Imagine you're a plant breeder. You're going out into a field at a breeding nursery. In this field you see tens, or even hundreds, of little plots that may be 1 by 10 feet or 4 by 8 feet. Each plot contains a different experimental crop cultivar. Your mission is to figure out which potential parents for your breeding population are superior.
To achieve your goal, you'll need to determine what to eliminate.
Look at the plots and measure them for the desired characteristics (type, yield, etc.). Next, ascertain whether each plot looks better because the cultivar has superior genes or because there's something better in the environment in that plot. Is there less pest pressure, better soil, better proximity to the water table, slightly better fertility?
Photo courtesy of Nathalie Dulex/sxc.hu.
Separate out the environmental effects from the important genetic effects, assess the data, and you will have your answers, presuming the experiment design you're using suits your site and resources.
The importance of experiment design
The design of an experiment is one of the most important tools a breeder has to determine which experimental lines are genetically superior. The simplest method is replication: planting the same thing in the field more than once.
One traditional breeding design, the Randomized Complete Block design, involves planting each test variety two or three times in different parts of the field. Each variety is planted once in the first block of the field and again in the second and third blocks of the field. If one cultivar looks good in all parts of the field, it probably really is superior.
Randomized Complete Block designs are easy to set up, analyze and understand, and they work well when conditions in the field are uniform. They also work well for breeders with ample space, labor and time to do multiple replications. Each variety needs to be examined several times while still in the field. This is easy for breeders working at land-grant universities with experimental fields.
Researchers working on farms in the developing world often experience highly variable field conditions. For example, a farm field in Mali (Africa) may have sandy soil on one end and clay on the other. In addition, resources may not be available to plant the same variety in replication. Breeders working on organic farms in the U.S. face similar challenges.
In conventional breeding and production, the soil and other environmental conditions don't matter as much; conditions can be amended with inorganic fertilizers, chemical pesticides, herbicides, etc. Since conventional breeding is done for the largest markets, it often occurs where the conditions are uniformly good.
Organic farming allows fewer environmental modifications to suit the crop, and requires more adjustments within the constraints of the environment. Since prime land is a rare find these days, a majority of new organic farmers are working on marginal farmland. In addition, commercial and educational funding for organic breeding is sparse.
Jared Zystro, a research and education specialist with the Organic Seed Alliance (OSA), says conducting off-farm breeding that's relevant to organics is not the most effective method. "We really believe that good breeding for organic farming needs to incorporate organic farms."
It also needs to incorporate alternative experiment designs that work within the constraints faced by organic farmers.
Amending designs for diverse breeders
OSA has been working with organic farmers and breeders in the U.S. for close to 10 years to introduce experiment designs that have been used in developing nations since the 1960s. Zystro says one of the challenges in moving away from the Randomized Complete Block design is comfort. "Everybody knows how to use it. As soon as you get into something else, you need someone on staff with statistical expertise to analyze the data," he explains.
Inspired by the work of Salvatore Ceccarelli, Zystro espouses three experiment designs for organic seed breeders and farmers who wish to breed on the farm. Numerous experiment designs exist, but Zystro says the following offer the best tools for breeders who are not statisticians.
1 The Moving Grid: When one is growing hundreds of breeding lines in an environment that is not very uniform, the size of the trial may make it difficult to control for changing soils within the field. To simplify the process of assessment and data collection, the breeder compares each plot only to its eight neighbors. If the plant in the center plot is tall and the plants in the neighboring plots are also tall, one can extrapolate that each plant does equally well in the same environmental conditions. If the center plot's plant is tall and the neighbors are short, the breeder gives extra credit to the tall one. The grid of nine plots is thus moved throughout the field until observations are made and data is collected about each plant.
2 Augmented Design: When space is limited, but the breeder needs to evaluate a large number of experimental lines, augmented design allows for accurate data collection without variety replication in the field. In this model, the breeder grows each variety once, but uses standard varieties as checks. By appropriately locating a set of standard varieties throughout the field, the breeder has standards against which to compare the experimental varieties in that same section of the field. If the checks react atypically for their variety, there might be something going on in that part of the field.
3 Alpha-Lattice: This takes the same traditional randomized complete block design and uses statistical methods that break up those large blocks and arranges them so that instead of assessing 200 at a time, the breeder assesses 20 at a time. This method makes it easier for the breeder/grower to account for the possible differences in conditions throughout the field, and for the differences between numerous varieties.
These techniques are most valuable when looking for relatively small differences (like yield), and when those differences are highly affected by environment. Zystro reports that farmers growing vegetable crops often easily see differences in key traits, such as color and leaf texture, without resorting to these techniques.
"With Swiss chard, what we care about is the color of the leaves; the contrast between the leaf and the stem, or petiole; how upright the leaf stems are from the ground when one picks them; and how curly the leaves are when one picks them. However, vigor is an important trait for us that will vary by field conditions," he says.
Zystro has found these alternative experiment designs to be simple and relatively cheap. "A little bit of work up front, putting things in the right places, and at the end analyzing the data ... There are relatively simple designs that can help us do better work in breeding for organic. It's just a matter of implementing those."
OSA is in the early stages of applying these techniques to organic agriculture in the U.S.
"We've got work to do to provide some good case studies for everyone else," says Zystro. "We are really just starting. We're optimistic that it will help our work, but we just haven't been doing it long enough to have any definitive answers."
The author is a freelance writer based in Massachusetts and a monthly contributor to Growing.