We will create and trial a genotype by environment prediction model imbedded in a public website that generates top performing small grain varieties based on a zip code. Farmers will plant randomized, replicated trials with 2 varieties recommended by the tool and share results with other farmers through presentations.
The next big trend in the Midwest for soil health and water quality is a market-based incentive model to grow corn and soybean in extended crop rotations with small grains plus legumes. Extended crop rotations, which kept living roots in the ground year round, were replaced in the 1960s by a short, “warm-season-only” system of corn and soybeans. This has resulted in increased nutrient pollution to surface waters leading to the hypoxic zone in the Gulf of Mexico and increased farmer dependence on purchased inputs like synthetic fertilizers and pest and weed control products. Although cover crops are critical to improving water quality, cover crops alone are not enough to sustain and improve the environmental quality and natural resource base on which agriculture depends because they are limited in their ability to break pest cycles and significantly reduce the need for synthetic fertilizers. Extended crop rotations do provide these additional benefits, making them crucial for soil and ecological health. These crop rotations additionally spread out labor throughout the year, enhancing the quality of life for farmers and repopulating communities due to more consistent labor demand. When small grains left the landscape in the 1960s, robust university breeding programs, extension efforts and farmer decision making tools shifted to focus on corn and soybeans only.Today corn and soybean farmers, along with their agronomists, use voluminous data to select high yielding seeds.
Our project – Treating Small Grains as a Cash Crop: stepping up small grain variety selection for Cornbelt farmers acts on the theory that small grains should have the same data-driven support as corn and soybeans and tests a technology platform to amplify the limited university small grains research that exists today. We propose to build a small grain decision-support tool for corn and soybean farmers, including organic producers, that will use a genetics by environment model calibrated with small grains variety trial data from around the Midwest to improve farmers’ confidence in variety selection. The tool will then be validated and further calibrated by randomized, replicated, on-farm research trials evaluating performance of the top two varieties selected by the tool. This publicly available and easy to use tool will result in improved small grain yield performance and profitability of these soil health boosting crops by helping farmers select the right variety for their climate.
Today, farmers primarily plant soybean opposite corn across the Cornbelt (states with >1% of the national corn production are Iowa 18%, Illinois 17%, Minnesota 10%, Wisconsin 4%, South Dakota 4%, North Dakota 2% and others from 2006-2010 USDA NASS data), but this has not always been the case. Previous to the 1960s the common crop rotation in states that drain into the Upper Mississippi River Basin included corn, a warm-season crop, planted opposite cool-season grasses and legumes. But since the 1970s, soybean, a warm-season legume, has replaced cool season crops in the rotation. Previous to the 1970s, small grains like oats in Iowa and northern Illinois; winter wheat in central Illinois, Indiana and Ohio; and barley in Minnesota and Wisconsin dominated the landscape. Small grains were traditionally established as nurse crops for cool-season legumes like red clover and alfalfa. Other times, small grains were sole-seeded and followed by leguminous summer cover crops. Planted acres of red clover in 1969 were 251,512 acres across 12 states in the Cornbelt. Alfalfa was planted on over 13.7 million acres in those same states (USDA NASS 1959-69). Today, nearly 36.3 million acres of soybeans are grown in Illinois, Iowa, Indiana, Minnesota, Ohio, and Wisconsin, occupying much of the acres grown opposite corn (USDA NASS 1982-2012).
The move to corn and soybean rotations has had a negative impact on water quality, soil health, and greenhouse gas production. Because soybeans and corn are both warm season plants, every winter the ground is left bare and vulnerable to soil erosion and nutrient leaching for four or five months, one third to one half of the year. This is resulting in a water quality crisis. Agriculture accounts for over 70% of the nitrogen (N) and phosphorus (P) that enters the Gulf of Mexico via the Mississippi River, nutrients that have already created a nearly 5000-km2 low-oxygen, or “hypoxic,” zone that threatens marine life in one of the nation’s largest and most productive fisheries (White et al. 2014). Cover crops in corn and soybean rotations have been widely promoted for their critical role in improving water quality, but cover crops alone are not enough to sustain and improve the environmental quality and natural resource base on which agriculture depends because they are limited in their ability to break pest cycles and significantly reduce the need for synthetic fertilizers. Another benefit is that yields of crops grown in extended rotation are typically 10% higher than those of crops grown in simple grain crop monocultures, and as much as 25% higher in drought years (SARE Crop Rotations Web; Bennett et al., 2012). Extended crop rotations provide these additional benefits, presenting a low cost, high impact (multiple co-benefits) practice to adopt relative to other water quality and soil health interventions.
With 2018 predicted to be the fourth year in a solid streak of extremely low corn and soybean prices, there has been a groundswell of interest in revisiting these traditional rotation crops. However, a significant barrier for these farmers to reintroducing small grains into crop rotations in the Cornbelt is a dearth of research into small grains breeding and best practices. Farmers have requested more small grains cost share than PFI can accommodate with our funding for 3,650 acres over Iowa, Minnesota and Wisconsin, demonstrating not only the interest in growing small grains, but the need for additional support to do so profitably. When small grains left the landscape in the 1960s, robust university breeding programs, extension efforts and farmer decision-making tools shifted to focus on corn and soybeans only. Today corn and soybean farmers, along with their agronomists, use voluminous data to select high yielding seeds while small grains farmers do not benefit from such tools and resources they can use to continuously improve small grains production on their farms. Without these tools, small grains yields and quality lag behind market demands and create slim, risky profit margins. Furthermore, 80 farmers and researchers at an August 2018 small grains conference and research meeting PFI hosted in Ames, IA identified that even when information such as variety trial data is available, it is difficult to translate into appropriate decisions for their farm because no decision making tools exist to filter and translate data into a usable form.
Our project will amplify the small grains variety trial data that exists and create the decision tool that is lacking to improve the profitability to farmers and associated agricultural businesses. Lucia Gutierrez Chacon with other researchers tested mixed models as a predictive tool for small grains performance in different sets of mega-environments by relying on data from 35 location-years of wheat variety trials. They found that the models were highly predictive achieving great accuracy compared to collected field data (Lado et al. 2016). We intend to expand this model to other small grain varieties using variety trial data from throughout the Cornbelt. This model will generate best performing varieties for barley, oats, wheat (spring and winter), winter rye, and winter triticale based on the mega-environment, located through a zip code. Our hope is that through this process we can effectively leverage the small amount of breeding work currently funded in the Cornbelt to inform small grains decision making over a wider area and, through this collaboration, small grains researchers from the six states can identify breeding gaps and align research programs with farmer needs for maximum impact. With appropriate research support and decision support tools farmers will be able to grow small grains more profitably, therefore increasingly the viability of returning these crops to the landscape and realizing the many environmental and natural resources benefits they provide.
Previous SARE projects have looked at small grain production, such as “Optimizing yield, milling quality, and disease management in spelt and other heritage grain production in the mid-Atlantic,” and “Developing best management practice for growing grain suitable for malt in the Northeast.” However, no projects provide a variety selection tool for the NCR region as this proposal.
- Bennett et al. (2012) Biological Reviews 87: 52-71
- Lado et al. 2016. “Modeling Genotype x Environment Interaction for Genomic Selection with Unbalanced Data from a Wheat Breeding Program.” Crop Sci. 56, pp. 2165-2179.
- n.d. Building Soils for Better Crops 3rd Edition – Crop rotations. Web.
- USDA NASS. (1959-69). Crops Harvested: 1969, 1964, 1959. Web.
- USDA NASS. (1982-2012). Historical Highlights: 2012 and Earlier Census Years. Web.
- White, M.J., Santhi, C., Kannan, N., Arnold, J.G., Harmel, D., Norfleet, L., Allen, P., DiLuzio, M., Wang, X., Atwood, J., Haney, E., and Vaughn Johnson, M. 2014. Nutrient delivery from the Gulf of the Mississippi River to the Gulf of Mexico and effects of cropland conservation. J Soil Water Conserv. 69(1), pp. 26-40.
We will build a variety selector tool for small grains for each geographical region in the Midwest by modeling the genotype by environment interaction to group environments with similar raking of varieties and then predicting the performance of all varieties for a specific geographic region using well established mixed models developed by Dr. Gutierrez’s group (Lado et al., 2016). We believe that due to large genotype by environment interaction that do not follow physical distances but is associated to environmental properties, this tool will predict better the performance of a line than using the information from the closest location regardless of the environmental conditions.
We coordinated small grain breeders from the North Central region housed at land grant universities to populate a central database with historical and on-going small grain variety trial information. Breeders provided ARS at Cornell variety trial results and added to the the historical multi-environment variety testing trials available at the Triticeae Toolbox (T3) data base (https://triticeaetoolbox.org) to evaluate the performance of small grain varieties in the Midwest and to characterize the genotype by environment interaction among varieties.
The oat data used in this study were retrieved from publicly available T3/Oat database (https://triticeaetoolbox.org/oat/). A total of 3,051 genotypes (i.e. advanced inbred lines and released cultivars) were grown in 69 locations across 10 states (IA, IL, IN, MN, ND, NE, NY, OH, SD, and WI) from 1996 to 2018. In Iowa two of the oat variety trials were hosted on PFI member farms. These experiments were primarily from variety testing trails and regional nursery evaluations by public institutions. Most of the oat genotypes were evaluated for agronomic-, quality-, and disease-related traits. Of the 3,051 genotypes evaluated, nearly 1,152 genotypes were genotyped using different genotyping platforms such as GBS and Infinium chip sequencing.
Estimating genotype x environment interaction and identifying mega-environments using biplot analysis requires two-way genotype environment tables. In this variety trial dataset, very few sets of similar lines were tested across locations within a year, leading to highly unbalanced multi-year and multi-location data. Therefore, it is challenging to characterize genotype x environment interaction as well as to delineate mega-environments. To overcome this challenge, we are using the recently published methodology of location grouping (LG) biplot (Yan 2019) that does not require common genotypes across years and utilize existing variety trail data to characterize genotype x environment interaction and identify mega-environments. LG biplot analysis includes two steps: the first step estimates the Pearson correlations among tested locations using tested genotypes and creates location by trial matrix. The second step involves the imputation of missing values in a location by trial matrix by using a singular value decomposition based iteration method (Yan 2013) and displayed in LG biplot. Five mega-environments were identified in the Midwest for the oat dataset. At this stage, we are further working on characterizing GxE and identifying stable mega-environments for each state in the Midwest.
Bread wheat (Triticum aestivum L.) is basically divided into five different wheat classes such as hard red spring, hard red winter, soft red winter, soft white and soft red wheats. The majority of wheat produced in the midwestern region consists of soft- and hard red winter wheat classes. Therefore, the current project was mainly focused on soft- and hard-red winter wheat type.
The wheat data used in this study were retrieved from publicly available T3/Wheat database (https://triticeaetoolbox.org/wheat/). For the evaluation of hard red winter wheat in the Midwest, a total of 1,123 lines were grown in 67 locations across six states (IA, KS, MN, ND, NE, and SD) from 2000 to 2018. These experiments were primarily obtained from varietal testing trails and regional nursery evaluations by public institutions. Most of these lines were tested for multiple-traits throughout the growing season such as days to heading, plant height, disease-pest severity scores, test weight, thousand grain weight, grain yield, and some quality traits such as protein content. Of these 1,123 lines evaluated, nearly 286 lines were genotyped using different genotyping platforms, and 3,121 consensus markers were generated from the consensus tool implemented in T3/Wheat.
Similarly, for the evaluation of soft red winter wheat across the Midwest, a total of 2,275 lines were tested in 43 locations across nine states (IL,IN, KS, MI, MO, NE, NY, OH, and WI) from 2000 to 2018. These experiments were also primarily obtained from varietal testing trails and regional nursery evaluations form public institutions. Most of these lines were tested for multiple-traits throughout the growing season such as days to heading, plant height, disease-pest severity scores, test weight, thousand grain weight, grain yield, and some quality traits such as protein content. Out of 2,275 lines, 110 lines have marker information available. At this stage, we are working on retrieving pedigree information for these lines so that it can be used in the prediction model.
We used the historical multi-environment trials to evaluate the performance of winter wheat in the Midwest and characterize the genotype by environment interaction among varieties using LG biplot methodology described above and GGE biplot methodology. Groups of environments with similar ranking of the varieties known as mega-environments were created using multiplicative models such as GGE biplot add LG biplot. For example, in hard red winter wheat dataset, at first GxE characterization was performed in each year using GGE biplot analysis and then second strategy is to use LG biplot to identify repeatable GE across locations and identify stable mega-environment. As an example, using GGE biplot method in 2016 year of dataset, we identified five group of mega-environments in 2016 growing season (Figure 5). The ME1 consists of only the Roseau (MN) location; the ME2 includes Alliance (NE) and Minot (ND); the ME3 consists of Casselton (ND), Winfield (KS), Williston (ND), and Crookston (MN); the ME4 consists of Wichita (KS), Brookings (MN), Sidney (NE), and Hayes (SD); and the ME5 consists of Lincoln (NE) and North Platte (NE). At this stage, we are looking at different approaches to characterize GGE and identify mega-environments.
At this preliminary phase, we performed genomic prediction using Nebraska hard winter wheat dataset. In Nebraska, we have seven locations evaluated across multiple years. From the GGE biplot analysis, we identified that these seven locations can be divided into two mega-environments (grouping of environments with similar rankings of genotypes). A total of 193 lines have genotypic and phenotypic information available, which was used as a training set for the genomic prediction models. Genomic prediction using GBLUP method (Lado et al., 2016) was performed on adjusted mean grain yield using 3,119 markers from 193 hard red winter wheat lines. The accuracy of genomic prediction model was estimated as the correlation between predicted and observed genotypic value using 5 fold cross-validation with 100 iterations. Genomic prediction was performed for each environment, mega-environment, and all environment combined, and prediction accuracies were compared to evaluate the performance of different models. For instance, genomic prediction in Nebraska was performed for individual environment such as Lincoln (LNK) and Clay Center (CC), mega-environments such as Lincoln and Clay Center combined (i.e. ME2), and all seven environments combined. As expected, the prediction of all locations combined was lower compared to either mega-environment or individual environment predictions. The predictability within mega-environment was better than individual environment and all environment combined predictions. This demonstrate the potential use of mega-environment prediction for identifying best ranking and stable genotypes across multiple locations within in mega-environment. However, we are working on optimizing prediction models by enriching the mega-environment delineation by using a large set of environmental characteristics from each tested environment including the abiotic (i.e. rainfall, temperature, soil properties, etc.) and the biotic (i.e. disease pressures, etc.) environments. This stage will consist of building a mixed model enriched with the environmental covariates that better explain the change in ranking of the varieties using partial-least square regressions (Vargas et al., 1999; Heslot et al., 2014).
Lado B, Barrios PG, Quincke M, Silva P, Gutiérrez L. Modeling genotype× environment interaction for genomic selection with unbalanced data from a wheat breeding program. Crop Science. 2016;56(5):2165-79.
Vargas M, Crossa J, van Eeuwijk FA, Ramírez ME, Sayre K. Using partial least squares regression, factorial regression, and AMMI models for interpreting genotype× environment interaction. Crop science. 1999;39(4):955-67.
Yan W. Biplot analysis of incomplete two-way data. Crop Science. 2013 Jan 1;53(1):48-57.
Yan W. LG biplot: a graphical method for mega-environment investigation using existing crop variety trial data. Scientific reports. 2019 May 9;9(1):7130.
Educational & Outreach Activities
Workshops and Field Days:
- Bhatta M. and L. Gutierrez. 2019. A small grains variety selector tool. Presented at Grain Production Systems Tour for Agronomy/Soil Field Day on Aug 28. Arlington, WI.
- Bhatta M. and L. Gutierrez. 2019. Variety selector tool for small grains farmers and breeders. Industry Field Day on August 09, Madison WI.
- Bhatta M. and L. Gutierrez. 2019. Variety selector tool for small grains in the Midwest. Small Grains Field Day on July 15, Arlington WI.
Nearly 300 people including research scientists, students, and growers visited during the entire field days.
Webinars, Talks and Presentations:
- In December 2019 the project was discussed on a “shared learning” conference call hosted by PFI with a primarily farmer audience. 21 attended.
Published press articles/newsletters: