Source: IOWA STATE UNIVERSITY submitted to NRP
IDENTIFYING SOYBEAN FIELDS AT HIGH RISK FOR SUDDEN DEATH SYNDROME THROUGH AERIAL PHOTOGRAPHY AND DNA-BASED TOOLS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
OTHER GRANTS
Reporting Frequency
Annual
Accession No.
1020997
Grant No.
2019-70006-30444
Cumulative Award Amt.
$325,000.00
Proposal No.
2019-02989
Multistate No.
(N/A)
Project Start Date
Sep 1, 2019
Project End Date
Dec 31, 2022
Grant Year
2019
Program Code
[ARDP]- Applied Research and Development Program
Recipient Organization
IOWA STATE UNIVERSITY
2229 Lincoln Way
AMES,IA 50011
Performing Department
Plant Pathology & Microbiology
Non Technical Summary
Sudden death syndrome (SDS), caused by Fusarium virguliforme, significantly limits soybean production in the United States. Management practices include the use of resistant plant varieties and commercially available, but expensive, seed treatments. We seek to improve farmer productivity and minimize unneeded pesticide use by developing pre-plant tools that enable farmers to make a priori SDS risk assessment of their fields. This will be achieved using remote sensing coupled with a machine learning algorithm and validating a qPCR based detection method, both developed by our group. We will utilize replicated SDS research plots, whereby soybeans of contrasting SDS resistance levels are combined with seed treatments, and commercial farm fields in Illinois, Iowa, and Michigan. F. virguliforme in soil, yield, and SDS severity data will be collected over a three-year period. Data will be used to optimize and determine the effectiveness of the SDS assessment. The use of the optimized tool will be shared with disease diagnostic clinics, and its utility and function shared regionally and nationally through multiple extension-based outlets.
Animal Health Component
100%
Research Effort Categories
Basic
(N/A)
Applied
100%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
4041820116050%
2121820116020%
2161820116030%
Knowledge Area
212 - Pathogens and Nematodes Affecting Plants; 216 - Integrated Pest Management Systems; 404 - Instrumentation and Control Systems;

Subject Of Investigation
1820 - Soybean;

Field Of Science
1160 - Pathology;
Goals / Objectives
This project aims to broaden the base of knowledge regarding SDS in soybean, specifically concerning the risk of SDS occurring in a field before planting, and to develop a diagnostic/detection test to help farmers, agronomists, and those involved in making crop management decisions.
Project Methods
We will collect freely available high-resolution (3 m) PS satellite imagery from Plant Labs.High-resolution PS satellite images will be acquired as a single RGB and NIR frame for our field experiment locations. Besides core multispectral RGB and NIR bands, the normalized difference vegetation index (NDVI) will also be used to evaluate its suitability to detect and discriminate among healthy and diseased soybean quadrats. NDVI is one of the most robust vegetation indices that precisely describes vegetation health. NDVI will be calculated for all obtained ortho scenes using the Raster Calculator tool, available as part of the Spatial Analyst Toolbox in ArcGIS Pro (ESRI 2018). Four spectral bands of RGB and NIR, along with calculated NDVI, will be tested to detect healthy and diseased patches in commercial fields and at research farms.The obtained information comprising both remote sensing and non-spectral, ground-based parameters will be used to develop a Random Forest-based model for SDS prediction.The model will be trained using 70% of the dataset and will be validated using the remaining 30% of the dataset. The model will be inspected for its training and validation statistics. We will use the confusion (error) matrix table to assess model precision, specificity, sensitivity, and overall accuracy. Besides these, we will also use the Kappa-index parameter to assess consistency and interrater agreement between trained model predictions. The Random Forest also provides the predictive importance of each variable used for training the model. Hence, all input explanatory variables will be inspected based on their SDS risk predictive importance (%). This information will also help to identify and prioritize variables that most contribute to SDS risk.We will validate the new predictions in small test plots and commercial production fields in year 2 and 3 (2021 and 2022) in Iowa, Illinois, and Michigan. We will select 10 locations with historical observations of SDS and 5 fields without SDS in each state. Each field will be scouted and yield will be collected. We will use the GPS coordinates of these fields to download historical PS imagery from previous soybean plantings. We will analyze the images with the trained model and compare the projected risk with disease information. The model performance will again be tested using the confusion matrix and Kappa statistics.Soil samples from Iowa, Illinois, and Michigan will be collected during the spring and fall in research plots established by the soybean checkoff to determine the most reliable time for detecting and quantifying F. virguliforme. In year 1 (2020 field season), areas of a field with low disease (0 to 20% foliar disease severity in the infected plants), moderate disease (21 to 50% severity) and high disease (50 to 100% severity) will be identified and marked with GPS coordinates. From each severity level, or "risk zone," 5 soil samples (3 cores for each sample) at 20 cm (8 in) depth will be collected during the fall (after soybeans are harvested) and before planting the following spring at 20 cm depth. A total of 90 samples (3 states x 3 risk zones x 2 sampling times x 5 replications), will be collected and processed in each year of the experiment. F. virguliforme levels will be quantified using an existing qPCR protocol.We will use defined threshold levels of F. virguliforme in soil to categorize a field as low or high risk based on the concentration of fungus in the collected soil samples from different risk zones. We will then validate the sampling technique in commercial fields in all three participating states. In years 2 and 3 (2021 and 2022), soil samples will be taken from the same 15 fields (10 fields with a history of SDS and 5 with no history of SDS) in each state. Following the SCN soil sampling protocol, fields will be divided into 8 hectare subsections and 20 soil cores per subsection will be collected from each field and bulked into a composite sample. DNA will be processed as described previously. Sampling and processing protocols may be refined each year based on findings. We will collect soil samples using the same protocol; extract DNA and run qPCR. Each field will be scouted to determine when and how much SDS appears while yield will be collected using a small-plot combine or a calibrated yield monitor. We will record GPS coordinates of these fields and download aerial imagery at R6 growth stage. This typically occurs in late August and is the time most frequently used when assessing foliar symptoms of SDS. Aerial images will help to map SDS distribution.Survey soybean farmers, agribusiness personnel, and checkoff organizations to determine the application and bottlenecks of adapting an "SDS Risk Test." We will work with the evaluation specialist for the Iowa State University IPMProgram, to survey farmers, agribusiness personnel, and soybean checkoff organizations. The "SDS Risk Test" would include the potential tools/services produced from Objective 1 and 2. We also will collect information on possible bottlenecks or hurdles from their perspective for using such a test. The survey will be completed in year 2. Engage with agronomists/precision ag experts (objective 1), plant disease diagnosticians, and soil testing companies (objective 2) to prepare for an "SDS Risk Test." Currently, many companies are using aerial imagery as part of their scouting process. We want to set up separate meetings for each objective to understand how our tools can fit into their business models or improve scouting of soybean fields. Also, diagnostic labs have a molecular-based quantification technique for SDS of infected plants, but not a soilborne SDS risk prediction tool. In year 3, we will set up a meeting with these groups within the North Central States (see letters) to discuss our findings, educate them on the process of testing for the SDS pathogen from soil, and to gather information on possible bottlenecks or hurdles of making these tools available to the end users. We will record the presentations on our research findings and how to identify SDS from aerial imagery and processing soil to determine SDS inoculum levels. These presentations will be posted on the Crop Protection Network website (www.cropprotectionnetwork.org) for those that cannot attend the meeting and for those who plan on using these tools in the future. Develop education materials to prepare to implement an "SDS Risk Test." We will work closely with the Crop Protection Network (CPN).

Progress 09/01/19 to 12/31/22

Outputs
Target Audience:The target audience includes farmers, crop advisers, and possibly diagnostians. Changes/Problems:This project did have a few changes/problems. First, treating and distibuting seed and data collection was a challenge during the COVID restrictions. The COVID restrictions also slowed processing of soil for Fusarium virguliforme at Michigan State.Second, we had two aerial imagery companies go bankrupt during this project, so we lost all imagery from one year. We were able to collaborate with Dr. Guiping Hu and her student Luning Bi at Iowa State University to help with identifying SDS with imagery. However, our research in this area was restricted to only a few fields and so there is still much to learn about assessing risk of SDS using imagery. Finally, our colleague, Dr. Nathan Kleczewski, resigned from his position at the University of Illinois, so the soil health component of this trial was not completed. We did adjust and work with a company from California, Pattern Ag, who analyzed our soil to try to identify levels of beneficial and harmful microorganisms. The results from Pattern Ag will be included in future models for risk assessment. What opportunities for training and professional development has the project provided?We presented the data at the NCERA 137 meeting in Pensacola, FL to update soybean pathologists across the U.S. and Canada about how different factors may play a role in SDS development. We also presented these data at an American Phytopathological Society meeting in 2021. How have the results been disseminated to communities of interest?Extension talks in several states were given. These include more than 15 talks across Iowa in 2021 and 2022 to more than 5,000 attendees. This same presentation was given at the South Dakota Agribusiness, Indiana Agribusiness, and University of Illinois Crop Clinic. A slide deck was created and shared with colleagues in all the states for which data was provided. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Each year of the project we conducted field experiments in Illinois, Iowa, and Michigan. In 2019 and 2020, we also had trials in Delaware, Indiana, Missouri, Nebraska, North Dakota, Kentucky, Arkansas, Kansas,Wisconsin, and Ontario, Canada to determine how fungicideseed treatments afffectSDS under different environments. Three separate field experiments were conducted in each state in 2019 to i) test the efficacy of seed treatment fungicides for SDS management ii) evaluate the efficacy of nematicides seed treatments against SCN and SDS and iii) develop integrated management plan for SDS.Data were collected on plant population, foliar SDS incidence and severity using standard protocols, and yield. Root rot also was collected only from few treatments. We also collected soil samples for soil fertility levels, SCN counts and HG tying at planting at each location. SCN counting and HG typing from those spring samples was processed at SCN diagnostics at University of Missouri, Columbia. DNA extraction of the SDS pathogen was done at Michigan State University. In all, 123 trials across the U.S. and Canada were completed for this study. Additional locations were supported by soybean checkoff and BASF.

Publications

  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Kandel, Y.R., Lawson, M.N., Brown, M.T., Chilvers, M.I., Kleczewski, N.M., Telenko, D.E.P., Tenuta, A.U., Smith, D.L., and Mueller, D.S. 2023. Field and greenhouse assessment of seed treatment fungicides for management of root rot and foliar symptoms of sudden death syndrome and grain yield response of soybean. Plant Disease. In press. https://doi.org/10.1094/PDIS-03-22-0527-RE
  • Type: Journal Articles Status: Accepted Year Published: 2023 Citation: Bi, L., Wally, O., Hu, G., Tenuta, A.U., Kandel, Y.R., and Mueller, D.S. 202X. A Transformer-based approach for early prediction of soybean yield using time-series images. Frontiers.
  • Type: Journal Articles Status: Published Year Published: 2020 Citation: Bi, L., Hu, G., Mueller, D., Leandro, L., Raza, M., and Kandel, Y. 2020. A Gated Recurrent Units (GRU)-based model for the detection of soybean sudden death syndrome with time-series satellite imagery at the field level. Remote Sensing. 12:3621. https://doi:10.3390/rs12213621.
  • Type: Other Status: Published Year Published: 2019 Citation: Kandel, Y.R., Mueller, D.S., Sisson, A.J., Adee, E.A., Bradley, C.A., Bond, J.P., Chilvers, M.I., Conley, S.P., Giesler, L.J., Kelly, H.M., Malvick, D.K., Mathew, F.M., McCarville, M.T., Rupe, J.C., Smith, D.L., Sweets, L.E., Tenuta, A.U., and Wise, K.A. 2019. Seed Treatment and Foliar Fungicide Impact on Sudden Death Syndrome and Soybean Yield. Crop Protection Network. CPN 5002. https://doi.org/10.31274/cpn-20191206-0.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Kandel, Y.R., Brown, M.T., Chilvers, M.I., Kleczewski, N.M., Telenko, D.E.P., Tenuta, A.U., Smith, D.L., and Mueller, D.S. 2021. Evaluation of seed treatment fungicides for management of root rot and foliar symptoms of sudden death syndrome and grain yield response of soybean. American Phytopathological Society. Virtual.


Progress 09/01/21 to 08/31/22

Outputs
Target Audience:This research benefits two groups - farmers and those who advise farmers. The research we do will also benefit the Extension community as we are increasing our understanding of if/when sudden death syndrome can be problematic. Changes/Problems:Nathan Kleczewski left Illinois so his component of the project was redirected. His technician still put out trials for us, but looking at the soil health component was redirected to a private company - Pattern Ag. What opportunities for training and professional development has the project provided?The research teams have been working closely with Pattern Ag, a company based in California that assesses soil mircobe levels based on DNA tests. We started working with this company as our colleague in Illinois, Nathan Kleczewski, left his position for industry. We wanted to see if any of the DNA profiles from Pattern Ag would help identify risk in soybean fields to SDS. How have the results been disseminated to communities of interest?We have presented preliminary findings at several venues, including the NCERA 137 Soybean Diseases meeting, several state Certified Crop Adviser meetings (Iowa, South Dakota, Michigan,Nebraska, Illinois, and Indiana), and a few online venues (e.g., CropsTV at ISU). We have three students working on their PhD dissertations - Ryan Hamilton at Michigan State and Edward Ernat at Iowa State are still working on their PhDs. Luning Bi at Iowa State was partially funded by the project to look at aerial imagery for identification of SDS in fields and he has completed his PhD. Several manuscripts are in progress. What do you plan to do during the next reporting period to accomplish the goals?We will continue to write manuscripts, make progress on dissertations, and present information to farmers/ag professionals as necessary.

Impacts
What was accomplished under these goals? We conducted 123 trials in Arkansas, Delaware, Illinois, Indiana, Iowa, Kansas, Kentucky, Michigan, Missouri, Nebraska, North Dakota, South Dakota, Wisconsin and Ontario, Canada during 2020 and 2021 where we evaluated the effect of seed treatment with Saltro or ILEVO in managing SDS in soybean. Resistant and/or susceptible varieties were planted for this trial. Further we collected soil physical properties, cultural methods, weather parameters and soil property quantification (Pattern-Ag) data and studied their association with occurrence of SDS disease and its severity. Meanwhile we also studied the effect of these biotic and abiotic factors on yield of soybean. Results revealed that soil pH, type, temperature, irrigation, SDS tolerance level of soybean, plant population, soil organic matter, total precipitation in June and July, spring SCN counts, and seed treatment are the most important predictors that contribute to the SDS incidence. This study indicated that certain soil and weather conditions and cultural practices pose a higher risk of SDS incidence in soybean. For example, soil pH 6.1-7, sandy or sandy-loam type of soil, low organic matter (< 2.5), irrigation, higher precipitation in June and July, etc. are potential risk conditions for incidence of SDS. But other cultural practices such as planting resistant varieties, seed treatments etc. reduced the risk of SDS in the field. Soybean seed treatment with ILEVO (fluopyram) and Saltro (pydiflumetofen) significantly reduced SDS severity and root rot but increased yield as controlled to check (base). Yield gain from fungicide application was greater when FDX level was greater than 10% where yield increased by 6.9% in ILEVO and 8.5% in Saltro as compared to check (base). Similarly, FDX decreased by 46.9% and 46.2% and root rot reduced by 19.1% and 17.8% in ILEVO and Saltro, respectively. There was no significant difference in yield among the treatments in resistant varieties, but ILEVO and Saltro had significantly more yield as compared to non-treated checks in susceptible varieties. This information contributes to an ongoing goal of identifying fields that are at a higher risk of SDS development.

Publications


    Progress 09/01/20 to 08/31/21

    Outputs
    Target Audience: The primary targetaudience from this project include soybean farmers, agribusiness, diagnostic clinics, agriculture scientists, and extension educators in each state where soybean is grown. Information will also be applicable to other soybean growing regions outside the United States where SDS is a threat, such as Ontario, Canada, South America, and Africa. Results of this research will reduce economic, environmental, and human health risks while minimizing unnecessary pesticide use. This will be accomplished by equipping farmers with the information they need to determine the necessity of using fungicide seed treatments, resistant varieties, or other management strategies. Changes/Problems:Image and soil sample collection was affected/delayed because of the delayed funding and covid-restrictions in 2020. In addition, Terravion, our contract companyfor aerial images, went out of the businessthus we could not get enough images in 2020.We have asked several collegues outside of the participating states (Illinois, Iowa, and Michigan)to get more aerial images in 2021. Soil samples processing was also delayed in 2020 because of the restrictions due to the pandemic. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?Some of the results from this project havebeen presented in farmer's meeting and workshops. Dr. Guiping Hu gave a presentation on "A gated recurrent unitsbased model for early detection of soybean sudden death syndrome through time-series satellite imagery" in Annual SoybeanFusarium working group meeting which was held virtually on February 12, 2021.Data from Dr. Chilvers' lab and Pattern Ag will be presented atnext annual American Phytopathological Society meeting. What do you plan to do during the next reporting period to accomplish the goals?Aerial images captured by satellite and/or UAV from some field trials in 2020 are being processed in Dr. Guiping Hu's research program. More aerial images will be collected from field trials in 2021. Different machine learning algorithms will betested to develop model to predict risk of SDS in the field. Soil samples that were collected in 2020 and 2021 season will be processedfor real-time PCR risk prediction analysis in Dr. Chilvers' lab and Pattern Ag. Data from these analysis will be summarized and presented in scientific meetings and farmers group meetings and workshops. To determine how SDS management effects onF. virguliformepopulation level in soil and soil health, every participating state collected soil samples from check plots and ILEVO treated plots of both resistant and susceptible cultivars from the field experiments that were establishedwith funding support fromcheck off dollars through north central reserach program.Soil samples were processed to the University of Illinois for assessment indicators of soil health related to microbial activity and disease suppression, as well as mycorrhizal colonization potential, and total nematode community assessment. Data are being analyzed and a manuscript is being processed. We will publish and present the result from this study to scientific community and soybean farmers.

    Impacts
    What was accomplished under these goals? Aerial images captured by satellite and/or UAV were collected from some field trials in 2020 however it was less than expected due to the unexpected obstacles and restrictions caused by COVID -19 pandemic. Those aerial images are being processed in Dr. Guiping Hu's, Department of Industrial and Manufacturing Systems Engineering, research program and different machine learning algorithms are being tested to develop models to predict risk of SDS in the field. To develop models to predict SDS severity and soybean yield loss using at-planting risk factors to integrate with current SDS management strategies soil samples were collected from field trials at Iowa, Michigan, Illinois, and other collaborating states in 2020 (fall spring) and spring 2021. In 2019, a manuscript "Predicting soybean yield and sudden death syndrome development using at-planting risk factors" has been published in plant disease (Plant Dis. 109:1710-1719). The result from the study showed that the distribution of F. virguliforme at-planting had a significant correlation with end-of-season SDS severity, and a significant correlation to yield. SCN distributions at-planting were significantly correlated with end-of-season SDS severity and yield. Prediction models developed through multiple linear regression showed that F. virguliforme abundance, SCN egg quantity, and growing season explained the most variation in end-of-season SDS, whereas end-of-season SDS and end-of-season root dry weightexplained the most variation in soybean yield. These models indicate that it is possible to predict patches of SDS severity using at-planting risk factors. Verifying these models and incorporating additional data types may help improve SDS management and forecast soybean markets in response to SDS threats. In 2020, Martin Chilver's lab received >400 unique soil samples collected from field trial locations. The 2020 collections included both spring and fall samples. There was a delay in shipment and processing of these samples, due in large part to the impact of COVID-19 restrictions on normal operations.The 2020 samples have been homogenized and subsampled ready for real-time PCR processing. To determine how SDS management effects on F. virguliforme population level in soil and soil health, every participating state collected soil samples from check plots and ILEVO treated plots of resistant and susceptible cultivars from the field experiments established with funding support from check off dollars through north central reserach program. Soil samples were processed to the University of Illinois for assessment indicators of soil health related to microbial activity and disease suppression, as well as mycorrhizal colonization potential, and total nematode community assessment. Soil health indicators was be assessed according to the Cornell Soil Health Assessment and include chemical, physical, and biological indicators: microbial biomass, enzyme activities, mineralizable C, active C, total soil protein, pH, EC, OM, available N, P, K and micronutrients, bulk density, and infiltration. Data are being analyzed and a manuscript is being processed. Soil samples were collected from diseased and healthy zones from two locations in 2019 and analyzed in collaboration with Pattern Ag (https://www.pattern.ag). Result showed a clear difference in F. virguliforme population in soil level in healthy and diseased zones. In 2020, we collected soil samples from the trial locations of each state at planting to determine baseline F. virguliforme population. Soil samples are being processed in Pattern Ag. In 2020, Pattern Ag received and processed a total of 550 soil samples from field trials by Michigan State University, Iowa State University, and collaborators. Samples were collected in spring and fall. SCN was detected in 53-71% of the samples and F. virguliforme was detected in 86-98% of the samples in each field. Spring soil samples have also been collected for Pattern Ag to quantify F. virguliforme, SCN and other soil microbes and being shipped for year 2021.

    Publications

    • Type: Journal Articles Status: Published Year Published: 2020 Citation: Bi, L.; Hu, G.; Raza, M.M.; Kandel, Y.; Leandro, L.; Mueller, D. 2020. A Gated Recurrent Units (GRU)-Based Model for Early Detection of Soybean Sudden Death Syndrome through Time-Series Satellite Imagery. 2020. Remote Sens., 12:3621. Online. https://doi.org/10.3390/rs12213621


    Progress 09/01/19 to 08/31/20

    Outputs
    Target Audience:scientists, agriculture industry Changes/Problems:It was interesting getting a new graduate student started and field plots established with the COVID restrictions. We decided to ask colleagues in several states to get more locations. We also received some additional funding from an ag business to pay for these additional locations. The biggest problem was the delay in funding. We did not recieve the funding from USDA until well into 2020 (many months after the project was started). This set us back on getting our contract with TerrAvion. Thankfully Pattern Ag waived costs of processing soil samples. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?We are still collecting data. Once all the data are collected, we can see if we can better predict SDS risk based on what has been collected.

    Impacts
    What was accomplished under these goals? We have established plots in 8 different states. In each of these fields, we have 6 treatments (2 varieties, 3 seed treatments). We are collecting soil from each field and analyzing them in four different labs (SCN at ISU, SDS pathogen at MSU, soil fertility at Midwest Labs, and genomic portofolio with Pattern Ag). We are also collecting aerial imagery from several of these fields with TerrAvion. Disease and agronomic data are being collected as well. An M.S. student, Edward Ernat, started at ISU in May as part of this project (partial support from this project).

    Publications