**2.5. Calibration process and data analysis**

**Scenarios of adjusted agricultural practices Average fertiliser application (kg elemental N/ha per year)**

B. 6 without corn (and with soya) – 76 – Integrated vegetable crop rotation 1 162 – – Integrated vegetable crop rotation 2 213 – – Integrated vegetable crop rotation 3 122 – –

C. 7 Grassland—four‐cut‐BMP 182 Grassland—three‐cut‐BMP 122 Grassland—two‐cut‐BMP 47 Grassland—one‐cut‐BMP 0 D. 14 Average cattle rotation 183 Average pig rotation 170 Average arable crop rotation 160 Average permanent grassland 434 E. 18 Cattle rotation no livestock manure 147 Cattle rotation no maize (with soya) 87 Pig rotation no livestock manure 153 Pig rotation no maize (with soya) 37 F. 22 Organic vegetable rotation 47 Organic field crop rotation 60 G. 24 WPZ I—cattle 134 WPZ II/WPZ III—cattle 177 WPZ I—pig 144 WPZ II/WPZ III—pig 170 WPZ I—arable crop 142 WPZ II/WPZ III—arable crop 160 WPZ I—permanent grassland 191

31 WPZ II/WPZ II—permanent grassland

120 Water Quality

regime); II, narrow WPZ zone; III, wider WPZ zone.

**Table 2.** Agricultural land management scenarios.

**Dobrovce Maribor Ptuj**

415

*Key:* ca, cabbage; tn, turnip; on, onion; po, potatoes; pe, peppers; or, oilseed rape; cl, clover; co, corn; ww, winter wheat; wb, winter barley; bw, buckwheat; fp, field peas; cas, cabbage for seeds; BMP, best management practices according to guidelines for the scientifically grounded fertilisation [20]; WPZ, water protection zone; I, narrowest WPZ zone (stricter

In the first set (A.) are three scenarios for Maribor and Ptuj, where fertilisation of basic rotation changed depending on the quantity of yield (A. Scenarios 1–3) and one with organic fertiliser (cattle slurry) introduced in to rotation with strictly mineral fertilisers (5). For the location of Dobrovce, organic fertilisers are replaced by mineral (Scenario 4). In the second set (B.) is one scenario for the Maribor rotation, where soya replaced corn (Scenario 6) and three scenarios for Dobrovce with alternative vegetable rotations, with one legume as a main crop and winter greening (Scenarios 11–13). In the third set (C.) are four scenarios including four‐cut, three‐cut, two‐cut and extensive one‐cut (no fertilisers) grassland (Scenarios 7–10). In the fourth set (D.) are average cattle/dairy, pig, arable rotation and for the research area typical permanent Model‐testing procedures were carried out on daily level for all three research locations. Simulation period was split on warm‐up, calibration and validation period. Warm‐up period was excluded from comparison due to model setting up the water and nutrient cycle balance. Model calibration and validation were performed with comparison of measured and simulated soil water content data at research location. These were the only available data for testing whether water cycle in the soil profile is functioning adequately. Calibration and validation periods are as follows: Ptuj December 2011‐March 2012 and April‐May 2012, respectively, Maribor November 2011 and December 2011, respectively, and Dobrovce July‐August 2011 and August to September, respectively. Parameters for automated and manual calibration were selected based on sensitivity analysis tool in ArcSWAT [21] and expert knowledge of the research area. Ten parameters were selected, including CN2, ESCO, GW\_REVAP, REVAPMN, CANMX, FFBC, SOIL\_BD, SOL\_AWC, SOL\_K and SOL\_ALB. Sensitivity analysis, calibration and validation procedures for these three locations are in‐depth explained in previous publication [22].

Model performance was determined with comparison of measured and simulated time series via graphical or visual comparisons and objective function called percent bias (*PBIAS*) [23]. It measures the average tendency (higher or lower) of simulated values to be different than observed ones. Negative *PBIAS* values mean excess water in simulation and positive values mean lack of water in simulation. Visual comparison was used due to important share of rocks in the soil which impact probes measurements of soil water content at the research locations. Simulated values were acceptable if they fall within minimum‐ and maximum‐measured values.

To obtain useful and informative results, simulation of base and alternative management scenarios was run for a period of 12 years (2000–2011) for all three locations. The first three warm‐up years (2000–2002) were excluded from result analysis. Results were analysed on the basis of hydrological response unit (HRU) obtained from SWAT OUTPUT.HRU data file on daily, monthly and annual level for the period of 9 years (2003–2011). Results include analysis of nitrogen balance for base and alternative agricultural land management scenarios. The main model output variables for nitrogen balance are nitrogen fertiliser applied (N\_APP), N added to soil profile by rain (NRAIN), N fixation (NFIX), fresh organic to mineral N (F‐MN), active organic to mineral N (A‐MN), active organic to stable organic N (A‐SN), denitrification (DNIT), plant uptake (NUP) and N leached from the soil profile (NO3L). All variables are expressed as kilograms of N per hectare (kg N ha−1).

Wilcoxon rank‐sum non‐parametric test was used for the detection of significant differences between base and alternative scenarios. We compared the average annual values of two independent samples of equal size (*n*1 *= n*2 *=* 9). The results of alternative agricultural land management scenarios are statistically significantly different from base situation, if the Wilcoxon test value exceeds 62 at *α* = 0.05 or 70 at *α* = 0.20.
