Progress 03/15/19 to 03/14/20
Outputs
Target Audience:The target audience for dissemination of the findings of this project are broad-acre vegetable growers and industry stakeholders, regional extension personnel, and other scientific peers. The advisory group for this project consists of key snap bean growers and industry stakeholders responsible for procuring snap beans for processing within New York and other states. All participants and additional growers provided access to their fields for the evaluation of white mold and viewed field trials conducted at research facilities at Cornell University, and equipment at RIT. In addition, the objectives and direction of the project were discussed at formal advisory group meetings at the NY Vegetable Research Association and Council Processing Vegetable Group (December 2019; Batavia, NY). Results were also prepared and published in scientific journals (see products) and in scientific conference proceedings. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The graduate student (E. Hughes) based at the Center for Imaging Science, Rochester Institute of Technology, has defended his PhD thesis (December 2019) and is conducting revisions before final submission. The student gained valuable advanced imaging science skills and a unique opportunity to participate in multidisciplinary research to tackle this critical issue in collaboration with extension educators. How have the results been disseminated to communities of interest?Results have been discussed with communities of interest including the advisory group for this project, extension specialists and crop scouts in the region and broad-acre vegetable growers which include snap bean in their cropping rotation. Results have been disseminated to the scientific community through several journal articles in plant pathology and agricultural science journals, and at imaging science conferences. What do you plan to do during the next reporting period to accomplish the goals?This project is in final extension. The graduate student will be completing his thesis revisions for final acceptance at the Rochester Institute of Technology and submitted two papers based on his imaging science research.
Impacts
What was accomplished under these goals?
Objective 1: Trial 1.We investigated an UAS-based, hyperspectral sensing approach to estimate leaf area index (LAI) for a broadacre disease-crop combination, namely white mold caused by the fungus, Sclerotinia sclerotiorum, in snap bean. This approach is based on preliminary analyses that have shown that structural metrics, specifically leaf area index (LAI), can be correlated with white mold incidence. We used linear and multivariate regressions, with spatial- and spectral-related features, to estimate ground truth LAI. We found that a single reflectance feature at ~510 nm yielded a coefficient of determination (CoD) of 0.76 and an root mean square error (RMSE) of 0.609. Other features, including pixel density, EVI, NDVI, and GNDVI had CoD and RMSE values that ranged between 0.42-0.57 and 0.817-0.942, respectively. Our multivariate regression obtained an adjusted CoD of 0.85 and RMSE of 0.390. This arguably represents a successful outcomes our objective: We could accurately assess canopy structural complexity. These findings will form the basis ofa comprehensive white mold risk model. Objective 1: Trial 2. The most important tactic for in-season management of white mold is fungicide application with the aim of protecting flowers from infection byS. sclerotiorumascospores. The most popular fungicide is thiophanate-methyl but for rotational purposes, boscalid and fluazinam are common although more expensive alternatives. Despite the application of fungicides, white mold often occurs. In the absence of knowledge on optimal fungicide timing, up to two applications may be made wth the first occurring at 10% of plants having at least one open flower, and the second approxmately seven days thereafter coinciding the full flowering and pin-pod. Six small plot, replicated field trials were conducted to identify optimal timings for the most commonly used fungicides (thiophanate-methyl, boscalid, and fluazinam) to reduced the incidence of white mold in two processing snap bean cultivars. Fungicides were applied at either early (10%) and/or late (100%) flowering and the effect on white mold incidence in pods and plants, green leaf area measured using canopy reflectance at 830 nm, and pod yield was evaluated. Application of thiophanate-methyl at early flowering was optimal for reducing the incidence ofwhite mold and a second application at late flowering was not beneficial. In contrast, if thiophanate-methyl was delayed until late flowering, white mold incidence was high and in some cases was not significantly different from nontreated plots. Timing of fluazinam or boscalid was more flexible and not significantly different between a single application at early or late flowering. Two applications of fluazinam or boscalid did not significantly reduce white mold compared to a single application. Delayed application of thiophanate-methyl to later in the flowering may be a contributing factor in the suboptimal management of white mold in processing snap bean. Optimal timing of fluazinam or boscalid appears less critical and offers growers flexibility when conditions are not optimal for application at early flowering. In addition, these products allow for rotation to different modes of action but with trade-offs in cost. This information will provide the economic basis for the use of spectral signatures to guide fungicide usage according to flower development. Objective 2.Disease risk management is one of the most pressing research questions in precision agriculture applications, with the ultimate goal of minimizing chemical inputs and maximizing yield. Unmanned aerial systems (UAS) and associated sensing payloads have come to the fore in this context as a viable option for assessing disease risk in broadacre row crops. Structural metrics, specifically leaf area index (LAI), have been shown to correlate with white mold incidence. We thus used linear and multivariate regressions, with spatial- and spectral-related features, derived from UAS-based imaging spectroscopy data, to assess ground truth LAI. Spectral indicators in the green and red edge portion of the spectrum exhibited coefficients of determination (CoD) greater than 0.7, while the spatial and spectral indices had CoDs and root mean squared errors (RMSE) ranging from 0.422-0.565 and 0.817-0.942, respectively. These results are highly encouraging for incorporating such LAI assessments in final white mold risk models. Objective 3. Findings of this project were discussed with the advisory board in this reporting period in December 2019 and featured in regional extension event programming throughout New York (Cornell Ag InService), and Pennsylvania (Mid-Atlantic Fruit and Vegetable Convention).
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Pethybridge, S. J., Gugino, B. K., and Kikkert, J. R. 2019. Efficacy of Double Nickel LC (Bacillus amyloliquefaciens D747 strain) for white mold control in snap and dry bean. Plant Health Progress 20:61-66.
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Shah, D. A., Dillard, H. R., and Pethybridge, S. J. 2019. Identification of factors associated with white mould in snap bean using tree-based methods. Plant Pathol. 68:1694-1705.
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Pethybridge, S. J., Gugino, B. K., and Kikkert, J. R. 2019. Optimizing fungicide timing for the control of white mold in processing snap bean. Crop Protection 125:104883.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Pethybridge, S. J., Kikkert, J. R., Gugino, B. K., Hughes, E., van Aardt, J., and Shah, D. 2019. Optimizing white mold management in processing snap bean in New York. American Phytopathological Society North East Division Meeting, State College, Pennsylvania. Pp. 29-30.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2020
Citation:
Hughes, E., Kikkert, J. R., Pethybridge, S. J., and van Aardt, J. 2020. Toward risk modelling in row crops the need for accurate fine-scale structural descriptors. IGARSS Conference, Pp. 4.
- Type:
Theses/Dissertations
Status:
Under Review
Year Published:
2019
Citation:
Hughes, E. 2019. Risk modelling in row crops with imaging science. PhD Thesis. Rochester Institute of Technology (Submitted; in review).
|
Progress 03/15/17 to 03/14/20
Outputs
Target Audience:The target audience reached during this period werebroad-acre vegetable growers and industry stakeholders, regional extension personnel, and other scientific peers. The findings of the project were discussed at formal advisory group meetings at the NY Vegetable Research Association and Council Processing Vegetable Group (December 2019; Batavia, NY), vegetable growers at the Empire Expo (January 2020; Syracuse, NY), and members of the advisory group via a webinar (March 2020).A graduate student also successfully defended his PhD thesis at the Rochester Institute of Technology. One journal article was accepted and several accepted for presentation at scientific conferences (see products). Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?Two graduate students based at the Center for Imaging Science, Rochester Institute of Technology, were involved in this research. One thesis has been accepted (March 2020) and we fully anticipate an additional two scientific manscripts to be published. The students were immersed in a multidisciplinary team and interacted with extension educators, agricultural scientists, and imaging scientists. The students are pursuing careers in private industry but have a strong appreciate of the importance of precision agricultural tools, challenges faced by farmers, and communication to a broad range of audiences. How have the results been disseminated to communities of interest?As discussed above in Accomplishments (Objective 3), results have been discussed with communities of interest including the advisory group for this project, extension specialists, crop scouts in the region and broad-acre vegetable growers which include snap bean in their cropping rotation. Results have been disseminated to the scientific community through several journal articles in plant pathology and remote sensing, and at scientific conferences in each discipline. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
Objectives 1 and 2. White mold caused by the fungus, Sclerotinia sclerotiorum is the most important disease affecting snap bean production in NY and elsewhere. A key part of the lifecycle of the fungus is infection by ascospores through the dead and dying flowers. Therefore, timing of disease management decisions such as the application of fungicides is critical to achieving optimal control and preventing crop loss.However, most of the flowers are located beneath the canopy and potentially hidden by foliage making detection by visual scouting and precision agriculture technologies such as RGB cameras problematic.The overarching objectives of this research therefore were to identify spectral signatures for the onset of fowering to optimally time the application of fungicide, investigate spectral characteristics prior to white mold onset in snap beans, and link the location of white mold with biophysical (spectral and structural) metrics to create a spatially-explicit probabilistic risk model for the appearance of white mold in snap bean elds. Spectral angle mapper (SAM) and ratio and thresholding (RT) were used to detect pure vegetation pixels, toward creating theflowering detection models. The pure pixels then were used with a single feature logistic regression (SFLR) to identify wavelengths, spectral ratio indices, and normalized difference indices that best separated the flowering classes. Features with the largest c-index were used to train a support vector machine (SVM) and were then applied to imagery from a different growing season to evaluate model robustness. This research found that single wavelength features in the red (600-700 nm, with a peak at 680 nm) discriminated and predicted flowering up to two weeks before visible flowering occurred, with c-index values above 90%. Structural metrics, such as leaf area index (LAI), have been proven to correlate with white mold incidence, so linear and multivariate regressions were used to ingest spatial- and spectral-related features, derived from the imaging spectroscopy data, and predict ground truth LAI data. These features included raw spectral reflectance values, pixel density, normalized difference index (NDVI), green normalized difference index (GNDVI), and the enhanced vegetation index (EVI). Indicators in the green and red-red edge portion of the spectrum exhibited coeffecients of determination (CoD) greater than 0.7. The spatial and spectral indices had CoDs and root mean squared errors (RMSE) ranging from 0.422-0.565 and 0.817-0.942, respectively. The top 28 features were used in a multivariate regression to predict LAI and the results showed a maximum adjusted CoD of 0.849, with aRMSE of 0.390. Future work should include raw reflectance values, LAI correlated spectral features, as well as auxiliary in-field measurements (degree days, average rainfall, average temperature) in the creation of a white mold risk model. Previous studies have shown that white mold occurrence is tightly coupled to denser canopy structures. Our boosted regression tree analysis also confirmed this relationship for NY snap bean fields (published in Plant Pathology, 2019). By using acombination of SAM and RT we identified pure pixels in order to create predictive models for flowering. We were able to not only discriminate, but also predict flowering at two weeks prior toflowering onset, using single red wavelength features, with 44 c-index values above 90%, using these methods. These wavelength/spectral features are known for their physiological coupling to plant health, photosynthetic activity, and inter-cellular structure. Our structural models models, in turn, were based on both direct, raw reflectance features, as well as spectral indices that have been shown to be correlated to plant structure, e.g., leaf area index (LAI). Plant-level leaf structure/layering (LAI) were measured in situ and then modeled by performing linear and multivariate regressions between spatial and spectral features. Our work found that a single reflectance feature at 510 nm yielded a CoD of 0.76 and an RMSE of 0.609. Other features, including pixel density, EVI, NDVI, and GNDVI had CoD and RMSE values that ranged between 0.42-0.57 and 0.817-0.942, respectively. We then included 28 spatial and spectral features in a multivariate regression to predict LAI,and obtained an adjusted CoD of 0.85 and RMSE of 0.390. This represents successful outcomes for both objectives: We could accurately detect and predict flowering onset and also assess canopy structural complexity. Both of these outcomes eventually can be used to develop a comprehensive mold risk model. Future work could include a spectral convolution study to investigate the robustness of a multi-spectral solution (band center at 680 nm, band pass width 40 nm), the placement of independent calibration panels in the field to assess the delity of the conversion to reflectance, and the incorporation of spectral reflectance values, LAI correlated spectral features, and auxiliary environmental factors (degree days, average rainfall, average temperature, etc.) into a single white mold risk model. The key take-aways from this study was that highly accurate actual and/or predictive (two week in advance) flowering maps can be created for pro-active white mold management in snap beans, along with proven, accurate assessments of plant canopy structure, for improved risk modeling. These results bode well for eventual implementation toward more directed, judicious application of fungicide, which in turn will have significant impacts in terms of a reduced environmental footprint and optimization of yieldt. It remains essential, however, that UAS-based solutions to pest management in crops focus on operational solutions, i.e., solutions that are distilled from expensive, research-grade equipment to accessible, low-risk, and cost-effective automated platforms. Objective 3. Findings of this project were discussed at meetings of the NY Vegetable Research Association and Council, the project advisory committee, and scientific peers during this final period. One manuscript was accepted and two more are in preparation following the acceptance of the student thesis at RIT. Overall, during this project, over 600 people were reached through dissemination of the progress and findings.
Publications
- Type:
Journal Articles
Status:
Accepted
Year Published:
2020
Citation:
Hassanzadeh, A., van Aardt, J., Murphy, S. M., and Pethybridge, S. J. 2020. Yield modelling of snap bean based on hyperspectral sensing: A greenhouse study. J. Appl. Rem. Sens. Accepted 20 May 2020.
- Type:
Journal Articles
Status:
Submitted
Year Published:
2020
Citation:
Pethybridge, S. J., Bowden, C., Murphy, S., Klein, A., and Kikkert, J. R. 2020. Efficacy of fungicides for white mold control in succulent bean, 2019. Plant Dis. Manage. Rep. Submitted 13 May 2020.
- Type:
Conference Papers and Presentations
Status:
Awaiting Publication
Year Published:
2019
Citation:
Hassanzadeh, A., van Aardt, J., Pethybridge, S. J., and Murphy. S. P. 2019. Yield modelling and harvest scheduling of snap bean using remote sensing: a greenhouse study. AGU Conference, San Francisco, CA.
- Type:
Conference Papers and Presentations
Status:
Awaiting Publication
Year Published:
2020
Citation:
Hassanzadeh, A., Murphy, S. P., Pethybridge, S. J., and van Aardt, J. 2020. Toward maturity assessment of snap bean crops: A best-case greenhouse scenario. IGARSS Conference. Pp. 4.
- Type:
Conference Papers and Presentations
Status:
Awaiting Publication
Year Published:
2020
Citation:
Hughes, E., Kikkert, J. R., Pethybridge, S. J., and van Aardt, J. 2020. Toward risk modelling in row crops the need for accurate fine-scale structural descriptors. IGARSS Conference, Pp. 4.
- Type:
Conference Papers and Presentations
Status:
Awaiting Publication
Year Published:
2020
Citation:
Zhang, F., Hassanzadeh, A., Kikkert, J. R., Pethybridge, S. J., and van Aardt, J. 2020. Toward a structural description of row crops using UAS-based LIDAR point clouds. IGARSS Conference, Pp. 4.
|
Progress 03/15/19 to 09/14/19
Outputs
Target Audience:The target audience for dissemination of the findings of this project are broad-acre vegetable growers and industry stakeholders, regional extension personnel, and other scientific peers. The advisory group for this project consists of key snap bean growers and industry stakeholders responsible for procuring snap beans for processing within New York and other states. All participants and additional growers provided access to their fields for the evaluation of white mold and viewed field trials conducted at research facilities at Cornell University, and equipment at RIT. In addition, the objectives and direction of the project were discussed at formal advisory group meetings at the NY Vegetable Research Association and Council Processing Vegetable Group (December 2019; Batavia, NY). Results were also prepared and published in scientific journals (see products) and in scientific conference proceedings. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The graduate student (E. Hughes) based at the Center for Imaging Science, Rochester Institute of Technology, has defended his PhD thesis (December 2019) and is conducting revisions before final submission. The student gained valuable advanced imaging science skills and a unique opportunity to participate in multidisciplinary research to tackle this critical issue in collaboration with extension educators. How have the results been disseminated to communities of interest?Results have been discussed with communities of interest including the advisory group for this project, extension specialists and crop scouts in the region and broad-acre vegetable growers which include snap bean in their cropping rotation. Results have been disseminated to the scientific community through several journal articles in plant pathology and agricultural science journals, and at imaging science conferences. What do you plan to do during the next reporting period to accomplish the goals?This project is in final extension. The graduate student will be completing his thesis revisions for final acceptance at the Rochester Institute of Technology and submitted two papers based on his imaging science research.
Impacts
What was accomplished under these goals?
Objective 1: Trial 1.We investigated an UAS-based, hyperspectral sensing approach to estimate leaf area index (LAI) for a broadacre disease-crop combination, namely white mold caused by the fungus, Sclerotinia sclerotiorum, in snap bean. This approach is based on preliminary analyses that have shown that structural metrics, specifically leaf area index (LAI), can be correlated with white mold incidence. We used linear and multivariate regressions, with spatial- and spectral-related features, to estimate ground truth LAI. We found that a single reflectance feature at ~510 nm yielded a coefficient of determination (CoD) of 0.76 and an root mean square error (RMSE) of 0.609. Other features, including pixel density, EVI, NDVI, and GNDVI had CoD and RMSE values that ranged between 0.42-0.57 and 0.817-0.942, respectively. Our multivariate regression obtained an adjusted CoD of 0.85 and RMSE of 0.390. This arguably represents a successful outcomes our objective: We could accurately assess canopy structural complexity. These findings will form the basis ofa comprehensive white mold risk model. Objective 1: Trial 2. The most important tactic for in-season management of white mold is fungicide application with the aim of protecting flowers from infection byS. sclerotiorumascospores. The most popular fungicide is thiophanate-methyl but for rotational purposes, boscalid and fluazinam are common although more expensive alternatives. Despite the application of fungicides, white mold often occurs. In the absence of knowledge on optimal fungicide timing, up to two applications may be made wth the first occurring at 10% of plants having at least one open flower, and the second approxmately seven days thereafter coinciding the full flowering and pin-pod. Six small plot, replicated field trials were conducted to identify optimal timings for the most commonly used fungicides (thiophanate-methyl, boscalid, and fluazinam) to reduced the incidence of white mold in two processing snap bean cultivars. Fungicides were applied at either early (10%) and/or late (100%) flowering and the effect on white mold incidence in pods and plants, green leaf area measured using canopy reflectance at 830 nm, and pod yield was evaluated. Application of thiophanate-methyl at early flowering was optimal for reducing the incidence ofwhite mold and a second application at late flowering was not beneficial. In contrast, if thiophanate-methyl was delayed until late flowering, white mold incidence was high and in some cases was not significantly different from nontreated plots. Timing of fluazinam or boscalid was more flexible and not significantly different between a single application at early or late flowering. Two applications of fluazinam or boscalid did not significantly reduce white mold compared to a single application. Delayed application of thiophanate-methyl to later in the flowering may be a contributing factor in the suboptimal management of white mold in processing snap bean. Optimal timing of fluazinam or boscalid appears less critical and offers growers flexibility when conditions are not optimal for application at early flowering. In addition, these products allow for rotation to different modes of action but with trade-offs in cost. This information will provide the economic basis for the use of spectral signatures to guide fungicide usage according to flower development. Objective 2.Disease risk management is one of the most pressing research questions in precision agriculture applications, with the ultimate goal of minimizing chemical inputs and maximizing yield. Unmanned aerial systems (UAS) and associated sensing payloads have come to the fore in this context as a viable option for assessing disease risk in broadacre row crops. Structural metrics, specifically leaf area index (LAI), have been shown to correlate with white mold incidence. We thus used linear and multivariate regressions, with spatial- and spectral-related features, derived from UAS-based imaging spectroscopy data, to assess ground truth LAI. Spectral indicators in the green and red edge portion of the spectrum exhibited coefficients of determination (CoD) greater than 0.7, while the spatial and spectral indices had CoDs and root mean squared errors (RMSE) ranging from 0.422-0.565 and 0.817-0.942, respectively. These results are highly encouraging for incorporating such LAI assessments in final white mold risk models. Objective 3. Findings of this project were discussed with the advisory board in this reporting period in December 2019 and featured in regional extension event programming throughout New York (Cornell Ag InService), and Pennsylvania (Mid-Atlantic Fruit and Vegetable Convention).
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Pethybridge, S. J., Gugino, B. K., and Kikkert, J. R. 2019. Efficacy of Double Nickel LC (Bacillus amyloliquefaciens D747 strain) for white mold control in snap and dry bean. Plant Health Progress 20:61-66.
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Shah, D. A., Dillard, H. R., and Pethybridge, S. J. 2019. Identification of factors associated with white mould in snap bean using tree-based methods. Plant Pathol. 68:1694-1705.
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Pethybridge, S. J., Gugino, B. K., and Kikkert, J. R. 2019. Optimizing fungicide timing for the control of white mold in processing snap bean. Crop Protection 125:104883.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Pethybridge, S. J., Kikkert, J. R., Gugino, B. K., Hughes, E., van Aardt, J., and Shah, D. 2019. Optimizing white mold management in processing snap bean in New York. American Phytopathological Society North East Division Meeting, State College, Pennsylvania. Pp. 29-30.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2020
Citation:
Hughes, E., Kikkert, J. R., Pethybridge, S. J., and van Aardt, J. 2020. Toward risk modelling in row crops the need for accurate fine-scale structural descriptors. IGARSS Conference, Pp. 4.
- Type:
Theses/Dissertations
Status:
Under Review
Year Published:
2019
Citation:
Hughes, E. 2019. Risk modelling in row crops with imaging science. PhD Thesis. Rochester Institute of Technology (Submitted; in review).
|
Progress 03/15/18 to 03/14/19
Outputs
Target Audience:During this reporting period, the target audiences were snap bean and broad-acre vegetable growers, crop scouts and regional/county extension agents (that provide agronomic recommendations to growers), andcolleagues interested in applying digital agriculture tools to solve critical agricultural issues including disease management in other production systems in the scientific community. During 2018/19 the Advisory Group has been updated on project progress and provided feedback to facilitate the project on three occasions (March 2018; December 2018; June 2019). Other vegetable growers and extension specialists were also informed on research outcomes at conferences and events in November 2018, January 2019, March 2019. Scientific presentations were conducted in New York and at national and international events (as invited speakers in New Zealand and Canada) throughout 2018 and 2019. Changes/Problems:No changes have been made to this project. We are continuing to prepare two scientific manuscripts and conduct our final impact evaluation with the Advisory Group for the final report to USDA-NIFA. What opportunities for training and professional development has the project provided?The project has provided graduate level training for onestudent (E. Hughes) based in the Center for Imaging Science, Rochester Institute of Technology. This student has also gained valuable extension and public engagement skills by presenting the results of this research at an international precision agriculture conference in Canada, and local agricultural grower-focused conferences in New York (Empire Expo and commodity meeting/Advisory Board groups). Four additional undergraduatestudents have also been involved in the final year of the project gathering data on spectral reflectance and plant physiology in the field. The project has provided a unique opportunity to conduct multidisciplinary research between imaging scientists and agricultural scientists specialized in plant pathology and physiology, and extension specialists and provide the students broad-training in problem solving and logistics, apart from their specialized training in their imaging science discipline. How have the results been disseminated to communities of interest?Results of the second year of the project have been discussed with the communities of interest (advisory group represented by industry stakeholders; extension specialists and crop scouts; and broad-acre vegetable growers in New York) and the scientific community. The project was highlighted by an invited keynote presentation in New Zealand by the Project Director, and the graduate student was invited to talk on research within the first objective at an international precision agriculture conference in Canada. Results have been incorporated into multiple conferences and extension events within NY and surrounding states (Ohio and Pennsylvania) to broaden the reach of the project outcomes as detailed in the 'Other Products' section. What do you plan to do during the next reporting period to accomplish the goals?This project is in the final extension. We are preparing two additional scientific papers for submission to journals and will conduct a final advisory group meeting to discuss the results and obtain feedback on the way forward from our industry stakeholders and growers. The first paper will be focused on spectral signatures associated with snap bean flowering as a tool for optimizing fungicide application and will be submitted to the journal, Crop Science. A second paper will be focused on risk modelling using LIDAR data and will be submitted to an agricultural-focused remote sensing journal. The final report for USDA NIFA CARE project will be submitted within the timeframe.
Impacts
What was accomplished under these goals?
Objective 1. Trial 1.Research continued to evaluate results from both years of the project to quantify the spectral signatures associated with snap bean flowering.A DJI Matrice-600 UAS, boasting a Headwall Photonics Nano-imaging spectrometer (272 bands; 400 to1,000 nm) was utilized to collect the hyperspectral imagery (HSI) in each year. High frequency flights were flownat Cornell AgriTech at The New York State Agricultural Experiment Station, Cornell University, Geneva, New York, around days when portions of the snap bean fields were flowering. The HSI data then were converted into reflectance using the empirical line method, based on in-field black/white calibration panels. Samples of flowering and non-flowering snap bean spectra were selected from the HSI data for the July 2018 time period using spectral band ratios and thresholding (RaT), followed by selective panchromatic illumination thresholding (sPIT) to select high signal-to-noise vegetation signals. Single feature logistic regression then was used to determine which spectral ratio indices (RI), normalized difference indices (NDI), and wavelengths were critical for discriminating between flowering and non-flowering plants. Single wavelengths in the green, red, and near infrared separated the classes with average accuracies ranging between 85 to90%. RI and NDI separated the data with accuracies of 93% using spectral features that are correlated with plant stress and chlorophyll b. Next, the features with the highest c-index were used to train logistic regression (LR), support vector machine (SVM), and perceptron models to investigate deep learning approaches. The SVM had the highest average accuracy (>93%) with a single NDI feature. When the model was applied to flowering and non-flowering test data from July 2017, the average plant pixel probabilities were 86.3% and 41.1%, respectively. This finding suggests high accuracy and promise for our ability to develop UAS-based white mold risk models, since we can identify flowering plants, and furthermore, the spectral bands are located primarily in the cheaper, more operational silicon (Si) detector range. Objective 1. Trial 2.White mold (Sclerotinia sclerotiorum) is the most important fungal disease affecting processing snap bean production in New York (USA). The most important tactic for in-season disease management is fungicide application with the aim of protecting flowers from infection by S. sclerotiorum ascospores. The most popular fungicide is thiophanate-methyl but for rotational purposes, boscalid and fluazinam are common although more expensive alternatives. Despite the application of fungicides, white mold often occurs. In the absence of knowledge on optimal fungicide timing, up to two applications may be made with the first occurring at 10% of plants having at least one open flower, and the second approximately seven days thereafter coinciding with full (100%) flowering and pin-pod. Six small plot, replicated field trials were conducted over three years (2015 to 2017) to identify optimal timings for the most commonly used fungicides (thiophanate-methyl, boscalid, and fluazinam) to reduce the incidence of white mold in processing snap bean cvs. Huntington and Denver. Fungicides were applied at either early (10%) and/or late (100%) flowering and the effect on white mold incidence in pods and plants, green leaf area measured using canopy reflectance at 830 nm, and pod yield was evaluated. Application of thiophanate-methyl at early flowering was optimal for reducing the incidence of white mold and a second application at late flowering was not beneficial. In contrast, if thiophanate-methyl application was delayed until late flowering, white mold incidence was high and in some cases not significantly different from nontreated plots. Timing of fluazinam or boscalid was more flexible and not significantly different between a single application at early or late flowering. Two applications of fluazinam or boscalid did not significantly reduce white mold compared to a single application. Delayed application of thiophanate-methyl to later in flowering may be a contributing factor in the suboptimal management of white mold in processing snap bean. Optimal timing of fluazinam or boscalid appears less critical and offers growers flexibility when conditions are not optimal for application at early flowering. In addition, these products allow for rotation to different modes of action, but with trade-offs in cost. Objective 2.Results from the row spacing and plant density, small-plot replicated trials conducted in both years were interrogated with a larger dataset made available from the industry stakeholders to quantify white mold risk on two scales.Most processing snap bean fields are treated with fungicides at flowering to suppress white mould, one of the more significant diseases of this crop. Farmers would like to know when their fields are at sufficient risk of white mould, to plan fungicide applications or avoid spraying if the risk is below a tolerable threshold. In 2006, 2007 and 2008, observational data were collected from processing snap bean fields across western and central New York State, USA. White mould was found in 20% of fields. Boosted regression trees were used to model white mould presence or absence in a field (a binary response variable) as a function of agronomic and edaphic variables, and macro-scale drought indices. The five most important predictors were canopy openness during pod development, the number of days after planting, hydrologic soil group, canopy openness during bloom, and elevation. The risk of white mould increased by about 20% when canopy openness was less than 20 cm at the bloom stage and approximately 30% when canopy openness was less than 30 cm at the pod stage. The most relevant interaction was between canopy openness at the pod stage and hydrologic soil group. A random forest model for predicting the presence of white mould by bloom had an estimated classification accuracy of 73%. The extension of these results to remote forecasting of white mould in processing snap bean production is discussed. Incorporation of structural evaluation metrics gathered remotely such as LIDAR data, into these risk evaluations is the next step for this project. Objective 3.Findings of the project were discussed with the advisory board on three separate occasions, and included in regional extension event programming throughout New York (Empire Expo; New York Vegetable Research Association and Council; Cornell Ag InService)and Pennsylvania (Mid-Atlantic Fruit and Vegetable Convention). The number of attendees reached at each extension event are listed in the 'Other Products' section.
Publications
- Type:
Journal Articles
Status:
Under Review
Year Published:
2019
Citation:
Shah, D. A., Dillard, H. R., and Pethybridge, S. J. 2019. A boosted regression tree analysis of factors associated with white mould in New York snap bean fields. Plant Pathol. PP-18-351. Accepting pending revision.
- Type:
Journal Articles
Status:
Under Review
Year Published:
2019
Citation:
Pethybridge, S. J., Gugino, B. K., and Kikkert, J. R. 2019. Optimizing fungicide timing for the control of white mold in processing snap bean. Crop Protection CROPRO-D-19-00127. Accepted pending revision.
|
Progress 03/15/17 to 03/14/18
Outputs
Target Audience:Target audiences reached during this reporting period were snap bean growers and industry stakeholders, regional and county extension agents, and the scientific community. Key broadacre snap bean growers and industry stakeholders were involved in this study by participating in the Advisory Group. The overall objectives of the project were discussed with the Advisory group at a project initiation meeting in early spring 2017 which also facilitated the research in the first year of the project (2017). Members of the advisory group and other vegetable growers from across New York were also informed about the research project and outcomes to date at two annual research and developments, the New York Processing Vegetable Research Association and Council Advisory Group Meeting (Batavia; December 2017) and Empire Expo (Syracuse; January 2018). Changes/Problems:This project is proceeding as planned. No major changes are anticipated for conducting the project in the second and final year (2018). What opportunities for training and professional development has the project provided?One graduate student (PhD candidate) based in the Center for Imaging Science, Rochester Institute of Technology, has been employed by this project. Three undergraduate students also gained experience in remote sensing by accompanying UAS flights and use of the spectroradiometer in the field. The experience has provided these students training in advanced computer programming, imaging science techniques, and practical aspects of deployment of precision agriculture tools used in this project. The project also provided the graduate student the opportunity to refine their extension and engagement skills with farmers and industry stakeholders in interactions at the extension event (Empire Expo) in January 2018. These skills will be further refined as results of the project are discussed with the advisory group and the broader vegetable community. How have the results been disseminated to communities of interest?Results of the first year of this project have been disseminated to the advisory group and broader farming community at local and statewide extension events. Incorporating results from the second and final year of this project (2018) into subsequent presentation events will highlight the robustness of findings from this project. What do you plan to do during the next reporting period to accomplish the goals?Objective 1. Identify the spectral signatures associated with flowering in snap beans as a tool for optimal timing of fungicides for white mold control. Trial 1. In 2018, three plantings of snap bean (var. Huntington) will be established as planned with planting dates representing the early, middle, and late windows for New York snap bean growers, at the New York State Agricultural Experiment Station, Geneva, New York. Within each planting, 10 plots (5 m2) will be arbitrarily placed for monitoring of crop phenology and the placement of equipment. We will monitor 10 plants within each plot on a daily basis for growth stage including flowering. On each day, canopy height will be measured and growth stage noted according to the phenological growth stages. The time to 100% flowering, 100% pin-pod, and 100% marketable pods will be modelling according to degree days with base temperatures of 0 and 10 degrees Celsius. Hyperspectral reflectance data will also be collected from the UAS platform at regular intervals throughout the 2018 cropping season, as was conducted in 2017. Reflectance panels will be deployed to allow the conversion of the raw digital count imagery to reflectance allowing for variable atmospheric conditions and time of image capture. Light detection and ranging (lidar) data will also be collected to model canopy architecture. We also will evaluate structure-from-motion (SfM), i.e., 3D imaging from overlapping stereo-pairs of imagery, as an affordable viable alternative to expensive and complex lidar data, for the mapping of canopy architecture. Finally, multivariate analysis will be used to select wavelengths and structural metrics with the highest predictive ability from the hyperspectral ad lidar/SfM data, respectively. These metrics then will be used to develop UAS-based white mold risk models (risk "heat maps"). Trial 2. A small-plot replicated trial will be established in the late planting used for Objective 1/trial 1 to further identify the optimal timing of fungicides for white mold control. The trial design will be the same as that used in 2017. The entire trial will be inoculated with S. sclerotiorum ascospores at 10% flowering to ensure a high incidence of white mold and to enable treatment comparisons. To quantify the efficacy of fungicides on disease incidence and pod yield at harvest, the incidence of white mold on plants and pods, and marketable pod yield (weight of pods/row meter) will be assessed by removing entire plants from four 0.5 m sections from each plot at a growth stage equivalent to commercial harvest. Pods will be separated into diseased or healthy and the number in each category counted. The weight of healthy pods will be recorded to calculate the average weight of individual 'marketable' pods. UAS hyperspectral imagery and lidar data will be collected at each field visit (six times per season), in tandem with the flowering data collections discussed above. These data will be collected at 0.1m (hyperspectral) and >20 hits/m2 (lidar). An Accupar LP-80 ceptometer will be used to collect LAI data at 0.5 m intervals for all treatments; these measurements will serve as reference data for canopy structure assessment, based on remotely sensed UAS-based data. The efficacy of fungicide timings on white mold incidence on pod and plants and yield parameters will be assessed using generalized linear modeling (Genstat Version 17.1; Hemel Hempstead; United Kingdom). Canopy volume and LAI will be modeled with lidar density, penetration (ground vs. canopy hits), and vertical distribution independent variables, as well as narrow-band indices. SAS V.9.4 software will be used in a stepwise forward regression approach, with necessary variable transforms to ensure linearity; all estimates will be accompanied by root mean square error values to ensure that error propagation through models are tracked. These estimates will be scaled to plot-level, to serve as input parameters (along with flowering estimates) to the white mold risk model. Objective 2. Examine the effects of agronomic factors using remote sensing on white mold severity, test for trade-offs, and incorporate this information into a site-specific risk model for white mold in snap bean. Repeats of the small plot, replicated trials examining the effects of row spacing and plant density are planned for 2018 at the New York State Agricultural Experiment Station, Geneva. These factors are changed to manipulate canopy density and model the relationships between white mold incidence, yield and canopy density. It is envisaged that canopy density is a major driver in white mold risk and hence will be used as a remotely sensed proxy to determine justification for the investment in disease control tactics (i.e. fungicide application). These replicated trials will be also inoculated with S. sclerotiorum ascospores to ensure a high incidence of disease and enable treatment comparisons. The methods for evaluating the response variables will be similar to those describe for Objective 1/trial 2. Hyperspectral data and canopy architecture (through LIDAR) will also be collected at regular intervals to enable model development and links to the ground-truthing data. Objective 3. Disseminate research findings. Outcomes from the first year of the project have been discussed with broadacre vegetable growers and industry stakeholders over winter 2017/18. A meeting of the advisory group will be held in March 2018 to obtain feedback on trials planned for the second year of the project. The advisory group will also meet after the field season in fall 2018 to review findings and discuss next steps.
Impacts
What was accomplished under these goals?
Objective 1. Trial 1. The study area was located atCornell University (Geneva, NY). Snap bean (var. Huntington) plots (~ 0.5 acre) were established on three planting dates (3, 14 and 28 June). Ten plots (4 rows wide × 10 feet long) were designated in each planting.The average number of plants per row meter in the early, middle and late plantings was 28.8, 23.4, and 28.1, respectively. Phenological attributes of ten arbitrarily selected plants were recorded at 2 to 4 day intervalsuntil all plants in all plots had marketable pods. Cumulative degree days from the planting date were calculated to when all plants had at least one open flower, pin pod, and marketable pods using bases of 0ºC and 10ºC. Data was obtained from NEWA Cornell using the on-site weather station. The average (and standard deviation) degree days to 100% flowering was 1593 (32) and 823 (24) at 0 and 10 degrees Celsius. A DJI Matrice-600 UAS was utilized to acquire the imagery; the system boasts a camera platform with a high spatial resolution color (RGB) camera, Headwall Photonics Nano imaging spectrometer (272 color bands/channels; ranging between 400-1000 nm), and a Velodyne VLP-16 light detection and ranging (LiDAR) system. High frequency flights were executed when portions of the field started to bloom. Hyperspectral imagery from the Headwall Nano imaging spectrometer had to be ortho-rectified, calibrated into reflectance, and then mosaicked using GPS/IMU (inertial measurement unit) information. The empirical line method (ELM) was used to calibrate the radiance images into reflectance. Pure pixels, or pixels that contain a single, specific object, i.e., these pixels are not mixed spectra of various crop components, of snap bean plants were separated using a supervised classification process called Spectral Angle Mapper, followed by stepwise discriminant analysis. First stage classification models for detection of snap bean flowering were developed using linear discriminant analysis and principal components analysis (PCA). These models had an overall accuracy of 94% (kappa statistic = 0.88). Wavelengths showing the highest accuracy were 521 nm, 543 nm, 672 nm, 761 nm, and 785 nm.These wavelengths are in the relatively cheap silicon (Si) detector range, which implies that we could develop 5-6 band Si sensors for routine application toward detecting flowering. Trial 2. The objective was to identify the optimal timing of fungicides for white mold control using the industry standard, Topsin. A complementary objective was to use of hyperspectral data for detection of white mold. The field trial was conducted at Cornell University,Geneva, New York (RN20) in var. Huntington. Treatments were: Topsin 4.5 FL (30 fl oz/A), Omega (0.85 pt/A) and Endura (11 oz/A) applied at either 10% ('early') flowering, (+/-) second application at 100% flowering ('late'), including a nontreated control. The experimental design was a randomized complete block and each treatment was replicated four times. Treatments were applied either at 36 DAP (10% flowering) and/or 41 DAP (100% flowering).The effect of treatment was analyzed using analysis of variance. No significant (P > 0.5) differences in plant density were detected between plots. The incidence of white mold on plants (94.4%) and pods (43.4%) was high in the nontreated plots. All fungicides and timings significantly reduced the incidence of white mold on pods and plants. Application of Omega and Endura irrespective of timing at early or late flowering significantly reduced the incidence of white mold on pods and plants and were not significantly different between each other. No additional benefits in disease control were also detected from two applications (early and late) of these products. Application of Topsin at early flowering significantly reduced the incidence of white mold on plants compared to an application at late flowering. The incidence of white mold on plants in plots receiving Topsin at early and late flowering was not significantly different to when applied at early flowering only, and from all other products irrespective of timing. Application of Topsin at late flowering resulted in the incidence of white mold on pods that was not significantly different to nontreated control plots, while all other products and timings provided significant improvements in disease control that were not significantly different from each other. Fungicide timings had no significant effect on pod weight and individual pod weight. Objective 2. A small plot, replicated trial was conducted to quantify the effect of row spacing and plant density on white mold and test the trade-offs with yield. The trials were planted on 3 June 2017 using snap bean var. Huntington. The experimental design was a completely randomized block with four replications of each treatment including a nontreated control. For the row spacing trial, block treatments were the industry standard row spacing (30 inches) and double row spacing (60 inches), which was established by hand-thinning on 19 June 2017. For the plant density trial, the industry standard row spacing (30 inch row spacing) was used and plant density in some plots were reduced by half with hand-thinning.Plots were inoculated with 1.6 x 10^6 ascospores per ml on 25 July (10% flowering). In-row plant density was quantified in each plot by counting the number of plants in each of two 0.5 m sections/plot on 7 July. The effect of treatment on the incidence of plants and pods with white mold, pod yield, and average pod weight was quantified at harvest (9 August). Row spacing trial. Row spacing had a significant effect on the incidence of white mold on plants and pods. The incidence of white mold on plants and pods was significantly less in the wider (60 inch) row spacing plots compared to the industry-standard 30 inch row spacing (as expected). Pod yield and average pod weight was not significantly between these treatments. Inoculation with S. sclerotiorum ascospores significantly increased the incidence of white mold on plants and pods across all row spacing treatments. There was a low level of background inoculum which contributed to some white mold found in noninoculated plots. There was a significant interaction between row spacing and S. sclerotiorum inoculation on the incidence of white mold on plants. The incidence of white mold on plants was significantly higher in inoculated plots with a standard row spacing compared to inoculated plots with a 60 inch row spacing. Plant density trial. Plant density was significantly reduced by approximately half from hand-thinning of the treatments soon after stand establishment. The incidence of white mold on plants and pods was significantly less in the reduced density treatment compared to the standard plant density treatment (as expected). Pod yield and average pod weight was not significantly different between the density treatments. Inoculation with S. sclerotiorum significantly increased the incidence of white mold on plants and pods across all row spacing treatments. There was also a low level of background of inoculum which contributed to some white mold in noninoculated plots. Pod yield and average pod weight were not significantly affected by S. sclerotiorum inoculation. There was a significant interaction between plant density and S. sclerotiorum inoculation on the incidence of white mold on plants and pods. The incidence of white mold on plants was significantly higher in inoculated plots with a standard plant density compared to inoculated plots with the reduced density. Objective 3. Findings were highlighted in regional extension events held in December 2017 (New York Processing Vegetable Growers Association and Council, Batavia), January 2018 (Empire Expo, Syracuse). Best management practices for white mold were highlighted in extension talks distributed to attendees.
Publications
- Type:
Other
Status:
Published
Year Published:
2017
Citation:
Pethybridge, S. J., and Kikkert, J. R. Optimizing the fungicide-based management of white mold in two varieties of snap bean. Processing Vegetable Crops Advisory Meeting, Batavia, New York.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2017
Citation:
Hughes, E., Pethybridge, S. J., Salvaggio, C., and van Aardt, J. Towards a white mold risk model for snap beans can we see the flowers? Systems and Technologies for Remote Sensing Applications Through Unmanned Aerial Systems. 2017 Workshop. 20 October 2017, Rochester, NY.
- Type:
Other
Status:
Other
Year Published:
2017
Citation:
Pethybridge, S. J. Transforming white mold management in snap bean using remote sensing via unmanned aerial systems. USDA-NIFA Virtual Project Directors Meeting.
- Type:
Conference Papers and Presentations
Status:
Awaiting Publication
Year Published:
2018
Citation:
Hughes, E., van Aardt, J., Pethybridge, S. J., Kikkert, J. R., and Salvaggio, C. Progress in the application of remote sensing to white mold management in snap bean. Empire Expo, Syracuse, New York.
- Type:
Conference Papers and Presentations
Status:
Awaiting Publication
Year Published:
2018
Citation:
van Aardt, J., Hughes, E., Pethybridge, S. J., Kikkert, J. R., and Salvaggio, C. Transforming disease management through the use of unmanned aerial systems. Proc. Int. Congr. Plant Pathol. Boston, MA.
- Type:
Conference Papers and Presentations
Status:
Awaiting Publication
Year Published:
2018
Citation:
Hughes, E., Pethybridge, S. J., Kikkert, J. R., Salvaggio, C., and van Aardt, J. Snap bean bloom detection from UAS imaging spectroscopy. Int. Prec. Ag. Toronto, Canada.
|
|