Materials and Methods

Plant material and seed pretreatments.

Untreated seed of 17 heirloom dry bean (P. vulgaris L.) cultivars were sourced from commercial vendors (Osborne Family Farms, 2014; Purcell Mountain Farms, 2014; Seed Savers Exchange, 2014; Vermont Bean Company, 2014) (Table 1; Fig. 1). Heirloom cultivars were assumed to be heterogeneous populations of homozygous plants with low levels of outcrossing. These cultivars were selected because of their novel seed appearance, bush-type plant architecture, and potential as a cultivar for local niche markets. Seed size descriptors, including 100 seed weight (g) average seed length (mm), width (mm), and thickness (mm) were collected from original seed stock in 2012 and yield trial stock from 2013 trials (Table 1). In addition, seedcoat characteristics such as coat color, corona color, and seedcoat pattern were recorded (Leaky, 1988). Three commercial market class cultivars, Eclipse, Lariat, and OAC Rex (Michaels et al., 2006; Osorno et al., 2009, 2010) were included as commercial checks in the trial; commercial control cultivars were assumed to be pure lines with low levels of outcrossing. In addition, commercial checks were selected to represent the most widely grown market classes grown in the Midwest and commercial types that show potential in organic production. Germination of all seed stock was tested using the “between paper” method specific to common bean (Rao et al., 2006). Just before planting, seed was inoculated with a commercial source (Novozymes, Franklinton, NC) of nitrogen-fixing Rhizobium bacteria (Rhizobium leguminosarum biovar phaseoli) in a peat-based suspension.

Table 1.Descriptions (cultivar name, seed origin, and plant habit) of commercial and heirloom dry bean (Phaseolus vulgaris) cultivars included in the 2013–14 yield trials. Seed descriptors (seed size, shape, weight, coat color, coat patter, corona color, length, width, and thickness) were evaluated from 2012 original seed stock (n = 15) and 2013 yield trial stock (n = 15).

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

Experimental design.

Experiments to determine the yield and yield stability of heirloom dry bean cultivars were conducted during 2013 and 2014. Plots were established at four locations: Fresh Earth Farm in Afton, MN; Foxtail Farm in Osceola, WI; Gardens of Eagan in Northfield, MN; and The Cornercopia Student Organic Farm in St. Paul, MN (Table 2). The Northfield, MN, and St. Paul, MN, locations were United States Department of Agriculture certified organic, whereas the Afton, MN, and Osceola, WI, locations had been managed according to the UDSA guidelines for more than 10 years before the onset of the experiment (Paul Burkhouse, personal communication, 2013; Chris James, personal communication, 2013; USDA, 2014). The University of Minnesota’s Research Analytical Laboratory (http://ral.cfans.umn.edu/references/; St. Paul, MN) performed soil fertility tests (Table 3) in the fall of 2013 and 2014 using standard extraction methods for pH, organic matter (OM), P, K, Ca, and Mg. Previous crops and soil nutrient composition differed among locations (Table 3). No fertilizer, insecticides, or herbicides were applied during these experiments. All locations were maintained as nonirrigated, rain-fed sites.

Table 2.Description (location, geographic coordinates, elevation, rainfall, and growing degree days) of trial sites of the 2013–14 dry bean (Phaseolus vulgaris) yield trials. National Weather Service (NOAA Online Weather Data, 2015) estimates represent accumulated rainfall (mm) and growing degree days (‘GDD’ calculated with base ‘50’) during the Julian period 152–174 (1 June to 1 Oct.).

View Fullscreen

Table 3.Soil types, cropping history, and soil nutrient information for trial site locations included in the 2013–14 heirloom dry bean (Phaseolus vulgaris) yield trials. Soil sampling was performed in the fall of 2013 and 2014 across all locations and replications. Soil tests were performed by University of Minnesota’s Research Analytical Laboratory.

View Fullscreen

Each experiment was a randomized complete block design with three replications. A block consisted of 20 dry bean cultivars as treatments. Heirloom cultivar Lina Sisco’s Bird Egg was omitted from the 2014 trials as a result of improper planting. An experimental plot consisted of a single treatment planted in a 2-row plot that was 4.6 m (15 feet) long with 0.61 m (24 inches) row spacing. Seed count was adjusted for germination percentage plus 5% seedling mortality to obtain a target seeding rate of 215,186 plants/ha (87,120 plants/ac) or four plants per 0.304 m (1 foot). The soil had been prepared by chisel plowing and then finished by field cultivating or rototilling before seeding. Experiments were seeded between Julian days 152 and 161 (1 June to 10 June) in both seasons. Weed-free plots were maintained using a wheel hoe (Valley Oak, Chico, CA) in the early season and biweekly hand cultivation mid-to-late season.

Data collection.

Plots were hand harvested when all pods were beyond physiological maturity in late September or early October (Nowatzki, 2013). The inner 2.44 m (8 feet) of each 2-row plot was hand harvested. Plant populations within the harvested area were determined during the 2014 growing season. Harvested pods were placed in a low-temperature (35 °C) dryer overnight and then threshed with a small belt thresher (Agriculex, Guelph, Ontario, Canada). To maintain a marketable yield for direct-to-consumer markets, harvested seed was cleaned of debris, splits, and diseased seed before weighing. Reported per-plot yields represent weights at ambient relative humidity expressed in kg·ha⁻¹.

Statistical analysis.

Environments were defined as specific location–year combinations. To account for heterogeneity in variances between heirloom cultivars and commercial checks, yield data were subject to a square root transformation, on the basis of a Box–Cox test (Box and Cox, 1964). All data analyses were performed using R software (version 3.1.2) (R Core Team, 2014); models were executed using the “lme4” package (Bates et al., 2014) and evaluated using the “lmerTest” package (Kuznetsova et al., 2015). To account for the unbalanced nature of the yield data set, the following linear mixed effects model was fit using restricted maximum likelihood:where y_ijk is the measured yield observation of the i^th cultivar in the j^th environment (i = 1, 2, …, g; j = 1, 2, …, e; k = 1, 2, …, r_j), μ is the overall mean, α_i is the effect of the i^th cultivar, β_(k)j is the effect of the k^th replication nested within j^th environment, (αβ)_ij is the effect of the i^th cultivar with the j^th environment, and ε_ijk is the random error term. Fixed effects included μ and α_i whereas β_(k)j, (αβ)_ij, and ε_ijk were fit as random effects; random effects, β_(k)j, (αβ)_ij, ε_ijk were declared independent, normally distributed, and scaled toward zero with a variance of σ_β², σ_αβ², σ_ε², respectively. The significance of each random effect variance component was determined from the log-likelihood ratio test statistic, comparing models fit via maximum likelihood with and without the variance component in question, with a Chi square distribution and one df (Table 4). Effects were declared significant at α = 0.05. Profile confidence intervals (95%) were calculated for each fixed effect.

Table 4.Random effect variance components and their respective contribution to yield model used to evaluate the 2013–14 heirloom dry bean (Phaseolus vulgaris) yield trials conducted at four locations in southern Minnesota. Models were fit via maximum likelihood with and without the variance component in question and compared using a Chi square (χ²) distribution (df = 1).

View Fullscreen

Yield stability analyses.

Heirloom cultivar stability was evaluated according to the Type I (static) and Type II (dynamic) concepts of stability classified by Lin et al. (1986) and described further by Becker and Léon (1988). Stability under the Type I concept was evaluated by calculating a cultivar’s cv , whereby the square root of the cultivar’s variance among environments was divided by the grand mean yield (Francis and Kannenberg, 1978). To visually assess Type I stability, a cultivar’s cv was plotted on the x-axis against its estimated mean yield (kg·ha⁻¹) on the y-axis (Fig. 2). The biplot was divided into four quadrants based on the mean cv (23.5%) and mean observed yield (1525 kg·ha⁻¹) across cultivars. Stable cultivars within these trials, according to the Type I definition of stability, exhibited high yield, a low cv, and appeared in the upper left quadrant of the biplot.

View Figure

• Download as Powerpoint
• Download Figure

Visual assessment of Type II (dynamic) stability (Becker and Léon, 1988; Lin et al., 1986) was achieved by plotting an environmental regression coefficient (b_i) (x-axis) against the estimated yield (y-axis) of each heirloom cultivar (Fig. 3). The environmental regression coefficient was calculated from the regression of the predicted random effects of cultivar × environment (αβ)_ij against the predicted random effect of environment β_(k)j. A negative regression coefficient indicated superior performance in poor (negative β_(k)j) environments, a positive coefficient indicated superior performance in favorable (positive β_(k)j) environments, and a coefficient of zero indicated stability across all environments. A regression method of stability using an environmental index, presented here as β_(k)j, was first proposed by Eberhart and Russell (1966). This environmental index, however, was not directly compared with the yield in the present modified method. Rather, the cultivar × environment (αβ)_ij effects were first regressed on the environmental index (β_(k)j) as a means of evaluating stability without directly using yield. Much like the Type I stability biplot, the plot was divided into four quadrants by a vertical line drawn at b_i = 0.0 and a horizontal line drawn at the mean observed yield (1525 kg·ha⁻¹). Type II (dynamic) stability in this biplot was best illustrated by cultivars with high yield (i.e., those with a yield greater than the trial mean of 1525 kg·ha⁻¹) and regression coefficient (b_i) near zero.

View Figure

• Download as Powerpoint
• Download Figure

Results

The final yield model reported random effect variances of replication nested within environment, environment, and cultivar × environment interaction that significantly (P ≤0.001) contributed to the model’s variance (Table 4). Effect of environment and cultivar × environment both accounted for ≈22% of the exhibited variance, where just over 47% of the model’s variance was attributed to random error. No clear correlation between environmental data and yield results was observed, although some of the strongest environmental effects were observed at the Afton and Northfield locations in 2013. Although both locations exhibited very high levels of OM, lower rainfall precipitation in 2013 (Table 2), may have affected pod set and development. Low pH and P levels (Table 3) at the Afton location may have also contributed to lower yields, though, given adequate yields in 2014 at the Afton location, these may not be consistent causative factors in yield reduction.

Average yield of commercial control cultivars was 44% greater than average yield of heirloom cultivars (Table 5). Commercial control cultivar yields ranged from 2355 kg·ha⁻⁻¹ (‘OAC Rex’) to 2617 kg·ha⁻¹ (‘Lariat’), with an average yield of 2447 kg·ha⁻¹. Yields of heirloom cultivars differed (P ≤ 0.05) from one another, with yields ranging from 924 kg·ha⁻¹ (‘Dapple Grey’) to 2127 kg·ha⁻¹ (‘Peregion’) and an average yield of 1362 kg·ha⁻¹ across all heirloom entries. The yield performance of heirloom cultivar Peregion was 36% greater than the mean yield of all other heirloom cultivars and 530 kg·ha⁻¹ greater than the next highest yielding heirloom cultivar, Lina Sisco’s Bird Egg. Both commercial checks and heirloom cultivars did not attain target plant populations; commercial checks ( =199,853 plants/ha) and heirloom ( =177,443 plants/ha) cultivars were 7% and 18%, respectively, below target plant population levels. Lower plant populations were attributed to early-season weed competition, although uniformity of weed densities throughout field plots suggests that differential plant populations did not affect yield comparisons among cultivars.

Table 5.Ranked yield results (cultivar name, yield, 95% confidence interval, cv across locations, and environmental regression coefficient) from the 2013–14 dry bean (Phaseolus vulgaris) yield trials conducted at four locations in southern Minnesota. Trait averages for commercial control cultivars have been tabulated separately from heirloom cultivars.

View Fullscreen

The average cv for yield was similar for both commercial (23.5%) and heirloom cultivars (23.5%) (Table 5). ‘Jacob’s Cattle Gold’ had the lowest cv (15.4%) and third highest yield of the heirloom cultivars (1543 kg·ha⁻¹). In contrast, ‘Dapple Grey’ exhibited the highest cv (36.6%) and lowest yield (924 kg·ha⁻¹) of all heirloom cultivars. Five cultivars, Lariat, OAC Rex, Peregion, Jacob’s Cattle Gold, and Lina Sisco’s Bird Egg displayed Type I (static) stability, though only Jacob’s Cattle Gold and Peregion had a cv less than 20% within that group (Fig. 2). An additional seven heirloom cultivars had cvs less than the average cv (23.5%) across cultivars, though all seven were below the mean yield (1525 kg·ha⁻¹) of all cultivars in the trial.

Calculated environmental regression coefficients ranged from b_i = −0.568 (‘Lariat’) to b_i = 0.512 (‘Yin Yang’) (Table 5). Visual assessment of the Type II (dynamic) stability biplot identified three heirloom cultivars with an environmental regression coefficient near zero: ‘Jacob’s Cattle Gold’, ‘Tiger’s Eye’, and ‘Jacob’s Cattle’ (Fig. 3). Of those three, only ‘Jacob’s Cattle Gold’ was above the mean yield of all cultivars in the trial (Fig. 3). Three cultivars (Eclipse, Peregion, and Lina Sisco’s Bird Egg) were placed in the upper right-hand quadrant of the plot, indicating superior yield in favorable environments (Fig. 3). Two commercial cultivars, Lariat and OAC Rex appeared in the upper left-hand quadrant, suggesting a competitive yield advantage in unfavorable environments; ‘Eclipse’ exhibited the smallest environmental regression coefficient (b_i = 0.099) of all commercial cultivars. Though heirloom cultivars Steuben Yellow Eye and Koronis Purple yielded below the trial mean, their negative environmental regression coefficients suggest a yield advantage in unfavorable environments. In no instance, however, was a cultivar’s mean yield performance in unfavorable environments superior to its mean yield performance in favorable environments (Fig. 3).

Discussion

The assessment of heirloom cultivar yield performance in Minnesota was accented by yield stability evaluation. Examination of the 95% profile confidence intervals indicated a significant difference between the highest and lowest yielding heirloom cultivars, Peregion and Dapple Grey, but there were 11 cultivars with intermediary (1200–1500 kg·ha⁻¹) yield performance that were not readily distinct in their performance (Table 5). Without additional stability analyses, cultivars Koronis Purple, Steuben Yellow Eye, and Ying Yang would have been selected as superior on the basis of yield alone. These cultivars, however, did not appear readily stable according to either the Type I or Type II stability assessments.

Under the Type I (static) stability concept, a cultivar is considered stable if its among-environment variance is relatively small and its performance is unaffected by changing environmental conditions (Becker and Léon, 1988; Lin et al., 1986). In that sense, this stability parameter is the best estimate of a cultivar’s biological stability, independent of other cultivar treatments in the trial. Lin et al. (1986) indicated that a satisfactory Type I stability parameter (i.e., cv) is often associated with poor yield performance, which makes the results of the heirloom yield trials somewhat counterintuitive. Half of the heirloom cultivars in this trial exhibited low variation and (relatively) high yield (Fig. 2). This trend may be unique to heirloom dry bean cultivars, whereby greater plant-to-plant variation may exist within the cultivar and, as a result, lower variation across environments. Further breeding work is required to estimate variation within these populations, but, given that the mean cv of all heirloom cultivars was the same as the commercial control cultivars, it can be assumed that heirloom cultivars exhibit adequate static stability.

The Type II (dynamic) stability biplot (Fig. 3) is a unique way to use the predicted environmental effects in a linear mixed effects model as a means of simultaneously selecting a cultivar for high yield and stability across environments. This biplot depicts dynamic stability that incorporates the predicted effect of cultivar × environment regressed on an environmental index, which, in this case, was the predicted effect of each environment (Eberhart and Russell, 1966; Lin et al., 1986). The Type II (dynamic) stability biplot is convenient for identifying cultivars whose yields respond predictably to increased agronomic inputs and/or management practices (Ferreira et al., 2006). In addition, this biplot allows for visualization of cultivars that may retain some selective yield advantage to abiotic stressors or are better adapted to poor environmental conditions (i.e., those that appear in the upper-left quadrant).

Kelly et al. (1987) noted that factors relating to plant habit, growth rate, and center of domestication notably affect the yield stability among dry bean cultivars. In general, Andean (large-seeded) cultivars exhibit lower yield and stability than their Mesoamerican (small and medium-seeded) counterparts. As such, the yield and yield stability comparisons between commercial control cultivars, all three of which were of Mesoamerican origin, and Andean heirloom cultivars were slightly predisposed to favoring commercial control cultivars. Regardless, the yield of larger-seeded heirloom cultivars Jacob’s Cattle Gold and Tiger’s Eye were relatively stable from both biological and agronomic perspectives (Becker and Léon, 1988; Lin et al., 1986). The relatively high yield of ‘Lina Sisco’s Bird Egg’ and ‘Peregion’ also make them interesting cultivars to explore further.

Production of heirloom cultivars, however, is not without specific production and quality limitations. The large percentage of variation attributed to error in this study was most likely because of seed processing after harvest, in which all diseased and split seed was removed to represent a yield for direct-to-consumer markets. Diseases, particularly common bacterial blight and bean common mosaic virus, were noted throughout both years of trialing. Issues regarding consistent, clean seed supply will need to be addressed if small-scale producers intend to purchase seed on a yearly basis, as most specialty seed companies offer heirloom seed in one pound increments and typically have a limited supply. Researchers at the Organic Seed Alliance, University of Vermont, and Washington State University, have all begun to experiment with small-scale threshing and seed cleaning equipment suited to small-scale production (Colley et al., 2010; Harwood, 2011; Miles, 2015). In all likelihood, these cultivars will be produced on very small acreage (i.e., <1 acre), but concerted efforts will still need to be made to produce enough quality seed to fulfill market demand.

Heirloom cultivars, in particular Jacob’s Cattle Gold, Lina Sisco’s Bird Egg, Peregion, and Tiger’s Eye, are suitable for local organic production according to the yield and yield stability analyses performed herein. Our results are some of the first to highlight field-scale yield estimates for heirloom dry beans produced in the Midwest. Differences between the yield performance of commercial control cultivars and heirloom cultivars may seem drastic, but, when the economic incentives are considered, heirloom cultivars become a viable marketing option. Recent Minnesota surveys indicated that the organic dry beans could obtain $6.27·kg⁻¹ for nonheirloom and $10.52·kg⁻¹ of heirloom cultivars in direct-to-consumer markets (RSDP, unpublished data, 2014). Given that the mean yield of commercial checks (2,447 kg·ha⁻¹) was 44% greater than the mean yield of heirloom cultivars (1,362 kg·ha⁻¹), the price differential accounts for all but 6% of the discrepancy in estimated gross profit per hectare between commercial ($15,342·ha⁻¹) and heirloom ($14,348·ha⁻¹) cultivars grown in organic settings. Further, small-scale producers that sell their organic beans to farmer’s markets stand to make an even greater profit. These results lead to the conclusion that heirloom dry beans offer small-scale organic producers economic incentives and a niche in direct-to-consumer markets within the Twin Cities and Greater Minnesota regions.