ON-FARM PRECISION AGRICULTURE RESEARCH TO DRIVE VARIABLE RATE OF INPUTS

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: HATCH
• Reporting Frequency: Annual
• Accession No.: 1018721
• Project Start Date: Feb 19, 2019
• Project End Date: Jan 31, 2023
• Program Code: [(N/A)]- (N/A)

Recipient Organization
LOUISIANA STATE UNIVERSITY
202 HIMES HALL
BATON ROUGE,LA 70803-0100

Performing Department
School of Plant, Environmental, and Soil Sciences

Non Technical Summary
The use of drones, active sensors and soil EC meters are growing for a specific niche of high technically well trained farmer that have consultant advice, but even them have a lot of questions about how to use the data from those sensors to transform in agronomically management practice. Some researchers believe that Precision Agriculture (PA) will use new tools with combination with common knowledge in crop production. However, it seems that the end users think that agronomic knowledge have to be refined to support a more efficient and high technology-driven agriculture. The experiments have to change place from research stations' small-scale to large-scale farmer's fields, but with adequate design and procedures to serve from the small to the large farmer. Currently there is a strong need to improve how to make agricultural experimentation in the field to generate agronomic parameters to support PA practices and have different statistical analysis to be adopted to extract better information to be used by farmers, consultants and stakeholders. In this context the objective of this project are: (i) Conduct spatial experimental designs and procedures to collect large amounts of spatial data on-farm to drive variable rate of inputs; (ii) Incorporate plant and soil spatial variability in to agricultural experimental research using proximal soil sensors and remote sensing; (iii) Conduct a long term experimental field using a multidisciplinary Precision Agriculture approach. The expected products and outcomes from this project are: i) Generate agronomic parameters on-farm to drive variable rate technology to gain efficiency saving the environment; ii) Incorporate spatial layers of information to the LSU variety trial to support selection based on spatial variability to improve yield; iii) Generate a big dataset using a multidisciplinary precision agriculture approach to support decision making, crop modeling and deep learning techniques to increase farmer's profit.

Animal Health Component

80%

Research Effort Categories

Basic

0%

Applied

80%

Developmental

20%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
205	7210	2061	50%
205	5310	2020	50%

Knowledge Area
205 - Plant Management Systems;

Subject Of Investigation
5310 - Machinery and equipment; 7210 - Remote sensing equipment and technology;

Field Of Science
2020 - Engineering; 2061 - Pedology;

Keywords

digital agriculture

precision agriculture

cloud computing

deep learning

remote sensing

sustainable agriculture

Goals / Objectives
(i) Conduct spatial experimental designs and procedures to collect large amounts of spatial data on-farm to drive variable rate of inputs.(ii) Incorporate plant and soil spatial variability in to agricultural experimental research using proximal soil sensors and remote sensing.(iii) Conduct a long term experimental field using a multidisciplinary Precision Agriculture approach.

Project Methods
Farmers' fields will be selected to implement a specific spatial statistical design of large plots avoiding interference on producer's routine in season. Different than traditional plots, these large plots (planter width by 40 m) will be made with high precision GPS and GIS software's and implemented with variable rate machinery to apply the treatments. This means no flags, alleys or labor establish plots in the field will be needed. Before planting or after harvest the farmer field will be mapped for soil EC using a geophysical equipment at 3 different frequencies 3, 5 and 15 KHz or a contact rig with disc blades (examples are EM-38, Profiler and Veris), mainly to map soil organic matter, texture, CEC and elevation.During the growing season one image using quadcopter drones with multispectral cameras (near infrared - NIR, red and red edge) will be collected in a strategic growth stage to acquire vegetation indices maps (passive remote sensing). Active crop canopy sensors will be mounted on high clearance machines to map the whole field for a series of variables (normalized difference vegetation index - NDVI, normalized difference red-edge index - NDRE, plant height, leaf area index - LAI, canopy temperature) as a ground truth for the aerial image. By doing this, two different methods of remote sensing will be implemented, (active sensors, that do not require sunlight to gather reflectance from target) vs. passive sensor. The final yield will be mapped with commercial combines equipped with yield monitoring systems with GPS. The experimental design to be tested will be a 3x3 Latin square, for example, 3 N rates and 3 seeding rates. The idea is to have several repetitions of the treatments laid down in the same direction of the planting track crossing the whole field spatial variability. This can be accomplished also with randomized complete block design (RCDB) delineating blocks consecutively in the field, and this will be also tested depending on field size (Figure 1). With Latin square design, it is possible to make geostatistical analysis to generate maps and compare treatments across field spatial variability. All statistics will be analyzed using R software and QGIS.This design should be the same in all farmers' fields to have a systematic way to generate, collect and process data. Other than the traditional Latin square design, this procedure allow spatial analysis using geostatistical techniques to generate map comparisons between treatments in the same area, making sure that we have at least 15 points in the grid with each treatment for a good semivariogram modelling for data interpolation to generate maps by kriging. For example: you can compare uniform rate (producer practice) with variable rate of seeds or fertilizers/lime in the same field. In this case we can have 3 scenarios, farmer practice, variable rate 1 and variable rate 2. The major outcome after each season will be a big data stored accessible by cloud to the cooperator farmer and a web based platform to provide spatial information's and spatial analysis to support crop management decisions that consultants and farmers can use.LSU Ag Center is conducting on-farm trials yearly in farmers' fields comparing hybrids, varieties and cultivars regarding yield at several locations. The information generated is mainly the performance in terms of average yield in a region and specific cropping system in Louisiana. Several data have been collected and there is a need to incorporate more layers of information to these experiments to improve the analysis and delivery of products from this effort. The major idea is to map strip trials (core blocks) conducted on-farm regarding soil and plant characteristics using modern technologies available, such as drones, soil sensors, and optical plant canopy sensors. Soil spatial variability will be mapped using apparent soil EC one time during the growing season. Plant canopy information will be mapped with unmanned aerial systems or drones equipped with digital camera and active crop canopy sensors using 3 band sensors (NIR, red and red edge), in a strategic time during the growing season. The advantage to have more georeferenced layers of information on top of traditional experimentation is to be able to extract more information from such high-cost and labor-intensive experiments, sometimes even improving the analysis and outcomes from the same experiment.The experiment will be initiated next growing season and conducted intwo fields, one field where the farmer have soybean and corn rotation and another field with soybean and cotton rotation. The only requirement is that the farmer's combine harvester has to be equipped with yield mapping monitors and is conducting core block trials with LSU AgCenter. The treatments will be varieties by soil EC, NDVI and soil type. The generated data will be analyzed using traditional multivariate statistics such as principal component analysis and new techniques of machine learning, e.g. convolutional analysis. The final output is a database with varieties response in different zones of EC, NDVI and soil type. A secondary objective is to build functions to predict soil properties that drives yield and how different varieties are influenced by spatial variability. Based on this information by zones, the consultants and farmers can select an appropriate variety that fits their needs to build prescription maps.A long-term experiment establish at the LSU Agcenter Central Research Station (Ben Hur Farm) during Spring 2018 is being conducted to simulate a farmer field with a 4x4 Latin square, with 4 N rates (0, 45, 135, 200, 225 kgN.ha) and 4 plant populations (depending on hybrid, but the treatments will be -20%; -10%; rec and +10%). The statistical analysis will be the same as Objective I, to develop a better procedure to evaluate spatial statistical designs. The difference is that the whole field is going to be full of Latin square treatments and will be a long term study to simulate farmer's practice, including rotation and practices (Figure 2). This procedure will help us to study various types of spatial configurations to compare treatments considering spatial variability of several layers of information. There are 21 points in a regular grid for each of the 20 treatments that will be modelled and interpolated in Vesper using Local and Global Kriging depending on the data collected.The reason why a small plot size-experiment was established is to have ability to map several layers of information conducted by a multidisciplinary team working together in the same field to integrate faculty member interactions and expertize. It is expected to map several attributes in this field such as weeds, soil compaction and fertility, diseases, insects, etc. The field will be used also to calibrate APSIM model and test new prototypes and apparatus developed before they become commercial and easy to use on a larger scale in farmers' fields. The same equipment of Objective I will be used, but also new technologies will be tested on this field. Another reason to have a small plot study at Ben Hur is that consecutive readings during growth stage will be needed for some new sensors and logistically Ben Hur Farm is close to LSU Campus for demonstration purposes to students and farmers. Weed seedbank, soil microbiology diversity, soil bulk density and diseases will be mapped using grid sampling approach and drones, trying to correlate these different layers of information in a map based approach.

Progress 10/01/19 to 09/30/20

Outputs
Target Audience:The target audience is farmers, commodity groups, crop consultants, scientists and students that adopt precision agriculture technologies to improve crop management, but also small farmers that can benefit from publicly available remote sensing products to improve crop yield. With an extensive on-farm data collection due to the nature of precision ag research in which spatial variability is needed, direct contact with farmers, consultants, employees, and commodity groups was essential to run the project. Efforts to deliver science-based knowledge included formal classroom instruction teaching AGRO4002 - Precision Agriculture and AGRO4002 - Data-driven Farming (planned for Fall 2021), Servers for cloud computing and coordination of the Digital Ag Program at LSU. Changes/Problems:Unfortunately, due to Covid19 imposing limitations on travel by our staff we could not expand the procedure for our county agents to help with drone and soil mapping. Instead we decided to run two large trials in one reference farm and contracted a third-party drone service to fly in other fields What opportunities for training and professional development has the project provided?During this period, with the funding provided by LSU we created a Digital Agriculture Laboratory with powerful servers to perform cloud computing and storage of remote sensing products and precision ag vector data. The servers are mainly being used to train students in this area of expertise now generically called Digital Agriculture. We created the first annual Conference to be held by the Louisiana State University AgCenter with the objective of promoting the practical use of precision agriculture approaches. The main point of this event is to have a theme with a keynote speaker and interactions with farmers, consultants, and commodity groups. In this meeting, interactions with Extension agents and users of commercial platforms were stressed. This next year the training for extension agents and consultants on specific topics will be again promoted as was requested in the first meeting. Parallel meetings were performed mentioning the support from NIFA and other budget sources. The main ideas in this Hatch project are being carried out using several different sources of financial support, providing an opportunity to train students at all levels. A couple of interns and students graduated or are in training (2018 - Roberto Moreira; 2020 - Issa Flores, 2020 - Alex Munoz; Ongoing: Murilo Martins, Felippe Karp, Fagner Rontani, Phillip Lanza). A couple of them were awarded financial support from companies and during the first virtual ASA-CSSA-SSSA meeting in 2020 one student was awarded 3rd place in Graduate Student Competition. How have the results been disseminated to communities of interest?The results have been disseminated by field days, data collection in farmers' fields, planning meetings with consultants, farmers and companies to make the project operational and cost effective. Scientific papers and meetings such as ASA-CCSA-SSSA, Beltwide Cotton Conferences; Brazilian Cotton Congress, etc. What do you plan to do during the next reporting period to accomplish the goals?For the next reporting period the focus will be scientific publications and a strong target to the non-scientific community, such as extension circulars. A web and searchable platform such as the one run by University of Nebraska- Lincoln is being developed to enable producers to retrieve what is being done by the LSU Precision Ag Team and the idea is to have a platform for research and extension tied together to make more efficient the data and information transfer. Use of videos, graphs and other digital forms will be improved to expand extension efforts. Adjustments must be made to automate data collection on farm and extension personnel must be engaged and trained to absorb this new wave of Digital Agriculture techniques. As the Chair of the Precision Agriculture Systems Community in ASA-CSSA-SSA I will enhance and promote symposium and workshops with all these topics on the Agronomy meetings in 2021 to be held on Salt Lake City - Utah.

Impacts
What was accomplished under these goals? (i) A standard spatial design was developed and adopted to capture not only traditional treatment responses using a Latin square design from a field experiment, but also spatial data. Treatments were arranged spatially in such a way that the range of the semivariogram or the distance over which auto spatial correlation exists can adequately model the interpolated map surface and the physical separation of different treatments in the farmer's fields to allow a kriging interpolation to generate a raster map for each treatment response. Preliminary results showed that the design was appropriate to run on-farm PA experiments but a minimum of 17 acres is needed per treatment to be able to interpolate using kriging and produce a reliable map. We could determine the optimum plot size based on the characteristics of farm equipment errors. For example: The smallest plot for on farm trials should be at least three passes of the planter by 45 m long and 2 passes of the N fertilizer applicator by 45 m. Plot size smaller than this result in excessive experimental error due to operational issues during planting and 45m is the minimum distance that grain or cotton yield monitors can give a precise estimation of crop yield, since speed of harvest, grain elevator flow, and grain moisture directly affect the impact plate sensor, leading to inaccuracies. The best way to generate reliable data is to harvest at a fixed speed, with no variation in grain moisture, with constant flow of then grain elevator with constant flow. However, these conditions are very hard to keep constant during harvesting. Making plots longer is the only way to decrease coefficient of variation at the plot level. In plots in the LSU On Farm Network, the minimum plot size is 48 x 48m, and this plot size is used by the majority of farmers in our network in Louisiana. The smallest plot that can be used is the width of farm machinery implement. Smaller plots than this can introduce considerable errors in average yield per plot and fertilizer and seed rate to keep an optimum range when variable rate equipment is used to implement the experiment. Regarding experimental design the most significant aspect found was that the spatial location of plots is very critical and influence results more than the treatment itself. So, our semivariogram analysis using each treatment separately in very large fields showed that the plot size cannot be increased too much because of loss of model fitting and a nugget effect can compromise treatment interpolation as discussed before. On the other hand, if plots are smaller than 48 m the errors due to limitations in farm machinery introduce errors that surpass treatment effects. These findings are being analyzed to be published in a peer reviewed journal and they will benefit several initiatives that are starting in this area, perhaps being a benchmark on how the plots should be planned to match practical operational issues in on farm experimentation. (ii) To accomplish this second objective several drone flights using multispectral cameras and soil mapping using sensors were performed in the same locations in which LSU core block variety trials have beenconducted since the late 1990's. For the sake of this project we are adding LIDAR elevation DEM and NDVI before flowering using drones in the same trials that are already conducted by LSU. In 2019 we found that yield data is not the only parameter that should be used to rank a variety on a farm, since the plant and soil spatial variability can introduce a bias in the average of long strips harvested with a certain variety/hybrid. Variations of more than 30% of coefficient of variation were observed in yield monitor data inside the same strip. Part of this variance is explained using raw yield data without filtering for errors inherent from yield mapping systems and the other part is due to the plant and soil variability. Varieties that are more flexible, yielding well regardless of spatial variability along the strip (that usually is long and crosses the field from side to side) were found. Varieties were also found for which yields decreased or increased considerably in different soil types with high organic matter or concave areas even on leveled fields. In 2020 we decided to incorporate SSURGO soil type data with drone flight as a proxy for soil sensor mapping on LSU core blocks trials to be able to cover more trials. The results are promising for those farmers that have variable rate equipment and for those that want to select the right variety for a certain soil type that is predominant based on the spatial variability encountered in their fields. Similar results across years using different varieties were found. Even using larger fields, spatial variation in soil fertility where the trials were located had less effect on yield than the inherent yield for each variety, except for parts of the field where irrigation issues, nematodes and other crop establishment issues were present. The correlations between NDRE (NIR and Red Edge band) and yield were smaller than in 2019, showing that even with spatial variability in physical properties such as texture, 2 soiltypes and elevation were smaller compared to the field in 2019. There was an indication that variability in soil fertility is an important spatial aspect to be considered in a high-resolution experiment. Using SSURGO and Sentinel imagery only does not have ability to capture this small-scale spatial variability for variety selection purpose. This effort is being heavily funded by the Louisiana Soybean and Small Grains Research Board showing that there is a great demand by our producers in the state. The 2020 dataset is still under analysis to compose a large database on cloud to run machine learning algorithms. Benchmarks with Texas A&M Agrilife group led by Dr. Juan Landivar and Dr. Murilo Maeda are being initiated to increase the collaboration in this area of variety research, drone imagery and modeling.(iii) An experiment was conducted in 2018 and repeated in 2019 and 2020. Several sensors were used, such as Holland Scientific Phenom, Rapid Scan, Talon 1 and Raptor 2, the latter of which is the only active sensor made commercially to be mounted in drones to collect active sensor data. The biggest difference from our remote sensing program compared to others is that most drones collect passive data that requires sunlight and images have to be calibrated for time of the day, cloudiness, and radiometric and atmospheric issues. Our group was reported by Dr. Kyle Holland and Dr. James Scheper's as the first group to collect data and perform variable rate N fertilization in cotton using a Raptor 2 drone-based active sensor. Lately this initiative is being funded by the Patrick Taylor Foundation with a project called "Implementation / Demonstration of Best Management Practices on Model Farms in Louisiana"; Louisiana Rice Research Board and Soybean and Small Grains Research Board. These 3 years we had collaborations with different departments, research stations and several faculty from LSU (Dr. Dustin Harrell, Dr. Brenda Tubana, Dr. Stephen Harrison, Dr. Syam Dodla, Dr. Randy Price; Dr. Thanos Gentimis) and from abroad (Dr. Carlos Silva Junior, Dr. Christian Bredemeyer; Dr. Ziany Brandao). For 2021, this effort will be transferred to a reference farmer that conducts on farm trials every year using precision agriculture equipment for implementation to capture natural spatial variability at a larger scale not encompassed inside the field sizes normally used at LSU Experimental Stations.

Publications

• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: SHIRATSUCHI, L.S.; BRIANTE, P.; BRANDAO, Z.; BULLOCK, D.; SILVA JUNIOR, C.; RAMOS JUNIOR, E. On-Farm Precision Experimentation with Cotton to Generate Algorithms for Site-Specific Management of Nitrogen Fertilizers and Plant Growth Regulators Based on Airborne Imagery. ASA-CSSA Oral presentation.2019.
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: TREVISAN, R.G.; BARBOSA, A.O.G.; SHIRATSUCHI, L.S.; MARTIN, N.F. Cotton yield monitor values drift over time. Beltwide Cotton Conferences. 2019
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: SHIRATSUCHI, L.S.; BRANDAO, Z.N.; SILVA JUNIOR, C.A.; KARP, F.; MARTINS, M.S.; GENTIMIS, E. Spatial Nitrogen (N) Rich Strip Approach to Evaluate the N Demand for Cotton Based on Satellite Remote Sensing. ASA, CSSA. 2019.
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: BULLOCK, D.S.; BOERNGEN, M.; TAO, H; LUCK, J.; MAXWELL, B; SHIRATSUCHI, L.S.; MARTIN, N.F. Changing Agronomic Research through on-Farm Precision Experimentation. ASA, CSSA, 2019.
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: MARTINS, M.S.; KARP, F.; SHIRATSUCHI, L.S.; DODLA, S.K.; FROMME, D.; PADGETT, G.B. Comparison on Unmanned Aerial System (UAS) and Active Crop Canopy Sensor Derived Vegetative Indices to Estimate Rice Grain Yield Potential. ASA, CSSA, 2019.
• Type: Conference Papers and Presentations Status: Published Year Published: 2020 Citation: SHIRATSUCHI, L.S.; BRANDAO, Z.N.; RAMOS JUNIOR, E.U.; MAEDA, M. On farm precision experimentation with cotton. Beltwide Cotton Conferences. 2020
• Type: Conference Papers and Presentations Status: Published Year Published: 2020 Citation: SHIRATSUCHI, L.S. On farm precision experimentation to support variable rate technology. LSU Digital Agriculture Conference 2020
• Type: Other Status: Published Year Published: 2020 Citation: SHIRATSUCHI, L.S. Precision Agriculture Update (On farm data). Louisiana Consultant Association Meeting (LACA 2020).
• Type: Conference Papers and Presentations Status: Published Year Published: 2020 Citation: LANZA, P.; MARTINS, M.; KARP, F.; PRICE, R.; SILVA JUNIOR, C.; SHIRATSUCHI, L.S. A programmatic framework to evaluate crop yield response by soil characteristics in on farm precision experiments. ASA, CSSA, SSSA 2020
• Type: Conference Papers and Presentations Status: Published Year Published: 2020 Citation: MAEDA, M.; LANDIVAR, J.; CHANG, A.; JUNG, J.; OH, S.; ASHAPURE, A.; BHANDARI, M.; DUBE, N. SHIRATSUCHI, L.S. Cotton variety performance, how can we measure it? ASA, CSSA, SSSA 2020
• Type: Conference Papers and Presentations Status: Published Year Published: 2020 Citation: RONTANI, F.; KARP, F.; MARTINS, M.; LANZA, P.; SILVA JUNIOR, C.; MAEDA, M.; SHIRATSUCHI, L.S. Drone-based vegetation indices evaluation for variety differentiation. ASA, CSSA, SSSA 2020
• Type: Conference Papers and Presentations Status: Published Year Published: 2020 Citation: KARP, F.; HARRELL, D.; GENTIMIS, T.; MARTINS, M.; LANZA, P.; CARRIERE, M.; SILVA JUIOR, C.; SHIRATSUCHI, L.S. Yield map prediction for rice in Louisiana using machine learning. 2020
• Type: Other Status: Published Year Published: 2020 Citation: SHIRATSUCHI, L.S. Use of free elevation data for precision agriculture applications. LA Crops, 2020.
• Type: Journal Articles Status: Published Year Published: 2019 Citation: TREVISAN, R.G.; SHIRATSUCHI, L.S.; BULLOCK, D.S.; MARTIN, N.F. Improving Yield Mapping Accuracy Using Remote Sensing. Preprints (doi: 10.20944/preprints 201901.0287.v1), 2019.
• Type: Journal Articles Status: Published Year Published: 2020 Citation: SILVA JUNIOR, C. A.; TEODORO, L. P. R. ; TEODORO, P. E. ; BAIO, F. H. R. ; PANTALEAO, A. A. ; CAPRISTO-SILVA, G. F. ; FACCO, C. U. ; OLIVEIRA-JUNIOR, J. F. ; SHIRATSUCHI, L. S. ; SKRIPACHEV, V. ; LIMA, M. G. ; NANNI, M. R. . Simulating multispectral MSI band sets (Sentinel-2) from hyperspectral observations via spectroradiometer for identifying soybean cultivars. Remote Sensing Applications: Society and Environment, p. 100328, 2020.

Progress 02/19/19 to 09/30/19

Outputs
Target Audience:Target audience were farmers, commodity groups and crop consultantsthat adopt precision agriculture technologies to improve their cropmanagement, but also small farmers that can benefit from publicly available remote sensing products to improve crop yield. With a extensive data collection on farm due to the nature of precision ag research where spatial variability are needed, the direct contact with farmer, consultant, employees and commodity groups was essential to run the project. Efforts to deliver science-based knowledge included formal classroom instruction teaching AGRO4092 - Precision Agriculture, Servers for cloud computing and coordination of the Digital Ag Program at LSU. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?During this period,a Digital Agriculture Conference was held and this is now likely anannual event hostedby the Lousiana State University Agcenter to promotthe practical use of precision agriculture. The main point is to have a theme with a keynote speaker and interactions with farmers, consultants and commodity groups. The next year program will emphasize training for extension agents and consultants. How have the results been disseminated to communities of interest?The results have been disseminated throughfield days, and planning meetings with consultants, farmers and companies to make operational and cost effective decisions using precision agriculture. What do you plan to do during the next reporting period to accomplish the goals?More publications for scientific and non-scientific community. Improve automated data collection in support of research activities.

Impacts
What was accomplished under these goals? (i) Spatial designs were developed and strategic partnerships with private sector promoted the automated large amount of data on-farm to drive variable rates of inputs. Experiments were conducted on 440 farm acres in partnership withfarmersand a PhD student is working with spatial designs to conduct more experiments next season. (ii) Several drone flights and soil mapping were performed and the data still under analysis to compose a large database on cloud. (iii) Experiment was started in 2018 and repeated 2019. These initial 2 years involved collaborations with different departments, research stations and several faculties. An experimental site was selected to perform more research in years to come.

Publications

• Type: Journal Articles Status: Accepted Year Published: 2019 Citation: Bullock, D.S.; Boerngen, M.; Tao, H.; Maxwell, B.; Luck, J.D.; Shiratsuchi, L.S.; Puntel, L.; Martin, N. The Data-Intensive Farm Management Project: Changing Agronomic Research through On-farm Precision Experimentation. Manuscript ID AJ-2019-03-0165-OFR
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: Shiratsuchi, L.s.; Briante, P.; Brandao, Z.; Bullock, D.; Silva Junior, C.; Ramos Junior, E. On-Farm Precision Experimentation with Cotton to Generate Algorithms for Site-Specific Management of Nitrogen Fertilizers and Plant Growth Regulators Based on Airborne Imagery. ASA-CSSA Oral presentation.
• Type: Websites Status: Published Year Published: 2019 Citation: www.onfarmprecisionag.com