Comparison of uniform control and site-specific model-based nitrogen prescription in terms of grain yield, nitrogen use efficiency and economic aspects in a heterogeneous corn field

Vergleich einer einheitlichen (Kontrolle) und einer teilflächenspezifischen Modell-basierten Stickstoff-Düngeempfehlung hinsichtlich Ertrag, Stickstoffnutzungseffizienz und Nettoertrag in einem heterogenen Maisschlag

Materials and methods

Fig. 1. 30-year (1976–2005) mean rainfall (left) and mean temperature (right) and the given weather in the ex­perimental years 2006 (dark green), 2007 (light green), and 2008 (green) at the experimental station.

30jähriges Mittel (1976–2005) für Niederschlag (links) und Temperatur (rechts), sowie das Wetter auf der Versuchsstation in den Jahren 2006 (dunkelgrün), 2007 (hellgrün) und 2008 (grün).

The soil of the experimental site “Riech” is characterized as a heavy calcareous brown earth soils with high clay content (classified as silty clay and silty clay loam), see Fig. 2. The mean pH at the site was 7.2, the mean content of soil organic matter was 2.6%. The contents of phosphorous (32 mg P 100 g^–1 soil), potassium (29 mg K 100 g^–1 soil), magnesium (22 mg Mg 100 g^–1 soil) in the soil were above the optimum content class C defined by VDLUFA standards (Kerschberger et al., 1997; Baumgärtel et al., 1999; Landwirtschaftskammer Nord­rhein-Westfalen, 2012). This means the availability of those nutrients was sufficient and did not require ad­ditional fertilizer applications. The content of boron (0.5 mg B 100 g^–1 soil) was within the optimum content class C (Landwirtschaftskammer Nordrhein-Westfalen, 2011).

Fig. 2. Soil texture determined in 48 management units (grids) at experimental site “Riech”.

Erfasste Bodentextur in den 48 Managementeinheiten (Grids) des untersuch­ten Schlags „Riech“.

Based on the working width of the farm machinery, the experimental site was separated virtually in single management units (grids) with a size of 36 × 36 meters (1296 m²). The data collection was based on the determined management units. Soil samples for texture analyses were sampled in every management unit in three sampling depths (0–0.3 m, 0.3–0.6 m and 0.6–0.9 m). The texture analysis was based on the methods described by VDLUFA (1991). For the comparison of nitrogen management strategies, management units (1,296 m²) were split into two grids: a control treatment with a grid area of 432 m² (12 × 12 m), and a model-based treatment with a grid area of 864 m² (24 × 36 m). The nitrogen application was performed with a pneumatic fertilizer spreader (Rauch Aero 1112, Sinzheim, Germany) with a spreading width of 12 m. The arrangement of control and model-based treatment was random within each row of management units (Fig. 3). In order to avoid blending effects at the border of the management units, a distance of ten meters between each management unit was omitted for data analysis.

Fig. 3. Layout of the experi­mental design used for comparison of a uni­form (control) and a site-specific (model-ba­sed) nitrogen manage­ment strategy in 48 management units (grids).

Anlage des Versuchsde­signs mit 48 Manage­menteinheiten (Grids) zum Vergleich der ein­heitlichen (Kontrolle) und der teilflächenspe­zifischen (Modell-ba­sierten) Stickstoffmanagement-Strategie.

Mineral nitrogen was broadcast in both treatments as split application with the first application before seeding and the second application around 4^th leaf stage (BBCH 14 or Zadoks stage 14; Zadoks et al., 1974). At the first application 30 kg N ha^–1 calcium ammonium nitrate (CAN) was broadcast, at the second application urea was applied. The fertilizer spreader was controlled by a geo-referenced nitrogen application map to ensure and document the given amounts of nitrogen for each treatment and each management grid. Thus, the nitrogen application was performed offline as a mapping approach (Auernhammer, 2001).

The first nitrogen application rate of 30 kg N ha^–1 was broadcast uniformly for all grids within the site before seeding. For the second nitrogen application in the control treatment, a nitrogen application rate of 130 kg N ha^–1 was broadcast, according to the current farmer’s practice, disregarding existing heterogeneity within the site. In the site-specific treatment a model-based nitrogen application was broadcast, ranging from 100–210 kg N ha^–1 for single management grids (Fig. 4).

Fig. 4. Measured amount of nitrogen applied (kg N ha^–1) in uniform (con­trol) and site-specific (model-based) nitrogen treatment in investiga­ted years 2006, 2007, and 2008 (--- document mean N applied, whis­kers document the 5^th and 95^th percentile).

Erfasste Stickstoffgabe (kg N ha^–1) der einheit­lichen (Kontrolle) und der teilflächenspezifi­schen (Modell-basier­ten) Variante in den Jahren 2006, 2007 und 2008 (--- zeigen die mittlere Stickstoffgabe, Antennen zeigen das 5. und 95. Perzentil).

The site-specific model-based nitrogen prescription, which takes the long-term weather conditions into consideration, was developed with the crop growth model APOLLO (Batchelor et al., 2004). This process-oriented crop growth model calculates crop growth and development on a daily basis under consideration of current local conditions. Thus, grid and site based information, such as soil characteristics, weather, cultivar, management and historical yield data were used to calibrate the model for all grids within each site. During the calibration process, adjustments were made to selected soil properties to minimize the root mean square error (RMSE) between simulated and measured historical yield in all grids. Validation was performed by running the model for an independent season and comparing simulated and measured spatial yield, obtaining convincing results for the corn sites. For details and results on the model validation see Link et al. (2006c).

To develop the nitrogen prescription, the calibrated model was run for different nitrogen rates (0–200 kg N ha^–1 in increments of 10 kg) for each grid and for 30 different years of historical weather data (1976–2005) to generate information about the range of yield potential within the site over the long-term. The optimum site-specific nitrogen prescription for the three-year period was computed by maximizing marginal net return (MNR) in each grid within the site by the following equation (1):

MNR_n,t = Y_n,t P_W – N_n,t P_N

where MNR_n,t is net return (€ ha^–1) for grid n and year t, Y_n,t is corn yield (kg ha^–1), P_W is the price of corn (0.10 € kg^–1), N_n,t is the N application rate (kg N ha^–1), P_N is the price of nitrogen fertilizer (0.50 € kg^–1). Note that the computed optimum nitrogen rate depended on the chosen fertilizer and corn prices, as well as on the underlying weather data. In this study the historical weather data from 1976–2005 were taken into account.

For data analysis, collected data were differentiated in control and model-based nitrogen treatment and also categorized based on the underlying soil type. For this purpose, the four texture categories (TC), described in Tab. 1, were distinguished. After the growing season, nitrogen use efficiency (kg grain kg^–1 N available) was determined for both nitrogen management strategies in all management grids. The amount of N available (kg N ha^–1) was calculated for each sampling point on the following equation (2):

N_avail = N_min + N_fert

with N_min = soil mineral nitrogen (kg N ha^–1) before vegetation period in the management unit and N_fert = grid specific total nitrogen application rate (kg N ha^–1). The calculation of nitrogen use efficiency was based on the equation described by Moll et al. (1982), which defines NUE as the quotient of grain weight and nitrogen supply (equation 3):

NUE (kg grain kg^–1 N_avail) = yield (kg grain ha^–1)/
                                                                   N_avail (kg N ha^–1)

The net return (€ ha^–1) was calculated for all management grids to evaluate the different nitrogen management strategies over the three-year period on an economic basis.

Results and discussion

As shown in Fig. 5 the simulated optimum nitrogen application rate for each grid in the site-specific model-based nitrogen treatment was mainly driven by simulation of soil available water in each grid (R² = 0.47).

Fig. 5. Linear regression of si­mulated optimum nitro­gen application rate (kg N ha^–1) against the si­mulated amount of wa­ter availability (cm³ water cm³ soil^–1) in the corresponding manage­ment grid.

Lineare Regression zwi­schen der simulierten optimalen Stickstoff­­applikationsmenge (kg N ha^–1) und der simu­lierten Wasserverfüg­barkeit (cm³ Wasser cm³ Boden^–1) in der je­weiligen Managemen­teinheit.

Also the model intern parameter for water availability correlates with the site-specific nitrogen application rate (kg N ha^–1) over all investigated years (R² = 0.51), showing that the more water that was expected to be available in a management grid, the higher the simulated nitrogen application rate for this grid (data not shown).

The measured amount of available water during the growing season was mainly influenced by content of silt (R² = 0.65). This is consistent with the knowledge that soil texture has a direct impact on water availability in a field (Scheffer and Schachtschabel, 1989; Ehlers and Goss, 2003) and is considered by the authors Shahandeh et al. (2005) for an easy development of site-specific nitrogen prescriptions, as soil texture is easy to measure and consistent over time.

The fact that the model-based nitrogen prescription was mainly driven by soil available water in the site is also in line with results of Miao et al. (2007), who stressed that the optimum site-specific nitrogen rate in corn strongly depends on current weather conditions in each growing season. Thus, when calculating the site-specific nitrogen management, not only the maximum net return over the long-term should be considered (as described in the materials and method section), more attention should be drawn to current weather conditions and thus, on the expected water availability within the actual growing season.

Significant differences in grain yields between nitrogen treatments were not found in all investigated years (Fig. 6). In 2006, the control treatment resulted in slightly higher grain yields compared to the model-based treatment (5,436 and 5,332 kg ha^–1, respectively, P < 0.01), in 2007 and 2008 the situation was contrariwise (6,925 and 7,903 kg ha^–1, n.s.; 5,374 and 6,015 kg ha^–1, P < 0.05). Over the three-year period, the mean grain yield in the control nitrogen treatment (6,183 kg ha^–1) was significantly lower than in the model-based nitrogen treatment (6,749 kg ha^–1, the difference of means was 566 kg ha^–1, P < 0.05). This was, on the one hand, caused by the distinct although not significantly different mean values in grain yield between control and model-based nitrogen treatment in 2007. On the other hand, it might have been due to the fact that across all three years, the mean nitrogen application rate was about 6% lower in the control treatment compared to the model-based treatment. The tendency of slightly higher corn yields in site-specific nitrogen treatment compared to a uniform treatment was also found in other studies (Koch et al., 2004; Miao et al., 2007) indicating that site-specific nitrogen application fit the nitrogen demand of the plant. Also, concerning soil category (SC), significant differences were determined in corn grain yield over the three-year period. Over the whole site in loam management units (TC1), signifi­cantly higher grain yields were reached when compared to silty clay loam and silty loam (TC2) and silty clay (TC3) management units (Tab. 2). As TC1 represents a loamy soil, which is well known as a highly productive soil type (Ehlers and Goss, 2003), this result corroborates the expectations.

Fig. 6. Corn grain yields (kg ha^–1) in uniform (con­trol) and site-specific (model-based) nitrogen treatment in investiga­ted years 2006, 2007, and 2008 (--- document mean corn grain yield, whiskers document the 5^th and 95^th percentile).

Kornertrag (kg N ha^–1) der einheit­lichen (Kon­trolle) und der teilflä­chenspezifischen (Modell-basierten) Va­riante in den Jahren 2006, 2007 und 2008 (--- zeigen den mittleren Kornertrag, Antennen zeigen das 5. und 95. Perzentil).

For corn grain yield, there was no statistically significant interaction between year and nitrogen treatment (P = 0.24), year and TC (P = 0.10) or nitrogen treatment and TC (P = 0.46) (data not shown). The comparison of control and model-based nitrogen treatment within TCs indicated significant differences within silty clay loam and silty loam (TC 2) only. Here, the control treatment resulted in significantly lower grain yields when compared to the model-based nitrogen treatment (5,679 kg ha^–1 and 6,434 kg ha^–1, P < 0.05). This indicates that on silty clay loam and silty loamy soil, the soils with the highest amount of available water (Ehlers and Goss, 2003), the uniform nitrogen application rate was not adequate for corn plants.

However, for both nitrogen treatments only a slight correlation between corn grain yields and nitrogen application rate could be determined over all three years (data not shown). Also, between grain yields and nitrogen availability no strong correlations were obvious (data not shown). These results indicate that except for silty clay loams and silty loamy soil, nitrogen application rate and also nitrogen availability were not the main driving factor for corn grain yields. This is similar to the results of Scharf et al. (2006), who found that optimal nitrogen rates were seldom directly related to yield variability. They concluded that other factors as soil nitrogen supply, like in-season nitrogen loss and nitrogen uptake effi­ciency, also may play an important role for existing yield variability. Besides nitrogen, other nutrients such as P (Wiedenfeld and Provin, 2010), K, Mg are also described as major yield limiting factors in corn. Subedi and Ma (2009) identified weed infestation as a major yield limiting factor, followed by nutrient supply and seeding rate.

As described in the literature (Oberle and Keeney, 1990; Norwood, 2000; Calviño et al., 2003; Schmidt et al., 2011), water was considered as one of the most yield driving factors for corn yield; especially in dry seasons. However, in none of the investigated years, a significant correlation between measured soil moisture and grain yield could be determined. While in 2006 and 2008 the rainfall seems to be sufficient, in 2007 a very dry period at the beginning of the growing season was determined. The fact that soil moisture was not measured continuously, but only at a 14-day interval over the vegetation period, might be a reason that also in dry seasons these measurements seem to be of limited explanatory power for grain yield variability.

In year 2007, nitrogen use efficiency was significantly higher (40.81 kg grain kg^–1 N available) when compared to 2006 (29.48 kg grain kg^–1 N available) and 2008 (31.81 kg yield kg^–1 N available), respectively. A significant difference in nitrogen use efficiency could not be determined between control and model-based nitrogen treatment in any year. Also in 2007, the mean nitrogen use efficiency of the model-based nitrogen treatment was higher when compared to the control nitrogen treatment (39.14 and 42.47 kg grain kg^–1 N available); however the difference was not significant (Fig. 7). With regard to nitrogen use efficiency, no significant advantage of either one of the nitrogen management strategies was visible. However, a lower range in nitrogen use efficiency of the model-based nitrogen treatment compared to the uniform control treatment was visible in 2007 and 2008 (Fig. 7). Against the background that no significant differences in straw and grain nitrogen content was visible in 2006 and 2007 (data not shown) and in 2008 significantly higher nitrogen content in straw and grain were determined in the model-based treatment (straw: 0.65%; grain: 1.44%) compared to the control treatment (straw: 0.61%; grain: 1.24%), an adequate supply of corn plants in the site-specific nitrogen treatment was achieved.

Fig. 7. Nitrogen use efficiency (kg yield kg^–1 N availa­ble) in uniform (control) and site-specific (mo­del-based) nitrogen treatment in investiga­ted years 2006, 2007, and 2008 (--- document mean nitrogen use effi­ciency, whiskers docu­ment the 5^th and 95^th percentile).

Stickstoffnutzungseffi­zienz (kg Ertrag kg^–1 verfügbarem N) der ein­heitlichen (Kontrolle) und der teilflächenspe­zifischen (Modell-ba­sierten) Variante in den Jahren 2006, 2007 und 2008 (--- zeigen die mittlere Stickstoffnut­zungseffizienz, Anten­nen zeigen das 5. und 95. Perzentil).

Over the three-year period, the highest nitrogen use efficiency was determined in grids belonging to loamy grids (TC 1: 42.10 kg grain kg^–1 N available), the lowest efficiency in silty clay grids (TC 3: 32.80 kg yield kg^–1 N available). In grids with silty clay loam and silty loam soils (TC 2) and clay soil (TC 4) about the same nitrogen use efficiency with 34.03 and 34.93 kg grain kg^–1 N available, respectively, was reached. Thus, in this study the nitrogen use efficiency was not directly affected by the underlying soil type.

However, a dependency between measured soil moisture content and nitrogen use efficiency could be shown around BBCH 87 (R² = 0.49) in 2007, and around BBCH 30 (R² = 0.30) and 50 (R² = 0.63) in 2008. In 2006, a correlation was found between nitrogen use efficiency and nitrogen application rate (R² = 0.27), mainly driven by varying nitrogen rate in the site-specific nitrogen treatment. These results indicate that both nitrogen, and water availability effect nitrogen use efficiency in corn.

The highest net return over the whole site was achieved in 2007 (701 € ha^–1), while in 2006 and 2008, the net return was significantly lower (475 and 511 € ha^–1, respectively). In 2007 and 2008 a slightly higher net return was realized in model-based nitrogen treatment compared to the control treatment, but the differences were not statistically significant (Fig. 8). However, over the three-year period a significant difference between both nitrogen management strategies was visible (Tab. 2), indicating a higher net return for the model-based nitrogen treatment. Over all years in about 61% of the management grids higher net returns were achieved with model-based nitrogen treatment.

Fig. 8. Net return (€ ha^–1) in uniform (control) and site-spe­cific (model-ba­sed) nitrogen treatment in investigated years 2006, 2007, and 2008 (--- document mean net return, whiskers docu­ment the 5^th and 95^th percentile).

Nettoertrag (€ ha^–1) der einheit­lichen (Kontrol­le) und der teilflä­chenspezifischen (Modell-basierten) Va­riante in den Jahren 2006, 2007 und 2008 (--- zeigen den mittleren Nettoertrag, Antennen zeigen das 5. und 95. Perzentil).

The highest net return was achieved on loam soils (TC1), independent of the investigated year. In 2006, silty clay loam and silty loam soils (TC2) experienced the lowest net return, while in 2007 and 2008, silty clay soils (TC3) showed the lowest net return. Over the years and the TC, most of the site-specifically treated grids achieved a higher net return than uniformly treated grids (see section N application).

From solely assessing economic input and output (as described in eq. 1), the site-specific nitrogen management seems to therefore be beneficial, especially with regards to high yielding loamy soils. However, for the full statement of the facts, it is important to also state other costs, which are associated with the development of site-specific nitrogen prescriptions (data gathering, model development, and working hours) as well as machinery. Several authors (Babcock and Pautsch, 1998; Thrikawala et al., 1999; Liu et al., 2006) describe the cost associated with site-specific management, and most results indicate that the profitability of site-specific management depends on factors such as current management system, hetero­geneity of soil conditions, crop, variety, etc. Hence, for a final evaluation of economic feasibility of site-specific nitrogen application, many aspects other than purely assessing input and output factors need to be taken into account.

Concerning nitrogen use efficiency, no significant advantage of either one of the nitrogen management strategies was visible. The results indicate, however, that nitrogen use efficiency was mainly driven by nitrogen application rate and water availability in this field. The measured nitrogen content in straw and grain together with a lower range in nitrogen use efficiency for the model-based site-specific nitrogen treatment may point at an adequate nitrogen supply of corn plants across the site in the site-specific nitrogen treatment.

A similar picture was given for both nitrogen strategies concerning the economic aspect. Thus, neither of the nitrogen management strategies was superior when considering the selected criteria: corn grain yield, nitrogen use efficiency and net return. The situation may differ when focusing more on environmental aspects, and looking for example at nitrogen leaching potential of both strategies.

The model-based nitrogen prescriptions showed nonetheless a positive correlation to soil available water. Thus, the better the expected water availability in a grid, the higher the optimum nitrogen rate applied. Using this site-specific model-based strategy the nitrogen application rate was reduced compared to the control treatment in about 36% of the grids. Moreover, an over-fertilization of less productive areas on the site and the risk of nitrogen leaching have the potential to be avoided with the implementation of a site-specific model-based nitrogen treatment. On the other hand, areas of higher produc­tivity (about 74% of the grids) did receive more nitrogen, and thus, do not run into the risk of yield losses due to under-fertilization of these areas. This result indicates that the implementation of APOLLO for calculating site-specific nitrogen prescription for corn production seems to be quite beneficial in terms of utilising an efficient method for nitrogen input. The results indicate that the model-based nitrogen management was mainly driven by simulated water availability. Consequently, great attention needs to be given to current weather conditions and the expected water availability within the current growing season, potentially increasing the additional value of implementing model-based nitrogen prescriptions.

Overall, a further study identifying yield driving factors and analysing environmental aspects is required to come to a final evaluation of uniform and model-based site-specific nitrogen prescription in corn.