Progress 01/15/18 to 01/14/23
Outputs
Target Audience:The target audiences for predictive modeling effortsinclude county agents, crop consultants, shelling companies. As the initial buyers of farmer stock peanuts directly from growers, the shelling companies are a primary target for the knowledge and tools that will be developed from this project. They will be able to utilize models to inform their growers and mitigate risks in seasons with high aflatoxin potential. Changes/Problems:
Nothing Reported
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?
Nothing Reported
Impacts
What was accomplished under these goals?
During the final NCE year, we finalized analysis of VOC and began developing a manuscript. There is a large volume of data to analyze, synthesize and report, so efforts will continue beyond the timeframe of the award. Overall and as outlined in detail in previous reports, we were successful in measuring VOC and in initiating aflatoxin contamination through artifical inoculation with Aspergillus paraciticus.We were also successful in producing water stress to evaluate VOC production as a result, genotypic variation in VOC production and aflatoxin contamination as affectd by water stress and peanut genotype. In summary, we anticipate determining potential links between VOC and aflatoxin and plant water stress. We also anticipate understanding potential peanut genotypic variabitly in VOC production and wheither that variation impacts any potential relationship between VOC and afltoxin contamination. VOC may also provide insight into plant water stress that could be valuable in predicting irrigation strategies that would mitigate both drought stress and afaltoxin production. The effort to developa regional risk model has partnered with other projects in this space to provide more data on which to base modeling. In particular, we have recieved data from a commercial partner and have been able to couple that with existing models to refine and develop a more accuratemodel. Data from this study will be used to validate and refine the modelfurther.
Publications
|
Progress 01/15/21 to 01/14/22
Outputs
Target Audience:
Nothing Reported
Changes/Problems:
Nothing Reported
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?During the next reporting period we will develop a plan to organize, analyze data and summarize results for publicaiton.
Impacts
What was accomplished under these goals?
Task 2: The second field trial was conducted in 2020 at the University of Florida Plant Science Research and Education Unit (PSREU) in Citra, Florida. Peanut cultivar FloRun™ '331' (Tillman, 2021, Reg. no. CV-144, PI 680624) was planted to 48 field plots laid in a randomized completed block design. After plant emergence, Peanut plants were grown under full irrigation until 100 days after planting (DAP), after which these plants were treated with aspergillus, Aflaguard, or not treated - acting as the control. At 100 DAP, peanut plants were subjected to an end-of-season drought (covered by rain-out shelters, irrigation ceased) to induce aspergillus growth and aflatoxin production. Plant based volatile organic carbon (VOCs) were collected using Stir Bar Sorptive Extraction (SBSE, aka Twisters) before inoculation and weekly for 4 weeks after inoculation. After 10 DAP, peanut plants were harvested and pods were sent to GTRI for VOC collection and further analysis. In-shell and shelled pod VOCs were collected, respectively. Then the pod and kernel samples were sent to JLA to determine aflatoxin concentration. The VOCs analysis of Year 2 data was completed and primary findings were summarized as follows: VOCs collected from plants: The most successful classification model the team tried was Random Forest. The model was trained on peanut plant VOCs for each dataset individually and evaluated on the validation set reserved from the same dataset (replicates from the same block were both either in a training or validation set). The accuracies observed were: 79.2% (dataset collected on Sep 16), 95.8% (dataset collected on Sep 23), 91.7% (dataset collected on Figure 4. Linear Discriminant Analysis. Data Collected on September 9th, 2020 5 Oct 15), and 87.5% (dataset collected on Oct 22). This established model is better than the model for Year 1. VOCs collected from pods: The Machine Learning approach that uses a Random Forest classifier of VOCs was applied to the pods data. Similarly to the plant data, each of the three pods datasets was split into training (72 samples) and validation set (24 samples), making sure that the replicates remain in the same set. The classifier achieved the following accuracies on validation sets: 87.5% (in-shell peanuts), 95.83% (shelled peanuts), and 95.83% (shelled and dried peanuts). Classifiers trained on pods data didn't have a common list of important compounds, and didn't rely on the same compounds as the plant classifiers identified in previous quarter report. Analysis of VOCs - Aflaguard vs Aflatoxin comparison - based on 2020 field data: A list of plant based VOCs were identified and compared among the control group, aflatoxin group (inoculated with aspergillus) and the Aflaguard group (treated with a non-aflatoxin producing strain of aspergillus. Among these chemical compounds, 3-Hexen-1-ol, (Z) and cis-3-hexenyl acetate stand out and were selected as potential VOC markers to further evaluate the difference between aflatoxin group and the Aflaguard group. Aflatoxin Concentration Data - JLA Results: Results from the 2019 aflatoxin data indicates that plants exposed to the aspergillus produced kernels noticeably higher in aflatoxin concentrations with an average reading of ~20,000 ppb with some obvious outliers. The control had the second highest average aflatoxin concentration with ~900 ppb again with several outliers. Finally, the plants treated with Aflaguard saw the lowest contamination of aflatoxin with an average of ~30 ppb with an obvious outlier. Results from the 2020 aflatoxin data show some differences when compared to the 2019 data. The first notable difference is the lower aflatoxin contamination in the plants exposed to aspergillus. The average aflatoxin concentration for these samples was ~900 ppb with one sample containing an order of magnitude higher than the next highest sample. The 2020 data also indicates that plants exposed to the Aflaguard contained a higher average aflatoxin contamination compared to the control, ~160 ppb compared to ~36 ppb, when outliers are accounted. This data will be used to help refine Hierarchical Clustering Analysis and Linear Discriminant Analysis results through the removal of outlier samples
Publications
|
Progress 01/15/20 to 01/14/21
Outputs
Target Audience:Our target audience includes peanut producers, shellers, and manufacturers for the use of the model. We also expect scientific societies, colleagues in science, and graduate students to be target audiences for our basic scientific results related to sensor development, aflatoxin modelling, and image analysis techniques. Changes/Problems:Obtaining contaminated aflatoxin samples from local buying points has proven to be difficult due to the uncertainty of aflatoxin risk for a given season. In response we increased our collection efforts on samples from both greenhouse and field trials to maximize data obtainable. We also expect the large data sets available from Premium Peanut will help achieve our goals in this area. What opportunities for training and professional development has the project provided?Opportunities for professional development have been provided to the graduate students involved in the project by the formation of a joint journal group discussing relevant papers addressing approaches that combine agronomic and machine learning techniques. Submission of abstracts by one graduate student (McAmis) associated with the project has been completed and the paper presented. How have the results been disseminated to communities of interest?For 2020, dissemination efforts have been aimed at the oral presentations of project approaches by Rowland and Song. Project plans and goals were presented to the American Peanut Research and Education Society and the American Society of Agricultural and Biological Engineers. What do you plan to do during the next reporting period to accomplish the goals?In 2021 given our approved no-cost extension, we will focus on the following: Continued peak identification of BVOCs from field and pod samples - 2019 and 2020 samples Continued peak identification of BVOCs from field based plants and pod samples - 2020 samples Continued statistical analysis of all BVOC data collected from combined 2019 and 2020 samples and correlate it to aflatoxin concentration In-depth analysis of VOC data from all samples collected to date. Combined hyperspectral imaging and BVOC testing on single kernels to validate both signatures experimentally.
Impacts
What was accomplished under these goals?
Accomplishments occurred organized under each of the four tasks: Task 1: further refine and revise the aflatoxin regional risk model (ARRM) for peanut production in the north central Florida region that was developed in Year 1. Accomplishments Year 3: This modified task was 40% complete at the end of year 2. In 2020, Rowland worked with grad student, Justin Pitts, in collaboration with Co-PI Brym to collate current experimental data. The team initiated collaborative project with Premium Peanut, one of the primary shellers in Georgia, to begin collecting the larger volume of commercial production level data that is needed for model development. Machine Learning approaches were developed to predict the level of aflatoxin based on testing at the time of harvest and loading into the warehouse and the level of aflatoxin detected after approximately one year of storage at the stage of off-loading the warehouse. This was an effort to begin to identify the characteristics that would predict when aflatoxin levels would increase during storage. Task 2: Biogenic volatile organic compound (BVOC) sensors will be deployed in research plots with drought and control conditions implemented. Simultaneous to Task 1 activities, the second year of the field study was conducted under controlled conditions at the Plant Science Research and Education Unit in Citra, FL in a rainout shelter facility. Briefly, a replicated, randomized field trial was established with 48 plots. Plants were allowed to grow under normal conditions until 100 DAP, after which the plants were treated with Aspergillus parasiticus (using the same strain as was used in the greenhouse experiment), Afla-Guard (a commercial product containing atoxigenic Aspergillus), and no soil treatment - acting as our control. Shortly after inoculation, the rainout shelter was placed over all plots and irrigation ceased to induce aflatoxin production. Plant based BVOCs were collected from plants before inoculation, and weekly for 2 weeks post inoculation, as in Year 1 of this trial. SBSE were used for all collections and shipped to GTRI for analysis. After 140 DAP, the plants were harvested and the pods were sent to GTRI for VOC collection and analysis. In-shell pod BVOC samples were collected, as well as shelled samples. Once the BVOC samples were collected, the pod samples were shipped to JLA for aflatoxin concentration measurements. The initial analysis of BVOCs from Year 1 was completed in 2020 and analysis of Year 2 data from the field experiment is currently being analyzed. The following summarizes primary findings for 2020: BVOCs collected from plant: the 2019 field data was analyzed using XCMS Online methodology, providing a complete metabolomics workflow including feature detection, retention time correction, alignment, annotation and statistical analysis. Results of this analysis can be summarized as follows: All samples collected in 2019 after inoculation were properly organized in three classes: Aflatoxin, Afla-Guard and Controlled. A Linear Discriminant Analysis (LDA) was run. In order to validate the model a Leave-one-out cross-validation method was used, meaning that N separate times, the model was trained on all the data except for one point and a prediction was made for that point. Even though the main model could predict three classes (Aflatoxin, Afla-Guard and Controlled) with 100% accuracy, it failed to reliably indicate the appropriate class on the validating set (only 29.2% accuracy). This was primarily due to the limited number of data points. BVOCs collected from collected pods: In addition to plant BVOC's, the VOC's from peanut pods collected in the field were evaluated. All pods were classified into three categories: In Shell (peanut kernel is still in the shell); Shelled (peanut kernel is taken out of the shell but not dried); and Dry Shelled (peanut kernel is shelled and dried). The training model predicts three classes (Aflatoxin, Afla-Guard and Controlled) with 100% accuracy for In Shell and Dry Shell categories and with 77.1% accuracy for Shelled category. Validating set correctly classifies those classes in 45.5%; 36.5%; 33.3% for Shelled, In Shell and Dry Shell categories respectfully. The model's poorer performance on validation set can be explained by the fact that the number of identified features (ions) that contribute significantly to classification model is much greater than the sample size. Further downselection of representative compounds will be beneficial to improve the model. To support this effort, we also pre-imaged pods collected in the field with hyperspectal camera imaging. We applied the machine learning technique, MT-MIACE to small groups of kernels to determine if we could detect aflatoxin. Preliminary results indicated that there was a separation between samples with higher levels of aflatoxin than those without. However, further refinement would require single kernel testing for validation, a chemical aflatoxin testing approach that has not been available previously. We worked with JLA Global in 2020 to develop a test for single kernel aflatoxin and this was successful. We will deploy this testing in 2021 in conjunction with our HSI. This approach will give matched BVOC, HSI, and aflatoxin levels data. Tasks 3: Evaluation of ARRM performance utilizing commercial lots of peanut seed lots. As stated above, data is being collected in Tasks 1 and 2 and will be used to address this task; therefore, progress is gauged by the accumulation of this data in preparation for developing integrated system models. To further accelerate the results for Tasks 3, we have entered into a collaboration with Premium Peanut, a large commercial sheller. Initial model testing was done on one of their large, multi-year data sets that allowed us to develop Machine Learning approaches for predicting risk going into storage from the field. This was the initial purpose of the Task and represents a completion. The information from Task 3 will then be merged in Task 4 with additional collaborations with Premium Peanut to refine model development. Task 4: integration of data across ARRM, BVOC drought risk in-field, BVOC levels at harvest, and BVOC/hyperspectral analyses after storage. Accomplishments 2020: We have designed a field project with Premium in 2021 that will employ a full system analysis of aflatoxin risk. We have identified three commercial fields with matched sites of high and low aflatoxin risk, identified irrigation status. Micro-climate sensors including soil moisture, canopy relative humidity, and soil temperature will be installed in six replicated locations within both the low- and high-risk areas of these commercial fields. This environmental data will be collected season-long. At harvest, samples from each location will be collected, bagged and installed with temperature sensors. These bagged samples will be placed in warehouse storage and data logged until off-loaded, approximately one year later. Aflatoxin testing of samples as they are collected in the field and then after storage will also be collected. This data set should give us a full-chain representation of aflatoxin risk and help identify and inform further model development that is resulting from research plot data.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2020
Citation:
Rowland, D., B. Tillman, A. Zare, X. Guo, Z. Brym, D. Sabo, O. Kemenova, X. Song, M. Navaei, C. Heist, D. Britton, and W. Daley. 2020 (invited). Improved system assessment of aflatoxin risk utilizing novel data and sending approaches at points of vulnerability. American Peanut Research and Education Society, Annual Meeting, July 14-16, 2021 (virtual).
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2020
Citation:
Song, X., D. Britton, W. Daley, O. Kemenova, D. Sabo, M. Navaei, C. Heist, A. Zare, S. McAmis, R. Heim, I. Geedicke, J. Giuliani, and D. Rowland. Sensing and Managing Crop Drought Stress in Peanut: Plant BVOCs as an Effective CWS Sensor for Aflatoxin Risk. ASABE, July 13-15, 2021 (virtual).
|
Progress 01/15/19 to 01/14/20
Outputs
Target Audience:Our target audience includes peanut producers, shellers, and manufacturers for the use of the model. We also expect scientific societies, colleagues in science, and graduate students to be target audiences for our basic scientific results related to sensor development, aflatoxin modelling, and image analysis techniques. Changes/Problems:Obtaining contaminated aflatoxin samples from local buying points has proven to be difficult due to the uncertainty of aflatoxin risk for a given season. In response we increased our collection efforts on samples from both greenhouse and field trials to maximize data obtainable. We also added the measurement of thermal images for both the greenhouse and field experiments to determine if thermal signatures could be related to BVOC production or general plant stress. What opportunities for training and professional development has the project provided?Opportunities for professional development have been provided to the graduate students involved in the project by the formation of a joint journal group discussing relevant papers addressing approaches that combine agronomic and machine learning techniques. Submission of abstracts by one graduate student (McAmis) associated with the project has been completed and the paper presented. How have the results been disseminated to communities of interest?For 2019, dissemination efforts have been aimed at the oral presentations of project approaches by McAmis. Project plans and goals were presented to the American Peanut Shellers Association and the Florida Peanut Federation. What do you plan to do during the next reporting period to accomplish the goals?In 2020, we will focus on the following: The field trial utilizing existing rainout shelters will be repeated in the 2020 season using the refined information from our greenhouse and controlled condition experiments regarding BVOC collection, model components utilized for predicting AT contamination, and the Aspergillus infection process. Continued aflatoxin analysis of the 2018 field experiment will be completed and added to the model being developed by Co-PI Brym Continued peak identification of BVOCs from green house plants and pod samples Continued statistical analysis of all BVOC data collected from green house and field samples and correlate it to aflatoxin concentration In-depth analysis of BVOC data from all samples collected to date.
Impacts
What was accomplished under these goals?
?Task 1: further refine and revise the aflatoxin regional risk model (ARRM) for peanut production in the north central Florida region that was developed in Year 1. Accomplishments Year 2: This modified task was 40% complete at the end of year 2. Rowland worked with grad student, Shannon McAmis, in collaboration with other faculty at UF to refine the existing model algorithms that are included in DSSAT. In short, we finished the model that McAmis assembled and refined with twelve years of data collected by the USDA's National Peanut Research Lab in Shellman, GA regarding yield and aflatoxin concentration in peanut under different irrigation regimes. This data is now included in McAmis' MS thesis to be internally published in May 2020. This data is now being utilized by Co-PI Brym to create an alternative model from the DSSAT ARRM version. This will be evaluated against the DSSAT model with regards to the accuracy of predicting aflatoxin in the Shellman data and the aflatoxin data collected in 2019 in the greenhouse experiment and field experiments of the project. Task 2: Biogenic volatile organic compound (BVOC) sensors will be deployed in research plots with drought and control conditions implemented. Accomplishments Year 2: To more carefully refine previous information collected on BVOC signatures, one controlled condition experiments were conducted to address this task: one conducted at the Plant Science Research and Education Unit in Citra, FL in a greenhouse environment and one in the field using a rainout shelter facility. 1. PSREU, FL Greenhouse: In the 2019 season two runs of a greenhouse experiment were conducted in order to precisely examine the effects of the water stress, soil water status and soil temperature of the pod zone on aflatoxin concentration in peanut. The trial was conducted at a greenhouse located in the Plant Science Research and Education Unit in Citra, FL. The genotype FloRun 331 was tested under three different end of season irrigation treatments, consisting of 100%, 66% and 33%; with treatments beginning at 100 days after planting. Plants were arranged in a randomized complete block design with six blocks for a total of 36 plants across the two runs. A special pot with pegging pan apparatus was designed separate the pod and root zone allowing for different water management of the two zones to provide better control over the conditions of the pod zone. The design consisted of a PVC pipe (15 x 84 cm, D x H) fitted with a cap at both ends and a plastic tote (61 x 38 x 15 cm, W x L x H) acting as a pegging pan attached to the top cap. Soil moisture at 4 depths was measured using the Profiler Probe (Delta T). To quantify water uptake over a daily period, soil moisture readings were taken at 6 am and 6 pm and the difference was assumed to be due to root water uptake by the plant. These daily uptake measures were completed weekly in both a wet (12-15 hours after irrigation) and dry (48-60 hours after irrigation) condition. To ensure the presence of Aspergillus, a strain of Aspergillus parasiticus was purchased from the ATCC and cultured under sterile conditions and then used to produce a spore suspension with a concentration of 106. The soil of each pegging pan was inoculated with 100 mL of the spore suspension using the drench method. Soil temperature in the pod zone was measured by loggers on an hourly interval. A weather station was installed in the greenhouse to measure other environmental conditions. Aflatoxin analysis was conducted at the end of season. 2. PSREU Field: A replicated, randomized field trial was conducted under a rainout shelter in the field at the PSREU in Citra, FL. In this experiment, 48 plots were allowed to grow under normal conditions until 100 DAP, after which the plants were treated with Aspergillus parasiticus (using the same strain as was used in the greenhouse experiment), Afla-Guard (a commercial product containing atoxigenic Aspergillus), and no soil treatment - acting as our control. Directly after inoculation, the rainout shelter was placed over all plots and irrigation ceased to induce aflatoxin production. Plant based BVOCs were collected from plants before inoculation, and weekly for 2 weeks post inoculation. SBSE were used for all collections and shipped to GTRI for analysis. After 140 DAP, the plants were harvested and the pods were sent to GTRI for VOC collection and analysis. In-shell pod BVOC samples were collected, as well as shelled samples. Once the BVOC samples were collected, the pod samples were shipped to JLA for aflatoxin concentration measurements. The initial analysis of BVOCs was as follows: collected data were split into four categories and coded as follows using Background (BG); Green (VOCs collected from plants that received a 100% irrigation); Yellow (VOCs collected from plants that received a 66% irrigation) and Red (VOCs that received 33% irrigation). Preliminary analysis has shown that Background VOCs can be identified using several classification techniques: Hierarchical Clustering, Principal Component Analysis (PCA) and Linear Discriminant. Figure 4 illustrated PCA analysis performed on data aggregated from all 3 weeks. Other categories (VOCs from 100%, 66% and 33% irrigated plants) depend on the collection date and should be analyzed separately. 3. Additional testing at GTRI (Co-PIs Daley and Sabo) as follows: collected BVOCs from plants and harvested peanut pods was completed using SBSE. For plant collections, we placed the SBSE directly on the leaf, held in place with a small pill magnet, for 2 hours. The leaf was isolated using a plastic lined aluminum bag. BVOC collection of pods centered on the use a glass vial large enough to hold the entire sample. SBSE was then desorbed into a GC/MS using a thermal desorption unit. We are currently processing the data and identifying what chemical signatures relate to peaks of importance. Task 3: Evaluation of ARRM performance utilizing commercial lots of peanut seed collected at regional buying points, with evaluation of samples utilizing BVOC sensor systems for in-shell peanuts; repeated evaluation of samples after 6 months of storage utilizing BVOC sensor systems and a hyperspectral imaging system. Accomplishments Year 1: Aflatoxin levels appear to be high in 2019 and we are working with Florida Foundation Seed to attain samples from their fields. However, we are continuing to utilize our own samples from the greenhouse and field experiments to image using hyperspectral pod imaging with subsequent BVOC testing and aflatoxin quantification. Pod samples were taken from the PSREU greenhouse and field experiments and imaged using a hyperspectral camera system (HinaLea). These samples were then sent to Co-PI Daley and Sabo for BVOC analysis. BVOCs were collected on all plants using Stir Bar Sorptive Extraction (SBSE) material as follows: BVOC collection of pods centered on the use a glass vial large enough to hold the entire sample; magnets held the SBSE to the top of the jar for 18 hours. The BVOCs were extracted from the SBSE and passed through a GC/MS for separation and identification. Once BVOC sample collection was completed, the pods were shipped to JLA for aflatoxin concentration analysis. ?Task 4: integration of data across ARRM, BVOC drought risk in-field, BVOC levels at harvest, and BVOC/hyperspectral analyses after storage. Accomplishments Year 2: As stated above, data is being collected in Tasks 1-3 and will be used to address this task; therefore, progress is gauged by the accumulation of this data in preparation for developing integrated system models. Co-PI Brym has begun this task with the Shellman data and revising the DSSAT model. This revised model approach can then be used for the remainder of the project data once completed. Therefore, this task will not be fully addressed until Year 3 after experimental results are collected to allow for full risk model development.
Publications
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2019
Citation:
McAmis, S.K., D.L. Rowland, B.L. Tillman, K. Migliaccio, K. Boote, and G. Hoogenboom. 2019. Refinement of an aflatoxin prediction model using field and greenhouse data to elucidate physiological mechanisms of aflatoxin contamination in peanut. American Peanut Research and Education Society Annual Meeting, Auburn University, Auburn, AL July 2019.
|
Progress 01/15/18 to 01/14/19
Outputs
Target Audience:Our target audience includes peanut producers, shellers, and manufacturers for the use of the model. We also expect scientific societies, colleagues in science, and graduate students to be target audiences for our basic scientific results related to sensor development, aflatoxin modelling, and image analysis techniques. Changes/Problems:Efforts under Task 2 were modified to include an additional research trial and to conduct the two trials in controlled conditions as opposed to field conditions. One was conducted in a greenhouse with the other conducted in a growth room. This allowed for more precise control over drought conditions, Aspergillus inoculation, and BVOC measurement as opposed the precision that was possible in a field environment. This will improve our field measurements to be conducted in 2019. What opportunities for training and professional development has the project provided?Opportunities for professional development have been provided to the graduate students involved in the project by the formation of a joint journal group discussing relevant papers addressing approaches that combine agronomic and machine learning techniques. Submission of abstracts by two graduate students associated with the project has been completed with the plan for attendance in 2019. How have the results been disseminated to communities of interest?For 2018, dissemination efforts have been aimed at the oral presentations of project approaches to scientific audiences throughout the state of Florida within the UF system as well as industry groups. Four presentations regarding the approach of the project were given during Year 1 at Gulf Coast Research and Education Center, Tropical Research and Education Center, Indian River Research and Education Center, and Mid-Florida Research and Education Center. Project plans and goals were presented to the American Peanut Shellers Association and the Florida Peanut Federation. What do you plan to do during the next reporting period to accomplish the goals?In 2019, we will focus on the following: Two runs of the greenhouse experiment at PSREU will be completed. A field trial utilizing existing rainout shelters will be conducted in the 2019 season using the refined information from our greenhouse and controlled condition experiments regarding BVOC collection, model components utilized for predicting AT contamination, and the Aspergillus infection process. Additional samples of aflatoxin contaminated and non-contaminated peanut samples from the 2018 season are currently being collected for hyperspectral and spectrometer analysis; BVOC analysis will be conducted on these same samples to allow for correlation of AT concentration, hyperspectral signatures, and BVOC signature. Additional fine tuning collection procedures for VOC capture using SBSE; including the Twister placement on/around the plant Repeat of a 24 hour collection to confirm information on when the best time to collect VOCs Concentrating techniques such as Teflon bags over leaves or glass vials. Development of a protocol for larger scale test that will involve 50 plants. Evaluation of the GTRI environmental chamber and set up to handle live plants.
Impacts
What was accomplished under these goals?
Accomplishments occurredunder each of the four tasks: Task 1: develop and test an aflatoxin regional risk model (ARRM) for peanut production in the north central Florida region. Accomplishments Year 1: This task was 90% complete at the end of year 1. Co-PIs Rowland and Migliaccio worked with grad student, Shannon McAmis, in collaboration with other faculty at UF to modify existing model algorithms for predicting field production of aflatoxin. In short, McAmis assembled and refined twelve years of data collected by the USDA's National Peanut Research Lab in Shellman, GA regarding yield and aflatoxin concentration in peanut under different irrigation regimes. Using this data, McAmis was able to modify and refine an existing model within the DSSAT system in collaboration with Drs. Ken Boote and Gerritt Hoogenboom at the University of Florida. This data improved the functioning of the model and provides a field-based model that can be utilized by researchers and other specialists for delivering aflatoxin risk assessments for growers and regional production systems as proposed in our project. Specifically, data collected from an experiment conducted at the USDA-ARS Multi-crop Irrigation Research Farm in Shellman, GA was used be used to evaluate the performance of the CROPGRO-Peanut-Aflatoxin module of the Decision Support System for Agrotechnology Transfer (DSSAT). The original purpose of this experiment was to examine different irrigation and crop rotations on peanut quality parameters and is described in Lamb et al. (2010). The experiment consisted of three irrigation treatments consisting of 100%, 75%, 50% and a rainfed control in later years the irrigation was changed to treatments of 100%, 66%, 33%. A total of 14 treatments were used consisting of the various years and irrigation treatments that aflatoxin concentration were measured. Task 2: Biogenic volatile organic compound (BVOC) sensors will be deployed in research plots with drought and control conditions implemented. Accomplishments Year 1: To more carefully refine previous information collected on BVOC signatures, two controlled condition experiments were planned to address this task: one conducted at the Plant Science Research and Education Unit in Citra, FL and one conducted at the GTRI in Atlanta, GA. PSREU, FL: In the 2018 season the first run of a greenhouse experiment was conducted in order to precisely examine the effects of the water stress, soil water status and soil temperature of the pod zone on aflatoxin concentration in peanut. The trial was conducted at a greenhouse located in the Plant Science Research and Education Unit in Citra, FL. Two different genotypes, FloRun 331 and an experimental true breeding line that is hypothesized to be tolerant to Aspergillus infection were used. Three different end of season irrigation treatment were planned, consisting of 100%, 66% and 33%, these treatments were to begin at 90 days after planting. Plants were arranged in a randomized complete block design with six blocks for a total of 36 plants. A special pot with pegging pan apparatus was designed separate the pod and root zone allowing for different water management of the two zones to provide better control over the conditions of the pod zone. To ensure the presence of Aspergillus, a strain of Aspergillus parasiticus was purchased from the ATCC. Soil temperature and soil water content in the root and pod zone were measured by loggers on an hourly interval. A weather station was installed in the greenhouse to measure other environmental conditions. Aflatoxin analysis was to be conducted at the end of season, however complications arose with the fertility and irrigation of the plants and the trial was not taken to the end of the season, so no aflatoxin analysis was done for the 2018 season. Trial will be repeated twice with initiation of both runs occurring in February 2019. GTRI, Atlanta, GA: A small scale pilot test of VOC capture and water uptake dynamic measurements in our environmental chamber was conducted in preparation for larger scale testing in the environmental chamber at Georgia Tech. There were several goals for this test that included: a. Evaluation of the speed at which the sand would dry out due to the presence of plants in the buckets. b. Testing and evaluation of various VOC collection techniques and exposure times using absorbent stir bars known as Twisters. c. Ability to grow and manage peanuts in the environmental chamber at Georgia Tech. Peanut seeds, provided by Dr. Diane Rowland with the Agronomy Department of the University of Florida. lab at UF, were planted in 5 gallon buckets using river sand. We planted 3 seeds per bucket, 6 buckets total. Due to the fact that the plants are growing in sand, all macro and micro nutrients needed to be added. Normal application of these nutrients at the recommended rates was found to be not adequate for most of the nutrients, as they were easily washed away due to daily addition of water. We have since found that a double application of fertilizer is sufficient to overcome this deficit. These plants were grown outside of the chamber until the plants reached 30 DAP. This time also allowed us to continue to evaluate the environmental chamber for temperature and humidity control. The plants appeared healthy before they were moved into the environmental chamber. During this trial, analysis methods were evaluated and refined. In order to evaluate the sampling techniques several experiments were performed. They included: 1. Qualitative analysis, particularly usingGERSTEL Twister®/Stir Bar Sorptive Extraction (SBSE). headspace extractions to establish the extraction efficiency for each standard on the target list. 2. Quantitative analysis? In summary, these experiments evaluated the speed of sand dry down due to plant presence and develop VOC and relative water content collection techniques. Additionally, this year we have shown the capability for SBSE and our GC-MS system to collect and identify standards based on peanut plant VOCs. Task 3: Evaluation of ARRM performance utilizing commercial lots of peanut seed collected at regional buying points, with evaluation of samples utilizing BVOC sensor systems for in-shell peanuts; repeated evaluation of samples after 6 months of storage utilizing BVOC sensor systems and a hyperspectral imaging system. Accomplishments Year 1: Aflatoxin levels were very low in 2017 regionally, so availability of contaminated samples at peanut buying points was scarce for analysis in 2018. Samples were collected from the Florida Foundation Seed Producers producer fields, the Williston peanut company in Williston, FL, and from UF research plots focused on drought treatments. Subsamples from these field collections were arranged for imaging with a hyperspectral sensor as follows: seven batches consisting of 20-25 of non-contaminated kernels and eight batches of aflatoxin contaminated kernels. For each batch, spectra was collected at 90 degree intervals with a replication of images at each degree five times. The MI-ACE algorithm was used to analyze spectral bands; 1-100 and 2001-2151 bands were removed for de-noising. The algorithm was trained and then tested with various cross validation combinations (2/3 training, 1/3 testing). Separation of aflatoxin contaminated and non-contaminated kernels was achieved with 100% accurate classification. Task 4: integration of data across ARRM, BVOC drought risk in-field, BVOC levels at harvest, and BVOC/hyperspectral analyses after storage. Accomplishments Year 1: Data collected in Tasks 1-3 will be used to address this task; therefore, progress is gauged by the accumulation of this data in preparation for developing integrated system models. Therefore, this task will not be fully addressed until Year 3 after experimental results are collected to allow for full risk model development.
Publications
|
|