a wide range of values optimized model prediction. The input values, which were not stochastically varied are reported in Table 1. The average soil characteristics are reported in Table 2. As the LEACHN model was not intended to analyze long time series a maximum of 25 years, specific software was developed to use the values of the state variables at the end of a simulation year as initialization values for the following year. In this way, the model was applied to a series of 100 consecutive years. 2 . 2 . Variability associated to climatic data Daily minimum and maximum air temperature, humidity, rainfall and solar radiation meteorolog- ical data were available for a 23-year period in the sandy-loam soil site and for 15 years in the loam soil site. As the length of these series is not sufficient for an assessment of long term environmental effects Morari and Giupponi, 1995, a 100-year climatic data series was generated using the CLIMGEN version 3.05 model Sto¨ckle et al., 1998 for both sites. This weather generator is widely used as it is part of some well-known models Sto¨ckle and Nelson, 1994. The CLIMGEN simu- lator has shown simulations that are close to measured meteorological data in different sites throughout the world as well as in the Po Valley Acutis et al., 1998, 1999. The potential evapotran- spiration, required by LEACHN as input in the form of weekly total, was calculated according to Priestley and Taylor 1972. In the sandy-loam site, where a more complete set of meteorological data were available, the Priestley – Taylor method was compared to the FAO proposed Penman – Mon- teith formula FAO, 1990 and a good agreement was obtained considering weekly totals. Fig. 1. Calibration and validation of the LEACHM model for the sandy loam soil. A: Water section the water retention parameters were calibrated. B: Nitrogen section the humus mineralization coefficient and the nitrification rate were found to be the most influent parameters, and were therefore calibrated. Redrawn from Acutis et al. 1997a,b. Table 1 Input data used in the simulations Units Value Source a Parameters common to all simulations mm Dispersion coefficient l 200 Van Closter et al. 1995 − 3000 kPa Hutson and Wagenet 1992 Minimum root potential l kg − 1 Partition coefficient for urea K d 1 Lotse et al. 1992 l kg − 1 Partition coefficient for ammonia K d 4 Lotse et al. 1992 l kg − 1 Lotse et al. 1992 Partition coefficient for NO 3 K d 0.0008 Bresler ‘a’ coefficient Hutson and Wagenet 1992 adim 0.5 adim Johnsson et al. 1987 Synthesis efficiency factor f e adim Hurnification factor f h 0.2 Johnsson et al. 1987 2.3 Q 10 factor Ramos and Carbonell 1991 adim 0.08 m 3 m − 3 Hutson and Wagenet 1992 Air trapped at soil water saturation day − 1 Urea hydrolysis rate 0.43 Van Closter et al. 1995 day − 1 Denitrification rate 0.001 Tillotson and Wagenet 1982 0.0015 day − 1 Van Closter et al. 1995 Litter mineralization rate day − 1 Manure mineralization rate 0.04 Bergstrom and Jarvis 1991 day − 1 Ammonia volatilization rate 0.4 Lotse et al. 1992 10 mg l − 1 Johnsson et al. 1987 Denitrification half saturation constant 8 Limiting NO 3 NH 4 ratio in solution Johnsson et al. 1987 adim b Parameters for the sandy-loam soil Campbell retentivity function ‘a’ parameter − 6 a kPa Measured 2 a Measured Campbell retentivity function ‘b’ parameter adim 6,20E-02 day − 1 Acutis et al. 1997b Nitrification rate day − 1 Mineralization coefficient for humus 5,76IE-05 Acutis et al. 1997b 5.0 CN ratio for humus and microbic biomass Measured adim c Parameters for the loam soil Campbell retentivity function ‘a’ parameter − 11 a kPa Measured 5 a Measured Campbell retentivity function ‘b’ parameter adim 0.01 day − 1 Acutis et al. 1997b Nitrification rate day − 1 Mineralization coefficient for humus 1,131E-05 Acutis et al. 1997b CIN ratio for humus and microbic biomass 5.8 adim Measured d Parameters for maize Potential max N uptake 400 kg ha − 1 year − 1 Measured 1.1 Maximum value of Et c ET ratio Measured adim e Parameters for Italian ryegrass Potential max N uptake 220 kg ha − 1 year − 1 Measured 1 Measured Maximum value of Et c ET ratio adim a Average values. 2 . 3 . Variation of the hydrological parameters The scaling factor approach was used to insert a stochastic soil variability into the model. If the soil behaves like geometrically similar media, ac- cording to Miller and Miller 1956, the soil water characteristic curves and the hydraulic conductiv- ity at location i can be related to the h ref u and K ref u curves by: h i = a i − 1 h ref and K i = a i 2 K ref where h and K are, respectively, the water pres- sure and the hydraulic conductivity in location i or in the reference soil, and a i is the scaling factor, that is, the ratio of the microscopic characteristic length of the soil at location i to the same length of the reference soil. Simmons et al. 1979 and Warrick 1990 demonstrated that the scaling fac- tors are applicable even in soils which are not geometrically similar ‘ sensu strictu ’, if the rela- tive saturation is used in place of the volumetric water content. As relative saturation was used in this application, the scaling factor approach was surely appropriate for the present analysis, which was focused on ‘homogeneous soil’. Scaling factors were found to be log-normally distributed, and, if computed separately from pressure curves or from conductivity curves of the same sample, to be strictly correlated. The usual procedure for a stochastic model is to sample the distribution to obtain a number of values 500 – 1000 that are sufficient for a good statistical representation of the distribution. This procedure results in an excessive computational time the time required to simulate 1 year is 1 min with a Pentium II© 266 based computer. As proposed by Braud et al. 1995, it was here decided to use only ten scaling factors, dividing the distribution into ten equiprobabilistic intervals, and to per- form a simulation using only the central value of each class. The pedo-transfer functions of Rawls and Brakensiek 1985 were applied to a set of particle size analysis of eight plots of 1000 m 2 . These functions estimate the water content that corre- spond to seven tensions. In the sandy-loam soil, the mean and the variance of the scaling factors obtained from the use of the pedo-transfer func- tions were compared to those obtained from di- rect measurement of the hu curves, and no significant differences were found P \ 0.2; Ducco, 1997. The ability of pedo-transfer to take spatial variability into account at a decametric scale is also mentioned in Romano and Santini 1997. The values adopted for the mean and standard deviation of the logarithms of scale fac- tors were − 0.06 and 0.41 for the sandy-loam, and 0.055 and 0.42 for the loam soil, respectively. These scale factors were applied to parameters a and b in Campbell’s equations for the water reten- tion function and the hydraulic conductivity Campbell, 1974. 2 . 4 . Cropping systems On the basis of other researches Grignani and Acutis, 1994; Grignani, 1996 it was decided to simulate four maize cropping systems that are frequent in the intensive stock farming of the Western Po plain: 1 maize can be grown contin- uously, yielding a silage production without any return of straw to the soil MM; 2 it can be harvested for grain or spike production, therefore allowing the unharvested aerial part to return to the soil MG; 3 maize can be intercropped with Italian ryegrass LRM: in this case, the winter fodder crop is traditionally harvested in May, late harvest, thus delaying the maize seeding time by approximately 1 month; and 4 a more recent solution has been adopted to increase the forage quality of both crops, with an early harvest of Italian ryegrass April and a consequent anticipa- tion of the maize seeding ERM. Nitrogen up- take is higher in spring for LRM and in autumn for ERM. Fertilization levels were simulated at 200, 300 and 450 kg N ha − 1 for silage maize MM, and at 200 and 300 kg N ha − 1 for the other cropping systems MG, LRM, ERM. These simulated ni- trogen applications correspond to the range of regular agricultural practices in the studied area. In all cases 100 kg N ha − 1 was considered to be applied as urea in the middle of June and the complement computed considering the total amount of nitrogen in the manure was applied earlier as liquid manure in spring, just prior to soil tillage. Table 2 Physical and chemical characteristics of the two soils Loam Sandy-loam 2.3 Coarse sand 7.5 Fine sand 48.5 61.2 8.6 15.6 Coarse silt 17.6 18.2 Fine silt 4.5 16.0 Clay 6.03 pH water 12.5 7.88 0.90 Organic carbon 0.70 0.18 0.12 Total nitrogen 5.8 5.0 CN 1.39 Bulk density g cm − 3 1.25 All simulations were run by considering the crops to be irrigated at three input levels, as in Sto¨ckle et al. 1997: a ‘no irrigation’: rainfed maize; b ‘deficit’: maize was supplied with 50 mm when the difference between the cumulated Et and the rain reached this value; and c ‘full’: maize was irrigated through a rigid scheduling which distributed 50 mm every 15 days in July and August, as is often the case in the considered region, where water availability is high. A 100 years were simulated for every cropping system, in each of the ten equiprobable soil hy- drological conditions. The 1000 simulation results available for each cropping system were described in terms of aver- age leaching over the whole period, standard devi- ation and breakthrough probability curve of exceeding given levels of nitrate leaching.

3. Results and discussion