The Impact of Sowing Date on Soil Mineral Nitrogen Uptake Efficiency and Fertilizer N Uptake Efficiency for Winter Oilseed Rape (Brassica napus L.) in Ireland

Abstract

Nitrogen (N) application on winter oilseed rape (Brassica napus L.; WOSR) in the mild Atlantic climate of Ireland is based on a soil N Index system, which does not take into account any variations in crop N demand prior to the main spring dressing of N fertilizer. This study tests whether UK- canopy management (CM) principles for oilseed rape N fertilization are applicable for oilseed rape grown in Ireland. The tested principles included (i) final N uptake of unfertilized crops will be equivalent to the soil mineral N (SMN) and N in the crop measured at the end of winter, i.e., soil N uptake efficiency (SNUpE) = 1; and (ii) the (apparent) N fertilizer N uptake efficiency, (FNUpE) is 0.6 (60%). Three years of field trials were carried out from 2017/18 to 2019/20 on different sites based on a split-plot design. Three sowing dates (SD): mid-August (SD1), End-August (SD2) and mid-September (SD3) were set as main plots and five N application strategies namely, CM standard, CM higher yield, CM low N rate, Fix225 and zero-N were set at subplot level. Results from unfertilized plots demonstrated that the ratio of final crop N uptake at harvest to the combined post-winter SMN + crop N was 1.13 and 1.14 on the two early sowing dates but 1.68 for the latest sowing. Additionally, SMN was not systematically impacted by SD, except in site-year-2. Instead, crop N content in spring had the biggest influence when calculating soil N supply for the season. More mineralization occurred through the growing season resulting in SNUpE of above 100% on all SDs. This additional available N (AAN) was quantified as 11 kgN·ha⁻¹ on average for SD1 and SD2 and 38.6 kgN·ha⁻¹ for SD3. FNUpE values were calculated closest to the economical optimum N rate (Nopt) and were 0.534, 0.574 and 0.486, respectively for SD1, SD2 and SD3. The Nopts at each SD were 157 kgN·ha⁻¹ (SD1), 148 kgN·ha⁻¹ (SD2) and 175 kgN·ha⁻¹ (SD3) and the respective yields at each Nopt were 4.93 t ha⁻¹, 4.90 t·ha⁻¹ and 4.34 t·ha⁻¹. This study shows the UK–CM principles were applicable in the mild Irish climate; however, values of SNUpE and FNUpE differ from one site-year-SD to another. Mid-August to early September sowing dates produced higher post-winter crop N content due to SMN uptake, and this results in a lower Nopt than the later SD. Sowing later led to a higher Nopt due to a combination of lower post-winter N uptake and lower FNUpE, although the effect of these factors was partially offset by a greater AAN. Measurements of SNS post-winter, an estimate of N mineralization during the growing season, together with a consideration of the sowing date helped determine the scope for reduction in fertilizer while achieving high yield and high FNUpE. Among different N rate strategies, CM standard and CM high yield were closest to the Nopt for having lower N rates at the maximum economical yield point.

1. Introduction

Winter oilseed rape (WOSR) (Brassica napus L.) is an important broadleaf crop used in cereal rotations bringing benefits in disease, pest and weed control with consequent yield increases in subsequent cereal crops [1,2,3]. In Ireland, 9.5% of the agricultural area is used for annual crops, of which only 10.3% is used for non-cereal crops in rotations [4]. Increasing WOSR production as a break crop would bring greater rotational benefits and the crop has considerable growth and development potential in a wide range of climates [5]. WOSR has a high demand for N with the capacity to take up large amounts of soil mineral N (SMN) before the onset of flowering [6,7]. However, mineralization can hugely affect the crop’s SMN uptake as decomposition of organic N sources and mineralization are largely dependent on soil temperature and moisture [8,9]. Nitrogen use efficiency (NUE) of the crop is defined as the N offtake in harvested material (kg·ha⁻¹) divided by the supply of available N from both the soil and fertilizer (kg·ha⁻¹) [10]. It is important to keep NUE as high as possible to reduce the environmental impact of excess N being exposed to leaching loss and to control N costs [11,12]. Improving NUE is complex but achievable by reducing fertilizer N requirement while maintaining yield or increasing yield without increasing N requirement [13]. In Ireland, mild Atlantic weather and soils with high organic matter (OM) levels increase the likelihood of higher SMN release during the growing season [14]. Climate, soil and cropping history with wet autumns and mild winters in this region contrast with the continental European climate, where long periods of winter frost are typical. Despite this variability due to weather and soil conditions, N application advice for WOSR has been based on limited parameters. A soil N index system, determined by the previous cropping history of the site, is used as a basis for nutrient advice [15] and the maximum allowable N values in this publication are reflected in the implementation of the nitrates directive [16]. There is a scope for a more comprehensive system of N determination to be developed, to include soil and crop N content on a site-year basis, but this requires knowledge on synchronization of the crop N demand, soil N supply and fertilizer N uptake efficiency (FNU_PE) [17].

The sowing date of WOSR is chosen to optimize growth and development in different environmental conditions [18]. In continental Europe, early autumn sowing dates may correspond to dry conditions [19], and later sowing dates may not allow adequate growth for winter survival [20]. In conditions where significant winter frost occurs, WOSR is sown from 20 August to 10 September, and if it is sown closer to mid-August nitrate leaching will be reduced as the plants’ N uptake capacity in autumn is very high [21]. Delayed sowing date has a modest effect on seed yield in the UK due to its maritime climate [22], whereas a reduction in yield was reported in northern Germany [23]. To date, there have been no studies conducted in a mild Atlantic climatic region like Ireland which consider soil N uptake efficiency (SNUpE) and FNUpE in crops of different sizes and sown at different dates. This information would be essential in developing precise predictions of crop N needs.

The canopy management (CM) approach developed in the UK determines N fertilizer recommendations on the estimates of crop N demand, soil N supply (SNS)—defined as the crop N uptake without N fertilizer—and FNUpE—the percentage of (apparent) fertilizer N uptake efficiency [24]. Recommendations based on these factors are described in the AHDB Nutrient Management Guide [25] underpinned by research on the principles of CM in cereals [26] and WOSR [27,28,29]. The principles that the WOSR CM is based on include that the crops must achieve an optimum canopy size (measured as green area index, GAI) by the flowering of 3.5 units. Each GAI unit contains 50 kg N/ha. All SMN and crop N at the start of stem extension, i.e., SNS, is assumed to contribute to the N demand of the crop [26,27,28,29]. Nitrogen fertilizer is assumed to be taken up with an efficiency of 60% (=0.6) on most soil types. In addition, crops with an expected seed yield greater than 3.5 t·ha⁻¹ require additional N fertilizer at a rate of 60 kg N/ha per additional ton of seed [28,29]. While the UK’s climate is influenced by the continental-Atlantic weather systems where WOSR is grown, it is not known whether the CM system for estimating N fertilizer prediction is appropriate for regions with wetter autumns and milder winters such as Ireland.

To address the deficit of a specific N prediction system, this study aims to determine if soil and fertilizer N uptake efficiencies are different in a milder climate than in the UK and other regions; if sowing date influences soil and fertilizer N uptake efficiencies in a milder climate; the amount of additionally available soil N (AAN) that becomes accessible for the crops through the growing season; and whether this is affected by sowing date.

2. Materials and Methods

2.1. Experimental Design

Field experiments were carried out over a three-year period at Teagasc, Oak Park Carlow, Ireland (52°51′ N, 06°54′ E, 62 m above sea level). The WOSR trials were at different locations each year at Oak Park, and the WOSR crops were preceded by winter barley in a continuous cropping rotation. Site details are given in

Table 1. Detailed characteristics of sites, soils, sowing dates and post-winter measurements for each site-year.

Site-Years	Soil Texture 1	pH	Soil Organic Matter (%)	Sowing Date (SD)	Date of GAI Measurement and SMN Sampling
Site-Year-1 (2017/18)	Loam	7.1	4.4	SD1: 18 August 2017 SD2: 31 August 2017 SD3: 19 September 2017	Mid-February
Site-Year-2 (2018/19)	Loam	7.4	5.3	SD1: 15 August 2018 SD2: 28 August 2018 SD3: 14 September 2018	Mid-February
Site-Year-3 (2019/20)	Loam	7.2	4.6	SD1: 15 August 2019 SD2: 28 August 2019 SD3: 21 September 2019	Late-February

¹ Based on USDA textural soil classification.

.

The experiment was a split-plot design. The main plots were the three sowing dates (SD): mid-August, end-August and mid-September allocated in four replicate blocks, with various N strategies applied at a subplot level. Subplot dimensions were 21 m × 6 m, with seeding rate of 50 seeds /m² of the conventionally bred variety, Anastasia. The N management strategy treatments are outlined in

Table 2. Nitrogen application strategies: rates, and timing by year.

N Strategy	Site-Year-1			Site-Year-2			Site-Year-3			N Proportion and Timing 1
	SD1	SD2	SD3	SD1	SD2	SD3	SD1	SD2	SD3
	N Application (kgN/ha)
Fix225	225	225	225	225	225	225	225	225	225	1/3rd before stem extension and 2/3rd at early green bud
CMHiY	170	225	260	224	205	200	200	187	261	As above + 60 kgN/ha at yellow bud
CMStd	110	165	200	164	145	140	140	127	201	1/3rd before stem extension and 2/3rd at early green bud
CMLo	-	-	-	81	61	57	56	44	118
Zero-N	0	0	0	0	0	0	0	0	0

¹ Depending on weather conditions from early March to early April, N rates split more than two proportions within two-week intervals.

including zero-N (unfertilized plots) for all site-years/SDs. Calcium ammonium nitrate (CAN, 27%) was used as the N fertilizer for all experiments and was applied to individual plots using a full-width plot applicator. Other nutrients such as P and K were applied on the basis of regular soil tests for these elements with application rates determined by fertilizer advice from Wall and Plunkett [15]. All other crop inputs such as insecticides, herbicides and fungicides, were applied according to the WOSR reference guide (Teagasc Crop Report, 2019) [30] to prevent other factors from limiting yield.

“Fix225” is the recommended value of 225 kgN/ha for an index 1 soil, where the previous crop is either cereals or maize according to “Major and micronutrient advice” [15] (p. 114). For the CM approaches, spring SNS was calculated from the crop N measurements (from GAI) and SMN before the onset of spring growth. The standard CM, (CMStd), estimates the amount of N that is required for the plant to achieve 3.5 units of GAI or 175 kgN/ha by flowering according to the UK–CM system and targets. A high-yielding canopy management treatment, CMHiY, had an extra 60 kg·ha⁻¹ of N application on CMstd to support 1 t·ha⁻¹ yield over 3.5 t·ha⁻¹. A lesser N was applied to generate a GAI of 2.5 in the UK–CM system with 125 kgN/ha by flowering, namely CMLo which was applied in site-years two and three.

2.2. Measurements

Growth stages were defined using the BBCH growth stage system [31]. The total biomass weight of the separated plant components, leaves and stems (and pods from GS 79 to GS 85) was calculated as the sum of all plant fractions. In site-year two and three, depending on weather suitability for sampling, a total of twelve soil samples were taken separately in early spring (February): four replications at three SD and three depths of 0–30, 30–60 and 60–90 cm were taken. Mineralized N was assessed by determining NO₃–N (mg/kg) and NH₄–N (mg/kg) content using KCl solution extraction [32]. In site-year one, SMN was sampled in four replications at three depths to give an average figure for the whole site but not by individual sowing date.

As harvest approached and seeds changed color to brown, pre-harvest sampling was carried out using a 1m² grid in each plot prior to desiccation. Plants were cut at ground level, weighed and counted in the field. Next, a 20% subsample of plants was taken for weighing stems, leaves and pods separately for biomass. Materials were then oven-dried (75 °C, 48 h), pod walls were separated from seeds, and all components were weighed and ground. Analysis of the N content of all plant components was performed by the Dumas analysis (Rapid N Cube, Elementar Analysensysteme, Hanau, Germany).

Field desiccation with glyphosate reduced seed moisture from less than 30% moisture content (at GS 85 to 99) to an estimated 10% at harvest. Late sowing date crops were desiccated on a later date based on seed maturity. A 2.75 m wide full plot length strip was harvested from the center of each plot using a plot-combine harvester (Deutz Fahr 37.10), with an extended header and vertical cutting knives at both sides. Seed yield was calculated in t·ha⁻¹ at 91% dry matter.

2.3. Calculations and Analysis

Soil mineral N Uptake Efficiency (SNUpE) was measured on zero-N plots and calculated from the total N uptake at harvest as the seed, straw and pod wall N expressed as kg·ha⁻¹. The sum of early spring-measured crop N and SMN was used as SNS (1).

$$\mathrm{SNUpE}=\frac{\mathrm{Harvest}\mathrm{N}\mathrm{uptake}}{\mathrm{Crop}\mathrm{N}\mathrm{in}\mathrm{spring}+\mathrm{SMN}\mathrm{in}\mathrm{spring}\left(\mathrm{SNS}\right)}$$

(1)

To measure the apparent fertilizer N uptake efficiency (FNU_PE), total N uptake from zero-N plots was subtracted from the total N uptake in fertilized plots and divided by the specific N fertilizer rate (2).

$$\mathrm{FNUpE}=\frac{\mathrm{Harvest}\mathrm{N}\mathrm{uptake}\left(\mathrm{fertilized}\mathrm{plots}\right)-\mathrm{Harvest}\mathrm{N}\mathrm{uptake}\left(\mathrm{unfertilized}\mathrm{plots}\right)}{\mathrm{Fertilizer}\mathrm{N}\mathrm{Rate}}$$

(2)

ANOVA analysis was applied separately for each site-year using Genstat 19.0 [33]. A split-plot analysis was used for which sowing dates were considered as main plots and N strategies as sub-plots. Sowing dates and N strategies were counted as fixed factors and blocks were random factors. Heterogeneity of variance of all analyses caused by different sowing dates and site-years were included in the analysis, but no uniform standard error could be given; consequently, Bonferroni paired mean difference analysis was applied.

Analysis was conducted for each individual year and across three years for the core set of common N strategies including: Zero N, Fixed 225, CMStd, CMHiY and additionally CMLo for site-year two and site-year three. Yield and N rates were regressed using linear plus exponential (Lin + expo) curves [34] at separate years and grouped by SDs (3) to determine the economic Nopt for distinguishing the best defined FNUpE.

Y = A + BR^N + CN

(3)

where Y is the yield (t·ha⁻¹), “A”, “B”, “C” and “R” are fitted constants, and N is the nitrogen value. A stepwise regression process was followed for each year in four steps: (1) fitting a common curve on all SDs, (2) fitting separated parallel curves for each SDs (similar slopes) (3) fitting separate curves for each SD (similar intercepts), (4) fitting separate curves for each SD by allowing all parameters to vary. The root mean square error is explained at each stage for the improvement of the fitting procedure over the previous model. If there was no significant improvement between two stages, the previous was taken as the best-fitted lin + expo model.

The economic Nopt (4) was then used as the deviation of (3) for which K was the breakeven ratio, defined as the fertilizer N (price/kg of N) and seed (price/kg of seed) price ratio. A “K” value of 2.44 was used in this study as an average value of eight years (from 2011 to 2018).

$$\mathrm{Nopt}=\frac{\left[\ln \left(\frac{\mathrm{K}}{1000}-\mathrm{C}\right)-\ln \left(\mathrm{B}\left(\mathrm{lnR}\right)\right)\right]}{\mathrm{lnR}}$$

(4)

3. Results

3.1. Weather Data

Table 3. Monthly rainfall (mm) and mean air temperature (°C) at Oak Park during three years of experiment with LTA (1981–2010) with a final rainfall average and a summation of temperature across years.

	Total Rainfall (mm)				Mean Air Temperature (˚C)
	Site-Year-1	Site-Year-2	Site-Year-3	LTA-R	Site-Year-1	Site-Year-2	Site-Year-3	LTA-T
August	62.3	39.8	86.4	73.3	5	16	16.2	15.3
September	92.5	53.7	116	59.5	13	12.8	13.7	13.4
October	62.9	58.3	102	79	11.5	10	9.4	10.5
November	54.2	160	117	72.9	8	8.1	6.3	7.4
December	84.2	119	68	72.7	5.6	8.4	5.9	5.6
January	108	30.9	61.4	62.6	5.4	5.9	5.9	5.3
February	38.7	36.8	172	48.8	3.6	7.5	6	5.3
March	98.1	123	51.8	52.7	4.8	7.4	6.6	6.8
April	73	72.5	29.7	54.1	9	8.9	10.1	8.3
May	24.3	14.1	12.9	59.5	12.5	11	12.3	10.9
June	5.2	55	40.5	66.7	16.4	13	14.3	13.6
July	42.5	42.6	76.5	56.2	17.8	16.7	15.2	15.6
Average Rain (mm)	62.1	67	77.8	63.1	-	-	-	-
Temp. Summation (°C)	-	-	-	-	122	125	122	118

summarizes the precipitation and temperature patterns during the period 1981 to 2010 by comparing long-term average rainfall (LTA-R) and temperature (LTA-T) with weather conditions across the three years (2017/18 to 2019/20) of the experiment at Oak Park (Met Eireann) [35]. In site-year one (2017/8), from mid-autumn (September) to spring, rainfall was above average which was followed by higher than average temperatures in spring and summer. An exceptional summer drought with very low rainfall was experienced in Ireland during the summer of 2018.

IIn site-year two, above average precipitation occurred during November and December followed by a drier January and February (35 mm) with temperature being above average (5.9 to 7.5 °C). This was followed by a wet period with 100 mm of rainfall in March and April (on average), whereas May was the driest month of that year (14.1 mm).

In site-year three, rainfall was above average in autumn causing a delay in the third sowing date. October and November were slightly colder than normal, and February had a rainfall spike of >170 mm with subsequent, slightly warmer than average temperatures in the spring and summer of 2020.

3.2. Plant Establishment

Plant establishment in autumn and plant counts in pre-harvest were within the range of 21 to 35 plants per m² in site-year one for all sowing dates. The range of plant establishment values for different sowing dates in site-years two and three was between 35 and 46 plants per m². Sowing date did not significantly affect plant establishment. In site-year three, plots with severe slug damage were removed from the experiment, but four replications were still retained.

3.3. Crop N, SMN and SNS Measured in Spring

In

Table 4. Effect of sowing date and site-year interactions on early spring measurements (SMN, crop N and SNS), Harvest (final) N uptake and soil N uptake efficiency (SNUpE) for unfertilized plots.

Site-Years	Sowing Dates	Mean Values (kgN/ha)
		SMN 2	Crop N 2	SNS 2	Harvest N 2	SNUpE 2
Site-Year-1	SD1	{31.2} 3	108 a	142 a	125 a	0.888 c
	SD2		73.5 b	104 b	117 ab	1.14 b
	SD3		32.2 c	63.6 c	105 b	1.78 a
p-value		-	<0.001	<0.001	0.017	<0.001
Mean		31.2	71.2	102	113	1.25
Site-Year-2	SD1	21.4 b	55.2 ab	76.6	109	1.42 b
	SD2	25.2 ab	62.6 a	87.9	97.4	1.10 c
	SD3	30.7 a	36.8 b	67.6	114	1.69 a
p-value		0.021	0.031	ns	ns	<0.001
Mean		25.7	52.0	77.0	106	1.41
Site-Year-3	SD1	31.7	54.9 a	86.8 a	91.7 ab	1.06 b
	SD2	38.7	55.0 a	93.0 a	100 a	1.04 b
	SD3	30.1	18.8 b	48.8 b	78.0 b	1.65 a
p-value		ns 1	<0.001	<0.001	0.002	<0.001
Mean		33.0	43.0	76.3	89.4	1.29
All site-years	SD1	28.9	72.9 a	101 a	110 a	1.13 b
	SD2	31.1	63.1 b	94.2 a	108 a	1.14 b
	SD3	30.1	29.4 c	59.6 b	98.0 b	1.68 a
	SD	ns	<0.001	<0.001	0.017	<0.001
	Site-year	0.01	<0.001	<0.001	<0.001	ns
	SD × Site-year	ns	<0.001	<0.001	0.003	ns

¹ defines non-significant level (p-value > 0.05). ² Letters are allocated separately based on years, then based on SD for the three site-years mean values. ³ Average SMN value of the site-year.

, measured spring SMN and crop N, the sum of both as SNS, final (harvest) N uptake and SNUpE are shown for zero-N treatments. It was possible to assess the effect of sowing date on spring soil mineralization in site-years two and three. In site-year two, there was a significant 9.30 kg·ha⁻¹ of N content difference in SMN between SD1 and SD3 (p < 0.05), but no SD effect was found in site-year three (Table 4). Average values of SMN across site-years, separated by SDs, did not show any significant changes either from individual SDs or site-year by SD interactions; therefore, SMN was largely dependent on site-year effect, and its overall range was small from 21.4 kgN·ha⁻¹ in SD1 site-year one to 38.7 kgN·ha⁻¹ in SD2 site-year three. Unlike SMN, crop N contents in February (GS 20–25) varied from 19.3 kg N/ha (SD3/year 3) to 108 kg N/ha site-year one, SD1. In site-year one, crop N content was reduced as SD was delayed, and in site-years two and three, SD1 and SD2 had similar crop N contents with SD3 tending to have lower contents.

Consequently, spring SNS followed a similar trend to crop N content with a significant SD × year interaction (p-value < 0.001), with sowing date affecting SNS in years one and three but not in site-year two; however, it must be borne in mind that SMN was not measured on individual sowing date main plots in site-year one.

Final (harvest) crop N uptake was influenced by year, SD and their interactions. Sowing date made significant changes to harvest N in site-years one and three, and SD3 had a smaller final N uptake. No difference was recorded in site-year two.

The ratio of harvest N uptake to spring SNS in unfertilized plots, represented as SNUpE, is also summarized in Table 4. SNUpE was affected by SD with significant differences recorded each year and on all three site-years. The lower SNS values were associated with the delayed sowing dates, resulting in greater SNUpE values. Therefore, SD3 had significantly higher values.

3.4. Fertilizer N Uptake Efficiency (FNUpE)

When considering FNUpE, a number of interactions were recorded for individual years and also across treatments (

Table 5. The effect of sowing date and fertilizer application strategy on FNUpE over three years.

Sowing Date	N Strategies	Site-Year 1		Site-Year 2		Site-Year 3		Mean
		N Rate (kgN/ha)	1 FNUPE	N Rate kgN/ha	FNUPE	N Rate (kgN/ha)	FNUPE	2 Years (+CMLoY) SD×N	3 Years (−CMLoY) SD×N
SD1	Fix225	225	0.359 bc	225	0.249 d	225	0.743 ab	0.502 cde	0.449 abc
	CMHiY	170	0.411 ab	224	0.312 cd	200	0.603 bcd	0.461 def	0.444 abcd
	CMStd	110	0.451 a	164	0.550 ab	140	0.575 bcde	0.563 bcde	0.519 a
	CMLo	-	-	81	0.418 bcd	56	0.840 a	0.609 abc	-
Mean (SD1)			0.407 a		0.382 b		0.69 a	0.536 ab	0.471 a
SD2	Fix225	225	0.257 d	225	0.365 cd	225	0.547 cde	0.458 def	0.383 cd
	CMHiY	225	0.308 cd	205	0.401 bcd	187	0.729 abc	0.591 abcd	0.478 ab
	CMStd	165	0.255 d	145	0.419 bcd	127	0.632 bcd	0.543 bcde	0.426 bcd
	CMLo	-	-	61	0.651 a	44	0.751 ab	0.702 a	-
Mean (SD2)			0.273 b		0.469 a		0.665 a	0.573 a	0.427 b
SD3	Fix225	225	0.373 abc	225	0.314 cd	225	0.395 ef	0.355 f	0.363 d
	CMHiY	260	0.299 cd	200	0.597 ab	261	0.295 f	0.422 ef	0.369 cd
	CMStd	200	0.247 d	140	0.480 abc	201	0.493 de	0.487 cdef	0.393 bcd
	CMLo	-	-	57	0.569 ab	118	0.674 abcd	0.661 ab	-
Mean (SD3)			0.306 b		0.480 a		0.464 b	0.482 b	0.375 c
			p-value Site-Year 1	p-value Site-Year ear 2	p-value Site-Year ear 3	p-valueSite-Year 2, -3 (+CMLoY)	p-value All Site-years (−CMLoY)
		SD	<0.001	<0.001	<0.001	<0.001	<0.001
		N strategy	ns	<0.001	<0.001	<0.001	0.002
		Site-Year	-	-	-	<0.001	<0.001
		SD × N	<0.001	<0.001	<0.001	0.009	0.007
		SD × site-year	-	-	-	<0.001	<0.001
		N×site-year	-	-	-	<0.001	<0.001
		SD × N ×site-year	-	-	-	<0.001	<0.001

¹ FNUPE mean analysis of site-year 1 when SD × N were significantly different. Different letters indicate significant differences at p = 0.05 within a factor and parameter and are based on three-way significant interaction.

). Because CMLo was not included in site-year one, different N strategies did not result in significant changes in this year. In terms of SD in year one, FNUpE was significantly different in SD1 with 0.407 compared to 0.273 and 0.306, respectively, for SD2 and SD3. The highest fertilizer efficiency value was associated with CMStd with 110 kg·ha⁻¹ of N applied.

In site-year two, the addition of a lower N rate strategy (CMLo) resulted in a wider variation of FNUPE. Both SD2 and SD3 had higher FNUpE values compared to the previous year. For SD1, high N efficiency uptake was related to CMStd (164 kg·ha⁻¹ of N), which was not significantly different from CMLo (81 kg N/ha). Higher N rate strategies of CMHiY and Fix225, with >200 kgN/ha resulted in fertilizer efficiencies that did not differ statistically from one another within the range of 0.25 to 0.40 for SD1 and SD2, while for SD3, CMHiY (200 kgN/ha) had a statistically higher efficiency value than Fix225 with an FNUpE value of 0.6.

Fertilizer NUPE values in site-year three were the highest among site-years for SD1 and SD2 values with 0.690 and 0.665, respectively. In addition, CMHiY, on SD1 and SD2 had values of 0.603 and 0.729 with 200 and 187 kgN/ha of N content, respectively. Similarly, in this year Fix225 resulted in a higher FNUpE on the two early SDs.

Numerically, CMLo provided the highest efficiency of N uptake on all SDs with 56 kgN/ha, 44 kgN/ha and 118 kgN/ha on SD1, SD2, and SD3. The smallest fertilizer efficiency values were related to CMHiY and Fix225 in SD3. Mean FNUpE values across three site-years with only three N strategies (Fix225, CMHiY, and CMStd) were reduced as SD delayed. Lowest values were observed from the three N strategies on SD3 with Fix225, CMHiY and CMStd.

3.5. Optimum N Rates in Relation to FNUpE and Yield

To detect the Nopt, yield and N rates were regressed for individual year and separated by their SDs where significant (Figure 1a–c). On the same figures, FNUpE is represented on the secondary y-axis. Linear +expo curve parameters are presented in

Table 6. Linear plus exponential fitted curve parameters between yield and N rates (Y = A + BR^N + CN), S.E. is the standard error of the observation and D.F. is the degrees of freedom. Optimum N rate (Nopt), and yield at maximum economical point corresponded to Nopt; proportion accounted for variance as (adjusted R²), and the p-values of the stepwise regressed function.

Site-Years		Curve Parameter				S.E. (D.F.)	Nopt (kg·ha−1)	Max. Eco. Yield(t·ha−1)	adj. R2	p-Value
	SD	A	B	C	R
Site-Year-1	1	5.21	−1.52	−0.00098	0.981	0.025 (114)	113	4.91	0.91	<0.001
	2	3.81	−0.48	0.0052			undefined
	3	4.23	−1.61	0.0027			undefined
Site-Year-2		5.1	−1.96	−0.0020	0.989	0.253 (56)	143	4.37	0.77	<0.001
Site-Year-3	1	3.69	−0.065	0.014	1.01	0.288 (67)	190	5.63	0.93	<0.001
	2	3.75						5.69
	3	2.12						4.06
Across Site-Years	1	6.68	−3.03	−0.0046	0.993	0.465 (197)	157	4.93	0.74	<0.05
	2	7.57	−4.17	−0.0079			148	4.90
	3	6.70	−4.23	−0.0062			175	4.34

. This was to associate FNUpE with its most relevant Nopt where yield is economically maximized. In this case, it was possible to differentiate the supra-optimal values, or define how FNUpE values differs from the assumed 0.6 of UK–CM. Based on the parameters in Table 6, Nopt was shown by an arrow at 113 kgN/ha with a related FNUpE of 0.451 (Figure 1a). The closest N strategy to the modeled Nopt was 110 kgN/ha and corresponded to CMStd in site-year one, SD1 (

Table 7. Selected FNUpE closest to Nopt to define the best N strategy, N rates, and SDs for each separate year and SD.

			Nearest Related
Site-Years	Sowing Dates	Nopt (kg·ha−1)	N Strategy	N Rate (kg·ha−1)	FNUpE
Site-Year-1	SD1	113	CMStd	110	0.451
Site-Year-2	SD1	143	CMStd	164	0.550
	SD2		CMStd	145	0.419
	SD3		CMStd	140	0.480
Site-Year-3	SD1	190	CMHiY	200	0.603
	SD2		CMHiY	187	0.729
	SD3		CMStd	201	0.493

).

In site-year 2, SD was not an influential factor for defining yield and Nopt based on the stepwise regression procedure as shown in Figure 1b and Table 6. Hence, 143 kgN/ha calculated as Nopt corresponded mainly to CMStd N rates as summarized in Table 7.

In site-year 3, for which FNUpE were generally higher in SD1 and SD2 with 0.603 and 0.729, Nopt was 190 kgN/ha, shown as three parallel curves; SD1 and SD2 produced almost identical curves with very similar “A” coefficients (Table 6). The mutual Nopt of 190 kgN/ha was common across the three SDs and associated with CMHiY for SD1 and SD2 (200 and 187 kg·ha⁻¹ of N, respectively) and CMStd (201 kgN/ha for SD3).

Figure 1c also shows FNUpE points related to lower N strategies that appeared above the yield curves. Those were associated with CMLo, which notably, were not within the maximum economical yield point despite having higher FNUpE values.

4. Discussion

4.1. Soil N Supply and SNUpE

In this study, post winter SMN ranged from 21.4 to 38.7 kgN/ha but varied little with sowing date. When considering crop N at specific sowing dates and in specific years, variations from 19 to 108 kgN/ha were recorded. Spring SNS ranged significantly from 49 to 142 kgN/ha across SD and year combinations. Final crop N uptake at harvest in unfertilized plots was slightly more than spring measured SNS in SD1 and SD2 and this resulted in SNUpE values >1 of 1.13 and 1.14 across three site-years. Hence, there was a modest amount of additional soil N (13 to 14 kg N/ha higher than the measured SNS) captured by the harvest. In the case of SD3, SNUpE was similarly high at 1.65, 1.69 and 1.78 in each of the three site-years, respectively. The total harvest N uptake of these late-sown crops was either the same or less than that of earlier sowing dates, but this was achieved from a lower spring SNS. In a UK study by Kindred et al. [36], a wider range of SMN from 30 to 60 kgN/ha with similar range of crop N from 29 to 55 kgN/ha was measured with both crop N and SMN elements contributing equally to SNS. In another UK study by Berry and Spink (2009) [37], SNUpE was measured as 1.07 when regressing early spring SNS (as explanatory variable) and harvest N uptake (as the responsive variate), with an R² explaining the variables at 0.76. They reported a satisfying relationship in conditions where “no systematic difference in SNUpE was noticed between sites and seasons”, although, one variety (Castille) showed on average a higher ratio of final N uptake to spring SNS of up to 1.30. The authors explained that SNUpE of >1 could be related to sampling time variations or higher N mineralization after spring as SMN is a dynamic value. For fertilizer estimation purposes in the UK, SNUpE is practically assumed to be 1 (=100%). In the research reported here, spring SMN ranges were smaller and crop N uptake was comparatively higher than in studies performed in the UK. It is possible that low SMN in the current study made this value less reliable for calculating SNS in spring, as concluded by Walsh [38]. This author assumed that a high volume of mineralized N cannot be captured by the plant and could be lost through leaching or run-off, becoming immobilized or denitrified. When considering the SNUpE values measured on SD3, crops were shown to take up more N from the soil but did so later than early spring. It is therefore possible to debate that spring SNS with earlier SDs has been a more accurate estimation of final N uptake with a slight difference of 10 to 11 kg N/ha, whereas for SD3 it differed more significantly in all site-years. Late sown crops due to the less advanced growth stage and smaller root growth had been less able to capture SMN by early spring; however, these crops took up more mineralized N throughout the growing season. An alternative hypothesis for why late sown crops had a high SNUpE is that they experienced less loss of leaf and associated N after stem extension which could have resulted in an apparently higher SNUpE. This challenges the reliability of post winter SNS as an indicator of unfertilized N uptake by harvest for late sown crops in a mild climate with high soil OM levels. By contrast, earlier sowing crops (sown August to early September) had sufficient growth in the autumn and winter to capture more of the SMN and as a result, SNS calculation appeared to be a more reliable predictor of unfertilized harvest N uptake. However, where sowing in mid-September is inevitable, the capacity to quantify or estimate the AAN alongside SNS could help mitigate the risk of imprecise prediction in this climate. Spring SMN, which was measured in this study as a one-time snapshot, had a smaller contribution to SNS than crop N content for the three different SDs. It is appropriate to include both SMN and crop N content for spring SNS measurements as they both contribute to final N uptake at harvest. Post-winter N mineralization is difficult to predict, as it is subject to spatial–temporal variations, and the crop may not have the capacity to take up the mineralized N when it is available, leaving it at risk to loss pathways including leaching. In this study, the post-winter mineralized N which was taken up by the crop, referred to as additionally available N (AAN), was higher on SD3 crops (38.6 kgN/ha) but smaller on earlier SDs (11 kgN/ha). With early SDs, the lower AAN values allowed more accurate estimation of fertilizer N because the uncertainty of predicting AAN is less important. However, with SD3 the more significant contribution of AAN risked inaccurate estimation of fertilizer N requirements. Earlier sowing (SD1 and SD2) reduces the risk of inaccurate N application because a large proportion of the SNS has been taken up into the canopy by winter and can therefore be quantified with reasonable accuracy before the time spring N fertilizer must be applied.

4.1.1. Crop N, SMN Uptake in EUROPE

Studies from different European regions (Germany, France, Netherlands, Denmark and UK) on N supply to WOSR have shown that spring SNS (mainly focused on SMN) was considered a good indicator for N fertilizer prediction but not “satisfactory” [39]. The often-quoted “black box” in fertilizer N determination usually relates to the unknown soil resources and their availability during the growing season [40]. In the continental-Alpine climates of Germany and France, SMN dynamics also vary considerably during autumn and winter as mineralization in early spring depends on regional rainfall and temperature [41]. Henke et al. [42] concluded that autumn canopy N + SMN is a better indicator for adjusting N fertilizer rates compared to spring canopy N + SMN, because soil N availability in spring is less due to low soil temperature, little mineralization over winter and nitrate leaching [43]. The potential of WOSR to take up N during autumn in these regions was reported to be above 100 kgN/ha [43,44]. The knowledge of potentially mineralized N on a seasonal and site-specific basis could be a complementary factor for predicting soil N capture.

4.1.2. Fertilizer N Uptake Efficiency

The FNUpE figures in this work were generally lower than the UK standard assumption of 0.6 with an overall average of 0.53 but with high variation between site-years and sowing dates. In site-year one, the FNUpE was 0.45 for SD1 and could not be calculated for SD2 and SD3 this year. In site-year two, they ranged from 0.42–0.55 with no real evidence of a trend with the sowing date. In site-year three, there was a much greater disparity in FNUpE across the sowing dates with SD1, SD2 and SD3 having FNUpE values of 0.603, 0.729 and 0.493 respectively.

In UK literature, FNUpE also varied across site-years ranging from 0.23 to 0.63. In one example, at 100 kg·ha⁻¹ N application rate FNUpE was 0.67, whereas at 240 kgN/ha the value reduced to 0.43 [37]. It has also been shown that FNUpE varied based on the growth stage of the crop when N is applied: early application of N has been reported to be more efficiently taken up by early sown crops as the rapidly developing plants move to stem elongation when nitrate uptake increases [45]. In a study by Sieling et al. [46] conducted in northern Germany, the date of plant establishment also affected FNUpE: early sown plots (sown in the first and third week of August) had higher FNUpE than late sown ones (first and third weeks of September). Authors also showed that FNUpE reduced substantially from 0.46 to 0.28 with rising N application rates from 80 kgN/ha to 280 kgN/ha.

In this study, AAN values were estimated post-harvest using SNUpE on zero-N plots; yet, for practical on-site purposes, suggested anaerobic soil testing methods measuring potentially mineralizable ammonium during the growing season may be more feasible [47].

5. Conclusions

In this research, SNU_PE on all SDs was greater than 100% indicating the potential for more soil N being available than the assumed value in the UK–CM system. This additional available N was larger with late sown crops (38.6 kgN/ha) that had low SNS, than with earlier sown crops (about 11 kgN/ha) which had a larger SNS. Therefore, earlier SDs (SD1 and SD2) had a better capacity to measure SNS more accurately due to better N capture by the crops. Consequently, N loss risk was reduced by earlier SDs. FNUpE was measured as 0.550 for earlier sowing dates (SD1 and SD2) and 0.486 for the later sowing date; hence, FNUpE was lower for late sown crops than the standard of 0.6 assumed in the UK.

This research suggests that the measurements of SNS post-winter, an estimate of N mineralization during the growing season, together with a consideration of the sowing date, will help determine the scope for reducing fertilizer N while achieving high yield and FNUpE.