Progress 10/01/23 to 09/30/24
Outputs
PROGRESS REPORT Objectives (from AD-416): 1. Develop dynamic, robust, and resilient cropping systems that integrate conservation and science-based solutions to improve short- and long-term crop production systems, increase input efficiencies, and provide adaptability for changing climate and supply chain shocks. 1.A. Evaluate cover crop management and soil amendment effects on diverse ecosystem services in dryland crop productions. 1.B. Develop new and modern cropping systems that reduce or eliminate production inputs, increase yield and profit, and enhance environmental health. 1.C. Developing plant science-based solutions to improve crop resilience to climate change. 2. Develop, expand, and deploy high throughput data acquisition and analytics systems and platforms for multi-faceted data streams to improve the sustainability and relevancy of agricultural production systems and ecosystem services with an emphasis on soil health, production inputs, water conservation, water quality, and greenhouse gas (GHG) emissions. 2.A. Identify region-specific environmental health indicators for emissions, soil biology, and nutrient uptake by utilizing high throughput sequencing, infrared greenhouse gas analyzers, and unmanned remote sensing systems. 2.B. Identify secondary, unintended effects on ecosystem services from agricultural practices, utilizing high throughput data acquisition and analyses. 2.C. Develop and evaluate a multi-sensor platform technology for within- the-canopy data collection and machine learning models for high throughput approaches to soybean, pea, and dry bean crop development. 2.D. Soil carbon and GHG monitoring at the farm scale using unmanned aerial vehicle (UAV) and unmanned ground vehicle (UGV) systems. 3. Advance engineering and computational technologies for cropping system best management practices and ecosystem services through innovations in precision agriculture, digital transformations, advanced hardware and software technologies, autonomous systems, computer vision, and artificial intelligence (AI). 3.A. Develop and evaluate AI-enabled techniques and systems for in-field monitoring of crop growth status by multisource remote/proximal sensing and meteorological observations to provide data and information to regulate the performance of the cropping and animal production systems. 3.B. Determine the �essential data resolution� needed to develop effective models to quantitatively estimate crop resiliency at the genotype, environment, management, and its interactions by utilizing statistical analytics, machine learning and artificial intelligence methods. 3.C. To develop a variety of microwave (radio frequency - RF) sensors from small Unmanned Aircraft Systems (UAS) platforms, evaluate their use for water utilization and yield estimation in irrigated and rainfed farming, and create artificial intelligence-enabled algorithms for UAS- based precision agriculture. 3.D. Create the next-generation predictive and prescriptive tools for selection and deployment of climate-resilient cultivars adapted to the region. Approach (from AD-416): Big data, artificial intelligence, and machine learning are powerful tools that rely on high quality data input, particularly when working with complex, interconnected datasets. Agroecosystems, and their sustainable production and maintenance of environmental health, are known for their interconnected complexities. The role that agronomic management plays in these systems is key towards feeding an ever-growing population, thus ensuring production for decades to come, particularly with increasingly volatile weather systems. Agroecosystems in the southeastern and northern U.S. are the economic platform for a largely agriculture- based society, with a key focus on corn, wheat, soybean, cotton, and animal agriculture. To maintain ecosystem health and to extract all yield potential requires an understanding of all, or many, of the various systems at work (e.g. biology, chemistry, and physics). To maximize the potential of these systems, we must employ every relevant tool, such as fertilizers, cover crops, industrial byproducts, and the confluence of these inputs. No single agronomic plan is a fit for every region, thus this project plan aims to study disparate regional systems to develop best management plans addressing multiple regional conditions such as soil structure, weather, and availability of agronomic inputs. This �systems based� plan addresses the problems, solutions, and impacts of modern agroecosystems. We address these problems via dynamic, data-driven acquisition of large amounts of multi-faceted data streams. Farm, field, experiment station, and laboratory/greenhouse-based experiments will be employed to address these issues. The project datasets comprise biology, chemistry, physics, emissions, remote sensing via unmanned ground and aerial vehicles utilizing the latest data acquisition technologies. Big data analysis and management provide readily accessible data to the public at large, which facilitates transparency and further collaboration. This project plan brings together a large team with varied expertise to develop novel, flexible, and targeted best management practices for sustainable agricultural systems. Objective 1: Research continued at Mississippi State, Mississippi, to address management methods and their effect on yield, resilience, and soil health. Sub objective 1A: Broiler litter was applied following the fall harvest and cool-season cover crops (CC) were planted. Leachate water was collected after each rain event and analyzed for nitrogen (N). The CC were terminated using chemical and rolling/crimping methods. Above-ground CC biomass was collected for dry matter and N. Soil samples were taken and analyzed for pre-planting available N. The decomposition rate of CC residues during growing seasons was monitored in situ using the Litter Bag Technique. Soil moisture and temperature were continuously recorded. Additionally, plant characteristics were measured along with corn and cotton yields. Finally, after the harvest, soil physical, hydrological, chemical, and microbiological parameters were assessed. Coal char, biochar, and broiler litter were applied along with winter CC. Leachate water was collected after each rain event and analyzed for nutrient content. The CC was terminated chemically and by rolling/ crimping. Prior to termination, above-ground biomass dry weight was recorded. Cotton was planted and plant characteristics were measured along with yields. Post-harvest soil samples were collected and analyzed. Sub objective 1B: Edge-of-Field runoff and subsurface drained water were collected and analyzed from a field in Brooksville, Mississippi. Five field strips were selected within the farm and applied with broiler litter. Additionally, soil amendments were applied atop the broiler litter, both in the presence and absence of winter CC. The field strips were equipped with water sampling units. Automated runoff was collected and recorded and analyzed for water quality components. The 4 selected CC were drill-planted using a custom-built grain drill. In February 2024, the cereal rye had a good stand establishment. Of the legumes, only crimson clover and vetch had some emergence but not the others due to drought. Corn was planted into the strips and corn and cover crop grew simultaneously. This feature allowed the CC to accumulate biomass, providing additional N. The CC were killed after taking CC biomass. Preliminary results show that corn received more N from the vetch and crimson clover. Cereal rye suppressed weeds. A study was carried out at two locations with different soil types with low native soil potassium (K) levels. Cotton was fertilized with variable rates of K and poultry litter. One of the fields is a no-till field and the other is a conventionally tilled field. All plots received comparable N and phosphate fertilizations so that cotton is affected only by the level of applied K. Sub objective 1C: This study screened different CC subjected to stress treatments to affect plant traits. The biomass of CC species varied across the treatments. Dry conditions negatively affected root and shoot biomass. Brassicas had the highest shoot biomass, followed by legumes and cereals. Winter peas had similar shoot biomass as purple top turnip and hairy vetch and produced high root biomass. Hairy vetch had relatively high above-ground biomass in wet conditions. Objective 2 Research continued at Mississippi State, Mississippi, to address unintended effects on agroecosystems. Sub objective 2A: The portable mid-infrared leaf scanner system attachment clip was designed and developed. The current clip is attached to a spectrometer via fiber-optics to capture leaf spectra. The leaf clip design was optimized to find the best optical arrangement and is being tested. This leaf scanner will later be used to obtain leaf-level spectra to serve as ground truth data for unmanned aerial vehicles (UAV) imaging, to be used to build nutrient deficiency detection models. Samples were collected throughout field plots identified from Objective 1 for moisture content, nutrient status, and DNA extraction and genetic analysis. DNA was processed from prior year samples. Soil enzyme analysis was conducted on samples. Soil health genes assessed from a controlled greenhouse experiment were modeled using artificial intelligence (AI) and machine learning (ML) to predict gene levels in the root zone soil of cotton plants. Soil health genes, enzymes, and the microbiome were assessed from localized dry spot in grass systems. Greenhouse gas emission measurements began in the corn/cotton rotations of objective 1. Measurements included nitrous oxide (N2O) and carbon dioxide (CO2) flux complimenting other discipline measurements. Biweekly N2O and CO2 flux was measured for coal char and biochar treatments. Analysis is underway of the UAV-CO2 data. Sub objective 2B: Wildlife fecal DNA was collected throughout Mississippi and were analyzed for the presence of antimicrobial resistance genes and other bacteria and pathogens as part of the Salmonella Grand Challenge. Samples collected from vultures, wild turkey, and duck were processed. New beetle traps were created to capture beetles from the environment. Sub objective 2C: Options for the sensor platform including the camera model, quantity, and orientation were considered for a single row, bicycle-style, hand-pushed platform which was developed carrying an Intel RealSense D405 camera. Soybean plants facilitated preliminary testing and evaluation. Using this system and YOLOv8 showed that video is a better option for real time pod counts for the AI based pod count. Sub objective 2D: The proposed UAV-greenhouse gas prototype was further advanced with a CO2 sensor mounted to an aircraft selected for its max capacity and a sensor selected to maximize sensitivity and accuracy. Additionally, we mounted temperature, humidity, air pressure, windspeed, direction, dual band GPS, and laser rangefinder. Using this prototype, we conducted regular in-situ flights over corn and cotton cropping systems. Objective 3 Research continued at Mississippi State, Mississippi, and Fargo, North Dakota to utilize AI and ML techniques to augment smart agriculture. Sub objective 3A: Complete system designs for data and algorithm integration with preliminary system prototype to include standalone and internet-based spatial visualization of the remote sensing image products. Algorithm integration includes various machine/deep learning algorithms and their deployment from data processing, algorithm implementation, result analytics, and model reuse. Crop and weed multi-species imagery data were successfully collected for AI algorithm development. The robot platform was successfully updated with solar panels and a spraying tank for weed control. Soil sites were identified. A soil probe was successfully designed and installed on the robot platform and is ready to conduct field research. Sub objective 3B: An open access Python library was developed to create and visualize vegetation indices from UAV collected hyperspectral and multispectral data. Unsupervised ML models have been implemented and we are currently in the process of testing them on UAV data. A new data management environment to support data collection and modeling tools was developed and is available. The environment includes a web hub to consolidate login access to field experiment planning, data storage, querying and analysis tools in one place. Data collection from barley and dry bean breeding trials commenced and will be used for beta testing. Sub objective 3C: Daily UAV-based flights were performed to collect Global Navigation Satellite Systems-Reflectometry (GNSS-R) data. Five soil moisture probes were added to the outside of the study field. Measured soil moisture and GNSS-R from this area will be used for radio frequency signal power normalization. In addition, a static GNSS-R receiver was placed in the open area to measure the direct signal level. We continued to develop Ground Penetrating Radar based UAV and unmanned aerial systems. A GPS unit was added to synchronize collected signals and geolocations. To increase the penetration depth, an RF power amplifier was added. Sub objective 3D: The plant breeding programs were successfully transitioned over to the commercial Genovix software. Tablet computers were integrated and deployed for field collection of agronomic and quality data. A beta version of PredictPro was created to run genomic prediction models which includes multiple functionalities. Development was continued on the initial version of a software tool called AgSkySight which provides an efficient workflow for UAS data handling. Data collection for the current project plan utilized ARS data integrity as a core objective. Raw data collection implementation is the major focus which focused on 3 collection engines. The ARS �Farm Management� and �Research Tool� apps were used for field data collection; historical data collection remains a challenge. A relational database for laboratory data collection and connection with Microsoft Azure platform is planned. Five years of raw data collection for the unit is almost complete. Site Scan for ArcGIS from Esri was evaluated as a raw drone data collection system. Artificial Intelligence (AI)/Machine Learning (ML) Machine learning (ML) and Unmanned Aerial Vehicle (UAV) imagery were used to streamline the cotton breeding process by reducing manual labor associated with phenotypic measurements, typically taken by hand. Artifical Intelligence (AI) and ML were successfully utilized to predict soil moisture in a field using imagery-processed soil moisture data, validated with traditional ground measurements. Going forward, data from UAV sensors and in-situ observations will be analyzed using deep-learning models. Algorithm integration is being developed to identify and deploy AI algorithms from ML and deep learning algorithms to be used for modeling and analysis of crop field experiments. This is in conjunction with data integration which designs and prototypes a functional standalone/internet- based computing platform for big data management, visualization and interpretation of multisource remote sensing data, especially UAV remote sensing data. AI methods were used for deep learning algorithms (YOLOv8) for object detection (soybean pods) using the transfer-learning technique, and for Convolutional Neural Network (CNN), Visual Group Geometry (VGG16), and Residual Network (ResNet50) deep learning architectures were used to build weed classification models. AI benefited this project by allowing real-time object detection (soybean/ dry beans pods) and counting with acceptable accuracy, which can be used for yield estimation. Such a task would be unfeasible to carry out manually, given both the number of pods per plant and the number of plots in the field. Such an approach will improve efficiency in the breeding programs and assist breeders to identify the best-performing genotypes in their trials. AI also enabled technology for site-specific weed management in real-time in both greenhouse and field conditions. Weed and crop plants must be accurately classified. ACCOMPLISHMENTS 01 Hyperspectral data sets can be used to predict soil health genes. The soil microbiome, and particularly a disturbance in the microbiome leading to �unhealthy� soil, is difficult to predict given the inherent biases associated with spatial sampling. However, it is well understood that subtle differences in the human microbiome can mean the difference between healthy and not healthy; the same can be true of the soil microbiome. ARS researchers at Mississippi State, Mississippi location, in conjunction with Mississippi State University have successfully utilized hyperspectral data collected form water stressed cotton plants grown in a greenhouse to predict soil health genes in the rootzone. Water stress induced a spike in the presence of genes for general bacteria and fungi, as well as genes involved in biogeochemical processes. During the peak of water stress, the hyperspectral data was successfully modeled utilizing ML algorithms along with the empirically collected soil health genes to predict gene levels in the soil. Predictions were successfully validated, and thus the model data set has potential for further application at the field level and ultimately to provide remote sensing diagnosed soil microbiome problems in the field. 02 Analytics application developed for complex �OMICS data. User friendly analysis of complex genomic or phenotypic data is necessary to push - agricultural science forward. Scientists at the North Dakota State University, through a cooperative agreement with ARS researchers at Mississippi State, Mississippi location, designed a beta version of a new automated tool, PredictPro. This tool is a user-friendly analytics software application designed to make sense of complex OMICS data (i.e., genetic, phenotypic, phenomics and other types of OMICs data) for utilization in plant breeding programs. The initial capabilities in this beta version include metadata storage; data ingestion and validation processes; multi-omics model creation; validation result diagnostics; visualization and reporting functionalities. 03 Rye cover crop reduced soil nitrous oxide flux during winter. Agroecosystems account for the largest source of anthropogenic nitrous oxide (N2O), a powerful greenhouse gas (GHG). Cover crops (CC) generate many positive effects on soil health but have inconsistent impacts on GHG emissions. ARS researchers at Mississippi State, Mississippi location, investigated seven cover crop treatments following a corn rotation in a no-till field, and found that Elbon rye reduced soil nitrate the most and soil nitrate correlated to N2O flux. This is the first study to consider CC effects on soil N2O emissions in southeastern United States cropping systems during the cover crop season and will be the basis for prescribing comprehensive resilient crop management strategies. 04 Effect of vegetation and surface roughness on Global Navigation Satellite Systems-Reflectometry (GNSS-R) base soil moisture estimation. Smart irrigation should apply proper water to the right place at the right time. Measuring soil moisture is the first step in this process. UASs equipped with microwave radar systems (such as GNSS-R) can perform these measurements quickly and with high resolution. ARS researchers at Mississippi State, Mississippi location, through a collaborative agreement with ARS at Mississippi State, Mississippi, have analyzed how the main crops (corn and cotton) attenuate the measured signal amplitude. Signal attenuation was correlated with plant height and normalized difference vegetation index (NDVI). The effect of surface roughness was shown on signal measurements. The study indicated that rough surfaces cause very high signal attenuation. However, surface roughness is mostly uniform across the experimental field, which means the attenuation effect will be the same throughout the study field. These analyses provide essential details for accurately estimating soil moisture. For GNSS-R-based soil moisture mapping, researchers can measure surface roughness at the start of the study and use the same correction factor throughout the season. The height of the crops or the vegetation index is another factor that can help correct soil moisture estimations. 05 Agricultural remote sensing data integration and algorithm integration. Remote sensing data integration provides a big data hub to process, organize, visualize, and interpret multisource remote sensing data, especially Unmanned Aerial Vehicle-based remote sensing images. ARS researchers at Mississippi State, Mississippi location, designed a system on a national geospatial framework to create and manage a project in any farmland. Accordingly, a Machine Learning (ML) algorithm integration is synchronized to provide user-friendly, high-performance modeling and analytics with remote sensing data and ground-truth measurements. The system is designed to use ML algorithms through benchmark statistical analysis in conjunction with remotely sensed images and in-situ data collected from various crop field planned experiments. This development will significantly improve the agricultural remote sensing data management for timely applications and greatly facilitate use of machine/deep learning to accurately characterize and analyze crop growth processes with other factors in management and environment for researchers and stakeholders. 06 Sustainable agricultural practices that optimize cotton growth and yield are increasingly crucial. Growers need viable options to utilize available resources effectively, improving soil health and productivity while maintaining profitability. Insufficient soil moisture and available nitrogen (N) are major limitations during cotton�s critical growth stage, resulting in reduced yields and economic losses. ARS researchers at Mississippi State, Mississippi location, in collaboration with researchers at Mississippi State University, evaluated the effects of tillage and fertilizer sources with cover crops on dryland cotton. Their findings revealed that long-term integration of organic fertilizer with cover crops in a dryland no-till system enhanced soil water storage capacity, increased available soil N, and improved cotton yield. These results can be applied in regions where cover crops can be grown during winter to reduce N leaching. Additionally, minimizing tillage intensity and N fertilization rates can lower energy costs while optimizing crop production, supporting growers� returns, and promoting ecosystem quality. 07 Cereal rye and mustard cover crops in conjunction with poultry litter improved aggregate stability. Improving aggregate stability in degraded or agricultural soil helps many aspects of soil health, including water holding capacity, biological functions, and nutrient stability. ARS researchers at Mississippi State, Mississippi location, showed that cover crops of cereal rye and mustard mixed with cereal rye, and poultry litter addition increased soil aggregate stability and mean weight diameter. Soil bulk density and pH were the most important predictors of soil aggregate stability and size distribution. High pH restricted soil aggregate stability and size distribution. These improvements demonstrate the effectiveness of multiple agronomic approaches to improving soil health, specifically aggregate stability.
Impacts
(N/A)
Publications
- Chang, T., Feng, G.G., Adeli, A., Brooks, J.P., Jenkins, J.N. 2023. Soil health assessment for different tillage and cropping systems to determine sustainable mangement practices in a humid region. Soil & Tillage Research. 233:1-14. https://doi.org/10.1016/j.still.2023.105796.
- Hu, J., Miles, D.M., Adeli, A., Brooks, J.P., Podrebarac, F.A., Smith, R.K. , Lei, F., Li, X., Jenkins, J.N., Moorehead Ii, R.J. 2023. Effects of cover crops and soil amendments on soil CO2 fluxes in Mississippi corn cropping system on upland soil. Environments. 10(2):19. https://doi.org/10. 3390/environments10020019.
|