ArticlePDF Available

Adaptability analysis in a participatory variety trial of organic vegetable crops

  • Organic Seed Alliance

Abstract and Figures

Successful organic farming requires crop varieties that are resilient to environmental variability. Assessing variety performance across the range of conditions represented on working farms is vital to developing such varieties; however, data collected from on-farm, participatory trials can be difficult to both collect and interpret. To assess the utility of data arising from participatory trialing efforts, we examined the performance of butternut squash ( Cucurbita moschata L.), broccoli ( Brassica oleracea L.) and carrot ( Daucus carota L.) varieties grown in diverse organic production environments in participatory trials in Oregon, Washington, Wisconsin and New York using adaptability analysis (regression of variety means on environmental index). Patterns of adaptation varied across varieties, with some demonstrating broad adaptation and others showing specific adaptation to low- or high-yielding environments. Selection of varieties with broad vs specific adaptation should be guided by farmers’ risk tolerance and on-farm environmental variation. Adaptability analysis was appropriate for continuous variables (e.g., yield traits), but less so for ordinal variables and quality traits such as flavor and appearance, which can be vitally important in organic vegetable crop variety selection. The relative advantages of adaptability analysis and additive main effects and multiplicative interactions are also discussed in relation to on-farm trial networks. This work demonstrated the unique challenges presented by extensive participatory vegetable trialing efforts, which, as compared to grain crops, require novel approaches to facilitating farmer participation as well as data collection and analysis. Efficient, precise and reliable methods for evaluating quality related traits in these crops would allow researchers to assess stability and adaptation across a wider range of traits, providing advantages for effective plant breeding and trialing activities within the organic sector.
Content may be subject to copyright.
Renewable Agriculture and
Food Systems
Research Paper
*Current address: University of British
Columbia, 2357 Main Mall, Vancouver, BC, V6 T
1Z4, Canada.
Cite this article: Lyon A, Tracy W, Colley M,
Culbert P, Mazourek M, Myers J, Zystro J, Silva
EM (2019). Adaptability analysis in a
participatory variety trial of organic vegetable
crops. Renewable Agriculture and Food Systems
Received: 11 December 2017
Revised: 21 May 2018
Accepted: 13 November 2018
Key words:
On-farm research; organic agriculture;
participatory variety selection; seed systems;
vegetable production
Author for correspondence:
Alexandra Lyon, E-mail: alexandra.lyon@ubc.
© Cambridge University Press 2019
Adaptability analysis in a participatory variety
trial of organic vegetable crops
Alexandra Lyon1,2,*, William Tracy1, Micaela Colley3, Patrick Culbert2,
Michael Mazourek4, James Myers5, Jared Zystro1and Erin M. Silva1
University of Wisconsin-Madison, Madison, WI, USA;
University of British Columbia, Vancouver, BC, Canada;
Wageningen University, Wageningen, Netherlands;
Cornell University, Ithaca, NY, USA and
Oregon State
University, Corvallis, OR, USA
Successful organic farming requires crop varieties that are resilient to environmental variabil-
ity. Assessing variety performance across the range of conditions represented on working
farms is vital to developing such varieties; however, data collected from on-farm, participatory
trials can be difficult to both collect and interpret. To assess the utility of data arising from
participatory trialing efforts, we examined the performance of butternut squash (Cucurbita
moschata L.), broccoli (Brassica oleracea L.) and carrot (Daucus carota L.) varieties grown
in diverse organic production environments in participatory trials in Oregon, Washington,
Wisconsin and New York using adaptability analysis (regression of variety means on environ-
mental index). Patterns of adaptation varied across varieties, with some demonstrating broad
adaptation and others showing specific adaptation to low- or high-yielding environments.
Selection of varieties with broad vs specific adaptation should be guided by farmersrisk tol-
erance and on-farm environmental variation. Adaptability analysis was appropriate for con-
tinuous variables (e.g., yield traits), but less so for ordinal variables and quality traits such
as flavor and appearance, which can be vitally important in organic vegetable crop variety
selection. The relative advantages of adaptability analysis and additive main effects and multi-
plicative interactions are also discussed in relation to on-farm trial networks. This work
demonstrated the unique challenges presented by extensive participatory vegetable trialing
efforts, which, as compared to grain crops, require novel approaches to facilitating farmer par-
ticipation as well as data collection and analysis. Efficient, precise and reliable methods for
evaluating quality related traits in these crops would allow researchers to assess stability
and adaptation across a wider range of traits, providing advantages for effective plant breeding
and trialing activities within the organic sector.
In organic systems, farmers must integrate ecological approaches into crop management. Crop
improvement through selection has emerged as an agroecological practice that allows for
greater success in low-input environments, providing a buffer against environmental stresses.
The development of crop varieties specifically for organic environments allows for more effect-
ive integration of this agroecological strategy by organic farmers. Developing varieties for
organic systems through plant breeding in conventional systems is a form of indirect selection,
in which the selection environment as assumed to be a sufficient representation of the target
environment. However, research in multiple crop species has found that organic agriculture is
distinct from conventional agriculture as a selection environment for plant breeding, and that
direct selection in organically managed environments will result in higher yielding varieties in
those systems than would indirect selection in conventional environments (Murphy et al.,
2007; Singh et al., 2011; Kirk et al., 2012; Lammerts van Bueren and Myers, 2012; Entz
et al., 2015). These findings suggest that plant breeding for organic production should include
organic trial sites either exclusively or in combination with sites under conventional manage-
ment (Burger et al., 2008; Renaud et al., 2014).
Breeding for organic agriculture, though still a small proportion of plant breeding efforts, is
now being conducted in the private, public and nonprofit sectors (Mazourek et al., 2009;
Myers et al., 2012a,2012b; Shelton and Tracy, 2015). Efforts to breed for organic agriculture
have highlighted the importance of on-farm evaluation and selection of crop varieties, a strat-
egy that has been successfully employed in participatory plant breeding (PPB) to produce var-
ieties that are well-adapted to local environmental stresses and farming communitiesneeds. In
contrast to centralized approaches in which breeding is carried out solely by trained plant
breeders on a research station or breeding facility, the decentralized approach of on-farm selec-
tion in PPB incorporates farmersknowledge, practices, and environmental contexts in the
selection process (Frossard, 2002; Soleri and Cleveland, 2002). Although PPB methods were
developed in the context of subsistence agriculture and marginal
environments (Ceccarelli et al., 1992; Ceccarelli, 1994), plant
breeding for organic agriculture can also benefit from on-farm
variety testing and selection (Baenziger et al., 2011) and farmer-
research collaboration (Chable et al., 2008). Organic farms tend
to experience higher farm-to-farm variation due to the lack of
synthetic inputs and emphasis on biological management, com-
pared with more environmental uniformity across conventional
farms (Dawson et al., 2011). Additionally, the smaller scale and
diversified nature of many organic vegetable farms fosters the
development of individualized cropping practices to adapted to
local environments and markets (Drinkwater et al., 1995;
Przystalski et al., 2008; Wolfe et al., 2008; Renaud et al., 2014).
On-farm evaluation allows for more accurate representation of
the broad range of environmental and cropping systems charac-
teristic of organic farms. In addition, some results have shown
that variety performance is less correlated between farms and
research stations in organic agriculture than in conventional agri-
culture (Singh et al., 2011). Whether a further reflection of the
higher degree of organic farm-to-farm variation mentioned
above, or a disconnect between station-managerspractices and
the practices of experienced organic farmers, this discrepancy
highlights the value of evaluation conducted on working organic
farms. Finally, on-farm evaluation facilitates farmer involvement
and leadership in determining the most valuable crop traits to
measure and characterizing variety performance, a critical aspect
to ensure that breeding and evaluation methods align with farm-
ers diverse organic seed and variety development priorities (Lyon
et al., 2015). For all of these reasons, organic agriculture stands to
benefit from on-farm crop evaluation, whether at the beginning of
a crop breeding project or after variety development to test fin-
ished varieties and determine the best varieties for specific regions
or market sectors.
Researchers have elaborated different strategies for on-farm
variety testing, many of which have focused on grain or pulse
crops (Snapp, 2002; Bellon et al., 2003; Ceccarelli et al., 2003;
Murphy et al., 2005; Dawson et al., 2011). On-farm evaluation
of grains and pulses can be considered more straightforward com-
pared to vegetable crops, as the harvested crop is more easily
transported or stored, allowing wider windows to accomplish
evaluation activities. This flexibility can be critical to allow farm-
ers to engage in these activities amidst their busy farming sche-
dules. Vegetable crops often have narrower temporal windows
in which evaluation must take place, and the timing of these
with farmersseasonal workloads, particularly on diversified
farms, can hinder farmersability to participate. Further, quanti-
tative yield measures such as weight are more straightforward in
grain and pulse crops which yield a single harvest, unlike many
vegetable crops with indeterminate growth patterns and multiple
harvests. Approaches to on-farm variety testing must consider the
feasibility of their implementation for farmers and these unique
characteristics of vegetable crops.
The response of varieties to environmental variation can be
characterized in terms of stability and adaptation, both of which
are commonly evaluated in variety trials (Abidin et al., 2005).
Stability can be assessed by linear regression of variety perform-
ance on an environmental index (EI). The EI of a given environ-
ment is defined as the mean performance of all varieties at that
environment. The resulting regression coefficient, b
, provides a
numerical description of stability. Finlay and Wilkinson (1963)
define a stable variety as one with b
approaching zeroin
other words, with similar variety performance across low- and
high-yielding environments. Alternatively, Eberhart and Russell
(1966) define a stable variety as one with b
= 1.0. A stable variety
in this definition has a relative performance that tracks with the
trial mean across lower-yielding and higher yielding environ-
ments, and that performance can be either consistently above or
below the trial mean in all sites.
Building on these two separate definitions, Becker and Leon
(1988) describe two concepts of stability: static/biological stability,
characterized by similar performance across all environments (the
FinlayWilkinson definition); and dynamic/agronomic stability,
characterized by increasing yields in more optimal environments
(the EberhartRussell definition). Selecting for dynamic stability
by regressing yield on EI, in combination with testing in a wide
array of marginal environments, contributed to the gains in
yield potential demonstrated in maize breeding in the 1960s
1990s (Bradley et al., 1988; Tollenaar and Lee, 2002).
Subsequent authors have proposed stability parameters to incorp-
orate multiple concepts of stability across genotypes (G), as well as
indicators of average yield and genotype by environment (GE)
interactions (Shukla, 1972; Francis and Kannenberg, 1978; Lin
et al., 1986; Yan and Kang, 2003). A comparison of the Finlay
Wilkinson regression coefficient, Shuklas stability variance,
Francis and Kannenbergs covariance and Yan and Kangs
Genotype (G) and Genotype × Environment (GE) Interaction, or
GGE biplot found the results generally to be in agreement
(Murphy et al., 2009). Additive main effects and multiplicative
interaction (AMMI) has also been widely used to analyze variety
trials. AMMI and GGE biplot are both based on singular value
decomposition, but while AMMI treats the effects of G and GE sep-
arately, GGE biplot combines them (Yan, 2014, p. 102). The relative
merits of AMMI vs GGE biplot for analyzing crop trials have been
debated: Gauch (2006) maintains that AMMI is more reliable across
datasets and that G and GE should be considered separately
because of their different implications for agricultural objectives
while Yan et al.(2007) argue that G and GE must be considered
together because specific adaptation is based on both.
While these methods of stability analysis offer strong potential,
their use in PPB and on-farm evaluations can be challenging due
to unbalanced data, high inherent variability and challenges to
implementing trial designs (Raman et al., 2011). In particular,
analysis of variance (ANOVA), AMMI and GGE biplot require
multiple replicates in each environment in order to provide
enough degrees of freedom to analyze the GE interaction terms.
By contrast, on-farm variety trials such as those in the Northern
Organic Vegetable Improvement Collaborative (NOVIC) project
frequently consist of single replicates on each farm, because mul-
tiple replicates can be prohibitively time consuming and expensive
to plant and evaluate in farmer-managed trials. For these situa-
tions, Hildebrand and Russell (1996) proposed adaptability ana-
lysis that employs the FinlayWilkinson regression but shifts the
interpretation of the resulting figures to identify specific adaptation
to high- or low-yielding environments rather than selecting for
general adaptation as it was used by Finlay and Wilkinson.
Decisions about which kind of adaptation is preferable, then rest
on farmerstolerance of risk and an evaluation of the farm envir-
onment (Fig. 1). Hildebrand and Russell contend that the wide-
spread use of stability parameters to identify varieties with broad
adaptation (agronomic stability), to the exclusion of varieties
better-adapted to poor environments, is motivated by an assump-
tion that farmers will use conventional agronomic practices that
suppress variability. For many of the worlds resource-limited
farmers, though, risk avoidance and average yields in poor
2 Alexandra Lyon et al.
environments (and seasons) are far more important than above-
average yields in all environments(Hildebrand and Russell,
1996, p. 23). This framing shows the potential of adaptability
analysis for the types of agroecosystems that also benefit from
PPB: those that tend to experience high farm-to-farm and
season-to-season variability, from organic farming to marginal
environments (Ceccarelli et al., 1994; Haussmann et al., 2012).
Snapp (2002) recommends Hildebrand and Russells approach to
stability analysis as a method to evaluate variety performance
across many on-farm environments in motherbaby trials, which
pair larger mothertrials at research stations with smaller baby
trials. The motherbaby trial design has been influential in partici-
patory variety trialing and selection, particularly in grain crops,
and has been implemented in variety trials for organic agriculture
including in our research. Yet while Snapp used adaptability ana-
lysis to assess adaptation to environment in motherbaby trials of
grain crops, this method had not been applied to vegetable crops,
which present unique challenges for the reasons discussed earlier.
In order to contribute to the further development of on-farm
trialing and the use of stability analysis for vegetable crops, we
assessed adaptability analysis as a method of evaluating variety
trials through the experience of the NOVIC. NOVIC is a multi-
institution participatory variety trialing project, involving
researchers and organic farmers in Washington, Oregon,
Wisconsin and New York. Beginning in 2010, variety trials were
conducted under certified organic conditions at research stations
and on working organic farms, in coordination with organic
breeding programs of partnering land grant universities and a
non-profit organization. The NOVIC trials used an adapted ver-
sion of the motherbaby trial discussed earlier (Snapp, 2002).
This design included the incorporation of on-farm evaluations
consisting of a single replicate on each farm and replicated
research station trials, allowing for a robust experimental design
while minimizing time requirements for farmers. In order to
address the question of the appropriateness of adaptability ana-
lysis as a tool to inform both plant breeders and farmers as to
the utility of a specific variety to be included in organic manage-
ment, we used selected data from the NOVIC trials in order to
achieve the following objectives:
(1) Assess differential responses of varieties included the NOVIC
trials grown in high- and low-yielding environments.
(2) Assess the feasibility of simple graphic methods to analyze
adaptation and performance in a participatory vegetable var-
iety trialing project.
(3) Inform recommendations on participatory trialing methods
that would be valuable to both farmers and plant breeders.
Thus, this paper describes a novel way in which both farmers
and plant breeders can assess the organic management suitability
of both new and existing vegetable germplasm, for which more
widely applied methods used for grain and cereal crops may
not be most appropriate or useful.
Experimental design and data collection
From 2010 to 2013, the NOVIC variety trials were conducted
across four production seasons on research stations in Ithaca,
NY; Madison, WI; Corvallis, OR and Sequim, WA; as well as at
five farm sites in New York, 10 in Wisconsin (including one in
Minnesota), seven in Oregon and 14 in Washington. Research sta-
tion trials consisted of three replicates of all varieties in rando-
mized complete block design, while on-farm trials consisted of
a single replicate of each variety. Although farmers received a
small stipend for participation, they were essentially volunteering
their time and land to participate in the variety trials while also
managing the rest of the production activities on their farms.
Planting and maintaining multiple replicates of the trial would
have been prohibitively time consuming for most of the famers
involved; single replicate on-farm trials allowed farmers to partici-
pate and provide feedback and data on trial varieties. Of the 36
total participating farms, some grew trials every year of the pro-
ject, while others participated for only some growing seasons.
The five core crops, which were grown on all research stations,
included broccoli (Brassica oleracea), sugar snap and snow peas
(Pisum sativum), carrots (Daucus carota subsp. sativus), sweet
corn (Zea mays) and butternut squash (Cucurbita moschata).
Farmers chose which of these they would trial each season on
their farms, depending on their interests and capacity.
Research station trials were managed by staff from each partici-
pating institution. On-farm trials were co-managed by farmers
and research staff, with farmers responsible for planting and
growing the crop as well as conducting mid-season and harvest
evaluations as their time permitted. Researchers distributed trial
seed and seedlings to farmers and assisted with some mid-season
evaluations, the majority of harvest evaluations, and all storage
evaluations. Data were collected in each state by researchers at
the research stations and researcherfarmer teams at the farm
sites, with some variation by state. For qualitative scores, evalua-
tors were given an explanation of the trait and then asked to view
or taste every plot and establish the extremes of the scale (1 and
5). In general, the same individual researchers were present for all
evaluations within each of the four states, but different farmers
were part of the evaluation team on each farm site. Data from
the four participating regions were compiled each year by
Fig. 1. Idealized example of three varieties plotted against EI. Variety B, with slope=
1, is broadly adapted. A is specifically adapted to high-yielding environments, while C
is specifically adapted to low-yielding environments. Adapted from Hildebrand and
Russell (1996, p. 6).
Renewable Agriculture and Food Systems 3
Organic Seed Alliance (a Washington State-based nonprofit
organization) staff to create a multi-site, multi-year dataset.
Statistical analysis
Crops were selected for inclusion in the analysis upon review of
the NOVIC dataset to identify the crops and variables with the
most complete data. Causes of missing data included failure of
all varieties of a crop, destruction of individual varieties by ani-
mals, farmers harvesting the crop before data was collected,
farmer reluctance to commit space for trialing the same varieties
over multiple seasons or other similar issues common to partici-
patory variety trials. With each of the crops satisfying these cri-
teria (squash, broccoli and carrots), we selected two variables to
include in the analysis based on completeness of the data and
relevance of the trait for crop improvement. We selected two con-
tinuous variables for squash: marketable number of fruit per plant
and marketable fruit weight per plant. For broccoli, we used one
continuous variable (head diameter), and one ratio (uniformity of
maturation). For carrots, we used two variables measured as 15
scores: root smoothness and sweetness (Table 1).
For each trait of interest, we used R software (R Core Team,
2014) to identify varieties within each crop that provided data in
the majority of environments tested (Table 2). Because farm sites
and varieties changed from year to year, we were not able to con-
sider the influence of years and locations separately. Rather, we
treated each year-by-location combination as a distinct environ-
ment. In order to analyze research station sites (which had three
replications of each variety) and on-farm sties (which had only
one replication) in the same analysis, we used the average of the
three replications from the research station sites. Given that a trade-
off existed between including more varieties and evaluating them in
more environments, we chose the combination that balanced both
factors. To ensure this approach was valid, we performed the ana-
lysis with fewer or greater environments to assess impact on results
on the variables we had previously selected. Using this approach,
analysis of squash variables included seven varieties in 11 environ-
ments and analysis of broccoli variables included seven varieties in
23 environments and analysis of carrot variables was conducted
with six varieties in 22 environments. Notable changes in results
with different inclusion of sites were only observed with the two
carrot variables that we had previously identified. For these vari-
ables analysis was conducted with six varieties in 22 environments
as an illustration of the limitations of this type of analysis for quali-
tative variables, as we will discuss.
ANOVA was conducted on environment (location × year) means
using the PROC GLM procedure in SAS® 9.1.3 (SAS Institute, Inc.,
. All effects were treated as fixed. Because only one replication
of each variety was available for the on-farm trials, insufficient
degrees of freedom prevented the inclusion of G × E interactions
when conducting ANOVA for the research station and on-farm
sites combined. For each response variable, we used the complete
dataset (research station and on-farm sites) to calculate environment
means (performance of all varieties within an environment), grand
variety means and least significant differences (LSD), treating each
environment as a replication and using the different environments
for each variety to calculate within-group error. Tukeysrangetest
was used to control for the effect of multiple comparisons. We
then created a linear regression of variety performance on EI,
which was calculated as the mean value of the trait of interest for
all varieties under consideration in that environment. The x-axis
in the resulting scatterplots thus represents a continuum from poor-
quality to high-quality environments as determined by observed
mean variety performance in that environment. We also calculated
the regression coefficient (β) and coefficient of determination (R
for each variety, and tested whether regression coefficients were sig-
nificantly different from a slope of one.
In order to fully utilize the replicated data collected at the
research station sites, we conducted a separate an AMMI analysis
of these data using the agricolae package for R (de Mendiburu,
2017). We produced ANOVA models including effects of G, E,
GE and within environment replicates, and biplots of the first
two principal components of the GE interactions as an indication
of stability.
Analysis of variance
The ANOVA model that includes both research station and
on-farm sites was significant for both squash yield variables (mar-
ketable fruit per plant and marketable weight per plant). For
Table 1. Variables evaluated through adaptability analysis of squash, broccoli and carrot varieties in participatory organic trials in NY, WI, OR and WA, 20102013
Crop Variable Measurement Varieties Environments Description
Squash Marketable fruit
per plant
Number of
7 11 Average number of squash per plant that were judged
marketable at harvest
Squash Marketable weight
per plant
kg 7 11 Average weight of squash per plant judged marketable at
Broccoli Head diameter cm 7 23 Width of widest part of broccoli head at harvest (average of
five heads per rep)
Broccoli Uniformity of
percent 7 20 Percent of total heads judged prime (vs over- or under
mature) on the day of harvest
Carrots Root smoothness 15 6 22 Score of root smoothness, which can be influenced by stress
and disease, where 1 = least smooth and 5 = smoothest
Carrots Sweetness 15 6 22 Score of sweetness determined by field tasting at harvest,
where 1 = least sweet and 5 = sweetest
An environment is one site in 1 year; the number of varieties and environments evaluated varies due to data availability from participating farms.
1. SAS and all other SAS Institute Inc. product or service names are registered trade-
marks or trademarks of SAS Institute Inc. in the USA and other countries. ®indicates
USA registration.
4 Alexandra Lyon et al.
squash marketable fruit per plant, environment was highly signifi-
cant (P< 0.0001) but variety was not (P= 0.099). For squash mar-
ketable weight per plant, both effects were significant, with P
0.0001. In other words, while environmental variation seems to
have had equal effect on both fruit number and fruit size, genetic
variation was important for fruit size but hardly at all for fruit
number. In broccoli, the ANOVA model as well as environment
and variety effects were significant (P< 0.0001) for both response
variables. Variety was a significant source of variation for carrot
sweetness (P= 0.035) but environment was not (P= 0.134). The
model was not significant for carrot root smoothness (P=
0.356) (Table 3).
Variety means
Though the ANOVA results provided strong evidence of a variety
effect on some traits (e.g., marketable weight per plant P= 0.001),
and marginal evidence for others (e.g., marketable fruit per plant
P= 0.099), when making pairwise comparisons of variety per-
formance using Tukeys LSD with a family-wise α= 0.05, no pair-
wise differences were significant. Though these differences did not
reach statistical significance, Metro,and Tianaperformed better
than other varieties in terms of average number of marketable
fruit per plant while Walthamwas a top performer for fruit
weight but not fruit number (Table 4). Tianaperformed better
than other varieties for both response variables. In broccoli,
Belstar,’‘Gypsyand Arcadiahad significantly larger average
head diameter than the three open-pollinated entries, OSU
Composite,’‘Solsticeand Common Ground Population.
Gypsydisplayed more uniformity in maturation (percent
prime on harvest day) than Common Ground Populationand
Solstice.Differences between carrot varieties also did not reach
statistical significance, Yayaand Bolerowere tied for the top rat-
ing for sweetness and Yayarated highest for root smoothness.
The carrot traits were evaluated using one to five scores, resulting
in data that were strongly clustered between two and four, pos-
sibly accounting for the lack of significant differences between
varieties and environments for these traits.
Environment means
In order to create the EI, which comprises the y-axis for regres-
sion in adaptability analysis, environments are arranged in
order of their respective mean variety performance (Fig. 2).
Error bars in this figure represent standard error, though it should
be noted again that quantitative variables (squash fruit number
and weight, broccoli head size) were measured as a sum per
plot rather than subsampling within plot, and qualitative variables
(broccoli uniformity, carrot sweetness and flavor) were evaluated
as a single rank per plot. Standard error here thus represents error
between plots only, rather than within plots. This figure demon-
strates that the rank of environments differed by crop and
response variable. For example, OR1-11 was a high-quality envir-
onment for squash in terms of marketable fruit number, but not
for marketable fruit weight, while WI1-11 was a high-quality site
for fruit weight but not fruit number. Except for two low-
performing New York sites (NY4-12 and NY3-12), broccoli
uniformity was strongly grouped between 0.25 and 0.75, charac-
teristic of variables measured as a proportion. For both carrot
variables there was a high amount of variation (standard error)
within environments and little separation between environments,
particularly for root smoothness.
Adaptability analysis
For squash marketable fruit per plant, linear regression against EI
suggested that Metrowas specifically adapted to high-yielding
environments, with a regression coefficient greater than and signifi-
cantly different from one (β= 1.4, P= 0.034) and relatively little
deviation from the regression (R
=0.9) (Fig. 3). Tiana,which
had similar overall average fruit numbers, was more broadly
adapted with somewhat more variable performance (β=0.99, P
= 0.61). Comparing the two varieties with average overall
performance, JWS 6823was very broadly adapted (β=1.0, P=
0.87, R
= 0.70) while Walthamdisplayed more specific adaptation
toward low-yielding environments (β=0.66, P=0.071, R
Bugle(β=1.2, P=0.27, R
= 0.82) and Pilgrim(β=0.98, P=
0.92, R
= 0.77) appeared to be more responsive to improvements
in environment than Early(β= 0.7, P= 0.0015, R
= 0.92). A cav-
eat is that the Oregon research station in 2011 (OR1-11) had very
high numbers of fruit in all varieties, so with only 11 environments
this site may have an exaggerated influence on the results.
Linear regression revealed different patterns of adaptation for
the two squash yield metrics (Fig. 4). Where Tianashowed
Table 2. Squash, broccoli and carrot varieties included in adaptability analysis
of participatory organic trials in NY, WI, OR and WA, 20102013
Crop Variety Type
Date of
Squash Early F1 1978
Squash Pilgrim F1 1999
Squash Bugle OP 1999
Squash Waltham OP 1970
Squash JWS 6823 F1 2010
Squash Tiana F1 2008
Squash Metro F1 2010
Broccoli Common ground
OP breeding
Broccoli Solstice OP 2011
Broccoli Oregon State University
(OSU) population
OP breeding
Broccoli Windsor F1 1998
Broccoli Arcadia F1 1985
Broccoli Belstar F1 2001
Broccoli Gypsy F1 2005
Carrot Rumba OP 2001
Carrot Bolero F1 1989
Carrot Spring market OP 1996
Carrot Scarlet nantes OP 1885
Carrot Yaya F1 2006
Carrot Nelson F1 2006
The two broccoli entries described as OP Breeding Stockwere discrete populations in the
late stages of a participatory breeding project led by Oregon State University. Dates of
release are based on American Society for Horticultural Science vegetable variety lists or
PVP documents but those that are preceded by are based on when they first appear in
seed catalogs on the internet. Introduction into the USA may happen later than when first
introduced into Europe.
Renewable Agriculture and Food Systems 5
broad adaptation for fruit number, in terms of weight it was more
specifically adapted to high-yielding sites (β=1.6, P= 0.088, R
0.74). Walthamdisplayed the opposite, with broad adaptation in
terms of marketable weight (β= 1.1, P=0.75, R
= 0.7). Among
the lower-yielding varieties, Pilgrim(β= 0.96, P=0.83, R
and Bugle(β=0.76, P= 0.094, R
= 0.79) appeared to be more
responsive to improved environments while Early(β=0.67, P=
0.0015, R
=0.91) and JWS 6823(β= 0.51, P= 0.039, R
appeared to be less responsive. For both yield metrics, the higher-
performing varieties displayed greater variation in terms of per-
formance across environments than lower-yielding varieties.
For broccoli head size, Gypsy(β=1, P= 0.93, R
= 0.51) and
Belstar(β= 0.85, P= 0.53, R
= 0.53) were broadly adapted, with
regression coefficients equal to or very close to one, though
Gypsyhad head sizes more consistently above site averages
(Fig. 5). Arcadiaand Windsoralso displayed broad adaptation.
Table 3. ANOVA of six variables in organic trials of squash, broccoli and carrot varieties in NY, WI, OR and WA, 20102013
Source df Sum of squares Mean square Fvalue Pr > F
Squash marketable number
Model 16 117.654 7.353 10.790 <0.0001
Environment 10 109.958 10.996 16.130 <0.0001
Variety 6 7.696 1.283 1.880 0.099
Residuals 60 40.890 0.681
Corrected total 76 158.544
Squash marketable weight
Model 16 103.628 6.477 9.200 <0.0001
Environment 10 83.586 8.359 11.880 <0.0001
Variety 6 20.041 3.340 4.750 0.001
Residuals 60 42.220 0.704
Corrected total 76 145.847
Broccoli head diameter
Model 28 783.843 27.994 7.770 <0.0001
Environment 22 425.996 19.363 5.380 <0.0001
Variety 6 357.847 59.641 16.560 <0.0001
Residuals 132 475.377 3.601
Corrected total 160 1259.220
Broccoli uniformity of maturation
Model 25 4.662 0.186 5.560 <0.0001
Environment 19 3.022 0.159 4.750 <0.0001
Variety 6 1.640 0.273 8.150 <0.0001
Residuals 114 3.820 0.034
Corrected total 139 8.482
Carrot sweetness
Model 21 24.031 1.144 1.720 0.044
Environment 16 15.586 0.974 1.470 0.134
Variety 5 8.445 1.689 2.540 0.035
Residuals 80 53.167 0.665
Corrected total 101 77.198
Carrot root smoothness
Model 21 28.367 1.351 1.110 0.356
Environment 16 11.540 0.721 0.590 0.881
Variety 5 16.827 3.365 2.760 0.024
Residuals 80 97.418 1.218
Corrected total 101 125.785
6 Alexandra Lyon et al.
The three open-pollinated entries, though genetically related, dis-
played markedly different patterns of adaptation. The variety
Solsticeshowed specific adaptation to higher-yielding environ-
ments (β= 1.4, P= 0.013, R
= 0.79), while Common Ground
Populationshowed broad adaptation across environments (β=
1, P= 0.86, R
= 0.86) and OSU Compositeseemed to display
better adaptation to more challenging environments (β=0.5, P
= 0.028, R
= 0.21).
Regression on EI for broccoli uniformity of maturity was less
informative than for the preceding variables. The variation
among varieties was small enough that we may not have had
the statistical power to detect differences between varieties with
this sample size. A contributing factor was that environmental
means were strongly grouped between 0.25 and 0.75. Measuring
variables as proportions may have obscured the variation between
environments. We have included regression analysis for broccoli
uniformity of maturity (Fig. 6) as a demonstration of what AA
looks like under these circumstances. Similarly, the use of
one-to-five rankings for both carrot variables resulted in data
strongly clustered around three and therefore not capturing a
wide range of average performance. Inclusion of more or fewer
environments in the linear regression for the carrot variables
demonstrated that the results of regression on EI were highly
influenced by which sites were included. Because of this and
because ANOVA showed environment not to be a significant
source of variation for either carrot variable, regression analysis
was not conducted.
AMMI analyses of replicated hub sites
Biplots of the first two principal components (with each variety
and environment plotted) for all traits, as well as stability statistics
and ANOVA have been included as Supplementary material.
When considering research station sites only (because sufficient
replication at those sites provided enough degrees of freedom to
analyze GE interactions), ANOVA models showed a significant
treatment (i.e., genotype and environment) effect for all traits.
In addition, the GE interaction was highly significant for squash
marketable weight (P= 0.0001), squash marketable number (P
= 0.0001), broccoli average head size (P< 0.0001) and carrot
root smoothness (P= 0.0001), and marginally significant for broc-
coli maturation uniformity (P= 0.0554). The GE interaction was
non-significant for carrot sweetness (P= 0.4317).
Because the number of GE interaction terms is the product of
the number of varieties and environments, it can be challenging to
interpret this large number of terms. AMMI reduces these interac-
tions to their principal components to simplify interpretation. If
the first two principal components capture >50% of the variability
in the GE interaction terms, an AMMI biplot may be interpreted
such that varieties plotted near the origin are considered to be
more stable than those plotted further from the origin. Varieties
close to the origin in AMMI biplots should therefore be expected
to correspond to regression slopes close to 1.0 in the adaptability
analysis. However, this was not always borne out in our analysis.
For example, the AMMI biplot for broccoli average head size
(Fig. 7) showed that Windsor,’‘Arcadiaand Gypsywere plotted
far from the origin, implying lower stability. In the adaptability
analysis (Fig. 5) however, none of these varieties had a slope sig-
nificantly different from 0 (P-values = 0.68, 0.45 and 0.93, respect-
ively). Squash marketable weight per plant had some agreement
between analyses. The biplot (Fig. 8) shows Tianaand JWS
6823 Butternutfar from the origin, and both of these varieties
had evidence of slopes different than 1.0 (P= 0.088 and 0.039,
respectively). The AMMI analyses did provide greater detail
regarding the relative environmental conditions of the research
sites. The example of broccoli head size demonstrates the strong
effect of temporal variation in this data, for example WI1-11
and WI1-12 were negatively correlated (in opposite quadrants)
although they are the same location in two consecutive years.
Our results show the responses of varieties to environments in the
NOVIC trials and provide insights into the advantages and limita-
tions of adaptability analysis in participatory variety trials. When
considering quantitative data and continuous variables, as we did
for yield characteristics in squash and broccoli, adaptability ana-
lysis proved to be a useful method to produce graphical
Table 4. Variety means by crop and trait for squash, broccoli and carrot in NY,
WI, OR and WA, 20102013
Squash marketable weight per
Squash marketable fruits per
Tiana 3.183 A Metro 2.95 A
Waltham 3.115 A Tiana 2.92 A
Metro 2.801 A JWS 6823 2.72 A
Pilgrim 2.374 A Waltham 2.55 A
JWS 6823 2.144 A Bugle 2.53 A
Early 2.044 A Pilgrim 2.21 A
Bugle 1.768 A Early 2.04 A
LSD (0.05) 0.846 LSD (0.05) 0.83
Broccoli head size (cm) Broccoli percent prime on harvest
Gypsy 12.664 A Gypsy 0.659 A
Belstar 12.504 A Windsor 0.582 AB
Arcadia 11.891 A Arcadia 0.529 AB
Windsor 11.02 AB Belstar 0.496 AB
OSU Com 9.524 B Common 0.384 B
Solstice 9.079 B Solstice 0.37 B
Common 8.97 B OSU Com 0.355 B
LSD (0.05) 1.323 LSD (0.05) 0.137
Carrot sweetness Carrot smoothness
Yaya 3.59 A Yaya 3.475 A
Bolero 3.59 A Scarlet
3.292 A
Nelson 3.26 A Bolero 3.289 A
3.17 A Nelson 3.013 A
Rumba 3.16 A Rumba 2.84 A
2.75 A Spring
2.243 A
LSD (0.05) 0.719 LSD (0.05) 0.895
Fishers least significant distance is displayed for each trait, and letters represent
significance groupings calculated using Tukeys range test was used to control for the effect
of multiple comparisons.
Renewable Agriculture and Food Systems 7
representations of adaptation patterns. We agree with Hildebrand
and Russell (1996) that graphs of performance regressed on EI
have the potential to allow farmers to visualize variety adaptation,
which could help them to make decisions based on the type of
adaptation they prioritize. Based on our experience of sharing
results with farmer audiences, we view these graphics as a useful
tool to help farmers judge whether a variety is more broadly or
narrowly adapted. Further evaluation of farmer experience in
the NOVIC project is planned, but as of this writing we have
not conducted a more formal evaluation to determine how well
farmers are able to understand and implement findings from
this method into their decisions. Further testing through direct
Fig. 2. Variety performance means by location 6 response variables in broccoli, carrot and squash. Environments represent year-by-location combinations. The last
two digits of each site code represent the trial year; WI1, NY1, OR1 and WA1 are all research station sites (average of three replications) and other codes are on-farm
sites (one replication). Broccoli head size, squash marketable weight and squash marketable fruit represent plot averages. Bars represent standard error.
8 Alexandra Lyon et al.
farmer feedback would be useful in assessing how this type of
analysis could best be applied to PPB.
Adaptability analysis proved less informative when working
with ratio variables and qualitative rankings because these types
of variables tended to be highly clustered on the x-axis, leading
to a scattershot pattern. This is important because such qualitative
rankings are commonly used to evaluate traits related to vegetable
quality, in our case flavor, appearance and harvest (maturation)
uniformity. Participatory research would benefit from improved
methods to evaluate these characteristics, while considering the
logistical challenges associated with collecting data from multiple
field sites at peak maturity for fresh vegetable crops. This is
Fig. 3. Linear regression of mean marketable fruit numbers of seven squash varieties on an EI composed of the mean of all varieties in each environment, from
participatory organic vegetable trials in NY, WI, OR and WA, 20102013. Varieties are shown from top left in order of mean performance in all sites. Dashed gray lines
represent a reference slope of 1; solid black lines represent the regression line a given variety on EI. Regression lines with slope >1 (e.g., Metro) show specific
adaptation to higher-yielding environments, while those with slope <1 (e.g., Waltham) show specific adaptation to lower-yielding environments. The null hypoth-
esis for the regression equation is that the regression coefficient (slope) is not significantly different from one.
Renewable Agriculture and Food Systems 9
particularly important in breeding for the organic market, where
culinary characteristics are of great importance.
Adaptation of NOVIC varieties
The contrast between Tiana,’‘Metroand Walthamin terms of
marketable squash fruit per plant illustrates the utility of
regression on EI for demonstrating differences in adaptation.
These varieties show the importance of considering multiple vari-
ables to characterize adaptation, as results differed depending on
whether yield was measured as marketable weight or marketable
fruit number. The desirability of these two yield metrics can
depend on marketing practices, as farmers growing for wholesale
and processing may benefit from increased net weight per plant,
Fig. 4. Linear regression of mean marketable weight of seven squash varieties on an index of 11 environments, from participatory organic vegetable trials in NY, WI,
OR and WA, 20102013. Varieties are shown from top left in order of mean performance in all sites. Dashed gray lines represent a reference slope of 1; solid black
lines represent the regression line a given variety on EI. The null hypothesis for the regression equation is that the regression coefficient (slope) is not significantly
different from one.
10 Alexandra Lyon et al.
while CSA growers often prioritize having more individual fruit to
put in customer boxes and tend to prefer medium-to-small sized
winter squash. However, a tradeoff has been observed between the
size of individual squash fruit and the number of fruit per plant
(Pessarakli, 2016, p. 14). Additional variables would create an
even more complex picture of adaptation; for example, maturity
is not considered here but probably influenced yield across
For broccoli head diameter, the broad adaptation and high
average performance of Belstarand Gypsyconfirmed farmers
observations based on experience with these widely used hybrids.
While this trial was not designed to compare adaptation of hybrids
Fig. 5. Linear regression of mean head diameter of seven broccoli varieties on an index of 23 environments, from participatory organic trials in NY, WI, OR and WA,
20102013. Varieties are shown from top left in order of mean performance in all sites. Dashed gray lines represent a reference slope of 1; solid black lines represent
the regression line a given variety on EI. The top four varieties in terms of head size all showed broad adaptation (slope not significantly different from 1). The null
hypothesis for the regression equation is that the regression coefficient (slope) is not significantly different from one.
Renewable Agriculture and Food Systems 11
and OP varieties, it is worth noting that the only broccoli entries
that deviated from the pattern of wide adaptation were OP entries
derived from participatory breeding methods, with Solsticeand
OSU Compositeshowing specific adaptation to high- and low-
yielding environments, respectively. The OSU Compositepopula-
tion, from which Common Ground Populationand Solstice
diverged, was developed through a cyclical convergentdivergent
selection process on organic farms predominantly on the east
and west coasts (Myers et al., 2012b). It was exposed to more
diverse environments than the other two populations during the
selection process, and had some germplasm introgression from
other broccoli varieties that farmers were growing in their
Fig. 6. Linear regression of maturation uniformity of seven broccoli varieties on an index of 20 environments, from participatory organic trials in NY, WI, OR and WA,
20102013. Varieties are shown from top left in order of mean performance in all sites. Dashed gray lines represent a reference slope of 1; solid black lines represent
the regression line a given variety on EI. The null hypothesis for the regression equation is that the regression coefficient (slope) is not significantly different from
12 Alexandra Lyon et al.
individual environments. Solsticeand Common Ground
Populationdiverged from the OSU Compositeat different stages
in development. Solsticeis the result of participatory breeding
undertaken by Jonathan Spero of Lupine Knoll farm in south-
western Oregon in collaboration with Prof James Myers of
Oregon State University, using the OSU Compositebreeding
stock (Myers et al., 2012b) and became isolated in the early
2000s. The growing environment at Lupine Knoll farm generally
has milder winters, earlier springs and significantly hotter sum-
mers that confine growth of cole crops to the spring and fall, in
contrast to the Willamette Valley OR and Olympia WA locations
where summer temperatures are generally not limiting to cole crop
production. In contrast, Common Ground Populationdiverged in
2009, where it was subject to strong selection in a single nitrogen-
limited PNW maritime environment. Though we hesitate to draw
conclusions based on the low number of environments, our ana-
lysis raises the question of whether Solsticeis better adapted to
relatively heterogeneous and potentially stressful environments
due to the conditions of the farm and region in which it was
selected. In other words, the environmental conditions and subse-
quent effect on natural and farmer selection at Lupine Knoll Farm
where Solsticewas selected, would more closely resemble the con-
tinental climates of WI and NY than cooler, wetter sites in the
Willamette Valley OR and Olympia WA where Common
Ground Populationwas selected.
The results for the two carrot variables demonstrate the limits
of qualitative ranking (i.e., one-to-five scores) as an evaluation
method in multi-farm variety trials. Because evaluation teams
varied from state to state, with different farmers participating at
each farm site, evaluator effects are impossible to separate from
environment effects for the carrot variables. Multiple evaluators
with varying taste perceptions may have resulted in too much
variation in the scores given to each variety to observe significant
separation between varieties. However, the difference between
Spring Market,which is a New Zealand overwintering carrot,
and the other varieties which are all Nantes types, shows that eva-
luators agreed about the difference in tastes between these variety
Adaptability analysis in on-farm trials
The adaptability analysis approach of regression of variety yields
on EI can serve as a simple, visual method for interpreting how
varieties respond to higher- and lower-quality growing environ-
ments and identifying patterns of adaptation. This method can
be used even when varieties were not grown in the same place
in consecutive years, and when on-farm trials consist of single
plots, an advantage for participatory trials. Characterizing adapta-
tion as demonstrated by these analyses can provide a better basis
for choosing varieties that suit specific purposes, environmental
conditions and farming strategies. However, cases where perform-
ance is primarily measured as a proportion or a qualitative score
(such as 1 to 5 or 1 to 9), will likely violate the requirement of
adaptability analysis that EI represent a wide range of average per-
formance. It may therefore be of limited use when evaluating
traits that are categorized using rating scales rather than measured
Whether narrow adaptation to specific environments (as
demonstrated by varieties like Metro) is preferable to broad
adaptation across environments (as demonstrated by varieties
like Gypsy) rests in part on the type of environmental variation
that farms experience. Varieties with specific adaptation to high-
Fig. 7. AMMI2 biplot (PC1 vs PC2) for broccoli average head diameter (cm) with seven genotypes (G) and nine environments (E) consisting of research station in OR,
WA, WI and NY in 20102012. In environment names, letters denote states and the last two digits denote year.
Renewable Agriculture and Food Systems 13
yielding environments are appropriate when farmers place a high
priority on maximizing yield, can reasonably assure optimal
growing conditions for the crop in question (e.g., access to irriga-
tion and high levels of fertility), and can tolerate a higher risk of
crop loss in the case of unexpected environmental challenges. On
the other hand, varieties with adaptation to low-yielding environ-
ments are preferable for farmers with lower risk tolerance and
who prioritize having some yield in challenging circumstances
over maximizing yields in optimal circumstances (Ceccarelli,
1994). Environments may be low-yielding for many reasons.
While varieties bred for specific high yielding environments
may excel when they are free from stresses but varieties with,
for example, broad quantitative resistance will perform well in a
broad range of environments. Additionally, while spatial variation
in farm environmental conditions can be targeted by specific
breeding for eco-regions, regions with less predictable weather
(high temporal variation) also call for more broadly adapted
Relative to humus-based organic farms (as opposed to indus-
trial organic), conventional farms have much more homogenous
environments, with varieties experiencing less variation for factors
such as nutrient availability, weed competition and pest and dis-
ease epidemics. Organic farms, on the other hand, require var-
ieties where the genotype of the crop is able to buffer against
these factors. In addition, many organic vegetable farmers practice
diversified (i.e., multi-crop) production; in a survey of Wisconsin
organic vegetable producers, the average number of crops grown
was 23 (Lyon et al., 2015). This type of production inherently
means that some crops are likely to be cultivated in suboptimal
conditions for that particular crop, meaning that diversified
farms are likely to benefit from varieties that can tolerate lower-
quality environments. It can therefore be argued that, due to
the need for greater stability and environmental buffering, crop
varieties intended for organic systems require broad agronomic
adaption over space and time, even when developed for single
environments or regions.
Limitations of adaptability analysis
The usefulness of adaptability analysis was severely limited when
considering non-continuous variables, which included all the
non-yield traits. In our study, these traits included flavor, appear-
ance and maturation uniformity, but other traits named as prior-
ities by organic farmers, such as disease tolerance, insect tolerance
and season extension, present similar challenges for evaluation. In
the case of broccoli, the use of a proportion to measure matur-
ation uniformity (head at peak maturity/total heads) may have
led to data being artificially grouped around on 0.5, producing
an EI with an insufficient range of values for meaningful evalu-
ation of adaptation. In the case of carrot, both traits were evalu-
ated as qualitative one-to-five scores by different evaluators at
each site, creating an evaluator effect as well as an environment
effect which likely explains the high site-to-site variability.
Although scoring is a common method for evaluating entries
in plant breeding programs, researchers have long been aware
that evaluators tend to disproportionately assign values of two
through four, and avoid values of one or five, effectively limiting
in the range of the scale (Coe, 2002). An extra layer of variability
is added in a project with multiple evaluators, who may interpret
the scale differently. However, quantitative measurement of these
traits can be much more expensive and laborious. Establishing
specific descriptors for each level of the scale may help make
this data more reliable. Alternative methods of evaluating quality
traits should be explored, including systematic approaches to
Fig. 8. AMMI2 biplot (PC1 vs PC2) for squash average marketable weight (kg) with seven genotypes (G) and nine environments (E) consisting of research station in
OR, WA, WI and NY in 20102012. In environment names, letters denote states and the last two digits denote year.
14 Alexandra Lyon et al.
recording farmerswritten or verbal observations (Ashby, 1990).
Efficient but precise and reliable methods for evaluating quality
related traits would allow researchers to assess stability and adap-
tation with regard to a wider variety of traits, an advantage in
breeding and trialing varieties for the organic sector.
Comparing AMMI and adaptability analysis
Although AMMI is frequently used to determine both stability
and adaptation of varieties to ecoregions, in our case, a compel-
ling story did not emerge. For some varieties, AMMI and adapt-
ability analysis were in agreement about stability, while for others
they presented conflicting results. This could be due to the low
number of environments or varieties we were able to include
from this dataset. The differences in results when comparing
AMMI and AA for broccoli may be explained by AMMI having
roughly half as many environments as were considered in the
broccoli AA. However, for squash, the data considered were
almost identical (nine environments in AMMI vs 11 for AA),
yet results between the two analysis still did not align.
Despite this inconclusive result, it was enlightening to conduct
both AMMI and regression analysis with this dataset, which
represented a heterogeneous network of participatory variety
trials on both research stations and on-farm sites. While the
potential of AMMI to delineate mega-environments for plant
breeding as well as assess the stability of varieties has been
demonstrated, the clear limitation of AMMI in our case was the
need for replications within environments in order to estimate
GE interaction. For this reason, AMMI is better suited to compar-
ing environments where researchers have sufficient involvement
to carry out replicated trial designs.
Vegetable crops require intensive management throughout the
growing season and have narrow maturity windows during which
yield and quality must be evaluated. Including working farms as
trial environments is of interest and importance to plant breeding
for organic agriculture, yet on-farm trials are often planted and
maintained by farmer volunteers because more than one or two
researcher visits per season is cost prohibitive. Expecting farmers
to plant and maintain replicated vegetable trials simultaneously
with the rest of their farm production is not realistic. It is therefore
our challenge as researchers to implement trial designs and ana-
lyses that can incorporate on-farm data while also making use of
replicated research station trials to provide a more detailed picture.
Our experience with on-farm trials in the NOVIC project showed
the potential of participatory trialing networks to provide insight
into variety performance on working organic farms across wide
geographic areas. Innovative organic farmers are interested in
new crops and varieties, and university supported on-farm trials
have the potential to mitigate the risk involved in planting a
new and untested variety. However, working with the NOVIC
dataset demonstrated the difficulties of gathering on-farm data
for vegetable crops due to the unique issues of harvest time and
workflow characteristic of working organic farms. These chal-
lenges call for rethinking trial designs, particularly with respect
to data collection strategies, and statistical analyses for participa-
tory trialing and breeding, as well as the roles of farmers and
researchers in research collaborations. For vegetable variety trials
specifically, there are two possible directions. One would be a
more decentralized approach, with farmers receiving seed, trial
design suggestions and evaluation worksheets, and being largely
responsible for data collection. Evaluations in this scenario need
to be simple and rapid to execute. Researchers receive data from
farmers and be responsible for analysis and distribution of the
findings. The opposite direction is a more centralized approach,
with data collection happening largely on research stations and
with farmers engaged in alternative activities as opposed to
on-farm trials. After several years of motherdaughter trials,
this more centralized approach has been adopted for sweet corn
trials in the current iteration of NOVIC.
Recent projects also provide examples of creative combinations
of these approaches. For instance, centralized trials on research
stations can be paired with on-farm trials in which evaluation is
centered on farmers anecdotal comments. Though this would
preclude statistical analysis of on-farm trials, a significant depth
of information could be captured through such experiential eva-
luations. Along these lines, trailing networks such as the
Culinary Breeding Network (Oregon State University) and the
Seed to Kitchen Collaborative (University of Wisconsin-
Madison) are exploring novel approaches to sensory evaluation
involving a wide range of participants from farmers to chefs
(Beans, 2017; Healy, Emerson, and Dawson, 2017; Kissing
Kucek et al., 2017). Alternately, on-farm trials can be limited to
a few, highly involved farms, with feedback and guidance from
a wider circle of farmers who do not directly grow the trial entries.
The potential of key relationships between individual, highly
motivated farmers and trained plant breeders has already been
demonstrated in the release of several participatory bred varieties,
such as Who Gets Kissed?sweet corn (Shelton and Tracy, 2015),
Peaceworkpepper (Mazourek et al., 2009) and Dark Starzuc-
chini (Prairie Road Organic Seed, 2018). This model of relation-
ships may serve as a more appropriate framework for vegetable
trials, than attempting to gather data from many farms, as has
been achieved with participatory trialing of grain crops.
However, as the experience discussed earlier of selecting
Solsticebroccoli demonstrates, selection on an individual farm
may result in a variety specifically adapted to that particularly
environment. Better delineation of eco-regions should therefore
be of particular concern for breeding for organic agriculture.
Resilient and well-adapted varieties are widely recognized as
priorities for organic and agroecological production systems, in
vegetables as well as other crops. Developing effective ways to
identify and evaluate varieties with these characteristics, particu-
larly in terms of quality traits, will benefit both organic breeding
and farmersability to judge the performance of existing varieties.
Whatever the arrangement, approaches to variety trialing that
facilitate collaboration between trained researchers and organic
farmers will ensure that research for the organic sector truly
reflects the environmental and cultural conditions of organic
farms. Adaptability analysis may be a useful tool toward both
these ends, to be combined with others in improving the product-
ivity and sustainability of organic vegetable production.
Supplementary material. The supplementary material for this article can
be found at
Author ORCIDs. Alexandra Lyon 0000-0001-8486-7688
William Tracy 0000-0002-9855-302X
Michael Mazourek 0000-0002-2285-7692
James Myers 0000-0003-0976-144X
Acknowledgements. We are deeply grateful to the network of NOVIC col-
laborators, including farmers, researchers, field crew and organizational staff,
Renewable Agriculture and Food Systems 15
who made this research possible and continue to carry it forward. In particu-
lar, we are indebted to field managers who coordinated the trials during the
timespan covered here, including Michael Glos (Cornell), Anne Pfeiffer
(UW-Madison), Lane Selman (OSU) and Laurie McKenzie (Organic Seed
Alliance). We are indebted to Tessa Peters for her advice in improving our fig-
ures and analysis. In addition, three anonymous reviewers provided feedback
which greatly improved this manuscript.
Financial support. This paper is based upon research that is supported by
the National Institute of Food and Agriculture, US Department of
Agriculture, under award number 2009-51300-05585.
Abidin PE, van Eeuwijk F, Stam P, Struik PC, Malosetti M, Mwanga RO,
Odongo B, Hermann M and Carey EE (2005) Adaptation and stability
analysis of sweet potato varieties for low-input systems in Uganda. Plant
Breeding 124, 491497.
Ashby J (1990) Evaluating Technology with Farmers. Cali, CO: Centro
Internacional de Agricultura Tropical (CIAT).
Baenziger PS, Salah I, Little RS, Santra DK, Regassa T and Wang MY
(2011) Structuring an efficient organic wheat breeding program.
Sustainability 3, 11901205.
Beans C (2017) Science and culture: vegetable breeders turn to chefs for flavor
boost. The Proceedings of the National Academy of Sciences 114, 1050610508.
Becker HC and Leon J (1988) Stability analysis in plant breeding. Plant
Breeding 101,123.
Bellon MR, Berthaud JB, Smale M, Aguirre JA, Taba S, Aragón F, Díaz J
and Castro H (2003) Participatory landrace selection for on-farm conserva-
tion: an example from the Central Valleys of Oaxaca, Mexico. Genetic
Resources and Crop Evolution 50, 401416.
Bradley JP, Knittle KH and Troyer AF (1988) Statistical methods in seed corn
product selection. Journal of Production Agriculture 1, 34.
Burger H, Schloen M, Schmidt W and Geiger HH (2008) Quantitative gen-
etic studies on breeding maize for adaptation to organic farming. Euphytica
163, 501510.
Ceccarelli S (1994) Specific adaptation and breeding for marginal conditions.
Euphytica 77, 205219.
Ceccarelli S, Grando S and Hamblin J (1992) Relationship between barley
grain yield measured in low- and high-yielding environments. Euphytica
Ceccarelli S, Erskine W, Hamblin J and Grando S (1994) Genotype by envir-
onment interaction and international breeding programmes. Journal of
Experimental Agriculture 30, 177187.
Ceccarelli S, Grando S, Singh M, Michael M, Shikho A, Al Issa M, Al
Saleh A, Kaleonjy G, Al Ghanem SM, Al Hasan AL et al.(2003) A meth-
odological study on participatory barley breeding II. Response to selection.
Euphytica 133, 185200.
Chable V, Conseil M, Serpolay E and Lagadec F (2008) Organic varieties for
cauliflowers and cabbages in Brittany: from genetic resources to participa-
tory plant breeding. Euphytica 164, 521529.
Coe R (2002) Analyzing ranking and rating data from participatory on-farm
trials. In Bellon MR and Reeves J (eds), Quantitative Analysis of Data
from Participatory Methods in Plant Breeding. Mexico, DF: CIMMYT,
pp. 4465.
Dawson JC, Rivière P, Berthellot J-F, Mercier F, Kochko P, Galic N, Pin S,
Serpolay E, Thomas M, Giuliano S and Goldringer I (2011) Collaborative
plant breeding for organic agricultural systems in developed countries.
Sustainability 3, 12061223.
de Mendiburu F (2017) Statistical Procedures for Agricultural Research using
R. La Molina, Lima: Universidad Nacional Agraria.
Drinkwater LE, Letourneau DK, Workneh F, van Bruggen AHC and
Shennan C (1995) Fundamental differences between conventional and
Organic Tomato Agroecosystems in California. Ecological Applications 5,
Eberhart SA and Russell WA (1966) Stability parameters for comparing var-
ieties. Crop Science 6.
Entz MH, Kirk AP, Vaisman I, Fox SL, Fetch JM, Hobson D, Jensen HR
and Rabinowicz J (2015) Farmer participation in plant breeding for
Canadian organic crop production: implications for adaptation to climate
uncertainty. Procedia Environmental Sciences, Agriculture and Climate
ChangeAdapting Crops to Increased Uncertainty (AGRI 2015) 29,
Finlay KW and Wilkinson GN (1963) The analysis of adaptation in a plant-
breeding programme. Crop & Pasture Science 14, 742754.
Francis TR and Kannenberg LW (1978) Yield stability studies in short-season
maize. I. A descriptive method for grouping genotypes. Canadian Journal of
Plant Science 58, 10291034.
Frossard D (2002) How farmer-scientist cooperation is devalued and revalued:
a Philippine example. In Cleveland DA and Soleri D (eds), Farmers,
Scientists and Plant Breeding: Integrating Knowledge and Practice.
New York, NY: CABI Publishing, pp. 137160.
Gauch HG (2006) Statistical analysis of yield trials by AMMI and GGE. Crop
Science 46, 14881500.
Haussmann BIG, Fred Rattunde H, Weltzien-Rattunde E, Traoré PSC, vom
Brocke K and Parzies HK (2012) Breeding strategies for adaptation of pearl
millet and Sorghum to climate variability and change in West Africa.
Journal of Agronomy and Crop Science. n/an/a. doi:
Healy GK, Emerson BJ and Dawson JC (2017) Tomato variety trials for prod-
uctivity and quality in organic hoop house versus open field management.
Renewable Agriculture and Food Systems 32, 562572.
Hildebrand PG and Russell JT (1996) Adaptability Analysis: a Method for the
Design, Analysis, and Interpretation of on-Farm Research-Extension,1st
Edn. Ames: Iowa State University Press.
Kirk AP, Fox SL and Entz MH (2012) Comparison of organic and
conventional selection environments for spring wheat: comparison of
organic and conventional selection environments. Plant Breeding 131,
Kissing Kucek L, Dyck E, Russell J, Clark L, Hamelman J, Burns-Leader S,
Senders S, Jones J, Benscher D, Davis M, Roth G, Zwinger S, Sorrells ME
and Dawson JC (2017) Evaluation of wheat and emmer varieties for arti-
sanal baking, pasta making, and sensory quality. Journal of Cereal Science
Lammerts van Bueren ET and Myers JR (2012) Organic crop breeding: inte-
grating organic agricultural approaches and traditional and modern plant
breeding methods. In van Bueren ETL and Myers JR (eds), Organic Crop
Breeding. Wiley-Blackwell, pp. 113.
Lin C-S, Binns MR and Lefkovitch LP (1986) Stability analysis: where do we
stand? Crop Science 26, 894900.
Lyon AH, Silva E, Zystro J and Bell M (2015) Seed and plant breeding for
Wisconsins organic vegetable sector: understanding farmersneeds.
Agroecology and Sustainable Food Systems 39, 601624.
Mazourek M, Moriarty G, Glos M, Fink M, Kreitinger M, Henderson E,
Palmer G, Chickering A, Rumore DL, Kean D, Myers JR, Murphy JF,
Kramer C and Jahn M (2009) Peacework: a cucumber mosaic virus-
resistant early red bell pepper for organic systems. HortScience 44, 1464
Murphy KM, Lammer D, Lyon S, Carter B and Jones SS (2005) Breeding for
organic and Low-input farming systems: an evolutionaryparticipatory
breeding method for inbred cereal grains. Renewable Agriculture and
Food Systems 20,4855.
Murphy KM, Campbell KG, Lyon SR and Jones SS (2007) Evidence of
varietal adaptation to organic farming systems. Field Crops Research 102,
Murphy SE, Lee EA, Woodrow L, Seguin P, Kumar J, Rajcan I and
Ablett GR (2009) Genotype×environment interaction and stability for iso-
flavone content in soybean. Crop Science 49, 13131321.
Myers J, McKenzie L, Mazourek M, Tracy W, Shelton A and Navazio J
(2012a) Breeding peas, sweet corn, broccoli, winter squash and carrots as
part of the Northern Organic Vegetable Improvement Collaborative
(NOVIC). in: Strengthening Community Seed Systems. Presented at the
Strengthening Community Seed Systems. Proceedings of the 6th Organic
Seed Growers Conference, Port Townsend, Washington, USA, 1921
January 2012. Organic Seed Alliance, pp. 4445.
16 Alexandra Lyon et al.
Myers J, McKenzie L and Voorrips RE (2012b) Brassicas: breeding cole crops
for organic agriculture. In van Bueren ETL and Myers JR (eds), Organic
Crop Breeding. Wiley-Blackwell, pp. 251262.
Pessarakli M (ed.) (2016) Handbook of Cucurbits. Boca Raton: CRC Press.
Prairie Road Organic Seed (2018) Zucchini: Dark Star [www Document].
Prairie Road Organic Seed: Northern Grown. Available at https://www.
seed-1-packet-25-seeds (Accessed 11 August 18).
Przystalski M, Osman A, Thiemt EM, Rolland B, Ericson L, Østergård H,
Levy L, Wolfe M, Büchse A, Piepho H-P and Krajewski P (2008)
Comparing the performance of cereal varieties in organic and non-organic
cropping systems in different European countries. Euphytica 163, 417433.
R Core Team (2014) R: A Language and Environment for Statistical
Computing. Vienna, Austria: R Foundation for Statistical Computing.
Raman A, Ladha JK, Kumar V, Sharma S and Piepho HP (2011) Stability
analysis of farmer participatory trials for conservation agriculture using
mixed models. Field Crops Research 121, 450459.
Renaud ENC, Lammerts van Bueren ET, Paulo MJ, van Eeuwijk FA,
Juvik JA, Hutton MG and Myers JR (2014) Broccoli cultivar performance
under organic and conventional management systems and implications for
crop improvement. Crop Science 54, 1539.
SAS Institute, Inc. (2000) SAS 9.4. Cary, North Carolina: SAS Institute, Inc.
Shelton A and Tracy W (2015) Recurrent selection and participatory plant
breeding for improvement of Two organic open-pollinated sweet corn
(Zea mays L.) populations. Sustainability 7, 51395152.
Shukla GK (1972) Some statistical aspects of partitioning genotype environ-
mental components of variability. Heredity 29, 237245.
Singh S, Terán H, Lema M and Hayes R (2011) Selection for dry bean yield
on-station versus on-farm conventional and organic production systems.
Crop Science 51, 621.
Snapp S (2002) Quantifying farmer evaluation of technologies: the mother and
baby trial design. In Bellon MR and Reeves J (eds), Quantitative Analysis of
Data from Participatory Methods in Plant Breeding.Mexico:CIMMYT,pp.917.
Soleri D and Cleveland DA (2002) Understanding farmersknowledge as the
basis for collaboration with plant breeders: methodological development
and examples from ongoing research in Mexico, Syria, Cuba and Nepal.
Farmers Sci. Plant Breed. Integrating Knowl. Pract. 1660.
Tollenaar M and Lee EA (2002) Yield potential, yield stability and stress tol-
erance in maize. Field Crops Research 75, 161169; Preface.
Wolfe MS, Baresel JP, Desclaux D, Goldringer I, Hoad S, Kovacs G,
Löschenberger F, Miedaner T, Østergård H and Lammerts van
Bueren ET (2008) Developments in breeding cereals for organic agriculture.
Euphytica 163, 323346.
Yan W (2014) Crop Variety Trials: Data Management and Analysis. Somerset,
NJ, USA: John Wiley & Sons, Incorporated.
Yan W and Kang MS (2003) GGE biplot Analysis: A Graphical Tool for
Breeders, Geneticists, and Agronomists. Boca Raton, FL : CRC Press.
Yan W, Kang MS, Ma B, Woods S and Cornelius PL (2007) GGE biplot vs.
AMMI analysis of genotype-by-environment data. Crop Science 47,643653.
Renewable Agriculture and Food Systems 17
... However, farmers' participation in harnessing the technology is a concept that needs to be incorporated into its application. In optimizing the roles of farmers and researchers, participatory research that shares some similarities with participatory plant breeding (PPB) can be conducted in two ways: mother trials and baby trials (Snapp et al., 2002;Merga, 2017;Najeeb et al., 2018;Lyon et al. 2020;Tyack et al. 2020). Researchers plant a predefined variety in the mother trial, while farmers are responsible for the breeding in the baby trial. ...
... Researchers plant a predefined variety in the mother trial, while farmers are responsible for the breeding in the baby trial. Generally, a mother trial is a principal experiment to evaluate the best variety or cultivation technology, and a baby trial is a supplement to this evaluation (Lyon et al. 2020;Tyack et al. 2020). It means that combining both trials can increase the effectiveness of selecting the best-adapted variety or cultivation and that analyzing the trial results can determine the optimal technology and varieties to develop in an area. ...
... However, P1, P4, and P5 were the corn cultivation technology packages with consistently good outcomes. These results support the previous statement by Lyon et al. (2020) and Tyack et al. (2020) that the baby trial can substantially increase the effectiveness of selection in the mother trial evaluation. Therefore, P1, P4, and P5 are recommended for corn production, especially in Tarowang Village, Galesong Selatan District. ...
Full-text available
An unmanned aerial vehicle (UAV), widely known as a drone, proves very effective in assessing cropping or crop cultivation. Its practical use in evaluating corn cultivation technology systems is feasible when based on farmer participation. UAV can generate the Normalized Difference Vegetation Index (NDVI) algorithm that reflects the greenness of leaves, which is a parameter related to photosynthesis and plant productivity. Therefore, the purpose of this study was to evaluate whether the participation-based UAV-derived NDVI could be effectively used to assess corn cultivation technology and determine the appropriate technology to be used in the cultivation. The research was conducted in Tarowang Village in Galesong Selatan District, Takalar Regency, South Sulawesi, Indonesia, using two plots, namely, mother trial and baby trial. The mother trial applied a randomized block design in which eight packages of corn cultivation technology were randomly assigned, whereas the baby trial consisted of eight corn plots cultivated by farmers. In the latter, each farmer received one package of the cultivation technology. The study results indicated that NDVI and yield could effectively evaluate corn cropping. Three packages, i.e., P1, P4, and P5, are recommended for corn cultivation, especially in the village observed. Nevertheless, they are expected to be also applicable to other districts in South Sulawesi to promote improvement in corn production.
... The Genotype x Environment (GE) interaction was significant (p≤0.05) for the characteristics AUDPC, RRW, CRW, RCT and NCR, confirming changes in the behaviour of the evaluated genotypes in relation to the time of evaluation, i.e. sowing, either in the spring or summer. The presence of the GE interaction in evaluating carrot genotypes is reported in the literature LYON et al., 2020). ...
... A interação Genótipo x Ambientes (GA) foi significativa (p≤0,05) para os caracteres AACPD, MRR, MRC, RCT e NRC, confirmando haver mudanças no comportamento dos genótipos avaliados em relação à época de avaliação, ou seja, semeadura na primavera e semeadura no verão. A presença de interação GA na avaliação de genótipos de cenoura é relatada na literatura LYON et al., 2020). ...
Full-text available
The carrot is one of the most important and preferred vegetables in Brazil due to its versatility and nutritional significance, especially as a source of vitamins. The selection of carrot genotypes should consider several agronomically important characteristics to allow more suitable and balanced gains. The aim of the present study was to estimate genetic gains in the carrot using different selection indices. Thirty-six carrot genotypes were evaluated in two experiments, the first conducted between 16 November 2017 and 1 March 2018, and the second between 5 March and 18 June 2018 in, Brasilia. In both trials. the experimental design was of randomised blocks with three replications. The plots were sown directly onto the seedbed, which had a total working area of 1.8 m2. The incidence of leaf blight was evaluated 78 to 98 days after sowing, at intervals of seven days, giving a total of four evaluations. The harvest was carried out at 105 days when the characteristics of the yield components were measured. An analysis of variance was applied, together with five selection indices: direct selection, classic, desired gain, weight- and parameter-free, and genotype-ideotype. There were differences between genotypes for most of the characteristics under evaluation. The genotype-ideotype index afforded the greatest gains for the characteristics being evaluated, with an economic weight equal to the coefficients of genetic variation; however other indices were also efficient and offered balanced gains, with the exception of the desired-gain index.
... Hinkel and Bisaro [30] suggested that analysis of variance (ANOVA) is mostly applied in adaptation analysis studies. ANOVA constitutes statistical models and variance and mean estimation procedures for the variables that induce different variations [61]. ANOVA applies the sample variance through the equation 2 ...
Full-text available
The world-shaking communicable coronavirus disease (i.e., COVID-19) has become a pandemic threat to a healthy built environment. This study aimed to develop the COVID-19-adapted multi-functional corniche street design (Ca-MCSD) assessment model. Accordingly, this study identified variables coordinating the local environmental, physical, social, cultural, and political mediations of multi-functional corniche street design. Secondly, it measured the weight of every single variable through confirmatory analysis, normalization, and standardization techniques, and an expert-input study then developed the MCSD model and Ca-MCSD model. This study validated the models through a case study (i.e., Al Wakrah corniche street in Dubai, Qatar) and conducted ANOVA regression analysis and global sensitivity analysis (GSA). The Ca-MCSD model evaluates the design quality of a corniche street across five criteria—inclusiveness, desirable activities, safety, comfort, and pleasurability—and forty-two sub-criteria. The regression analysis determined that the MCSD model and Ca-MCSD model are linearly and positively correlated (Y = 0.811777X + 0.383401), where the Pearson regression coefficient (r) equaled 0.903729, r2 equaled 0.816727, and the p-value was 0.025 with 95% confidence intervals. The research found that, before the COVID-19 pandemic, microclimate comfort (avWSc.3.4 = 7.880), community gathering places (Sc.2.1), availability of foods (Sc.2.4), appropriate maintenance and physical condition (Sc.3.6), and attractiveness of space (Sc.5.8) (avW = 6.000) played critical roles in designing a multi-functional corniche street. However, after the onset of the COVID-19 pandemic, the key drivers changed to microclimate comfort (favWSc.3.4 = 12.632), appropriate maintenance and physical condition (favWSc.3.6 = 9.618), physical/visual connection or openness to adjacent spaces (favWSc.4.1 = 4.809), and over-securitization (favWSc.4.1 = 4.287).
... Despite the farmer-engaged nature of this project, each cultivar was selected on only one farm. In some cases, selection in organic growing conditions has yielded cultivars that show increased performance on organic farms, but in other cases, selection on a single organic farm has resulted in cultivars with limited adaptation (Lyon et al. 2018). In addition, correlation between variety performance on individual farms and that on research stations was lower in organic than conventional conditions (Singh et al. 2011). ...
Full-text available
Participatory plant breeding and rapid sensory evaluation are effective techniques for organic cultivar development. Table beet is an important crop for organic growers, and geosmin, a volatile compound which confers earthy aroma, has been suggested as the attribute around which hedonic liking of beet is organized. Open pollinated table beet populations with diverse pigmentation and low (LGC) or high (HGC) geosmin concentration served as starting materials for the first PPB effort in table beet. This project sought to develop consumer-accepted specialty beet cultivars for organic systems and to gauge consumer perception of and preference for geosmin concentration in non-laboratory conditions. LGC and HGC initial populations were significantly different in mean geosmin concentration but not mean TDS. LGC populations diverged significantly in geosmin concentration over two cycles of selection for hedonic liking, due to drift rather than selection. PPB yielded cultivars ‘Evansville Ember’, ‘Snowglobe’, ‘Blushing Not Bashful’, ‘Evansville Orbit’, and ‘Moving Target’. Cultivar novelty and market development were strengthened by chef input and association with a publicly funded seed system development group. Geosmin concentration was not the central determinant of hedonic liking or perceived earthy flavor in table beet. Earthiness was inconsistently associated with geosmin concentration and hedonic liking. Sweetness and bitterness were positively and negatively correlated with liking, respectively, although sweetness was not associated with variation in TDS. Cultivars with a broad range of geosmin concentration were well accepted by consumers, and manipulating expectation—via appearance—may be as powerful as manipulating flavor compounds in influencing liking of table beet.
... The cultivars that performed better in favorable environments did not repeat the same in unfavorable environments and vice versa. Variation in adaptability of butternut squash, broccoli, and carrot varieties was observed at different organic production environments-Oregon, Washington, Wisconsin, and New York (Lyon et al. 2020). Some varieties showed broad adaptation while others were adapted to either high yielding or low yielding environments. ...
The exponentially growing population has triggered the need for more food. Lot of work is being done in the field of crop improvement and variety testing to develop high yielding varieties. But the question is whether modifying crops at molecular or gene level is the only thing required to improve yield? Is there any factor that may affect the yield of these elite varieties? Should we consider that factor in crop improvement and variety testing programs? The answer is yes. Soil heterogeneity is an important factor that may affect the yield of these elite varieties. Unfortunately, it is the least explored area in any crop improvement or variety testing program. Soil heterogeneity results in over- or under-application of inputs such as fertilizers or irrigation that affects crop growth and development. Same genotypes planted at different sites on a field varying in amount of nutrients or organic matter tend to perform differently. All these increases the complexity of agricultural research and slow down the process of varietal selection by breeders. Soil heterogeneity may occur at site, space, or time level. In the field condition, soil is rarely homogeneous. Variability exists and hence, should be explored in crop improvement and variety testing programs. There are several ways to do soil sampling—random sampling, grid sampling, or sensor-based to study soil variability but they are either not accurate, time-consuming, require resources, or complex to work with. UAV has emerged as a low-cost tool that provides near real-time data and is capable of mapping soil heterogeneity for moisture, nutrients, etc. It can be a great tool to consider in exploring the effect of soil heterogeneity on crop varietal growth and development and making decisions accordingly.
... As a local variety, the Altino ecotype is characterized by good adaptation and rusticity and is prone to sustainable cultivation systems under low-input conditions [4], e.g., reduced the total use of agricultural chemicals, especially nitrogen (N) fertilization. In any case, N management must be calibrated based on the critical N curve and the N use efficiency to reduce N losses and environmental pollution [5]. ...
Full-text available
Local varieties represent a heritage for plant biodiversity and, thanks to their resilience, are characterized by a better adaptation and rusticity to environmental variables. This work reports the morphological and physiological responses of a local ecotype of sweet pepper grown in Southern Italy, i.e., Altino, to N and light supplied at optimal and sub-optimal levels. In 2017, two open-field experiments were set up, comparing increasing N rates (0, 100 and 200 kg N ha−1, 0_N, 100_N and 200_N, respectively) and different percentages of shading and/or manipulations of the transmitted solar radiation, obtained through photoselective nets (red net, RN; black net, BN; unshaded Control). The unfertilized plants reduced growth (by 32% on average, at 83 days after transplanting, DAT), especially in terms of number of leaves, leaf area and effectiveness (chlorophyll content) of photosynthetic apparatus; no differences were recorded among fertilization treatments. On the other hand, the shading avoidance mechanisms resulted in higher biomass partitioned to leaves at the expense of sink organs, the building of more expanse (higher total leaf area) and thinner (higher specific leaf area and lower leaf-cuticular waxes) photosynthetic apparatus, and in a greater concentration of leaf pigments. The effects on yield and fruit quality of Altino ecotype deserve to be further explored.
Full-text available
Local varieties are known to cope more efficiently with stressful and/or low-input conditions than cultivated ones by activating secondary metabolisms and, hence, are supposed to have higher nutraceutical potential. In this work, we investigate the effects of N and light, supplied at optimal and sub-optimal levels, on the fruit yield and quality of a local ecotype of sweet pepper, Altino, grown in the Abruzzo Region (Southern Italy). In 2017, two open-field experiments were carried out, comparing increasing N rates (0, 100 and 200 kg N ha−1, 0_N, 100_N and 200_N, respectively) and different percentages of shading and/or manipulations of the transmitted solar radiation, obtained through photoselective nets (red net, RN; black net, BN; unshaded Control). Both N and light were preconditions to obtain stable yields in terms of both fruit number and fruit weights. However, BN significantly reduced the number of sunscalded fruits (0.39 vs. 3.38 and 2.59 fruit plant−1 for BN, Control and RN, respectively), leading to lower waste. N deficiency favoured higher total polyphenol (TPC) and flavonoid (TFC) contents in ripened fruits; on the other hand, shading significantly reduced TPC (−12.4%) in immature and TFC (−18.2%) in red fruits. The variations in nutraceuticals were also evaluated in terms of the variations of the single phenolic acids in fruits during ripening; the most interesting results were associated with light treatments. Further research should be directed to the in-depth study of nutrition regimes, in combination with other photoselective nets applications (i.e., pearl nets), that could be suitable for the Altino genotype to enhance the yield and nutraceutical potentials of its fruits.
Full-text available
The extent of intraspecific variation in trait–environment relationships is an open question with limited empirical support in crops. In organic agriculture, with high environmental heterogeneity, this knowledge could guide breeding programs to optimize crop attributes. We propose a three‐dimensional framework involving crop performance, crop traits, and environmental axes to uncover the multidimensionality of trait–environment relationships within a crop. We modeled instantaneous photosynthesis ( A sat ) and water‐use efficiency (WUE) as functions of four phenotypic traits, three soil variables, five carrot ( Daucus carota ) varieties, and their interactions in a national participatory plant breeding program involving a suite of farms across Canada. We used these interactions to describe the resulting 12 trait–environment relationships across varieties. We found one significant trait–environment relationship for A sat (taproot tissue density–soil phosphorus), which was consistent across varieties. For WUE, we found that three relationships (petiole diameter–soil nitrogen, petiole diameter–soil phosphorus, and leaf area–soil phosphorus) varied significantly across varieties. As a result, WUE was maximized by different combinations of trait values and soil conditions depending on the variety. Our three‐dimensional framework supports the identification of functional traits behind the differential responses of crop varieties to environmental variation and thus guides breeding programs to optimize crop attributes from an eco‐evolutionary perspective.
Full-text available
O objetivo do trabalho foi verificar o efeito da interação genótipo x ambiente e sua implicação na recomendação de genótipos de feijão na primeira e segunda safra em Santa Catarina, conjuntamente ao diagnóstico simultâneo de genótipos produtivos e estáveis. Os dados são provenientes dos ensaios estadual e sul brasileiro de linhagens e cultivares conduzidos no estado, totalizando vinte e quatro ambientes e quarenta e oito genótipos de feijão, que foram avaliados nas condições de primeira e segunda safra (12 ambientes cada), nos anos agrícolas de 2012/13, 2013/14 e 2014/15. As análises foram divididas nas duas condições de cultivo, utilizando o modelo de efeitos principais aditivos e interação multiplicativa. A primeira safra apresentou maior efeito da interação genótipo x ambiente na variação total, em comparação a segunda. Os genótipos apresentaram estabilidade a poucos ambientes, com associação favorável ou desfavorável do rendimento de grãos. O genótipo estável com maior produção na condição de primeira safra foi o 46 e na segunda o 26.
Full-text available
Genetic modification (GM) of crop plants is frequently described by its proponents as a continuation of the ancient process of domestication. While domestication, crop breeding, and GM all modify the genomes and phenotypes of plants, GM fundamentally differs from domestication in terms of the biological and sociopolitical processes by which change occurs, and the subsequent impacts on agrobiodiversity and seed sovereignty. We review the history of domestication, crop breeding, and GM, and show that crop breeding and GM are continuous with each other in many important ways, but represent a momentous break from domestication because they move plant evolution off of farms and into centralized institutions. The social contexts in which these processes unfold dictate who holds rights to germplasm and agricultural knowledge, shape incentives to effect particular kinds of changes in our crops, and create or constrict biodiversity. Presenting GM as a continuation of domestication puts forward a false equivalency that fundamentally misrepresents how domestication, crop breeding, and GM occur. In doing so, this narrative diminishes public understanding of these important processes and obscures the effects of industrial agriculture on in situ biodiversity and the practice of farming. This misrepresentation is used in public-facing science communication by representatives of the biotechnology industry to silence meaningful debate on GM by convincing the public that it is the continuation of an age-old process that underlies all agricultural societies.
Full-text available
Organic farming systems present unique challenges that include limitations in available soil nutrients and interference from weeds. They are also vulnerable to environmental fluctuations resulting from climate change. The generation of adapted cultivars with the capacity to evolve under changing conditions may help to meet these challenges. Previous work by our research group demonstrated that directed selection under organic production produces wheat varieties better suited than conventionally selected varieties to organic production. A further step in developing crop varieties adapted to organic production is to directly involve farmers in the selection process through participatory plant breeding (PPB). PPB is particularly effective in stressful or unique growing environments that are underserved by the traditional plant breeding sector. Crop variety development typically takes place under conventional production and organic farmers can likely derive benefits from PPB. We report on a pilot project where 11 organic farmers in Manitoba, Canada were provided with F3 populations of spring wheat. A plant breeder distributed seeds of 3 populations (5000 seeds per population) to the participating farmers, who planted them under their normal cultivation conditions. Farmers identified selection priorities and made selections each year based on their evaluation of the population. After three years of on-farm selection, the performance of wheat populations selected by farmers was compared with registered cultivars in a replicated experiment under organic growing conditions on a research farm. Results from the initial year of testing showed that as a group, the farmer selected populations yielded 107% of conventionally selected cultivars, displayed greater early vigour, were taller and reduced weed biomass compared to the conventional cultivars. By selecting directly in the target environment, farmers are selecting for wheat that is able to thrive under organic production systems. We discuss the implications of this work for organic wheat breeding, functional diversity, and climate adaptation.
This book reviews "participatory" or "collaborative" plant breeding and presents case studies. It also addresses three key issues: (1) what is the nature of plant breeding knowledge, in theory and practice?; (2) in what ways are farmers' and plant breeders' knowledge similar or different?; and (3) what are the implications for successful plant breeding initiatives? The purpose of this book is to examine the nature of and relationship between the knowledge of farmers and of scientists, and how these can be best integrated in plant breeding. In the past, farmers' knowledge of local biodiversity has often been underutilized, but currently there is an increased recognition of the importance of farmer participation or collaboration. It is divided into two sections, farmer plant breeders and collaboration (5 chapters) and scientific plant breeders and collaboration (6 papers).
This book reviews "participatory" or "collaborative" plant breeding and presents case studies. It also addresses three key issues: (1) what is the nature of plant breeding knowledge, in theory and practice?; (2) in what ways are farmers' and plant breeders' knowledge similar or different?; and (3) what are the implications for successful plant breeding initiatives? The purpose of this book is to examine the nature of and relationship between the knowledge of farmers and of scientists, and how these can be best integrated in plant breeding. In the past, farmers' knowledge of local biodiversity has often been underutilized, but currently there is an increased recognition of the importance of farmer participation or collaboration. It is divided into two sections, farmer plant breeders and collaboration (5 chapters) and scientific plant breeders and collaboration (6 papers).
Identifying varieties best suited to local food systems requires a comprehensive understanding of varietal performance from field to fork. After conducting four years of field trials to test which varieties of ancient, heritage, and modern wheat grow best on organically managed land, we screened a subset of varieties for bread, pastry, pasta, and cooked grain quality. The varieties evaluated were three lines of emmer (T. turgidum L. ssp. dicoccum Schrank ex Schübl) and eleven lines of common wheat (Triticum aestivum L.), including two modern soft wheat varieties, four soft heritage wheat varieties, four hard modern wheat varieties, and one hard heritage wheat variety. A diverse group of bakers, chefs, researchers, and consumers compared varieties for qualities of interest to regional markets. Participants assessed differences in sensory profiles, pasta making ability, and baking quality for sourdough, matzah crackers, yeast bread, and shortbread cookies. In addition to detecting significant differences among varieties for pasta, sourdough, and pastry quality, participants documented variation in texture and flavor for the evaluated products. By demonstrating which varieties perform best in the field, in the bakery, and on our taste buds, these results can support recommendations that strengthen the revival of local grain economies.
Tomatoes are a profitable direct-market crop for diversified Midwestern farmers. Unfortunately, many tomatoes with the flavor and quality characteristics consumers desire (such as heirloom varieties) lack agronomic traits important to organic farmers. Hoop-house production offers potentially higher yields and quality than field-grown tomatoes, and has become a popular option for organic farmers. This study compares 19 varieties of tomatoes in both organic hoop house and field conditions, to identify high-performing varieties for future plant breeding, and to characterize the effect of hoop houses on productivity and quality traits. We found that tomatoes grown in a hoop house had significantly higher yield, lower disease severity and higher °Brix (soluble sugars) than those grown in an adjacent field; and that management (hoop house versus field) had significantly more influence over those traits than other variables (variety, market class or year). This lack of varietal differences between management systems will simplify breeding efforts aimed at introducing varieties for hoop house production.