Introduction
The choice of crop establishment system can directly impact on plant density, which is an early determinant of crop yield potential (Wade et al., Reference Wade, Spink and Orson2006). Establishment systems can also influence later plant growth, soil structure, soil fauna, greenhouse gas emissions, soil C storage, weed population density and distribution, soil moisture status and machinery work rates and costs (Abdalla et al., Reference Abdalla, Jones, Ambus and Williams2010; Bekele, Reference Bekele2020; Cook et al., Reference Cook, Clarke, Hughes and Moss2006; Coulibaly et al., Reference Coulibaly, Aubert, Brunet, Bureau, Legras and Chauvat2022; Forristal and Murphy, Reference Forristal and Murphy2009; Nunes et al., Reference Nunes, Karlen and Moorman2020; Šarauskis et al., Reference Šarauskis, Buragienė, Romaneckas, Sakalauskas, Jasinskas, Vaiciukevičius and Karayel2012; Smith et al., Reference Smith, Andrén, Karlsson, Perälä, Regina, Rounsevell and Van Wesemael2005). Crop establishment systems are frequently presented as a limited number of discrete systems such as plough, min-till and direct drill; however, there is a continuum of systems incorporating soil cultivation and sowing operations that vary in tillage type, depth and intensity (Davies and Finney, Reference Davies and Finney2002). Establishment systems can be broadly divided into inversion and non-inversion systems. Inversion utilizes a mouldboard plough that inverts the soil, typically to a depth of 200–250 mm, followed by secondary tillage either prior to or in combination with sowing (Tebrügge and Düring, Reference Tebrügge and Düring1999). Non-inversion systems may mechanically cultivate the soil but do not invert it. The depth and intensity of tillage in non-inversion systems vary considerably (Tebrügge and Düring, Reference Tebrügge and Düring1999) with three sub-categories: minimum tillage (min-till), strip till and direct drilling normally identified. Min-till includes soil cultivation, but the type, depth and level of disturbance can vary, with reports in Ireland, for example, indicating depths varying from 75 mm to 200 mm being used (Jameson, McDonnell and Forristal, Reference Jameson, McDonnell and Forristal2024). Strip-till, a subdivision of min-till, involves cultivation of strips of soil and simultaneous seeding within the cultivated strip, leaving relatively undisturbed soil between the seeded rows (Pöhlitz et al., Reference Pöhlitz, Rücknagel, Koblenz, Schlüter, Vogel and Christen2018). Direct drilling (also referred to as No-tillage or Zero-till) systems place seed directly into the soil with very little disturbance (Derpsch et al., Reference Derpsch, Franzluebbers, Duiker, Reicosky, Koeller, Friedrich, Sturny, Sá and Weiss2014).
In Ireland, the most important cereal crops are spring barley (116,400 ha), winter barley (67,500 ha) and winter wheat (56,200 ha) (CSO, 2021). The climate is Atlantic influenced (Jameson et al., Reference Jameson, McDonnell, Alwarnaidu Vijayarajan and Forristal2024; Peel et al., Reference Peel, Finlayson and McMahon2007), which, in combination with a cropland history dominated by mixed grassland/spring cropping, gives Ireland a very high yield potential for cereals (Forristal and Grant, Reference Forristal and Grant2011) with Ireland having the second highest rainfed average cereal yield globally in 2022 (FAO, 2024). Prior to 2000, plough-based crop establishment was almost universally used, but in recent years, there has been an increase in the interest and use of non-inversion systems to establish crops. Growers have turned to these systems in response to reduced labour availability and increasing crop establishment costs and a desire to retain more soil carbon, improve their sustainability and increase work rates (Forristal and Murphy, Reference Forristal and Murphy2009).
Crop establishment systems research in drier climates frequently suggests that there are benefits associated with non-inversion systems relating to improved soil moisture retention, comparable or improved yields, lower establishment costs and improved soil carbon retention (Alvarez and Steinbach, Reference Alvarez and Steinbach2009; Derpsch et al., Reference Derpsch, Friedrich, Kassam and Hongwen2010; Kassam et al., Reference Kassam, Friedrich, Shaxson and Pretty2009; Soane et al., Reference Soane, Ball, Arvidsson, Basch, Moreno and Roger-Estrade2012; Triplett Jr. and Dick, Reference Triplett and Dick2008; Zarea, Reference Zarea and Lichtfouse2011). However, limited Irish research has indicated some concerns around the suitability of these systems in wetter climates, such as increases in critical grass weeds and herbicide-resistance development risk (Alwarnaidu Vijayarajan et al., Reference Alwarnaidu Vijayarajan, Fealy, Cook, Onkokesung, Barth, Hennessy and Forristal2022; Alwarnaidu Vijayarajan et al., Reference Alwarnaidu Vijayarajan, Forristal, Cook, Schilder, Staples, Hennessy and Barth2021), inconsistent crop yields (Brennan et al., Reference Brennan, Hackett, McCabe, Grant, Fortune and Forristal2014), poorer crop establishment (Brennan et al., Reference Brennan, Hackett, McCabe, Grant, Fortune and Forristal2014; Byrne et al., Reference Byrne, McCabe and Forristal2022) and reduced suitability for spring cropping (Brennan et al., Reference Brennan, Forristal, McCabe and Hackett2015). For winter cereals, as autumn progresses, there is generally a risk of poorer establishment as a result of wet weather, especially for non-inversion systems, and many growers using these systems sow earlier when weather conditions are more favourable to reduce this risk (Jameson et al., Reference Jameson, McDonnell and Forristal2024). However, in a climate already conducive to considerable grass weed pressure, earlier sowing usually exacerbates grass weed issues (Moss, Reference Moss2017), effectively negating integrated weed management approaches, questioning the suitability of these systems for this climate. Currently, results from studies conducted in different climates and grower anecdotes are contributing to growers’ decision to adopt non-inversion systems (Jameson et al., Reference Jameson, McDonnell, Alwarnaidu Vijayarajan and Forristal2024). There are clear gaps in knowledge that need to be addressed before the role of non-inversion systems in an Atlantic climate can be determined.
Crop establishment systems, in addition to incorporating variations in tillage practice, can include differences in residue management, rotation, cover-cropping and traffic management. The multifaceted nature of these systems lends them to being considered as complex from an innovation adoption viewpoint (Alskaf et al., Reference Alskaf, Sparkes, Mooney, Sjögersten and Wilson2019). Researching agricultural systems using replicated field experiments is challenging, as plot size can be limiting and some factors involved in these systems can be difficult to replicate at a small plot scale (Wainwright et al., Reference Wainwright, Parsons and Abrahams2000). Furthermore, the contribution to any performance differences of the component elements of a system may be difficult to determine. Growers and advisors at dissemination meetings and as part of stakeholder groups indicate a strong preference for evaluations that apply the crop establishment elements similar to what occurs in commercial fields with larger implements, sequential cultivation passes at different angles to the first pass, forward speeds in excess of 10 km/h and wide turning areas. This results in very large experimental areas being required for replicated experiments with multiple treatments. Additionally, to determine the impacts of the individual components of a system, which may include management differences such as sowing date, weed control, etc., experimental designs become larger, more complex and difficult to implement. While farm studies as an alternative to controlled experiments may give growers more confidence, they have analysis challenges usually due to the lack of replication at individual locations, single treatment level per site and variations in management, soils and meteorological conditions between sites. On-farm research study types vary, with un-replicated split farm studies (Cooper et al., Reference Cooper, Hama-Aziz, Hiscock, Lovett, Vrain, Dugdale, Sünnenberg, Dockerty, Hovesen and Noble2020), tramline trials (Kindred et al., Reference Kindred, Clarke, Roques, Hatley, Marchant and Sylvester-Bradley2018), un-replicated large observational studies (Alesso et al., Reference Alesso, Cipriotti, Bollero and Martin2019), un-replicated multiple on-farm detailed case studies (Alwarnaidu Vijayarajan et al., Reference Alwarnaidu Vijayarajan, Fealy, Cook, Onkokesung, Barth, Hennessy and Forristal2022) and simple case studies (Bechini and Castoldi, Reference Bechini and Castoldi2009) used. There are challenges with some on-farm research methods; for example, split farm studies and tramline trials require growers to implement management changes/treatments to areas within farms or fields, which can be a participant deterrent. Larger observational studies, multiple on-farm detailed case studies and simple case studies overcome this issue. However, the numbers of experimental units needed in larger observational studies can be difficult to achieve as the resources needed to collect data on a large number of farms can be prohibitive. Simple case studies are practical and feasible to implement but usually only include a single or a small number of farms, making statistically valid comparisons difficult. Multiple detailed studies of a relatively low number of farms operating the systems of interest would be feasible and could facilitate detailed measurement and subsequent analysis, but variations in grower management and site characteristics remain a challenge. However, such study types have not been evaluated. There is very little knowledge concerning the role of on-farm research methods, particularly in comparison with conventional replicated small plot trials.
The objectives of the research described in this paper were:
-
To assess alternative crop establishment systems in an Atlantic-influenced climate by monitoring the performance of winter wheat crops grown following a break crop.
-
To assess the utility of an on-farm research model incorporating multiple detailed case studies of a number of farms operating different crop establishment systems compared with a replicated small plot experiment assessing the same systems.
This research should help growers to make more informed decisions about crop establishment system adoption and operation, while assessing alternative evaluation methodology for systems.
Materials and methods
The research reported here was conducted on crops of first wheat, that is, crops grown following a break crop, grown with three different crop establishment systems using two different research methods: a replicated trial and multiple on-farm case studies, over three growing seasons.
Replicated trial
Site description
The replicated experiment was conducted during the 2021 (year 1), 2022 (year 2) and 2023 (year 3) growing seasons on winter wheat (T. aestivum L. em Thell.) crops grown after winter winter oilseed rape (Brassica napus L. var. napus) at the Teagasc Crops Research Centre Knockbeg site (52°86_87 N, −6°94_14 W, 55m a.s.l), County Carlow, Ireland. The soil type is a Haplic Luvisol (humic epidystric) and is a medium textured clay loam of the Mortarstown series (Brennan et al., Reference Brennan, Hackett, McCabe, Grant, Fortune and Forristal2014; Conry, Reference Conry1987) (Table 1). Mean long-term (1980–2010) annual rainfall and temperature are 840.2 mm and 9.8 °C, respectively (Coonan et al., Reference Coonan, Curley and Ryan2024a; Coonan et al., Reference Coonan, Curley and Ryan2024b). The experiment was located on an existing trial site with a long history of establishment system studies. The site was annually cropped since 1992, and a replicated trial with establishment system treatments was established in 2001 to compare plough and min-till establishment. In 2014, the treatments were altered to include plough, min-till and strip till (with strip till and min-till on previously min-till plots) with sub-plots of rotational crops. Direct drilling was substituted for strip till at the start of the reported experiment (2021).
Table 1. Soil characteristics of the replicated trial experimental site (Knockbeg site) a
Notes: aSoil texture class – Clay Loam (UK/ADAS Classification). World reference base classification: Haplic Luvisol (humic epidystric); OM, organic matter.
b Adapted from Brennan et al. (Reference Brennan, Hackett, McCabe, Grant, Fortune and Forristal2014).
Treatments and experimental design
The study was part of a larger field trial that assessed the impact of crop establishment system and rotation in combination on the performance of a range of crops (full trial layout shown in Supplementary Figure 1). The crop establishment systems plough, shallow plough, min-till and direct drill were the main plot treatments on 30 m × 30 m plots. These plots were replicated four times in a randomized block design. All five individual crops of a 5-year rotation – winter oilseed rape (WOSR), winter wheat (WW1), winter oats (WO), winter wheat (WW2) and winter barley (WB), along with a fixed continuous wheat treatment (WW3) – were grown as sub-plots (30 m × 5 m) within each main plot. This study used the winter wheat grown after winter oilseed rape (WW1) in each of the plough, min-till and direct drill crop establishment treatments.
The plough plots were ploughed in mid-September in year 1 and in early October in years 2 and 3, to a depth of 225 mm. Secondary cultivation and sowing were carried with a combined cultivator and drill unit, using a 2.5 m rotary power harrow with vertical tines, working to a depth of 75 mm and a cereal box-type drill fitted to the harrow with seed coulters spaced at 125 mm (Amazone D9; Amazonen-Werke H. Dreyer SE & Co. KG, Germany). The plots were sown on the 14th, 11th and 10th of October, respectively, for years 1, 2 and 3. The min-till plots were cultivated with a tine-type stubble cultivator (Horsch Terrano FX; Horsch Maschinen GmbH, Germany) to a depth of 75–100 mm in the second half of August or the first days of September and consolidated with a ring roller as part of a stale seedbed process. New weed growth was sprayed with glyphosate three to five weeks later. Sowing was on the same date as the plough plots using a cultivator drill with disc coulters (Vaderstad Rapid; Vaderstad AB, Sweden). The direct drill plots were sown with a double disc direct drill using commercial coulter units (Weaving GD; E F Weaving Ltd., UK) on the same day as the plough and min-till plots. For all treatments, sowing depth was 30–40 mm, and straw from the previous crop was removed.
Crop management
The crops were managed according to normal commercial practice, based on Teagasc guidelines (Collins and Phelan, Reference Collins and Phelan2020), with all crop establishment treatments receiving the same management inputs. Wheat (cv. Graham) grown following a crop of WOSR was sown at 367, 360 and 360 seeds m2 on the respective sowing years. Fertilizer nitrogen (N) was applied at a total of 210 kg N/ha in three applications: at Zadoks decimal growth stage (GS) 25 (0.4 of the total N), GS 31 (0.4 of the total N) and GS 37–39 (0.2 of the total N). Phosphorus and potassium were applied in accordance with soil test results. Disease control consisted of a three-spray programme, with fungicides applied when leaf 3 was fully emerged (BBCH 32–33), when the flag leaf was fully emerged (BBCH 39) and at the start of flowering (BBCH 61) (Lancashire et al., Reference Lancashire, Bleiholder, Van den Boom, Langelüddeke, Stauss, Weber and Witzenberger1991). For all systems, weed control consisted of application of residual herbicides (e.g. flufenacet and diflufenican) early post-emergence (late October/early November) for broad-leaved and grass weed control. This was supplemented by spring application of mesosulfuron-methyl and propoxycarbazone-sodium targeting Bromus species and pinoxaden targeting Avena fatua. An insecticide (lambda-cyhalothrin) was applied to all treatments in October/November for aphid control to prevent barley yellow dwarf virus (BYDV) infection.
Data collection
Following crop emergence, plant population density was determined each year by counting plants present along a predetermined length of crop row in six locations within each plot. The area sampled differed by treatment based on drill row spacing; row spacing was 125 mm for plough and min-till treatments and 166 mm for the direct drill treatment. For plough and min-till treatments, four 1 m lengths of a crop row were counted, resulting in area sampled being 0.125 m2. For the direct drill system, three 0.75 m lengths of a crop row were counted, resulting in area sampled being 0.1245m2. As an indicator of crop growth and biomass development, the fraction of photosynthetically active radiation intercepted (FIPAR) by the crop was determined at growth stages 21–25, 30, 32 and 37. FIPAR was determined by measuring photosynthetically active radiation, both above and below the crop canopy, using a Sunscan Canopy Analysis System (Delta-T Devices, Cambridge, UK) (Gower et al., Reference Gower, Kucharik and Norman1999).
$$FIPAR = 1 - \left( {{{PARBelowCanopy} \over {PARAboveCanopy}}} \right)$$
At crop maturity, samples of a given area (outlined above for establishment counts) were hand harvested (clipped at ground level and secured in hessian bag) from which ears per m2, grains per ear and harvest index were subsequently determined. Plots were harvested using a Deutz–Fahr plot combine (SDF Group, Treviglio, Italy) fitted with a Harvestmaster weighing system (Juniper Systems & Harvestmaster Inc., USA) with harvest dates of the 16th, 8th and 4th of August for the 3 years, respectively. Plot yields were recorded and adjusted to 15% moisture content with samples taken for grain quality analysis.
Statistical analysis
Data analysis was carried out using R, version 4.1.0 (R Core Team, Reference Core Team2024). Linear mixed effects models were fitted using plant population density, FIPAR and grain yield data for the 3 years with the ‘lmer’ function from the ‘lme4’ package (Bates et al., Reference Bates, Mächler, Bolker and Walker2015). The models included system and year and interaction of system and year as fixed effects. Random effects were included to account for variation associated with the interaction between year and block, and a nested random effect was included to account for variation among treatment plots within each block (block and column location taken into account). The significance of fixed effects and interactions (P values) was calculated using type III sums of squares with the ‘Anova’ function from the ‘car’ package (Fox and Weisberg, Reference Fox and Weisberg2018). Where significant effects were detected, estimated marginal means were obtained using the emmeans package (Lenth, Reference Lenth2025), and pairwise comparisons were performed using the least significant difference (LSD) method. The same analysis method was also carried out on the same data at an individual year level with system included as a fixed effect and block and column location included as random effects.
To ensure the validity of the linear mixed effects models used in this study, key statistical assumptions were assessed. An analysis of residuals was performed to evaluate normality, homogeneity of variances and independence. Normality of residuals was evaluated visually using quantile-quantile (Q-Q) plots. Model residuals were visually inspected through residuals versus fitted values plots to confirm homogeneity of variances. The assumption of independence of residuals was checked by examining residual plots for patterns that might indicate temporal or spatial autocorrelation.
On-farm field scale case studies
Field selection and experimental design
To determine the impact of crop establishment system on ‘first’ wheat crops (i.e. those grown subsequent to a break crop) grown on farms, growers using plough, min-till and direct drill systems with a suitable cropping sequence, were recruited from the networks of Teagasc (Irish state-funded research and farm advisory organization) advisors and commercial crop advisors/consultants. Study fields were located within a 65 km radius of Teagasc Oak Park research centre, Carlow, Ireland, which is in the main cereal growing area of Ireland (Figure 1). Twenty-one growers participated each year; seven using each crop establishment system (plough-based, min-till and direct drill). This study was conducted during the same growing seasons as the replicated trial: 2020/2021, 2021/2022 and 2022/2023. A different field was monitored each year with each grower due to rotation constraints, giving a total of 63 monitored fields over the three years. If a grower didn’t have the target crop available, a new grower was subbed in. Soil texture and rotation sequence information is given in Table 2. All fields were managed under their respective crop establishment system for a minimum of 5 years for plough and min-till and 3 years for direct drill in advance of the reported study. The minimum number of years was reduced for direct drilling compared to other systems as there were very few growers in Ireland with a history of direct drilling that was greater than 3 years at the time of the study. The study fields also had combinable crops (cereals, oilseeds and grain legumes) grown for a minimum of 10 years prior to the experimental period. The mean field size was 11.86 ha (range: 2.06–50.40 ha). Plough-based establishment on farms was very similar to that implemented in the replicated trial. However, cultivation depth for min-till growers ranged from 50–150 mm, and most direct drill growers used a very shallow (∼20–30 mm) stubble cultivation/straw harrow operation to encourage germination of weed seeds and volunteers before glyphosate application in advance of planting.
Figure 1. Study site area: all sites were within the marked area.
Table 2. Soil texture analysis and preceding crops for on-farm fields
a Soil texture based on UK soil texture classifications (Natural England, 2008).
b Oats is considered a break crop in Ireland as it is not susceptible to the same take-all (Gaeumannomyces graminis) disease strain as wheat or barley, conferring a disease break to the following wheat or barley crop.
Data collection, establishment and growth
At the start of each year, 10 positions within each study field were selected using a ‘W’ pattern, with all points adjacent to crop tramlines and located subsequently by their satellite-determined co-ordinates and a field marker. Crop information was collected at these positions throughout the growing season. Field headlands were avoided due to variability in both management and crop performance associated with these areas. Following crop emergence (autumn), plant population density was determined by counting plants present along a predetermined length of crop row at each of the 10 points in each field; the length of row sampled depended on the row spacing of the drill used to establish that field. The FIPAR by the crop was determined at each of the 10 points at growth stage 21–25, 30, 32 and 37 as outlined previously where data collection methodology of the replicated trial is described.
Yield estimation in growers’ fields
As it was not feasible to use a plot combine to assess yields in each of the 21 fields due to the crop ripening in a short time period at crop maturity across all the locations, a detailed manual yield estimation technique was used, which allowed a wider sampling time window. The technique developed by Ward et al. (Reference Ward, Forristal and McDonnell2020) was adapted to account for differences in row widths, meaning a predetermined number of row lengths, adopted for the row width of the drill, were harvested to give a precise sample size in the range: 0.25–0.3 m2. Whole crop samples were harvested from the sample positions. The samples were dried in a glasshouse, weighed, heads removed and threshed (Saatmeister Kurt Pelz, Maschinenbau, Germany) and the resulting grain cleaned (Laboratory seed winnower, type 4111.10.00, Seed Processing Holland, Enkhuizen, the Netherlands). All yield components were weighed and dry matters determined. This technique generated a yield figure (t/ha) for each of the in-field sample points. These yield estimates tend to be higher than total field yield estimates as headland areas are avoided and no allowance for uncropped tramlines is made (Ward et al., Reference Ward, Forristal and McDonnell2020). Harvest parameters including ears per m2, grains per ear and harvest index were determined.
Crop management
As this study was carried out in commercial fields, crop management and husbandry practices varied both within and among crop establishment system groups. To account for potential effects on crop performance these variations in crop management and husbandry practices may have had, detailed records of all management actions, including input types and quantities, application rates and timings and all machinery operations were obtained from growers.
Statistical analysis
Comparisons between the different crop establishment system groups must be considered carefully, due to the limited number within each group, variations in management practice and site differences.
To determine if plant establishment was more variable in certain systems, the coefficient of variation of the plant density figures from the ten positions within each study field was used. The coefficient of variations of the different systems were then compared by carrying out an analysis of variance using the ‘aov’ function from the ‘stats’ package (R Core Team, Reference Core Team2024). Analysis of variance was also carried out on FIPAR and yield component (harvest index, ears/m2, grain/ear, thousand-grain weight [TGW], HL weight and protein) data to determine if there were significant differences between the different systems. For these analyses of variance, for the all-years data, system, year and the interaction of systems and year were included as effects. When significant effects were found, a post hoc Tukey multiple comparison of means was used (TukeyHSD function from the ‘stats’ package, R Core Team, Reference Core Team2024). Analyses of variance were also carried out on the data at an individual year basis with system only included as an effect. Prior to conducting the analysis of variances, the data were examined to ensure compliance with the key assumptions of normality, homogeneity of variances and independence. Normality of residuals was evaluated visually using quantile-quantile (Q-Q) plots. Homogeneity of variances was assessed by visual inspection of residuals versus fitted values plots, and independence of residuals was verified by examining residual plots for patterns that might indicate temporal or spatial autocorrelation.
To determine the factors that may impact on crop yield, a linear mixed effects model was fitted using 3 years data as described previously for the replicated trial. System, year and the interaction of system and year were included as fixed effects. Nitrogen application rates and fungicide expenditures were grouped into bands of 10 units (e.g. kg N/ha for nitrogen, €/ha for fungicides), with growers assigned to categories based on their respective input levels. These groupings were included as random effects along with soil texture (its effect was allowed to vary by year). As previously described, p-values were computed for the final model, and where significance was detected, pairwise comparisons were carried out using the LSD method. The same analysis method was also carried out on the same data at an individual year level with system included as a fixed effect and nitrogen application rate, grower fungicide spend and soil texture included as random effects. Prior to conducting the linear mixed effects model analysis, the data were evaluated to ensure that key assumptions were met as previously described.
A nitrogen application rate per hectare for each grower was established by calculating the nitrogen supplied by chemical and organic fertilizers. These nitrogen application rates were then adjusted based on the amount of nitrogen that was estimated to have been supplied by the preceding crop (e.g. legume or oilseed rape vs oats) type based on the guidelines in the Teagasc Winter Wheat Guide (Lynch et al., Reference Lynch, Spink, Doyle, Hackett, Phelan, Forristal, Kildea, Glynn, Plunkett, Wall and Hutton2016). A fungicide spend per hectare for each grower was established by multiplying the fungicide product rates used by growers by the recommended retail prices for these products.
With growers free to manage their crops as they normally would, there were inevitable differences in management between growers. If these differences were associated with establishment system, then they could be confounding the results. Therefore, key yield influencing inputs, in this case nitrogen application rate and grower fungicide spend, were tested to see if they were confounded with crop establishment system as failing to identify confounders can bias results and attribute effects to system that are actually due to other factors. To robustly assess potential confounding, we employed several complementary methods to test whether input application rates were confounded with system. The Kruskal–Wallis test (kruskal.test function) was used to test for associations between crop establishment system and input levels; kruskal.test function is from the ‘stats’ package (R Core Team, Reference Core Team2024), and associations of nitrogen application rate and grower fungicide spend with the system would indicate possible confounding. Spearman correlations (cor.test function) were used to test whether nitrogen rate and fungicide spend were correlated with yield; correlations of these factors with yield would indicate possible confounding, and cor.test function used is from the ‘stats’ package (R Core Team, Reference Core Team2024). A linear fixed effects model was fitted using the lm function (Lenth, Reference Lenth2025) with nitrogen application rate and grower fungicide spend as fixed effects to test whether these effects were significant predictors of yield; effect significance was calculated using type III sums of squares as previously described. These factors being significant predictors of yield would indicated possible confounding. Another approach was to fit 2 linear fixed effects models, one including system, year, the interaction of system and year, nitrogen application rate and grower fungicide spend as fixed effects and another that included system, year, the interaction of system and year as fixed effects, these models were then compared to assess how the estimated effect of system changes when nitrogen application rate and grower fungicide spend were included. A change by more than 10% in the estimated effect of system would indicate possible confounding. This multifaceted approach is consistent with recommendations in the literature to confirm confounding from multiple perspectives (Mickey and Greenland, Reference Mickey and Greenland1989; Maldonado and Greenland, Reference Maldonado and Greenland1993; Brookhart et al., Reference Brookhart, Schneeweiss, Rothman, Glynn, Avorn and Stürmer2006).
Results
Weather conditions
Cumulative rainfall within growing seasons (September to August) during the study varied considerably; year 1 and 3 were above average, while year 2 was below average. December, February and May of year 1, October and December of year 2 and September, October, March and July of year 3 were marked by heavy rainfall (Figure 2). In year 3, this heavy rainfall occurred in September and October, which is during the establishment period of winter cereals. Mean annual rainfall for the Republic of Ireland was 1202.6 mm (Coonan et al., Reference Coonan, Curley and Ryan2024a).
Figure 2. Monthly rainfall amounts for each month in the experimental period in comparison with long-term average (LTA – 1980–2010).
Monthly mean air temperatures were close to average during the growing season (March to harvest) in year 1 with the exception of July, which was 2.1 ˚C above average while in year 2, May, July and August were each 1.6˚C above average and in year 3, May and June were 2.1 ˚C and 2.9 ˚C above average, respectively (Figure 3). Mean annual temperature for the Republic of Ireland was 9.1 ˚C (Coonan et al., Reference Coonan, Curley and Ryan2024b).
Figure 3. Mean monthly air temperatures for each month of the experimental period in comparison to the long-term averages (LTA – 1980–2010).
Replicated trial
Plant population density and light interception
The effects of crop establishment system on plant population density are presented in Table 3. In years 1 and in the all-years analysis, the direct drill system had significantly lower plant population densities than the plough-based and min-till systems, indicating poorer crop establishment, as all plots were sown with the same seeding rate. In year 3, direct drill system had significantly lower plant population densities than the min-till system which had significantly lower densities than the plough-based system. The effects of crop establishment system on growth, as measured by light intercepted, are presented in Table 3. When all years were considered, the direct drill system was intercepting significantly less light at growth stages 30 and 32 than the plough-based and min-till systems, indicating less biomass accumulation at these stages. For individual years, the direct drill system was intercepting significantly less light than the other systems at growth stages 21–25 (year 1), 30 (in years 1 and 2), 32 (in years 1 and 3) and 37 (in year 3).
Table 3. Plant population density and fraction of photosynthetically active radiation intercepted (FIPAR) over three growing seasons in the replicated trial
1 Data from growth stage 30 missing due to an instrument data logging malfunction. Pr>F values are from analysis of variance.
Grain yield and yield components
The effects of crop establishment system on grain yield and yield components are presented in Table 4. In year 2, the direct drill system produced a significantly lower grain yield than the plough-based system but not the min-till system, and in year 1, there was also a trend that direct drill yields were lower but not significantly so. The direct drill system had significantly lower number of ears per m2 in year 3 and in the all-years analysis as establishment system was shown to significantly impact on ear numbers. Min-till had a significantly higher number of grains per ear in year 2 than the plough-based system, and the direct drill system had a significantly higher number of grains per ear in year 3 (Table 4). There was no significant difference in TGW or protein levels between systems in any of the study years. Year also had a significant impact on some parameters, with grain yield, TGW, HL weight and protein levels varying significantly with year.
Table 4. Grain yield and yield components of three establishment systems over three growing seasons in replicated trial
HI, harvest index; proportion of grain to straw dry matter; TGW, thousand-grain weight at 85% dry matter; HL weight, hectolitre weight at 80% dry matter. Pr>F values are from an analysis of variance.
On-farm field scale case studies
Plant population density and light interception
There was a large range in plant population densities both within and across systems in the on-farm study in all study years with the range of individual farm values within their system shown graphically in Figure 4 and the system means (7 farms in each) shown in Table 5.
Figure 4. Jitter plot of plant population densities for individual fields (n = 63) in the on-farm study for each of the study years.
Table 5. Plant population density and fraction of photosynthetically active radiation intercepted (FIPAR) for on-farm study
1 Planting date is given in the form of days after first crop in study planted; with first crop planted being day 1, first crops were planted on 25/09/2020, 23/09/2021 and 26/09/2022 in years 1, 2 and 3, respectively.
CV, coefficient of variation.
The plant population density between the individual farms ranged from 116 to 435 plants per m2, with high variation evident within each of the establishment system groupings. There was no difference in plant populations that could be attributable to crop establishment systems as indicated by P value > 0.05 for each of the individual years, or all years combined. There was a significant year effect on the plant populations. The coefficients of variation of the ten measured plant density values in each field were used as an assessment of crop establishment variability. There was no significant difference in variability of crop establishment between systems (Table 5). Plough-based growers planted their crops later than min-till growers and direct drill growers on average over the 3 years (Table 5). Also, the seed rate used by direct drill growers was higher than that used by the other groups.
Using light interception as an indicator of canopy development and crop growth, there was an expected trend in increasing interception as the plants progressed through their growth stages (Table 5 and Figure 5) and there were differences between years at certain growth stages, however there was no significant difference in light intercepted by the different establishment systems at any of the growth stages in any of the study years or in the all-years analysis.
Figure 5. Light intercepted for individual fields in on-farm study for each of the study years.
Notes: FIPAR – fraction of photosynthetically active radiation intercepted.
Grain yield and yield components
The effects of system on grain yield and yield components are presented in Table 6. The distribution of the individual farm yield data points is presented by establishment system in Figure 6.
Table 6. Grain yield and yield components for on-farm study
HI, harvest index; TGW, thousand-grain weight at 85% dry matter; HL weight, hectolitre weight at 80% dry matter.
Figure 6. Jitter plot of grower grain yields in the on-farm study.
The establishment system used significantly impacted on grain yield in years 1 and 3, with direct drill yields lower than the other systems in year 1 only. In year 3, while the impact of system was significant, post hoc testing could not show systems to be significantly different to one another. There was also a trend in the all-years analysis that system impacted on grain yield, with direct drill fields having numerically lower yields; however, this was not significant (Tables 6 and 7).
Table 7. Grower fungicide spend and nitrogen application rates in on-farm study for study years
Notes: P values from Kruskal–Wallis – to test for association between system and potential confounder.
Confounding analysis on the data from all years confirmed that nitrogen rate and fungicide spend were confounded with establishment system. The Kruskal–Wallis test confirmed that nitrogen rate and fungicide spend varied significantly between systems (p < 0.001). Spearman correlation analysis showed significant correlations between these variables and yield (p < 0.01), and in a linear model with nitrogen rate and fungicide spend as explanatory variables and yield as the dependent variable, fungicide spend was a significant predictor of yield (p < 0.01), while nitrogen rate showed a weaker effect (p = 0.071). Furthermore, models with and without these covariates showed a mean change of 18.37% in system effects, exceeding the 10% threshold for confounding, highlighting the impact of input differences on observed yield variation. Confounding analysis on individual years’ data followed the same trend as above, confirming confounding of nitrogen rate and fungicide spend with establishment system on an individual year basis but with year 2 showing slightly weaker support for confounding than the other years. The fungicide spend and nitrogen rate per group are displayed in Table 7.
The treatments with numerically lower yields had lower ear numbers and little compensation by increases in either grains per ear or TGW. Crops from direct drill fields had significantly lower harvest indices than min-tilled fields (Table 6). System also significantly affected harvest indices, with the direct drill fields having lower harvest indices in the all-years analysis, and direct-drilled fields also had significantly lower ear counts than the other systems in year 3 and significantly lower ear counts than the plough-based but not the min-tilled fields in the all-years analysis. For TGW, system had a significant effect in year 1 and in the all-years analysis with direct-drilled and plough-based fields having significantly lower TGW than min-tilled fields, following the trends observed with treatment yields.
Discussion
Results from both the replicated trial and the on-farm studies are considered together in this section. The limited number in each establishment system group in the on-farm studies, coupled with the large differences in site, soil and particularly crop management practices, requires a degree of caution to be exerted when interpreting differences observed in the farm data. The replicated trials allow fair and useful comparisons to be made and, having been carried out in the same seasons, provide a useful basis for interpretation of the results of the on-farm study.
While there were considerable differences in weather across the three study years, of most interest was the above average rainfall in September in year 2 and in both September and October in year 3. This was associated with lower plant population densities in direct drill in the replicated trial in year 3. Year 3 also had lower yields across all systems. To give an indication of the frequency of the occurrence of such weather events, in an earlier replicated trial on the same site, above average post sowing rainfall in one out of three study seasons led to lower grain yields but also a difference between crop establishment systems (Brennan et al., Reference Brennan, Hackett, McCabe, Grant, Fortune and Forristal2014), where min-till had lower plant population densities and subsequent lower yields. Furthermore, from the period 2005–2023 (19 years) at the replicated trial site, there were 3 years where October and 4 years where November rainfall was over 50% higher than the long-term average rainfall for these respective months. However, weather was not the only factor impacting establishment as the direct drill plant population density, was also significantly lower in year 1 perhaps due to less favourable seedbed conditions reported in uncultivated soils such as reduced soil to seed contact as a result of surface residues in the seed slot (Sprague, Reference Sprague and Triplett1986), increased soil strength resulting in restricted root growth and vigour of emergence (Ball and O’Sullivan, Reference Ball and O’Sullivan1982) and lower soil temperatures and slower soil warming as a result of increased soil moisture (Alvarez and Steinbach, Reference Alvarez and Steinbach2009); resulting in slower seedling emergence (Blake et al., Reference Blake, Spink and Dyer2003).
In the replicated trial, the plant densities recorded with the plough-based and min-till systems were higher than the direct drill system throughout the experiment, suggesting that these systems produced more suitable conditions for crop establishment in this region. Despite the differences in plant densities in years 1 and 3, there was no statistical difference in grain yields, indicating that adequate plant densities were still achieved to maintain yield potential in winter wheat. The lower grain yield recorded for the direct drill system in year 2 compared to the plough-based but not the min-till systems was not accompanied by a significant difference in establishment or light interception at flag leaf emergence. Considering light interception, in the replicated trial, while the low growth height at GS 21–25 makes measurement of crop canopy differences using light absorption difficult, there were differences in at least one of the later three growth stages measured with direct-drilled crop absorbing less light at all stages where differences were significant. The reduction in light absorbed by the direct-drilled crop at the earlier growth stages may be associated with poorer establishment where less plant numbers result in slower canopy development but this may also be compounded by some of the conditions that cause poorer establishment such as increased soil strength, lower soil temperature and slower soil warming; which have been shown to result in slower early season growth in direct-drilled crops (Alvarez and Steinbach, Reference Alvarez and Steinbach2009; Ball and O’Sullivan, Reference Ball and O’Sullivan1982; Blake et al., Reference Blake, Spink and Dyer2003). Slower growth in the direct drill system was confined to early in the season in years 1 and 2; however, in year 3, the direct drill system was still intercepting significantly less light at flag leaf emergence, but while this was associated with numerically lower grain yields, this was only a trend and not statistically significant. In year 1, reduced plant densities in the direct drill system were compensated for by increased tillering as evidenced by similar ear numbers at harvest; this is similar to findings from Brennan et al. (Reference Brennan, Hackett, McCabe, Grant, Fortune and Forristal2014) and Wade et al. (Reference Wade, Spink and Orson2006). However, in year 3, the direct drill system had significantly lower ear numbers but compensated by producing a significantly higher number of grains per ear. Despite this compensatory production of an increased number of grains per ear, yields were numerically lower than the other systems but not significantly so.
In the on-farm study, there was no statistical difference in plant densities or canopy development between the systems, in contrast to the replicated trial. Those using direct drill systems on farms used higher seed rates and earlier sowing dates to counteract the risk of poorer establishment. While sowing earlier has benefits from an establishment point of view, it may bring some other challenges, as it can increase grass weed (Moss, Reference Moss2017; Andrew and Storkey, Reference Andrew and Storkey2017; Cook et al., Reference Cook, Clarke, Hughes and Moss2006), disease (Hardwick et al., Reference Hardwick, Jones and Slough2001; Polley and Thomas, Reference Polley and Thomas1991; Fitt et al., Reference Fitt, Fraaije, Chandramohan and Shaw2011) and pest (Kennedy and Connery, Reference Kennedy and Connery2001; Foster et al., Reference Foster, Blake, Tones, Barker and Harrington2004; Mc Namara et al., Reference Mc Namara, Gauthier, Walsh, Thébaud, Gaffney and Jacquot2020) pressures which may result in more pressure on chemical control measures, increasing the risk of pesticide resistance development and yield impacting infestations. This is likely to be more challenging in wetter Atlantic-influenced climates, where delayed sowing of winter crops is suggested as a good integrated control measure for grassweeds (Alwarnaidu Vijayarajan et al., Reference Alwarnaidu Vijayarajan, Forristal, Cook, Staples, Schilder, Hennessy and Barth2020; Byrne et al., Reference Byrne, Spink, Freckleton, Neve and Barth2018), barley yellow dwarf virus (Walsh et al., Reference Walsh, Lacey, Doyle, Gaffney and Mc Namara2022) and foliar diseases (Kildea et al., Reference Kildea, Sheppard, Cucak and Hutton2021). Plough-based growers sowed later, following current advice and best practice. In contrast, min-till and direct drill growers chose to sow earlier to mitigate the risk of poor establishment due to adverse weather, which becomes more likely with later sowing dates. Anecdotally, growers believe crop establishment is more variable in non-inversion systems; however, there was no significant difference found in variability of establishment between systems using the coefficients of variations of the plant population densities recorded at 10 positions in each field.
In the on-farm study, results varied with year. In year 1, crops in the direct drill fields yielded significantly lower than those in the plough-based fields; in year 2, there was numerical differences in grain yields between the systems, but these differences were not statistically significant, and in year 3, there was a significant system effect on yield detected. However, despite the numerical differences in yields, post hoc testing could not differentiate between systems. Again, in the all-years data analysis, there were large numerical differences in yields with direct drill mean yields appearing to be substantially lower than the other systems; however, this was only a trend and did not reach significance. Since individual growers were free to apply different management strategies, the possibility of confounding was suspected. In observational studies, external variables not accounted for in the experimental design can influence both the independent factors and the observed outcomes; this dual influence may bias the estimated treatment effects, ultimately obscuring the true relationship between the variables (Greenland, Reference Greenland1989). In field trials, variability in management practices can confound treatment effects, with factors such as nitrogen application and fungicide use having previously been shown to independently influence yield (Olesen et al., Reference Olesen, Jørgensen, Petersen and Mortensen2003; Lollato et al., Reference Lollato, Pradella, Giordano, Ryan, Simão and Soler2022; Koua et al., Reference Koua, Baig, Oyiga, Léon and Ballvora2021). Our observations, which show that these variables are significantly associated with yield, agree with previous findings. This study also, through the detection of variables which were confounded the factor of interest; establishment system; demonstrates that this on-farm study methodology is unsuitable for isolating the performance of each establishment system, as key input levels were not controlled. The treatment effects of system on yield and yield components in the on-farm study are likely to be at least partly masked by the influence of the confounding variables, nitrogen rate and fungicide spend. This reinforces the importance of using controlled experiments to accurately assess system performance.
Differences in nitrogen and fungicide use by growers suggest that they may have been anticipating a lower yield from the use of non-inversion, specifically direct drill systems (Table 7). Another theory is that these non-inversion growers may have been using the establishment cost savings associated with these non-inversion systems as part of a larger effort to reduce their overall production costs. Having similar spends on fungicides and use of nitrogen between groups, or similar degrees of variation in spend and use among groups, may have overcome this challenge. However, this was not possible to achieve as this was an observational study. Given that similar yields were achieved by all systems in the replicated trial, where input levels were similar across all systems, the reduced use of inputs by growers may have had a negative impact on direct drill yields on the study farms. Furthermore, the reduced spend on nitrogen and fungicides in the direct drill fields in the on-farm study coupled with lower crop establishment costs may partially or fully compensate for the numerically lower yields observed in these fields, warranting further exploration in an economic analysis.
The characteristics of this study, which combined a replicated field trial with an on-farm study, across three seasons, allowed unique knowledge to be gathered. The replicated trial results showed that in two of the three years, there was no significant difference in first wheat yields between the three establishment systems, but there was a difference in 1 year, indicating the importance of season (and weather) and confirming the need for long-term studies in this area. The replicated trial also contributed significantly to the explanation of the on-farm study results. The differences in yield between systems recorded in the on-farm study cannot be solely attributed to the crop establishment system, as growers, perhaps anticipating yield differences, had altered their management, which was also a significant factor in explaining the yield differences. In both studies, there was a tendency for direct drilling to have some negative impacts on first wheat performance, indicating the need for climate–specific studies on a range of crop species and spring-sown crops, which may not have the same capacity as wheat to compensate for establishment challenges. While the on-farm studies indicated the range of management and crop performance found on farms, the scale and format of this study did not allow the factors that influence yield to be definitively determined, indicating that multi-factor, controlled, replicated trials are necessary to quantify the effects of different factors and their interaction. The utility of on-farm studies, as used in this trial, may best be limited to the determination of the factors that need to be studied in controlled experimentation. In this study, sowing date, seeding rate, disease control strategy and nitrogen fertilizer use were identified as being of interest when comparing establishment systems. Without these detailed studies, growers will not know how best to manage crops established with different systems. Before promoting the adoption of a crop establishment system in Ireland, further research is needed to assess how crop establishment systems affect the crop, economic and environmental performance of all major arable crops. Any recommendation must then consider the benefits and trade-offs of adopting a given system within the context of Irish cropping systems.
Conclusions
This study combined a replicated trial and on-farm study, showing that the effects of establishment systems on wheat yield are strongly influenced by season and management and that replicated controlled trials are essential to accurately estimate these effects. Future research should quantify establishment system impacts across crops, sowing seasons and environments to inform future management approaches in Irish cropping systems.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0021859625100245.
Acknowledgements
The authors would like to thank Martin Walsh, Frank Ryan and Kevin Murphy for maintaining the field trial over the course of the experiment, the Teagasc advisors and specialists who helped recruit grower participants for the on-farm study, Richie Hackett who provided technical support on the value of organic manure applications and Senan Forristal, Angelique Collin, Martin Bertrand, Antoine Vantal, Cyril Crismer and Gerry Nolan for their assistance harvesting and processing samples from the on-farm study. Special thanks must also go to the growers who participated in the on-farm study and generously gave up their time to provide data.
Author contribution
Conceptualization: J.J., P.D.F and K.M.; data curation: J.J. and P.D.F; formal analysis: J.J., P.D.F and K.M.; funding acquisition: P.D.F and K.M.; investigation: J.J., P.D.F and K.M.; methodology: J.J., P.D.F and K.M.; project administration: J.J., P.D.F and K.M.; supervision: P.D.F and K.M.; visualization: J.J., P.D.F and K.M.; writing – original draft: J.J.; writing – review and editing: J.J., P.D.F and K.M.
Funding statement
This research was funded under the Teagasc Walsh Scholarship Program, Teagasc project no. 0822.
Competing interests
The authors declare there are no conflicts of interest.
Open access
