Comparing the reliability of maize variety data from on-farm trials and experimental stations
Vergleich der Konsistenz von Sortenergebnissen bei Silomais in On-Farm- und Exaktversuchen
Journal für Kulturpflanzen, 66 (11). S. 389–395, 2014, ISSN 1867-0911, DOI: 10.5073/JfK.2014.11.03, Verlag Eugen Ulmer KG, Stuttgart
While current variety recommendations are based on replicated small-plot trials in a specific region for which the trial is thought to be representative, genotype × environment interactions, genotype × management interactions, and increasing weather fluctuations make it increasingly difficult to predict which variety will be best in a given environment. An additional approach is therefore to decentralize variety trials and place them on working farms. However, although on-farm trials offer potentially more relevance for direct variety selection on site, they are also likely to be subject to more noise and trial entries can often not be fully replicated. To evaluate the relative merit of on-farm trials vs. fully replicated trials conducted at experimental stations, we tested 6 maize varieties at four farms and at two stations in a region dominated by sandy soils. The variance of variety rankings over the years within each site was used as proxy to evaluate the consistency of variety information gained at each location. For dry matter yield, on-farm trials showed both the highest and the lowest consistency of variety ranking, with the consistency being intermediate at the experimental stations. For some quality parameters, namely non-fibre carbohydrate and starch content, the majority of on-farm trials showed more consistent variety ranking over the years than the more consistent of the two replicated trials. This suggests that in terms of year-on-year reliability of maize yield and quality, on-farm trials may have the potential to complement replicated variety trials. For both types of trials, however, there is also scope for decreasing technical sources of variation.
Key words: Genotype environment interactions; on-farm research; stability
Sortenempfehlungen basieren in der Regel auf regionalspezifischen Exaktversuchen. Dabei können die Wechselwirkungen zwischen Sorte und Umwelt sowie zwischen Sorte und Management zu steigenden Problemen bei der Sortenwahl führen. Eine ergänzende Möglichkeit sind daher dezentralisierte Sortenversuche in Landwirtschaftsbetrieben. Für die direkte Sortenwahl bieten diese On-Farm-Versuche eine hohe Praxisrelevanz, da die tatsächlichen Einflussgrößen im Betrieb getreuer abgebildet werden können, als es Exaktversuche vermögen, wenn sie nur an wenigen, zum Teil eingeschränkt repräsentativen Standorten durchgeführt werden. Andererseits ist die Aussagefähigkeit von On-Farm-Versuchen häufig durch Störgrößen und fehlende Feldwiederholungen begrenzt. Anhand von Maissortenversuchen mit sechs verschiedenen Sorten wurden auf vier Brandenburger Landwirtschaftsbetrieben sowie auf zwei Versuchsstationen die Sortenleistung in On-Farm- und Exaktversuchen auf Sandböden geprüft. An jedem Standort wurde die Varianz der Sortenrankings über die Jahre genutzt, um Informationen über die Konsistenz der Sortenleistung an jedem der Standorte zu gewinnen. Für den Trockenmasse-Ertrag zeigten die On-Farm-Versuche sowohl die höchste, als auch die niedrigste Konsistenz der Rankings. Einige Qualitätsparameter, wie die Nichtfaser-Kohlenhydrate (NFC) und der Stärkegehalt, zeigten in den On-Farm-Versuchen eine bessere Konsistenz des Sortenrankings über die Jahre als in beiden Exaktversuchen. Dies legt nahe, dass On-Farm-Versuche das Potenzial haben, die regionale Leistungsprüfung von Sorten zu unterstützen.
Stichwörter: Genotyp-Umwelt Interaktion, On-Farm- Versuche, Stabilität
Model-based climate scenarios predict an upward shift in the mean temperature and changes in the distribution of precipitation for the coming decades (Stocker et al., 2013). In addition, it has been predicted that climate change also involves a rise of weather variability, i.e. increasing deviations from the mean (Schär et al., 2004; Motha and Baier, 2005; Hansen et al., 2012), although recent research has contested this view (Huntingford et al., 2013). In any case, however, climate change is likely to affect the frequency of at least some types of extreme weather events, and this will have impacts on the growth of terrestrial plants (Reyer et al., 2013).
Such effects may be of particular importance when the capacity of the site on which plants are grown is low. This is the case in the East German region of Brandenburg where yearly precipitation is less than 600 mm and light sandy soils prevail, showing a low buffer capacity against shortages of water. In this region, climate scenarios predict decreasing evapotranspiration in particular in the months of May to July (Gerstengarbe et al., 2003). In face of increasing frequency of extreme weather events, coupled with a poor ability of the soil to buffer against water fluctuations, the ability of arable crops to produce high and stable yields under highly variable environmental conditions becomes more and more important (Döring et al., 2010; Urban et al., 2012). One possibility for adaptation to climate change is to cultivate varieties with a higher climatic tolerance against e.g. temporal drought stress. In addition, it is necessary to create new variety selection strategies, which are more adapted to the local natural conditions and take the regional climatic differences into account.
Generally, variety choice can be based on fully replicated field trials conducted on experimental stations, or on variety trials conducted on working farms. Experimental stations offer standardized techniques of gathering data and, because of replication, ensure that data can be analyzed statistically. However, such trials are typically characterized by small plot sizes, which could potentially result in substantial edge effects. Furthermore, small-scale fully replicated trials are managed with specialised machinery adapted for plot trials (e.g. for sowing and harvesting) that differ from farm scale machinery. On the other hand, there is currently no direct evidence for differences between plot-scale and farm-scale machinery on the same site. In addition, farmers have been shown to be sceptical about the relevance of information from experimental stations, because transferability of results to local conditions on the farm remains questionable in their view (Rzewnicki, 1991). Therefore, both farmers and researchers have advocated on-farm research to obtain more locally relevant information about variety performance in a decentralized way. While progress has been made in the past in terms of how to make on-farm trials more robust (Piepho et al., 2011; Thöle et al., 2013), constraints on the farm often mean that requirements suggested by statisticians cannot be met in practice. With these complementary benefits and drawbacks of trials conducted on-farm and on experimental stations, there is a need to know which type of trial provides more reliable information for on-farm variety selection. However, to our knowledge there is currently no quantitative evidence about the relative merits of the two different types of experimentation in Central Europe, although the topic has already been subject to legal disputes (OLG Düsseldorf, 2011).
This study therefore aimed to compare on-farm trials and fully replicated trials in the region of Brandenburg to support variety choice on agricultural farms, with a focus on the four main crops grown in the region, namely maize, rye, wheat and oilseed rape. Here we report results from the maize trials, including yield and quality data. The trial series was run as part of the INKA-BB project (Innovation Network of Climate Change Adaptation Brandenburg Berlin). It was organized by a farmer-researcher network established in the INKA-BB subproject “Variety Strategies to Adaptation on Climate Change”.
The on-farm trials and the fully replicated plot trials were conducted over three years under different local conditions in cooperation with four farms in the study region.
The regional situation of the farms represents the range of geoecological conditions in Brandenburg with a gradient from north to south relating to soil, climate and weather (Tab. 1). The locations Groß Schönebeck and Trebbin are characterized by very light sandy soils and low yield potential (average 28 soil points). Passow and Groß Gastrose represent locations with comparatively better soil conditions and higher yield potential (average 51 soil points). In both the replicated and the on-farm trials, row width was 75 cm. On-Farm trials were set up as unreplicated strips, and two rows were harvested from four pseudo-replicated subplots on each strip. The average length of these plots in the on-farm trials was 16.4 m (min: 8.1; max: 33.3), resulting in an average harvested area of 4 x 16.4 m × 1.5 m = 98 m² for each variety. The field trials on the experimental stations in Berlin-Dahlem and Thyrow were set up in a randomized complete block design with four replicates in each year and a plot size of 10 m × 3 m. In these trials, the central two rows were harvested from a length of 8 m, i.e. 1 m at each end of the plot as well as the outer two maize rows were not included in the analysis in order to reduce edge effects. This resulted in a harvested area of 4 x 8 m × 1.5 m = 48 m² for each variety. At the beginning of the project, mobile weather stations were installed on all four farms, in order to assess the results in relation to locally prevailing weather conditions.
Table 1. Experimental stations and on-farm locations in Berlin and Brandenburg. ‘German soil rating index’ represents a German system of classifying general productivity of an arable site, with 1 and 100 being the minimum and maximum respectively (Finnern et al., 1996)
Versuchsstationen und On-Farm-Standorte in Berlin und Brandenburg. Die Einteilung in Bodenpunkte ist ein deutsches Klassifizierungssystem, mit dessen Hilfe die Produktivität einer Ackerfläche angegeben wird. Die Einteilung erfolgt von minimal 1 bis maximal 100
Type | Name | Acronym | Region | Coordinates (latitude °N, longitude °E) | German soil |
Experimental stations | Berlin-Dahlem | DAH | Berlin | 52.46629,13.29924 | 35 |
Thyrow | THY | Teltow-Fläming | 52.25418,13.23679 | 28 | |
On-farm- | Passow | PAS | Uckermark | 53.14035,14.10801 | 50 |
Groß Schönebeck | GSB | Barnim | 52.91136,13.52784 | 28 | |
Trebbin | TRE | Teltow-Fläming | 52.19847,13.24494 | 27 | |
Groß Gastrose | GRG | Spree-Neiße | 51.88270,14.64833 | 51 |
Variety selection for the trials was based on the recommendations of the “State Agency for Rural Development, Agriculture and Land Reassignment Brandenburg” (LELF). In the years 2010–2012 the maize varieties ‘Kalvin’ (S 220), ‘LG30218’ (S 220), ‘Mazurka’ (S 240), ‘Torres’ (S 250), ‘Aabsolut’ (S 260), and ‘Ingrid’ (S 260) were cultivated at all six locations. The values given in brackets indicate the silage maturity grouping of maize. The German Federal Plant Variety Office specified these values based on the dry matter content at the point of harvest. The values are classified in three groups. Values of S 220 and below represent the “early” group varieties, the group “mid-early” contains varieties with values from S 230 – S 250, and in group “mid-late to late” are values larger than S 260 (BSA, 2013). Thus, in each of the three maturity groups two varieties were tested in the trials. In addition to the on-farm trials the varieties were cultivated in fully replicated trials in Berlin-Dahlem and Thyrow, which are locations of the Training and Research Station of the Faculty of Life Sciences of Humboldt-University Berlin.
In the on-farm trials the maize was harvested in the second half of September of each year, using experimental harvest technology of the Training and Research Station of the Faculty of Life Sciences of Humboldt-University Berlin. In this way it was possible to quantify the maize yield of different varieties and locations more exactly than if farm machinery had been used for harvesting. For all locations the analyses of maize dry matter yield and quality of harvested plant samples were conducted in each of the three years. The analysis of the quality parameters was carried out in the Laboratory of the State Control Association Brandenburg (LKVBB) according to established laboratory standards (VDLUFA, 1976).
A common way to compare crop varieties across different locations is to use rankings of their performance within each test environment (e.g. Huehn, 1990; Vlachostergios and Roupakias, 2008). Here, our aim was to compare the two different trial set ups (replicated trials at two locations and on-farm trials at four locations) in terms of their reliability of variety data.
The rationale was to assess for each of the six locations how much the variety rankings varied across the three study years. In this case a large variance of the variety rankings over the three years suggests that reliable information about which varieties performed best (or worst) at a particular location was difficult to obtain; conversely, low variance indicates high reliability of the information gathered about variety performance. Put differently, the variance of variety rankings over the years within each site was used as proxy to evaluate the consistency of variety data gained at each location. At the same time, these variances correspond to Huehn’s stability parameter Si(2) (Huehn, 1990).
In addition, the maize data were analyzed with a mixed model approach using site and year as random factors and variety as fixed factor to compare variety performance. Varieties were compared to ‘Ingrid’ as a control variety using Dunnett’s test, since ‘Ingrid’ showed the highest mean dry matter yield.
The variety testing showed considerable site-specificity of the differences among the varieties. Therefore, no common variety recommendation could be given for the study region as represented by the set of trial sites.
Despite the underlying differences in soil quality at the six different trial sites (Tab. 1), dry matter yields, averaged over the three study years showed only relatively small differences between the sites (Tab. 2). For instance, despite the low yield potential at Thyrow (28 soil points) the mean yield achieved there was nearly the same over the three years as obtained at the location with the generally higher yield potential (Berlin-Dahlem, 35 soil points).
Table 2. Mean silage maize yields (t ha–1 DM) of six different varieties in on-farm (GRG, GSB, PAS, TRE) and fully replicated trials (DAH, THY), means over three years (2010–2012)
Mittlere Silomais-Erträge (t ha–1 TM) von sechs Sorten in On-Farm (GRG, GSB, PAS, TRE) und Exaktversuchen (DAH, THY), Mittelwerte über drei Jahre (2010–2012)
Variety | DAH | THY | GRG | GSB | PAS | TRE | Mean |
Aabsolut | 18.8 | 19.0 | 19.6 | 16.5 | 16.2 | 18.1 | 18.0 |
Ingrid | 18.7 | 19.1 | 19.7 | 16.1 | 17.2 | 18.6 | 18.2 |
Kalvin | 17.5 | 17.5 | 16.0 | 16.2 | 16.5 | 18.4 | 17.0 |
LG30218 | 16.8 | 16.3 | 18.1 | 14.4 | 16.2 | 16.1 | 16.3 |
Mazurka | 17.0 | 15.9 | 16.8 | 15.7 | 16.4 | 16.4 | 16.4 |
Torres | 18.2 | 18.4 | 18.6 | 17.1 | 18.2 | 18.8 | 18.2 |
Mean | 17.9 | 17.7 | 18.1 | 16.0 | 16.8 | 17.7 | 17.4 |
However, as reported elsewhere (Klepatzki et al., 2013), the yield fluctuations over the years at the locations were related to soil quality, i.e. the light sandy soils showed higher yield variability than the better sites. Dry matter yields of varieties ‘Aabsolut’ and ‘Torres’ were not significantly different from control variety ‘Ingrid’, whereas differences between ‘Ingrid’ and the other three varieties were significant.
Among the quality parameters, relatively large differences among varieties were observed for crude lipid (CL, 35.6% difference between maximum and minimum) and starch content (ST, 26.1%), whereas differences between varieties were small for usable crude protein (UCP, 5.3%), metabolisable energy (ME, 4.8%) and enzyme digestible organic matter (ELOS, 5.0%) (Tab. 3). Differences between individual varieties and the control variety ‘Ingrid’ were significant for ‘Kalvin’ (all parameters except RA and NEL), ‘LG30218’ (all parameters except CP, CL, and RNB), ‘Mazurka’ (all parameters except RA and NFC), and ‘Torres’ (all parameters). In contrast, differences between ‘Aabsolut’ and ‘Ingrid’ were non-significant for all parameters except ST.
Table 3. Comparison of quality parameters in g kg–1 DM (means over three years and all locations)
Vergleich der Qualitätsparameter in g kg–1 TM (Mittelwert über drei Jahre und alle Standorte)
Variety | CL | CP | UCP | ST | RNB | CF | ME | NEL | oNDF | NFC | ELOS | RA |
Aabsolut | 19.5 | 67.4 | 125.1 | 275.8 | –9.2 | 200.9 | 10.5 | 6.3 | 480.1 | 393.7 | 700.3 | 39.3 |
Ingrid | 18.8 | 67.6 | 123.6 | 248.2 | –9.0 | 210.7 | 10.4 | 6.2 | 494.3 | 379.6 | 686.1 | 39.7 |
Kalvin | 21.2 | 74.3 | 128.6 | 296.6 | –8.7 | 184.8 | 10.7 | 6.5 | 461.8 | 404.1 | 710.2 | 38.7 |
LG30218 | 20.2 | 69.9 | 127.8 | 313.1 | –9.2 | 187.4 | 10.7 | 6.5 | 459.7 | 413.0 | 720.6 | 37.1 |
Mazurka | 21.8 | 76.1 | 128.6 | 292.7 | –8.4 | 190.0 | 10.7 | 6.4 | 472.6 | 389.3 | 703.8 | 40.2 |
Torres | 25.5 | 71.2 | 130.1 | 302.5 | –9.4 | 184.8 | 10.9 | 6.6 | 456.2 | 411.4 | 720.3 | 35.7 |
Mean | 21.2 | 71.1 | 127.3 | 288.2 | –9.0 | 193.1 | 10.7 | 6.4 | 470.8 | 398.5 | 706.9 | 38.5 |
Parameters: Crude lipid (CL), crude protein (CP), usable crude protein (UCP), starch (ST), ruminal nitrogen balance (RNB), crude fiber (CF), metabolisable energy (ME), net energy content for lactation (NEL), organic neutral detergent fiber (oNDF), non-fiber carbohydrate (NFC), enzyme digestible organic matter (ELOS), raw ash (RA) | ||||||||||||
Variety rankings of dry matter yield varied both between sites and between years (Tab. 4). Across all locations and years, varieties ‘Aabsolut’, ‘Ingrid’ and ‘Torres’ were consistently better than the other three varieties. In terms of reliability of the dry matter yield data gained at each location, the on-farm trials showed both the highest (∑Si(2) = 3.7) and the lowest (∑Si(2) = 18.3) consistency of variety ranking, with the consistency being intermediate at the experimental stations (Tab. 5).
Table 4. Variety rankings of six maize varieties at six locations over three study years
Sortenrangfolgen von sechs Silomaissorten an sechs Standorten über drei Untersuchungsjahre
Site | |||||||||||||||||||||||
DAH | THY | GRG | GSB | PAS | TRE | ||||||||||||||||||
Year* | 1 | 2 | 3 | 1 | 2 | 3 | 1 | 2 | 3 | 1 | 2 | 3 | 1 | 2 | 3 | 1 | 2 | 3 | |||||
Aabsolut | 1 | 3 | 1 | 1 | 2 | 3 | 3 | 1 | 1 | 2 | 2 | 3 | 2 | 2 | 6 | 1 | 4 | 4 | |||||
Ingrid | 2 | 1 | 3 | 3 | 1 | 1 | 1 | 2 | 2 | 5 | 3 | 2 | 1 | 6 | 1 | 2 | 3 | 3 | |||||
Kalvin | 3 | 5 | 4 | 4 | 4 | 5 | 5 | 6 | 5 | 4 | 1 | 4 | 4 | 5 | 4 | 4 | 1 | 2 | |||||
LG30218 | 6 | 6 | 5 | 5 | 6 | 4 | 4 | 4 | 3 | 6 | 6 | 6 | 5 | 4 | 5 | 5 | 5 | 6 | |||||
Mazurka | 4 | 4 | 6 | 6 | 5 | 6 | 6 | 5 | 6 | 3 | 5 | 5 | 6 | 3 | 3 | 6 | 6 | 5 | |||||
Torres | 5 | 2 | 2 | 2 | 3 | 2 | 2 | 3 | 4 | 1 | 4 | 1 | 3 | 1 | 2 | 3 | 2 | 1 | |||||
* Years:1 = 2010, 2 = 2011, 3 = 2012 | |||||||||||||||||||||||
Table 5. Mean ranks and Huehn´s Si(2) of the six maize varieties at the six locations: data for dry matter yield
Mittlere Ränge und Huehn´s Si(2) von sechs Silomaissorten und sechs Standorten: Trockenmasseertrag
Variety | DAH | THY | GRG | GSB | PAS | TRE | |
Mean ranks | Aabsolut | 1.7 | 2.0 | 1.7 | 2.3 | 3.3 | 3.0 |
Ingrid | 2.0 | 1.7 | 1.7 | 3.3 | 2.7 | 2.7 | |
Kalvin | 4.0 | 4.3 | 5.3 | 3.0 | 4.3 | 2.3 | |
LG30218 | 5.7 | 5.0 | 3.7 | 6.0 | 4.7 | 5.3 | |
Mazurka | 4.7 | 5.7 | 5.7 | 4.3 | 4.0 | 5.7 | |
Torres | 3.0 | 2.3 | 3.0 | 2.0 | 2.0 | 2.0 | |
Si(2) | Aabsolut | 1.3 | 1.0 | 1.3 | 0.3 | 5.3 | 3.0 |
Ingrid | 1.0 | 1.3 | 0.3 | 2.3 | 8.3 | 0.3 | |
Kalvin | 1.0 | 0.3 | 0.3 | 3.0 | 0.3 | 2.3 | |
LG30218 | 0.3 | 1.0 | 0.3 | 0.0 | 0.3 | 0.3 | |
Mazurka | 1.3 | 0.3 | 0.3 | 1.3 | 3.0 | 0.3 | |
Torres | 3.0 | 0.3 | 1.0 | 3.0 | 1.0 | 1.0 | |
Sum | 8.0 | 4.3 | 3.7 | 10.0 | 18.3 | 7.3 |
Tab. 6 shows the variability of the rankings for the quality parameters in g kg–1 dry matter for the examined locations. The parameter non-fiber carbohydrate (NFC) in all on-farm trials showed a more consistent variety ranking over the years than in the best of the replicated trials. For the other parameters we obtained a mixed picture, but in 10 out of 12 quality parameters at least one on-farm trial showed lower variability of the variety rankings than the best of the replicated trials. For two quality parameters, namely non-fiber carbohydrate (NFC) and starch content (ST), the majority (i.e. 3 or 4 out of 4) of on-farm trials showed more consistent variety rankings over the years than the more consistent of the two replicated trials.
Table 6. Variability of variety rankings with regard to the quality parameters, calculated as Si(2) summed over all six varieties (cf. Table 5). ‘Best rep’ represents the replicated trial with the lowest ∑Si(2), ‘Median rep’ is the median value for the two replicated trials
Variabilität in den Sortenrangfolgen in Bezug auf die Qualitätsparameter, kalkuliert als Summe von Si(2) über sechs Sorten (siehe Tabelle 5). ‘Best rep’ repräsentiert den Exaktversuch mit dem geringsten ∑Si(2), ‘Median rep’ ist der Median der beiden Exaktversuche
Site | Yield | CL | CP | UCP | ST | RNB | CF | ME | NEL | oNDF | NFC | ELOS | RA |
DAH | 8.0 | 6.3 | 6.3 | 9.3 | 13.7 | 6.3 | 10.7 | 11.0 | 12.7 | 14.7 | 21.3 | 14.7 | 11.7 |
THY | 4.3 | 13.0 | 9.3 | 11.0 | 11.0 | 11.3 | 18.3 | 16.3 | 18.3 | 18.3 | 21.0 | 19.3 | 19.7 |
GRG | 3.7 | 4.7 | 4.0 | 7.3 | 12.7 | 10.7 | 14.7 | 16.0 | 16.0 | 16.7 | 18.7 | 16.7 | 14.0 |
GSB | 10.0 | 8.3 | 15.3 | 10.7 | 6.0 | 22.0 | 6.0 | 5.7 | 7.0 | 5.3 | 8.3 | 21.3 | 21.7 |
PAS | 18.3 | 8.3 | 9.0 | 4.7 | 8.3 | 6.7 | 8.0 | 6.3 | 6.3 | 6.7 | 9.0 | 7.3 | 16.7 |
TRE | 7.3 | 13.3 | 9.0 | 11.3 | 10.3 | 12.7 | 14.0 | 13.7 | 13.0 | 15.3 | 11.3 | 10.3 | 17.7 |
Median | 7.7 | 8.3 | 9.0 | 10.0 | 14.0 | 11.0 | 12.3 | 12.3 | 12.8 | 15.0 | 15.0 | 15.7 | 17.2 |
Best rep | 4.3 | 6.3 | 6.3 | 9.3 | 11.0 | 6.3 | 10.7 | 11.0 | 12.7 | 14.7 | 21.0 | 14.7 | 11.7 |
Median rep | 6.2 | 9.7 | 7.8 | 10.2 | 12.4 | 8.8 | 14.5 | 13.7 | 15.5 | 16.5 | 21.2 | 17.0 | 15.7 |
Parameters: Crude lipid (CL), crude protein (CP), usable crude protein (UCP), starch (ST), ruminal nitrogen balance (RNB), crude fiber (CF), metabolisable energy (ME), net energy content for lactation (NEL), organic neutral detergent fiber (oNDF) non-fiber carbohydrate (NFC), enzyme digestible organic matter (ELOS), raw ash (RA). | |||||||||||||
Further, 2 out of 4 on-farm trials showed more consistent variety rankings than the more consistent of the replicated trials for the usable crude protein (UCP), crude fiber (CF) metabolisable energy (ME), net energy content for lactation (NEL), organic neutral detergent fiber (oNDF), and enzyme digestible organic matter (ELOS). Thus, our results suggest that for a considerable number of parameters replicated plot trials at experimental stations do not necessarily outperform on-farm trials in terms of consistency of variety rankings.
Usually, variety trials on experimental stations are characterized by relatively small plot sizes. In addition, sites of experimental stations are typically selected for homogeneous soil conditions. Thus, underlying heterogeneity of soil conditions is expected to be low in such trials. In contrast, on-farm trials, with their larger plots size, the use of pseudo-replications and potentially less careful site selection, can be expected to show comparatively large underlying soil heterogeneity within a trial. As a consequence on-farm trials would be predicted to show smaller reliability than trials conducted at experimental stations. However, the results of this study indicate that in terms of year-on-year reliability of maize yield and some quality parameters, on-farm trials may have the potential to complement, or even to outperform replicated variety trials.
One possible reason for this outcome is that year × variety interactions might have been stronger than interactions between variety and soil conditions within sites, so that the effects discussed above may just not be relevant. Underlying mechanisms for the observed results, however, remain speculative. In general, our results are preliminary in that they are based on a relatively small data set. Therefore, we suggest that more research with a larger number of stations and on-farm locations should be conducted over a longer period of time.
In summary, we have shown that on-farm experiments can generate valuable information about variety performance and adaptation to site conditions in arable systems on comparatively marginal sites. Thus we have demonstrated that on-farm trials allow a practically relevant complementation of regional variety testing. This is of particular importance when resources for state-funded, i.e. official variety testing are being cut, so that regionally relevant and independent information on variety performance is increasingly difficult to obtain for farmers. Adaptation of agricultural production to climate change will require coordinated strategies. Our study supports the view that it is useful to build a regional network of on-farm trials when using variety selection as one component of these efforts. Such networks are likely to be instrumental for mastering the multiple challenges lying ahead.
However, it must also be taken into account that on-farm variety trials can usually only cover a relatively small proportion of the available varieties. Special attention must therefore be paid to designing the process of selecting varieties that are to be tested on-farm. Further, on-farm trials and their coordination will also entail costs. In our case, however, the participating farmers showed strong interest in continuing this investment for generating relevant on-farm knowledge.
BSA (Bundessortenamt), 2013: Beschreibende Sortenliste – Getreide, Mais, Öl- und Faserpflanzen, Leguminosen, Rüben, Zwischenfrüchte. Hannover, Bundessortenamt (Federal Plant Variety Office).
Döring, T.F., M. Wolfe, H. Jones, H. Pearce, J. Zhan, 2010: Breeding for resilience in wheat – Nature’s choice. In: Goldringer, I., E. Lammerts van Bueren (Eds.): Breeding for resilience: a strategy for organic and low-input farming sytems? Eucarpia 2nd Conference of the Organic and Low-Input Agriculture Section. 1–3 December 2010. Paris, France; pp 45-48.
Finnern, H., W. Grottenthaler, D. Kühn, W. Pälchen, W.-G. Schraps, H. Sponagel, 1996: Bodenkundliche Kartieranleitung. Stuttgart, Schweizerbart'sche Verlagsbuchhandlung.
Gerstengarbe, F.-W., F. Badeck, F. Hattermann, V. Krysanova, W. Lahmer, P. Lasch, M. Stock, F. Suckow, F. Wechsung, P.C. Werner, 2003: Studie zur klimatischen Entwicklung im Land Brandenburg bis 2055 und deren Auswirkungen auf den Wasserhaushalt, die Forst- und Landwirtschaft sowie die Ableitung erster Perspektiven. Potsdam, PIK.
Hansen, J., M. Sato, R. Ruedy, 2012: Perception of climate change. Proc. Nat. Acad. Sci. 109, E2415-E2423.
Huehn, M., 1990: Nonparametric measures of phenotypic stability. Part 1: Theory. Euphytica 47, 89-194.
Huntingford, C., P.D. Jones, V.N. Livina, T.M. Lenton, P.M. Cox, 2013: No increase in global temperature variability despite changing regional patterns. Nature 500, 327-330.
INKA BB, 2013: Innovationsnetzwerk Klimaanpassung Brandenburg Berlin. Availaible at http://www.inka-bb.de, 23.06.2014.
Klepatzki, J., J. Sayer, T.F. Döring, F. Ellmer, 2013: Ertragsstabilität von Silomais in Brandenburg – Ergebnisse von On-Farm-Versuchen. Mitteilungen der Gesellschaft für Pflanzenbauwissenschaften 25, 335-336.
Motha, R., W. Baier, 2005: Impacts of Present and Future Climate Change and Climate Variability on Agriculture in the Temperate Regions: North America. In: Increasing Climate Variability and Change. Springer Netherlands, pp. 137-164.
OLG (Oberlandesgericht – Higher Regional Court) Düsseldorf, 2011: Urteil des OLG Düsseldorf vom 22.Februar 2011. Availaible at http://openjur.de/u/448118.html, 22.06.2014.
Piepho, H.-P., C. Richter, J. Spilke, K. Hartung, A. Kunick, H. Thöle, 2011: Statistical aspects of on-farm experimentation. Crop and Pasture Science 62, 721-735.
Reyer, C.P.O., S. Leuzinger, A. Rammig, A. Wolf, R.P. Bartholomeus, A. Bonfante, F. de Lorenzi, M. Dury, P. Gloning, R. Abou Jaoudé, T. Klein, T.M. Kuster, M. Martins, G. Niedrist, M. Riccardi, G. Wohlfahrt, P. de Angelis, G. de Dato, L. François, A. Menzel, M. Pereira, 2013: A plant's perspective of extremes: terrestrial plant responses to changing climatic variability. Glob. Change Biol. 19, 75-89.
Rzewnicki, P., 1991: Farmers’ perceptions of experiment station research, demonstrations, and on-farm research in agronomy. Journal of Agronomic Education 20, 31-36.
Schär, C., P. Vidale, D. Lüthi, C. Frei, C. Häberli, M. Liniger, C. Appenzeller, 2004: The role of increasing temperature variability in European summer heatwaves. Nature 427, 332-336.
Stocker, T.F., D. Qin, G.-K. Plattner, L.V. Alexander, S.K. Allen, N.L. Bindoff, F.-M. Bréon, J.A. Church, U. Cubasch, S. Emori, P. Forster, P. Friedlingstein, N. Gillett, J.M. Gregory, D.L. Hartmann, E. Jansen, B. Kirtman, R. Knutti, K.K. Kumar, P. Lemke, J. Marotzke, V. Masson-Delmotte, G.A. Meehl, I.I. Mokhov, S. Piao, V. Ramaswamy, D. Randall, M. Rhein, M. Rojas, C. Sabine, D. Shindell, L.D. Talley, D.G. Vaughan, S.-P. Xie, 2013: Technical Summary. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. NY, USA, Cambridge University Press, Cambridge, United Kingdom and New York, pp. 33-115.
Thöle, H., C. Richter, D. Ehlert, 2013: Strategy of statistical model selection for precision farming on-farm experiments. Precision Agriculture 14, 434-449.
Urban, D., M.J. Roberts, W. Schlenker, D.B. Lobell, 2012: Projected temperature changes indicate significant increase in interannual variability of U.S. maize yields. Climatic Change 112, 525-533.
VDLUFA (Verband Deutscher Landwirtschaftlicher Untersuchungs- und Forschungsanstalten), 1976: Method book III "The chemical analysis of feedstuffs" of VDLUFA. VDLUFA: Association of German Agricultural Analytic and Research Institutes.
Vlachostergios, D.N., D.G. Roupakias, 2008: Response to conventional and organic environment of thirty-six lentil (Lens culinaris Medik.) varieties. Euphytica 163, 449-457.
Suchen