*3.2. Nitrate Status Classification Performance*

According to the contingency table shown in Section 2.4 (Table 2), 85% of the assessed SW rivers are classified as True Positive (*TP*), indicating that simulated and observed values match in a good nitrate status; whereas 4% are classified as True Negative (*TN*), which indicate river sections with poor status in observations and simulations. The remaining SW rivers do not coincide in the classification of nitrate status in the simulated and observed data series (Figure 4).

**Figure 4.** Classification of the median nitrate concentration in surface water bodies in the Júcar RBD using the contingency table.

The indices obtained using the contingency table are summarized in Table 3. The Accuracy (*ACC*) ranged from 0.70 to 0.99 and was close to the optimal, indicating that the model can reliably represent the nitrate status. The BIAS indicator showed that the nitrate status in 78% of the systems is unbiased or slightly biased. The greatest BIAS was obtained in Vinalopó. The Success Ratio (*SR*) shows the proportion of *TP* and ranged from 0.90 to 1.0 for all systems, except for Vinalopó. According to the calibration in this system, the models tend to underestimate nitrate concentration, therefore, the *FP* rate is high and there is a low *TP* rate. In contrast, the highest *TP* rates were obtained in Mijares and Palancia. This indicates that SW rivers in these systems are properly classified in good status in the simulation.


**Table 3.** Indexes obtained from the 2 × 2 contingency table for the water resource systems (*ACC*: Accuracy; *SR*: Success Ratio; *SP*: Specificity).

The *SP* indicator shows the rate of SW rivers correctly simulated as poor status (*TN*). Values of *SP* between 0.22 and 0.46 were obtained in Mijares, Turia, Júcar, and Serpis; whereas this indicator was zero (the worst value) in the Palancia and Vinalopó. In the case of Palancia, this is attributable to the fact that there are no SW rivers in poor status, whereas in Vinalopó 15% of the SW rivers are impacted in the observed data series, which were not properly represented in the simulation.

Integration of the PATRICAL and RREA models accurately simulated the SW rivers with good and poor nitrate status in Mijares, Palancia, Turia, Júcar, and Serpis. In Vinalopó, the simulation did not represent the SW rivers in poor status, meaning that the simulated skill of the models must be improved to increase the *TN* rate. The difference between simulated and observed data may correspond to unassigned discharges to water bodies, since the simulations are influenced by the number of associated water bodies and the availability of data in small basins.

These results highlight that the contingency table is a useful method to evaluate the behaviour of the models in the classification of the pollutant status in a catchment, since an appropriate classification is more important than an accurate simulation of the pollutant concentration. If the indicators obtained from the contingency table are far from the optimal values, the simulation is not representing the real status of the water bodies.

## *3.3. Nitrate Transfer from GW into Rivers*

The contribution of nitrate transfer from GW into the rivers network (Figure 5c) was characterized by the GW discharge into the river (Figure 5a) and the nitrate concentration in GW (Figure 5b). Modelling the interception behaviour of streams, aquifers, lakes, wetlands, and springs allowed identifying aquifers that discharged or not to the surface. As a result, it was found that 9% of the district aquifers provided a high nitrate transfer to the rivers. The Júcar and Turia are affected by the presence of aquifers with concentrations above 25 mg NO3 <sup>−</sup>/L and discharges to rivers from aquifers over 5 hm3/year. The areas with the highest nitrate transfer in the district are in the middle zone of Júcar (Mancha oriental aquifer); lower zone of Júcar (Caroch Sur and Plana Valencia aquifers); and upper and middle zones of Turia (Alpuente aquifers). The coastal strip of the Júcar RBD is one of the most affected, due to high volume (20 hm3/year) and heavily polluted (NO3 − > 50 mg/L) discharges from aquifers.

**Figure 5.** Groundwater discharge into surface water (SW) (**a**), nitrate concentration in groundwater (**b**), classification of the contribution of groundwater (GW) nitrate to surface water flows, and nitrate concentration status in surface water (**c**).

Nitrate transfer to rivers was classified as medium in 7% of the aquifers. The middle and downstream part of the Vinalopó River presented discharges lower than 5 hm3/year with a concentration in the aquifer above 25 mg NO3 −/L. Discharge from GW can be up to 25% of the total flow per year due to low streamflow in the river. More than half of the aquifers (63%) provided a low nitrate transfer to rivers, although the discharge volume to the river is high, the concentration is below 25 mg NO3 −/L. The remaining 21% of the aquifers provide extremely low or no nitrate transfer to rivers. The influence of GW on nitrate concentration varies from low to none in Marina Baja and Marina Alta because many SW rivers are considered losers.

The monthly mean nitrate concentration in the SW rivers and GW along the main axes in the Júcar and Turia rivers are shown in Figure 6a,b, respectively. The Júcar River has the largest catchment area and the greatest flow contribution of the whole district, with a total length of 509 km for the main axis (Figure 6a). In the upstream and midstream (headwaters—438 km), the nitrate concentrations in aquifers and rivers (observed and simulated data) are below the threshold for good status. In the downstream (454 km—mouth in the Mediterranean Sea), the median nitrate concentration in the river increases near the threshold and is exceeded in some SW rivers. Simulated and observed concentrations in the third and fourth quartiles are above the threshold in the SW rivers. Simultaneously, there is a sharp increase in the median nitrate concentrations in the aquifer (Plana Valencia), reaching a poor nitrate status.

The Turia River is the second with the largest area and flow contributions of the Júcar RBD. In the upstream and midstream, the nitrate concentration is below the threshold of good status in the SW rivers and aquifers. In the downstream, the mean nitrate concentration in SW rivers rises abruptly without exceeding the threshold of good status. However, the concentrations obtained in the third and fourth quartiles do exceed them in some sections. Concurrently, a sharp increase in the median nitrate concentration in the aquifers Plana of Valencia and Liria-Casinos reached a poor status (Figure 5b). This behaviour is similar to the Júcar River. In Júcar and Turia, a simple linear regression between nitrate concentration in SW and GW was performed (Figure 6c,d), considering that the two variables are measured independently. For this purpose, the median of these variables was obtained for each SW-river with a gaining relationship between river and aquifer in the main river axis. This regression was useful to adjust parameters and improve the suitability between observed and simulated indicators.

A direct correlation was found between nitrate concentration in the river and aquifers in Júcar (r<sup>2</sup> = 0.9; Figure 6c) and Turia (r2 = 0.8; Figure 6d). This finding supported the classification of the contribution of GW nitrate to SW presented in Figure 5c. The median nitrate concentration in the main course of the Júcar and Turia rivers is considerably higher in the aquifer (29.7 mg NO3 −/L and 23.3 mg NO3 −/L, respectively) than in the river (5.8 mg NO3 −/L, and 7.8 mg NO3 −/L, respectively).

Most of the SW−rivers in poor status (NO3 − > 25 mg/L) have a high to medium nitrate transfer from aquifers. Therefore, in these areas of the Júcar RBD, there is a direct correlation between nitrate transfer from GW and poor nitrate status in rivers. However, the proportion of this correlation depends on the GW discharge into the river, the nitrate concentration in GW, and the relationship between SW-GW. The effects of the nitrate transfer from the aquifer to the rivers have been previously analysed in Mediterranean areas [50] and other parts of the world [24,51], where an increase of nitrates was found in rivers located in areas with high discharge from polluted aquifers. This demonstrates the need to use simulation models that include SW−GW interactions, what is particularly important in arid and semi−arid areas, such as the Júcar RBD.

Simulation suitability adequately represented changes in the median nitrate concentration along the river length in both simulated and observed datasets. However, the first and third quartiles did not always fit, suggesting a change in the model parameters to adjust the minimum and maximum for the representation of extreme events. Finally, nitrate concentrations in the rivers and aquifers displayed a tendency to increase from the upstream to the downstream, except with the Júcar system midstream (also polluted), as presented by the authors in the Refs. [36,52,53].

**Figure 6.** Monthly nitrate mean concentration observed in Júcar and Turia rivers (box squares without including outliers), simulated in rivers (continuous line, first and third lower and upper shaded quartiles, respectively), and observed in aquifers (continuous line with dot markers) in the main river course of the Júcar (**a**), and Turia (**b**) rivers. Linear regression for variables NO3\_SW and NO3\_GW in the gaining SW rivers in the Júcar (**c**) and Turia systems (**d**).
