Practical Biodiversity: Keeping Oat on Farms by Helping Farmers Enhance Disease Resistance

Final Report for LNC05-259

Project Type: Research and Education
Funds awarded in 2005: $111,548.00
Projected End Date: 12/31/2009
Region: North Central
State: Iowa
Project Coordinator:
Expand All

Project Information


Increased domestic production of oat with its economic and environmental benefits is impeded by oat susceptibility to crown rust (Puccinia coronata Corda). Simple variety blends have recently been successful in rice, wheat, and barley: blend components were all top agronomic varieties and combining them reduced disease. The epidemiological mechanisms of blend effects entail that this approach be tested in very large plots as are used in on-farm research.

We evaluated 19 commercially available oat varieties for their resistances to specific pathogen isolates and chose to test the blend of three that were divergent. In three years of on-farm testing at eight locations per year (four in Minnesota and four in Iowa) the blend experienced significantly less rust severity and yielded significantly more than the average of the component varieties grown in pure stands. While we believe that this simple ecological diversity technology could benefit farmers who grow oats, the infrastructure for collecting and disseminating the necessary information and knowledge are not trivial.


The loss of crop diversity in the Upper Midwest has lead to increased pathogen and insect pest problems in the remaining crops. Though Midwest agriculture has turned away from oat (Avena sativa L.), it has a long history on farms in the U.S. and remains a critical component of most sustainable farms. An oat crop breaks weed and pathogen cycles, builds soil tilth, protects soil from erosion, scavenges nutrients left from other crops, can serve as a nurse crop for the establishment of forages, and provides an irreplaceable feed ingredient for young livestock. Disruption of corn and soybean pests by oat reduces input costs and increases yields of these crops and improves returns to farmers while oat also diminishes the row crops’ potential negative environmental impacts.

Increased domestic production of oat with its economic and environmental benefits is impeded by oat susceptibility to crown rust (Puccinia coronata Corda). While crown rust is not always severe, it can be devastating, and this variability may dissuade farmers from planting oat. Yield statistics understate the economic impact of rust epidemics, which can lower grain quality below market grade. Breeders have sought to enhance oat crown rust resistance using single-gene resistances introgressed from wild relatives. The pathogen, however, has rapidly evolved virulence to all such single-gene approaches. Thus, the use of genetic resistance for disease management poses two problems. First, how can resistance be used to minimize disease severity in the crop? Second, how can the evolution of pathogen virulence be managed? Increasing host-crop diversity by planting variety blends rather than single varieties can reduce disease severity and delay the evolution of pathogen virulence and has been long recognized but insufficiently explored. In the 1950s, an ecological approach was pursued against crown rust: “multiline” oat varieties were bred for diversity in disease resistance but uniformity in other traits. Though multilines managed both disease severity and virulence evolution, they were not adopted because they performed less well agronomically and were expensive to maintain. Simple variety blends, however, have recently been successful in rice, wheat, and barley: blend components were all top agronomic varieties and combining them dramatically reduced disease. The epidemiological mechanisms of blend effects entail that this approach be tested in very large plots as are used in on-farm research.

A significant proportion of variety improvement efforts are devoted to increasing crop resistance. We review crop-pathogen interactions, at the level of the plant then, from an ecological perspective, at the level of the plant population. We discuss approaches to improve resistance within the constraints imposed by the biology and genetics of resistance.

Plants possess mechanisms of resistance that are either simply inherited (monogenic resistance), or that derive from multiple genes and their interactions with each other and the environment (polygenic resistance). In single gene resistances, the gene in the host plant enables it to recognize the product of a single gene in the pathogen; this form of resistance is called gene-for-gene resistance. A salient feature of single-gene resistances is the dramatic differences in virulence that they elicit between pathogen races. In particular, pathogen races that carry the gene whose product the host recognizes are avirulent, while genotypes that do not carry such a gene can be highly virulent. There are no accepted models for the mechanisms that underlie polygenic resistances to disease. These resistances differ in their phenomenology in two important ways from single-gene resistances. First, the resistance is generally not complete, but partial. Second, differences among pathogen races in virulence are not as great against polygenic as against monogenic resistances. In addition to these plant-level resistances, a population-level mechanism contributes to preventing explosive disease epidemics. Natural populations of a plant species exist as complex blends of genotypes in which components differ in their single-gene and polygenic resistances. There are many mechanisms through which blends protect components from disease, which we will detail below.

These characteristics of resistances have had important consequences on breeding approaches to disease resistance. Incorporating monogenic resistances provides the easiest and most rapid solution. Single-gene resistances have therefore been used longest and most widely by plant breeders. The reliance on single-gene resistances has met with short-term success but long-term failure. When a novel single gene is introduced into a variety, the dramatic resistance that it provides often leads that variety, or others possessing the same gene, to be grown over large areas. At the same time, pathogen races that are virulent against that gene have a very strong selective advantage and proliferate, “defeating” the resistance gene. Shifts in crown rust populations that defeat resistance genes have been swift: single genes, once released in varieties, have lost effectiveness in less than five years.

In response to these pathogen population shifts, crop breeders have developed strategies that include gene pyramiding, selection for partial resistance, and the development of multilines. Gene pyramiding involves the introgression of two or more effective resistance genes into a single pure-line variety. Gene pyramids have sometimes retained effectiveness and sometimes have been defeated. Pyramids are more difficult to breed than single gene introgressions. Selection for partial resistances involves handling disease resistance as a quantitative trait, seeking to increase the frequencies of the genes that have small but favorable effects. The heritability of partial resistance is generally reported to be low. Selecting partial resistance in the presence of segregating single-gene resistances is difficult.

Two approaches to increasing host diversity have been considered, variety blends and multilines. Variety blends involve physically mixing the seed of varieties bred as pure lines. Multilines involve selecting a single agronomically adapted variety and creating several sublines of that variety, with each subline containing a distinct single-gene resistance. Each subline derives from the simple but time-consuming process of crossing the variety to a carrier of the single-gene resistance of interest, then repeatedly back-crossing progeny to the original variety. The subline becomes genetically almost identical to the original variety, with the exception of the new resistance gene. The multiline variety is then also produced by physically mixing the seed of the sublines. Given these different approaches, the individual oat plants of a field planted to a multiline will be virtually identical for their polygenic resistances. In contrast, the individual plants of a blend will differ in both their monogenic and polygenic resistances, and added diversity that may confer overall resistance benefits.

Variety blends enjoy several advantages over multilines. In the 1960’s the Iowa Experiment Station released two multiline varieties that were quite effective in preventing economic losses from crown rust but never became popular because, by the time the multiline had been created, the single variety forming the basis of the multiline had become agronomically obsolete. A major breeding advantage of variety blends is that current agronomically competitive varieties can be used. Other advantages include that variety seed sources are easier and less expensive to maintain than multiline seed sources; creation of multilines are a burden to breeding programs that must also develop pure-breeding varieties; and most importantly, variety blends are controlled by farmers, who can tailor them to their farms, while multilines are controlled by breeding programs.

Biological Bases for Variety Blend Efficacy

The reduction in disease severity in variety blends may arise from several factors. Probably the most important cause of reduced disease is the reduction in the proportion of susceptible host plants available for infection. Because only a fraction of the pathogen population will be virulent on any component of a blend, pathogen infection efficiency decreases. The reduced frequency of susceptible plants also means that susceptible plants are physically further apart, which slows the spread of disease from plant to plant in the blend. Resistant host plants may also physically block the dispersal of inoculum between susceptible plants. Compensation and competition among host plants may also play an important role. Disease will affect competitive ability of blend components, leading to increased production by resistant components. Thus, the impact of an equal amount of disease may be less in the blend than in a monoculture.

Another possible mechanism of the blend effect in reducing disease may be induced resistance. An incompatible reaction around an infection site may prevent subsequent infection by an otherwise compatible isolate of the pathogen. Since blends may maintain more variable pathogen populations and high proportions of races avirulent to blend components, they may also enhance induced resistance effects. Competition between pathogen races may also play a role in the blend effect. Modeling of competition between races of stem rust (Puccinia graminis Pers. sp. tritici) of wheat has shown that the more competitive races may not be the most virulent. Finally, in a variety blend, even if two components lack major gene resistances, they will have different polygenic resistances to which some virulent races will not be adapted. Thus blends may have an important biological advantage over multilines: they provide diversity for polygenic and single-gene resistance. Similarly, the variety blend may also beneficial diversity for other traits. Diversity in resistance to other pests (e.g., insects) will make it more difficult for any pest population to adapt to the variety blend environment. Diversity in tolerance to weather-related stresses may make blends more stable in the face of unpredictable weather conditions. Compensating growth and yield of the blend components tolerant of the stress experienced in a specific season may cause a blend to out-perform or attain greater stability than a multiline from purely agronomic perspectives.

The importance of epidemiological factors in the effectiveness of blends has consequences for the empirical study of blend effects. In particular, these effects are most prominent when they can be assessed on large areas. In small plots, evaluation of the blend effect on the build up of the disease over the season can be swamped by border effects. Rather than assessing how the disease inoculum builds up in the blend itself, small plots ultimately show only how much disease built up elsewhere and blew in. This fact provides an important rationale for conducting the experiments we propose in large plots on farms. The most compelling examples of blend benefits come from studies over large areas.

Project Objectives:

Stated project targets were:
In the short term, this project will:
1) determine the extent to which variety blends can control oat crown rust, caused by Puccinia coronata Corda, and increase the profitability and stability of the crop;
2) provide the data on the disease resistance specificities of Upper Midwest oat varieties that farmers will need for variety selection; and
3) increase farmer content knowledge and foster communication to and among producers on the value and use of this information.

In the intermediate term, project outreach will: 1) provide farmers and seed houses procedural knowledge of how to increase and stabilize oat yield and quality by combining complementary commercial varieties;
2) increase discussion among oat breeders on how to develop varieties with diverse disease resistance specificities; and
3) by improving oat profits, increase the diversity of Upper Midwest cropping systems, enabling longer, environmentally healthier, and more profitable rotations.

In the long term, this project will:
1) contribute to highlighting the value and increasing the adoption of within-crop genetic diversity in sustainable farming systems by showing its effectiveness in oat and
2) encourage research on within-crop diversity approaches in other crop/ pest systems.


Click linked name(s) to expand
  • Marty Carson
  • Derrick Exner
  • Carmen Fernholz
  • Linda Grice
  • Roger Lansink
  • Mike Natvig
  • Deon Stuthman
  • Ted Tews
  • Verlan and June Van Wyk
  • Dan Wilson
  • Ray Yokiel


Materials and methods:

Variety Resistance Specificities and Blend Component Selection

The Cereal Disease Lab collected 47 and 139 single-pustule isolates of the crown rust fungus in the upper Midwest in the summers of 2004 and 2005, respectively. In the fall of 2005 each of these isolates was used to inoculate 19 commercially available oat varieties. Rust reactions were scored as susceptible = -3; moderately susceptible = -1; moderately resistant = 1; resistant or highly resistant = 3. The functional difference between pairs of varieties was assessed by the covariance of their reactions across the 186 isolates. A low (possibly negative) covariance indicated functional diversity in a pair of varieties. For three varieties (say A, B, and C) there were three pairwise combinations (AB, AC, and BC). A three-variety blend was therefore be characterized by the sum of three pairwise covariances. Blends with a low sum of pairwise covariances were blends with the most functionally diverse components. Based on this analysis, a blend of the varieties ‘Blaze’, ‘Kame’, and ‘Spurs’ was chosen for further study.

Measuring Disease Development in Field Plots and Overall Blend Performance

Approximately one-half hectare plots of the blend and of each of its three components were planted in eight environments (three farms and one experiment station in each of Minnesota and Iowa) in 2006, 2007, and 2008. One experimental block consisted of one plot of each of the pure line varieties and two plots of the blend, for a total of five plots. Blocks were replicated twice on station at Ames, IA but only once elsewhere. Experimental replication was therefore obtained mainly across environments rather than within environments. The farms and stations ranged from central MN to central IA, the farms and were primarily in southern MN and northern IA. The high number of environments over the study ensured that blends were challenged with a range of weather and initial inoculum conditions. Plots were planted using full-scale farm equipment.

The percent disease severity on the flag and flag – 1 leaves of 20 plants along a transect in the center of plots was estimated twice during the grain-filling period. Percent of leaf area infected was scored visually by a modified Cobb’s scale (Peterson et al., 1948). Measurement of plot yield varied from environment to environment and depended farm equipment available. On the cooperating farms, the oat was cut and windrowed, the center windrows of each plot were harvested with a farm combine, and weigh wagons or drive-on scales were used to measure yields. Windrows were at least 14 m. wide and 66 m. long. Seed moisture was measured and yield adjusted to a standard moisture of 13.5%. Yields are reported in bushels per acre. With a standard oat bushel of 32 lb, one bushel per acre is 35.8 kg ha-1.

Data Analysis

For rust severity analysis, we retained only environments where rust was detected on at least ten percent of flag leaves sampled in the environment; four, four, and seven environments were retained for 2006, 2007, and 2008, respectively. Percent severity was arcsine square root transformed. Yield data were lost from one environment in 2008. Yield measurements were untransformed. Data were analyzed using the following model: Yijk = Ti + Ej + ETij + Rk(j) + e

Where Yijk is the measurement of treatment i (treatments were pure line Blaze, Kame, or Spurs, and the blend), Ti was the effect of treatment i and was considered fixed, Ej was the effect of environment j (random), ETij was the effect of the interaction between environment j and treatment i, Rk(j) was the effect of replication within environment j (random), and e was an error term (random). This linear model was fit in a Bayesian framework using WinBUGS (Lunn et al. 2000). In the Markov chain Monte Carlo analysis implemented by WinBUGS, random effects were sampled from normal distributions with zero mean and effect-specific variances. Those effect-specific variances were sampled from inverse gamma distributions with shape parameter equal to 1 and scale parameter set so that the expectation of the variance was equal to variances calculated from ANOVA. A single chain was run with 500,000 burn-in iterations and 500,000 sampled iterations. The blend effect was calculated as (TBlaze + TKame + TSpurs) / 3 – TBlend. Probabilities given below are the frequencies with which a particular condition (e.g., a positive blend effect) occurred over the sampled iterations of the Markov chain. To analyze the stability of the blend relative to the pure line varieties, a further analysis was run in which environment by treatment interaction variances were estimated separately for each of the treatments. In this analysis, the blend effect was calculated as (vExTBlaze + vExTKame + vExTSpurs) / 3 – vExTBlend, where vExTi denotes the treatment-specific interaction variance.

Research results and discussion:

Blend Effects on Rust Severity

Each of the three component oat varieties showed different levels of rust severity. In prior assays of rust resistance against single-spore isolates, the ranking of the components had differed, with Kame being resistant to the highest percentage of isolates (95.7%), followed by Spurs (53.2%) and Blaze (48.4%). In this case, at least, seedling assays of resistance were not highly predictive of rust severity under field conditions. The probability that the blend had a rust severity that was below the average severities of the pure components was 97%, indicating strong support for the fact that some mechanism reduced rust severity below what might have been predicted from observation of the components. Furthermore separate estimation of the variety by environment interaction variance for rust severity for the blend (4.7 × 10-3) had a 98% probability of being lower than that of the pure line components (13.6 × 10-3). Thus, the blend was also more stable in the face of different environments than were the pure lines. One hypothesis for this greater stability is that at least one component of the blend was resistant to a greater fraction of the rust inoculum arriving into the field. Another hypothesis is that induced resistance played a greater role in reducing rust severity in the blend than in the pure component fields. We can view these results from two perspectives. First, if we assume that we knew at the outset which component was the most resistant, would it nevertheless be worth planting a blend? Here, we consider the difference between the blend and Spurs, which was in our trials the component with the lowest rust severity. Under this assumption, it seems clear that we would be better off not making a blend, and just planting the more resistant single component. The assumption that we might know which component is most resistant, however, is difficult to justify. In our case, for example, even extensive preliminary resistance assays indicated that Kame was more resistant than Spurs.

Blend Effect on Yield

In the case of yield, the blend performed equally to the best pure component. The probability that the blend had a higher yield than the average of the pure components was 98%, again indicating strong support for the exitence of a mechanism increasing yield above what might have been predicted from observation of the components. Again, the blend showed greater stability to environmental variation than did the pure components: it’s variety by environment interaction component (25.5) had a 96% probability of being lower than that of the pure components (69.4). The mechanisms for the increased yield and stability might simply be a consequence of the greater rust resistance of the blend relative to the pure components. Other environmental factors, however, could also play a role: if one component responded well to a factor while others responded poorly, gains from the former could compensate for losses from the latter.

[To view figures from this report, contact the NCR-SARE office at]

Research conclusions:

The impacts of this project related primarily to its mid- and long-term goals. At this point, project short term goals have mostly been met. The variety blend we studied reduced crown rust severity to levels to close to that of the most resistant variety in the blend while also providing yields equal to the highest yielding variety in the blend. The blend exhibited greater stability for both rust severity and yield than the pure lines. We have developed protocols to obtain disease resistance specificities needed for variety selection in composing blends.
Four field days (three in Iowa and one in Minnesota) increased knowledge of more than 100 oat producers who participated in them. Three cooperator workshops were part of this project and the farmers who participated gained more detailed knowledge yet. Some of the research results have been disseminated in the “gray” literature such as the Practical Farmers of Iowa general research report.

The project, however, has two major failings that reduce its intermediate and long term impact. First, we have thus far not published our findings in the peer-reviewed literature. Second, though this was far beyond the scope of our project, no ongoing mechanism has been developed to institutionalize the annual collection and dissemination of the information necessary for seed companies or farmers to blend oat varieties as a regular practice.

Economic Analysis

No economic analysis was performed in this research.

Farmer Adoption

No information was collected on farmer adoption of the use of blending oat varieties. Further information would be necessary for farmers to do so. An extension bulletin should be published giving the rationale for blending and providing detailed procedures for choosing blend components. Regular information on the rust resistances of new oat varieties would need to be published, possibly through the Uniform Oat Performance Nurseries run by the USDA.

Participation Summary

Educational & Outreach Activities

Participation Summary

Education/outreach description:

Four field days were held, reaching about 100 farmers.

Three cooperator workshop meetings were held, reaching a further 20 farmers.

Results as of 2007 were published in the Practical Farmers of Iowa general research report.

Project Outcomes


Areas needing additional study

In this study, we find blend effects that are quite impressive. Favorable blend effects for yield are not necessarily found for two- and three-way blends (Helland and Holland, 2001). Further research is therefore needed on the design of blends, and whether the particular design approach we propose allows consistent identification of blends with superior performance.

More care needs to be given to consistent and sustainable means to extend the information on variety blends to farmers.

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture or SARE.