Nitrogen Agronomic Efficiency and Estimated Losses in Potato with Enhanced-Efficiency Fertilizers

ABSTRACT The objectives were to i) assess the effect of enhanced-efficiency nitrogen (N) fertilizers (EENF) [maleic itaconic acid copolymer (NSN) or 3,4-dimethylpyrazole phosphate (DMPP)] and urea on potato (Solanum tuberosum L.) tuber yield, N agronomic efficiency (NAE), N recovery efficiency in tuber (NREtuber) and plant (NREplant), N physiological efficiency, residual inorganic N in soil at harvest (Nresidual) and N losses (Nlosses) and ii) determine the impact of the amount of drainage water on NAE and Nlosses. On-farm experiments were conducted in seven sites with two fertilizers (EENF and urea) and two N rates (0 and 100 kg N ha−1). A N mass balance was used to calculate Nlosses. At two sites (one with NSN and one with DMPP), tuber yield response to N was greater with EENF than urea (avg. 5.0 Mg ha−1). NAE, NREtuber and NREplant were 17%, 31% and 25% higher with EENF than urea, respectively. These efficiencies decreased as drainage water increased. The estimated Nlosses were 12% lower with EENF, being the reduction particularly relevant with increasing drainage water. Therefore, using EENF in combination with irrigation management that ensures low drainage water amounts is essential for maximizing the fertilizer use efficiency and minimizing Nlosses in potato production systems.


Introduction
Potato (Solanum tuberosum L.) is the fourth most consumed crop worldwide after rice (Oryza sativa L.), wheat (Triticum aestivum L.) and maize (Zea mays L.) (FAO 2020). Potato is characterized by a high nutrient requirement, especially nitrogen (N) (Zebarth and Rosen 2007). Root mass of potato plants is concentrated within the upper 30 cm of the soil (Stalham and Allen 2001). Due to the shallow root system development, N use efficiency of the fertilizer (i.e. N agronomic efficiency, NAE) and especially the N recovery efficiency in the plant (NRE plant ) and tuber (NRE tuber ) are usually low (Zvomuya et al. 2003;Zebarth and Rosen 2007).
Nitrogen losses (N losses ) to the environment occur by three main mechanisms: leaching, denitrification and volatilization (Zebarth and Rosen 2007;Jat et al. 2012). Heavy rainfall events during the growing season and an excess of irrigation can reduce the fertilizer recovery efficiency (Ziadi et al. 2011;Giletto and Echeverría 2013) due to increase in N losses mainly by NO 3 − leaching. In potatoes produced under irrigated conditions, NO 3 − leaching is the main mechanism of N losses , representing approximately 85% of the total N lost (Zebarth and Rosen 2007). Also, excessive N applications CONTACT Claudia M. Giletto cgiletto@mdp.edu.ar; Nahuel I. Reussi Calvo reussicalvo.nahuel@inta.gob.ar Facultad de Ciencias Agrarias, Universidad Nacional de Mar del Plata, Balcarce, Buenos Aires, Argentina Supplemental data for this article can be accessed here increase the residual NO 3 --N content in soil at harvest (N residual ), as well as N losses , increasing the risk of environmental pollution (Jat et al. 2012;Giletto et al. 2019). Therefore, N management strategies are required to optimize crop yield and minimize N losses , achieving sustainable production systems.
The best nutrient and fertilizer management practices involve applying the correct rate, source, placement and timing (Bruulsema et al. 2008). Selecting the proper source of N fertilizer would be a promising alternative to improve the synchronization between N supply and demand during the potato growing season. Urea is the main N source used in potato and is characterized by its susceptibility to losses from the soil-crop system. Therefore, the use of enhanced-efficiency nitrogen fertilizers (EENF) could increase the NRE and NAE by improving the synchrony between N supply and demand (Hyatt et al. 2010;Ba et al. 2021) and consequently reducing N losses (Zebarth and Rosen 2007). The EENF have different formulations, and some of the commonly used are maleic itaconic acid copolymer (NSN; Nutrisphere-N®) and 3,4-dimethylpyrazole phosphate (DMPP). The NSN inhibits the urease activity by forming nickel and copper ion complexes required for the formation and function of the enzyme (Hopkins et al. 2008;Franzen et al. 2011); thus, NSN reduces ammonia (NH 3 ) volatilization. The DMPP acts on the first step of nitrification process, the oxidation of ammonium (NH 4 + ) to nitrite (NO 2 − ). Particularly, the DMPP inhibits ammonium oxidation for 6 to 8 weeks, having a bacteriostatic effect on Nitrosomonas (Jat et al. 2012).
The urea with NSN (urea-NSN) has been evaluated in crops such as wheat (Franzen et al. 2011), sugarbeet (Beta vulgaris L.) (Norton 2011) and maize (Barbieri et al. 2010) with contrasting results regarding its effect on yield and NAE. In potato, Hopkins et al. (2008) demonstrated that using urea-NSN increased potato yield for Idaho conditions. Even though there is some evidence of a better performance of urea-NSN over urea on potato N nutrition, the information regarding the impact of urea-NSN on NAE and N losses is scarce, especially for Mollisols of humid regions.
For urea with DMPP (urea-DMPP) in irrigated maize, lower nitrification rates were reported (Barrios et al. 2012) and a reduction in NO 3 − leaching compared to urea (Díez-López et al. 2008). In potato, studies conducted in tropical sandy soils from Brazil showed that the use of urea-DMPP increased tuber set and yield compared to urea (Souza et al. 2019). However, information regarding the use of urea-DMPP in Mollisols is limited, especially concerning N losses and their relationship with drainage water.
The scarcity of studies testing the effect of EENF on NAE and its components [NRE plant and N physiological efficiency (NPE)] and explaining variations in N losses due to water excesses led to the establishment of the current work. Moreover, most of the studies on EENF in potato were conducted on coarse-textured soils with low (<30 g kg −1 ) organic matter contents (Zvomuya et al. 2003;Wilson et al. 2009;Cambouris et al. 2016;Souza et al. 2019). Soils from the Argentinean Southeastern Pampas are characterized by medium textures, with relatively high (>40 g kg −1 ) organic matter contents (Rubio et al. 2019). Under these conditions, a lower risk of NO 3 − leaching is expected, and consequently, there may be lower benefits on yield and NAE by using EENF (Zebarth and Rosen 2007). As far as we know, there is no information on the use of EENF in medium-textured Mollisols of humid regions, which needs to be addressed for selecting the best fertilizer management practices in potato cropping systems with the mentioned characteristics.
The objectives of this study were to i) assess the effect of EENF (urea-NSN and urea-DMPP) compared to urea on potato tuber yield, NAE, NRE tuber , NRE plant , NPE, N residual and N losses and ii) determine the impact of the amount of drainage water on NAE and N losses .

Field experiments
A total of 7 on farm field experiments were conducted during 2009/10, 2010/11, 2012/13 and 2018/ 19 seasons (Table 1). The experiments were located in the Argentinean Southeastern Pampas (35-39° S, 58-59 °W) on Typic Argiudoll soils (USDA 2014). The climate in the area is humid and mesothermal. The annual mean precipitation is 950 mm, and the mean air temperature is 13.5°C. Textural classes of the evaluated soils varied from loam (240 g kg -1 clay, 330 g kg -1 silt and 430 g kg -1 sand) to sandy clay loam (207 g kg -1 clay, 238 g kg -1 silt and 555 g kg -1 sand). The geographical area studied allowed us to explore a wide range of climatic and edaphic conditions, including different N availability levels (Table 1).
In each experiment, two N sources (EENF and urea) at a rate of 100 kg N ha −1 and a N-unfertilized control were evaluated. The N rate applied represents the average of minimum N rates that maximize tuber yield in the studied area (Giletto et al. 2007;Giletto and Echeverría 2013). Treatments were arranged in a randomized complete block design with three replications (plot size 6.8 m long by 5 m wide; ridge size 0.70 m wide by 0.25 m height). At sites 1, 2 and 3, urea-NSN (Specialty Fertilizer Products, LLC, Leawood, KS, USA) was used, while at sites 4, 5, 6 and 7, urea-DMPP (Profertil S.A., Buenos Aires, Argentina) was the fertilizer tested. In all situations, the product was sprayed uniformly on the urea at 1.9 L Mg −1 for NSN and 3.3 L Mg −1 for DMPP, based on the recommendation from the manufacturer. All the N fertilizers were surface banded on both sides of the row in two moments, 50 kg N ha −1 at planting and 50 kg N ha −1 at tuber initiation, following the recommendation for the studied area. After both fertilizer applications, 10 mm of irrigation water was added.
At planting, 50 kg P ha −1 calcium triple superphosphate (0-20-0) was applied to all plots to avoid phosphorus deficiency (P-Bray ranged from 14.2 to 86.2 mg kg −1 ). Potassium fertilizer was not applied since it was not a limiting nutrient in the studied area (exchangeable soil potassium was greater than 400 mg kg −1 ; Sainz Rozas et al. 2019). Planting was performed with a two-row planter at a density of five cuts per meter. In all experiments, the cultivar used was Innovator. Disease and insect control were performed by applying specific products, and chemical and mechanical methods were used to control weeds. Irrigation was applied by center-pivot from the critical period (45-50 days after planting) until maturity to ensure adequate water supply. The irrigation schedule was defined by the farmers based on daily evapotranspiration and soil water deficit estimation. Additional information on crop management and some soil characteristics are described in Table 1.

Soil and plant sampling and measurements
Soil samples composed of 20 subsamples were taken in each block at planting at depths of 0-20, 20-40 and 40-60 cm. Samples were oven-dried for 72 h at 40°C and grounded (2 mm). At a depth of 0-20 cm, pH (1:2.5 soil:water ratio), organic carbon (Walkley and Black 1934) and N mineralized in a shortterm (7 days) anaerobic incubation (Nan) (Keeney 1982) were determined. This last method consisted of incubating 10 g of soil in a stoppered tube under waterlogged conditions for 7 d at 40°C. The released NH 4 + -N was quantified by steam microdistillation, and the initial soil NH 4 + -N content was subtracted from NH 4 + -N determined after the incubation. Only NO 3 --N and NH 4 + -N concentrations (Keeney and Nelson 1982) were quantified in the three soil layers (i.e. 0-20, 20-40 and 40-60 cm).
The NO 3 --N and NH 4 + -N content in the soil at planting (N initial ; kg ha −1 ) in each experiment was calculated using an estimated bulk density following the model proposed by Hollis et al. (2012). The N mineralization potential (N 0 ) was estimated from Nan, according to Echeverría et al. (2000), using the equation N 0 =1.37×Nan+83.17 (r 2 =0.65). Subsequently, N 0 was corrected by temperature and soil water content to calculate the field N mineralization (N mineralized ) as N mineralized =N 0 × k × Y, where k and Y integrate the effects of temperature and soil water content throughout the cropgrowing season, respectively (Echeverría et al. 1994). In detail, k was calculated with the equation k = 10 (8.38+log 0.008)-(2580/°K) , where °K represents the average daily mean air temperature for the evaluated period in degree Kelvin and Y was calculated as Y = 4.7+0.93×[(W-W 0 )/(Wmax-W 0 )], where W is the mean gravimetric water content (0-20 cm) for the evaluated period, estimated from the soil water balance, and Wmax and W 0 are the gravimetric water content at −0.01 and −4 MPa, respectively. Wmax and W 0 were determined for each site by using a tension table and pressure plate extractor, respectively.  N mineralization during the crop-growing season estimated from Nan (Echeverría et al. 1994(Echeverría et al. , 2000. a 0-20 cm b 0-60 cm At physiological maturity, plant samples were harvested at 3.2 m 2 from the two central rows. The samples were separated into tubers and shoots, and each fraction was dried and weighed. The samples were ground (0.84 mm), and N concentration was determined on each fraction with a LECO TruSpec CN analyzer (LECO Corp., St. Joseph, MI, USA). In addition, soil samples were taken at 0-20, 20-40 and 40-60 cm depth in each plot to quantify the residual NO 3 --N and NH 4 + -N content (N residual ) as previously mentioned.

Nitrogen attribute calculations
N losses during the crop-growing season up to a soil depth of 60 cm were estimated by the partial N mass balance as previously used by Giletto and Echeverría (2013), Meisinger et al. (2008) and Giletto et al. (2019), where N losses is the sum of N lost through leaching, volatilization and denitrification, N initial is the NO 3 --N plus NH 4 + -N content in soil (0-60 cm) at planting, N fertilizer is the amount of N applied through fertilization, N mineralized is the field N mineralization, N irrigation is the amount of N applied via irrigation (avg. NO 3 --N concentration was 4 mg L −1 ; Costa et al. 2002), N plant is the N accumulated in the whole plant (shoot plus tuber) and N residual is the NO 3 --N plus NH 4 + -N content in soil (0-60 cm) at physiological maturity. All the variables considered in Eq. 1 were expressed in kg ha −1 . Nitrogen contribution from the previous crop was contemplated by N initial and N mineralized , calculated from Nan. It was considered that N concentration in rainwater is insignificant at all sites and has little relevance for agricultural production systems (Álvarez et al. 2015). The roots represented less than 5% of the total biomass of the crop and less than 4% of the total N uptake (Coraspe-León et al. 2009). Therefore, it was concluded that N obtained from the rain and N accumulated on roots were compensated and were of small amounts for the N balance.
At each site, drainage water was determined by the water balance suggested by Della-Maggiora et al. (2002) and validated in different crops (Reussi Calvo and Echeverría 2006;Nagore et al. 2014). The water balance was calculated at the effective root depth (60 cm), Drainage water = rainfall + irrigation -ETc -variation of soil water. (Eq. 2) Rainfall data were obtained from meteorological stations located in the experimental sites. The irrigation was supplied by center pivot, and the amount of water applied was recorded. Rainfall and irrigation data are presented in Supplemental Figure S1. The estimated crop evapotranspiration (ETc) was calculated by correcting the potential evapotranspiration (ET 0 ) by the crop coefficient (Kc) determined for potato (Della-Maggiora 1996). The ET 0 records were obtained from the National Institute of Agricultural Technology meteorological stations and were calculated with the Penman-Monteith model (Allen et al. 1998). The initial soil moisture was measured gravimetrically. The maximum (field capacity) and minimum (wilting point) limits of soil water content were 180 and 60 mm, respectively, typically used for the evaluated soils (Della-Maggiora et al. 2002).
The NAE and its components were calculated as suggested by Dobermann, 2007 where +N and -N represent the N-fertilized and unfertilized treatments, respectively.

Statistical analysis
The data set was split into two based on the type of EENF (sites 1, 2 and 3 with urea-NSN and sites 4, 5, 6 and 7 with urea-DMPP). Then, two analyses of variance (ANOVA), one for each EENF, were performed to evaluate the effect of treatments on tuber yield, N tuber , NAE, NRE tuber , NRE plant , NPE, N residual and N losses using PROC MIXED procedure of SAS (SAS 2002). Treatments and sites were treated as fixed effects, and blocks were nested within sites as random effects. The normality of evaluated variables was tested using the Shapiro-Wilks test, and the homogeneity of variances was evaluated and confirmed by the Levene-test (p > 0.05). Mean comparisons for significant differences were performed with an LSD test (p ≤ 0.05). Orthogonal contrasts were performed to evaluate differences between fertilizers (urea vs. EENF) and between N rates (control vs. N-fertilized). Linear functions were adjusted using the PROC REG procedure (SAS 2002). Significant differences between coefficients of the linear regressions were determined by the parallelism and coincidence test, using a regression model containing Dummy variables. In the regression, the Dummy variable is a binary classification variable with a value of 0 or 1 that is used to identify subgroups in the data set (Kuehl 2000).

Weather and soil conditions
Total rainfall during the potato-growing season ranged from 305 to 507 mm, while irrigation ranged from 108 to 478 mm, depending on the site (Table 1; Figure S1). As a result, drainage water varied between 40 and 400 mm ( Table 1). The daily mean temperature ranged from 11°C to 22°C, and the estimates of ETc during the growing season ranged from 443 to 529 mm. The soil pH varied between 5.4 and 7.0, while soil organic matter, N initial , and Nan ranged from 40 to 64 g kg −1 , 16 to 118 kg ha −1 and 73 to 116 mg kg −1 , respectively. The N mineralized ranged from 148 to 193 kg ha −1 (Table 1).

Yield and nitrogen response
Tuber yield showed a significant interaction between treatment and site (p < 0.05). Thus, each site was analyzed separately (Figure 1). Yield ranged from 45 to 60 Mg ha −1 in the control and from 49 to 70 Mg ha −1 in the N-fertilized treatments. Except at site 7, N fertilization increased the tuber yield, on average, by ~16% (8.4 Mg ha −1 ). At two sites (site 1 with urea-NSN and site 5 with urea-DMPP), the EENF increased the yield by 9% (5 Mg ha −1 ) compared to urea.

Nitrogen agronomic efficiency and its components
There were significant differences (p < 0.05) among sites and treatments on N tuber ( Table 2). The N tuber ranged among sites from 108 to 193 kg ha −1 . On average, N fertilization increased N tuber by 40%. When comparing fertilizers, a significant increase in N tuber was observed with EENF compared to urea (avg. 7.5%), exhibiting the same effect for both products (NSN and DMPP). The NRE tuber , NAE, NRE plant and NPE varied among sites from 0.38 to 0.67 kg kg −1 , 10.7 to 22.4 kg kg −1 , 0.51 to 0.80 kg kg −1 and 20.1 to 33.1 kg kg −1 , respectively ( Table 2). The NRE tuber , NAE and NRE plant in EENF treatments were, on average, 31%, 17% and 25% greater than in urea treatments, respectively (Table 2). In contrast, the NPE was not affected by the N source. To deeply explore differences between N sources, N agronomic and recovery efficiencies were plotted by site relating EENF to urea (Figure 2). Thus, the models fitted for the three variables (NAE, NRE plant and NRE tuber ) had an intercept different from zero and a slope close to 1, indicating that EENF increased NAE, NRE plant and NRE tuber compared to urea.
The NAE, NRE plant and NRE tuber decreased as drainage water increased (Figure 3). The negative linear models fitted between the N efficiencies (i.e. NAE, NRE plant and NRE tuber ) and drainage water varied according to the N source used. That is, the slopes of the regressions were not different between N sources, but in all the situations, the intercept was significantly greater with EENF compared to urea. Therefore, independently of the amount of drainage water, EENF allowed achieving greater efficiencies.

Residual nitrogen
At physiological maturity, N residual in soil (0-60 cm depth) ranged across sites from 17 to 85 kg ha −1 (Table 2). Nitrogen fertilization increased the N residual by 34% (13.5 kg ha −1 ), and there were no significant differences among N sources on this variable. Additionally, N residual in the unfertilized treatment was positively associated with the N mineralized (y = 0.8x-136.5; R 2 = 0.63; p < 0.05).

Nitrogen losses
Significant variations in the estimated N losses were observed among sites and treatments ( Table 2). The N losses ranged from 36 to as high as 158 kg ha −1 across sites. The N losses were, on average, 54% (39 kg ha −1 ) greater with N fertilization compared to the control. Concerning N sources, significant differences in N losses were only observed with urea-DMPP, which reduced the N losses by 11% (18 kg ha −1 ) compared to urea. Interestingly, the urea-NSN also reduced N losses (12%), but the difference was not significant, probably due to the small magnitude (9.4 kg ha −1 ). When N losses from both N sources (EENF and urea) were related, we found a linear relationship between both variables with a slope significantly different from 1 (p = 0.024; Figure 4). Therefore, a general 12% reduction on N losses was obtained for the entire dataset due to EENF. The sites with greater N losses (sites 3, 4, 5, 6 and 7) also have greater drainage water (≥ 190 mm; Table 1). Thus, N losses were strongly associated with drainage water (R 2 ≥ 0.94; Figure 5). According to the Dummy test, the N losses variations with drainage water increments (i.e. slope of the models) were similar among treatments. On average across treatments, N losses increased by 35 kg N ha −1 for every 100 mm of drainage water. For the same amount of drainage water, N losses were greater in the N fertilized treatments than in the control. In addition, the slope of the linear model tends to be greater with urea than with EENF (38 vs. 33 kg N ha −1 every 100 mm of drainage water). As a result, N losses among sources were similar at low drainage water amounts, but they tended to be reduced with EENF when the drainage water increased.

Discussion
Tuber yields and yield responses to N fertilization were within the range previously reported for Argentina (Giletto et al. 2019(Giletto et al. , 2020 and other potato production regions (Hyatt et al. 2010;Cambouris et al. 2016;Ghosh et al. 2019b). In our study, EENF increased tuber yield compared to urea in two of the seven sites. We could not find a clear explanation of the factors controlling the improvement in tuber yield caused by EENF at these sites.  increments due to EENF, but only in some of the evaluated sites and years. Despite the lack of widespread benefit in yield, our study showed that EENFs are alternative N sources that could increase potato tuber yield in medium-textured Mollisols of humid regions. Additional studies will be needed to understand the role of soil and weather conditions that produce a better performance of EENF over urea.
The EENF generated a greater increase in N tuber compared to urea. Similarly, Wilson et al. (2009) andGhosh et al. (2019b) reported for coarse-textured soils with a lower organic matter and greater N accumulation in the whole plant when using EENF. Differences between N sources on N plant and, consequently, on NRE suggest that EENF achieve a better synchrony between the crop N uptake and N release from the fertilizer.  The values of NRE plant we observed (avg. 0.66 kg kg −1 ) were slightly higher than those reported for sandy soils (avg. 0.55 kg kg −1 ) receiving similar N rates and water availability (Shoji et al. 2001;Zvomuya et al. 2003;Cambouris et al. 2016). The difference in NRE between studies suggests that soil texture could be one of the main variables affecting N losses from fertilizers in potato production systems. Therefore, soil textural class should be considered in N management strategies to optimize fertilizer N use efficiency. In addition, Kelling et al. (2011), Cambouris et al. (2016 and Souza et al. (2019) determined for sandy soils that NAE was improved with EENF compared to urea, which could be explained by a more appropriate balance of N forms in the soil (Souza et al. 2019). Our results demonstrated that using EENF is also a suitable option to improve NAE in medium-textured soils (loam to sandy clay loam) from Mollisols of humid regions.
The NAE, NRE plant and NRE tuber decreased as drainage water increased. Furthermore, EENF improved these three parameters compared to urea, independently from the amount of drainage water. Thus, while leaching seems to be the main mechanism explaining differences in N losses (i.e. great impact of drainage water), other minor losses such as denitrification and volatilization may be involved. In contrast, Zvomuya et al. (2003) and Cambouris et al. (2016) determined in sandy soils that NRE with EENF was higher than urea only with excessive water drainage. The risk of NO 3 − leaching increases with coarser textures and low soil organic matter contents due to reduced soil water-holding capacity (Jiang et al. 2011). This could explain the differences between our results and those observed by Zvomuya et al. (2003) and Cambouris et al. (2016) in sandy soils with low organic matter. Therefore, the NAE via the recovery efficiency is affected by the irrigation management and the soil textural class, both influencing N losses.
Residual soil NO 3 − during potato harvest could be greater with EENF than urea due to the lower N release rate (Zvomuya et al. 2003). However, in agreement with the results shown by Ziadi et al. (2011), we did not observe differences between N sources on N residual . From the N mass balance presented in Eq. 1 could be deduced that, when comparing N sources at each site, variations in N residual are explained by N losses and N accumulated in the plant, as all the other parameters were constant. Therefore, the tradeoff between NRE and N losses explained the lack of differences between N sources on N residual (i.e. NRE was high, and N losses were low with EENF, while the opposite occurred with urea).
Nitrogen losses ranged from 36 to 158 kg ha −1 , and in five out of the seven sites, N losses were above 50 kg N ha −1 , threshold reported by Giletto et al. (2019) to optimize tuber yield contemplating the environmental risk. Also, N losses were reduced by ~12% when using EENF. Three main mechanisms could explain differences between N sources and sites on N losses : i) Volatilization: Ghosh et al. (2019a) reported NH 3 volatilization losses from 1.3% to 8.3% with urea and 0.8% to 2.6% with EENF. Even though NH 3 volatilization amounts could be significant in potato production systems, we consider that this N loss mechanism was irrelevant in our study since 10 mm of irrigation water was applied after fertilization, minimizing NH 3 volatilization losses (Jat et al. 2012 (2001) reported in irrigated maize that denitrification losses with urea were small (~2 kg N ha −1 ), even for high N rates (210 kg N ha −1 ). Therefore, the magnitude of N losses by denitrification is low for N mass balances and could have little impact on the differences in N losses we observed between N sources. Nevertheless, we consider it relevant to keep assessing N sources effect on N 2 O-N emissions due to the high impact of this greenhouse gas on global warming. iii) Leaching: When the soil water-holding capacity is exceeded, the water movement below the root zone increases NO 3 − leaching (Zebarth et al. 2012). Woli et al. (2016) estimated N losses from 80 to 230 kg ha −1 and indicated that the highest values occurred when high N rates were applied in combination with excess irrigation water. In our study, some possible irrigation mismanagement from farmers combined with unpredictable rainfall events after irrigation caused a significant volume of drainage water at some sites. This allowed us to explore a wide range of drainage conditions under real potato production systems, which we consider a strength of our study. Using simulation models, Jensen et al. (1994) determined that NO 3 − leaching represented 85% of the total N lost to the environment in sandy soils receiving average precipitation of 850 mm per year. The strong relationship we observed between N losses and drainage water confirmed that, in our conditions, NO 3 − leaching is the main mechanism explaining N losses in irrigated potato production systems.
Both EENF tested (i.e. NSN and DMPP) delay the transformation of urea to NO 3 − in different steps of this process, and thus, could be effective in reducing NO 3 − leaching losses (Jat et al. 2012). We observed that the reductions in N losses caused by EENF tended to be higher as drainage water increased. Similarly, Zvomuya et al. (2003) found higher NO 3 − leaching losses with urea compared to EENF, especially in situations with a water excess. The NO 3 − lost through the soil profile is accumulated in the groundwater table and is one of the biggest environmental problems generated by agricultural activity. Thus, as stated in our results, the use of EENF and appropriate irrigation management should be considered to reduce N losses without compromising environmental quality.
Our study provided a framework for assessing EENF in field-grown irrigated potatoes and exploring opportunities to adjust current N management practices in cropping systems from Mollisols of humid regions. Even though we provide evidence of NAE improvements and reductions in N losses with EENF, future studies should focus on measuring NH 3 volatilization, denitrification and NO 3 − leaching to confirm our estimations. Also, the information should be validated with different N rates and under several productive conditions such as interacting nutrient limitations (e.g. phosphorus and sulfur), contrasting genetic materials, management practices, soil types and/or environmental situations.

Conclusions
For potatoes grown on medium-textured Mollisols of humid regions and under irrigated conditions, EENF increased N tuber , NAE, NRE tuber and NRE plant compared to urea. Moreover, in two of the seven sites, the tuber yield was also improved by EENF. The NAE, NRE plant and NRE tuber decreased as drainage water increased, independently of the amount of drainage water. On average, N losses were 12% lower with EENF compared to urea, and this reduction tended to be more relevant as drainage water increased. That is, EENF had no significant impact on N losses when the amount of drainage water is low (i.e. well-managed irrigation). Therefore, using EENF in combination with appropriate irrigation management is essential for optimizing the fertilizer use efficiency and minimizing N losses , reducing the potential negative impact of N fertilization on the environment in potato production systems.

Disclosure statement
No potential conflict of interest was reported by the author(s).