For 2020 and newer grants, please go to https://grants.ipmcenters.org/
Home       Current RFAs       PD User Guide       Projects       Login

Funded Project

Funding Program: IPM Partnership Grants

Project Title: Early Detection of Potato Leafhopper Damage Using Unmanned Aerial Systems

Project Directors (PDs): Chandi Witharana [1] Ana Legrand [2] Shuresh Ghimire [3]

Lead State: CT Lead Organization: University of Connecticut

Cooperating State(s): Massachusetts

Undesignated Funding: $49,783

Start Date: Apr-01-2020 End Date: Mar-31-2022

No-Cost Extension Date: May-22-2023

Pests Involved: Potato Leafhopper

Site/Commodity: Potato

Area of Emphasis: Advanced Production Systems

Summary: Early detection of disease and insect infestation within crops is essential to lower production losses, reduce environmental risk, and promote environmentally conscious management practices. Innovative pest detection and monitoring methods that are inexpensive while being highly efficient can increase pest management decision-making based on estimates of pest population size. There is a growing interest in adaptation of remote scouting methods that are centered on remote sensing (RS) technologies, to produce low-cost, real-time/quasi real-time, repeatable, and spatially-explicit analytics for IPM applications. Unprecedented advances in unmanned aerial system (UAS) technology and the development of robust, autonomous and lightweight sensors present a unique opportunity for enabling RS technologies for IPM use. UASs are rapidly evolving into standalone RS systems that deliver information of high spatial and temporal resolution in a non-invasive manner. UAS platforms can rapidly survey areas and can be deployed where and when needed. Both the cost and complexity of the UASs have been reduced to the point where an individual can afford a drone and use it in the field with minimal technical expertise. In spite of the potential benefit, very few studies have been conducted to exploit the potentials of UAS for early detection of pest infestation. This reflects a clear knowledge and methodological gap between IPM science and UAS technology. Without a concerted cross-disciplinary effort - remote sensing science, computer science, and IPM science - to build bridges between the IPM community and this new UAS-enabled future, we will never fully capitalize on the plethora of possibilities afforded by centimeter-scale imagery. The proposed research will investigate the practicality of off-the-shelf UAS outfitted with lightweight multispectral and hyperspectral sensors - as remote scouting instruments- in early detection and discrimination of crop infestation by potato leafhopper Empoasca fabae in potato. We propose a two-year project, deploying UAS remote sensing coupled with proximal remote sensing to develop and validate models for assessing and distinguishing early damage by potato leafhopper using UAS spectral reflectance data. Objectives: Our goal is to develop a Unmanned aerials system (UAS) based system for early detection of potato leafhopper damage in potato. Four specific objectives are to; [1] Identify specific and sensitive spectral wavelengths to leafhopper damage and to soil nitrogen. [2] Model the relationship between the leafhopper damage and spectral responses. [3] Investigate the applicability of UAS spectroscopy to detect leafhopper damage and characterize the level of severity. [4] Develop and implement an extension program for early detection of potato leafhopper damage using unmanned aerial systems.

Final Report:

Outputs Field Experiments: Potato leaf hopper (PLH) field experiments of Summer 2021 and Summer 2022 took place in the Research and Education Facility of University of Connecticut located in Storrs, Connecticut. The Summer 2021 comprised two trails (Trial 1 (planted late May) and Trail 2 (planted early August)). Both trials followed randomized complete block design (RCBD) with twelve plots. Each plot was 1.8 x 1.8 m with nine potato seeds planted 8 inches apart. In between each plot was a 2 m buffer plot to separate the treatments. Each trial consisted of three treatments and replicated four times based on PLH and different Nitrogen levels (Standard Nitrogen with PLH, Standard Nitrogen without PLH, Low Nitrogen with PLH and Low Nitrogen without PLH). The 2022 Summer field campaign consisted of 36 plots total with 6 treatments ordered in a RCBD. Treatments included No Nitrogen without PLH (NoFert_NoPLH), No nitrogen with PLH (NoFert_PLH), Low Nitrogen without PLH (LowN_NoPLH), Low Nitrogen with PLH (LowN_PLH), Standard Nitrogen without PLH (StdN_NoPLH), and Standard Nitrogen with PLH (StdN_PLH) in a randomized complete block design). Data Collection: A DJI Inspire 2 UASs was deployed with an attached Micasense Red-edge MX multi-spectral sensor to capture canopy imagery. Flights were held weekly at times between 9 am to noon to eliminate shadowing. The DJI Inspire 2 was flown at an altitude of 30 meters with a 75% overlap and a speed 6 m/s, totaling a 6-minute flight time and approximately 3 cm resolution. Reflectance was calibrated using the provided calibration panel from Micasense. In situ measurements of the plants were taken weekly. A handheld leaf spectrometer from CID Bio-Science CI-710s SpectraVue was used to measure the spectral reflectance of the potato plants at the leaf scale. Chlorophyll content readings were measured using the company atLeaf’s standard chlorophyll meter. Leaf reflectance and chlorophyll measurements were taken from the same leaves from plants observed in on-foot scouting. Potato leafhoppers were monitored on plants weekly using a systematic sampling route. The on-foot insect count for PLH was to examine every plot of potatoes with observations of leaf damage as well as count the number of adult and nymphs present. The on- foot insect count route was the same per each potato plot, but selected plants were changed every week to avoid overlap. Additional notes of observations were recorded – especially observations of symptoms of hopperburn. A damage scale was developed to assess PLH feeding damage to field plants and the level of damage based on plant structural and visual characteristics, throughout the treated plots. During summer 2022 trials, a rating scale was developed upon visible PLH feeding damage to the potatoes. This provided real-time visible observations of the feeding damage by PLH to potato plants. Observing the structural and visual changes also helped correspond the drone, spectrometer, chlorophyll meter, and scouting data to real-time observations. Greenhouse Experiments: Greenhouse experiments took place at the Agricultural Biotechnology Lab greenhouse range, Univ. of Connecticut, Storrs, Connecticut. The greenhouse trial was conducted in October 2022. There was a total of 90 plants with 30 plants per treatments of infested vs. control followed a factorial design of the three days of which PLH infestation occurred. Each treatment consisted of a small potted potato plant with one leaflet inserted into an insect cage where three PLHs were applied to feed for 24, 48, or 72 hours. Spectral readings were taken in the day 1, day 2, and day 3 intervals with the CID Bio-Science CI-710s SpectraVue handheld leaf spectrometer. After each day of collecting spectral reflectance, the plant was no longer used, and leafhoppers were placed back in their incubated colony. Data Processing and Analysis: The DJI Inspire 2 drone imagery data was processed in Pix4D Fields to create stitched true color, NDVI, and NDRE index orthomosaics using MicaSense’s center wavelength and bandwidth from the camera system. The orthomosaics were then uploaded to ArcPro software where the treated plots were individually digitized with shapefiles that extracted the index values from each week of flights. The handheld CID Bioscience spectrometer was used in field season 2022 trials. The spectrometer automatically calculated NDVI and NDRE indices during measurement readings. Vegetation indices derived from spectral reflectance data (proximal and drone) were used as predictor variables in a linear mixed effects statistical model. The significance of predictors is evaluated based on Wald tests in an Analysis of Variance (ANOVA). The model was intended to observe the mean responses of the indices measured based on the level of fertilizer rate and PLH infestation. Summer 2021 Trial 1 no LME analysis was conducted due to a high precipitation volume causing a Fusarium fungus throughout the field. The T-test analysis was conducted to infer greenhouse experimental data. Outcomes: Considering the overall results of the study, it shows that UASs have potential for high resolution data, which can detect field health and spot treatments throughout a period of time. From the linear mixed effects model (LME) in summer 2021 trial two shows a p < 0.001 significance level for weeks 1 and 2 for NDRE drone, NDVI drone, and chlorophyll content. This level of significance for the early weeks shows that time is a factor in index changes due to PLH feeding. The early weeks of PLH feeding determine the decrease in index values. This is also shown in the significance for PLH total of p = 0.008 for NDRE drone, p < 0.001 for NDVI drone. Meaning that PLH presence impacts the drone indices. There was also a significant interaction between fertilizer treatments and PLH total for chlorophyll content of p < 0.1. This could potentially mean that fertilizer and PLH feeding impact chlorophyll over time. In summer 2022 trial one, the linear mixed effects model shows p < 0.01 for weeks 1 and 2 in NDRE spectrometer, NDRE drone, NDVI spectrometer, and NDVI drone. This also shows that the early weeks of data collection and digital scouting are important for detecting PLH feeding damage. For fertilizer treatments NDRE drone shows significance of p < 0.001 for LowN and p < 0.001 for StdN treatments. This could mean that fertilizer played an effect to the NDRE index readings collected by the drone. NDRE may be sensitive to fertilizer applications over time along with NDVI drone. The NDVI index captured by the drown shows significance of p = 0.001 for LowN and p < 0.01 for StdN treatments. There was observed significance for PLH total of p = 0.08 in NDRE spectrometer and p = 0.03 NDVI spectrometer. This could mean that the handheld spectrometer is sensitive to PLH presence over time. For summer 2022 trial two, there was significance at the first two weeks with values of p < 0.001 for NDRE drone, NDVI drone, chlorophyll content, and p = 0.05 for NDRE spectrometer. Showing that time is a factor for digital scouting and data collection during PLH feeding. There was significance for fertilizer treatments with NDRE spectrometer having a p-value of p = 0.07 for LowN and p = 0.005 for StdN, NDRE drone has p < 0.001 for LowN and StdN, and NDVI drone also has p < 0.001 for LowN and StdN. The rate of nitrogen also is a factor for sensitivities in NDRE and NDVI indices. Results from the greenhouse experiment suggested that by Day 1 of infestation (24 hours) multiple vegetation indices were sensitive to PLH attack. The indices that show the decrease in means at Day 1 include: NDRE, WBI, SRPI, and PSRI. Results from the two sample t-test showed that the SRPI and NDRE do not show a significant difference in means for Day 1. However, the WBI and PSRI do showed a significant reduction. By Day 2 (48 hours of infestation), NDVI, WBI, SRPI, and G values of infested plants showed marked reduction compared to the controls

Outcomes Findings of this study advance our understanding on the challenges and opportunities associated with drone technology adaptations in IPM applications. We presented our findings to a broader audience from researchers, extension educators to grower. One of the key impact point was presenting the findings to growers in Connecticut via the UConn Extension's 2023 Vegetable and Small Fruit Growers Conference.

Report Appendices 9999369_0000002.pdf [PDF]

Close Window

Northeastern IPM Center 340 Tower Road Cornell University Ithaca, NY 14853 NortheastIPM.org		Developed by the Center for IPM © Copyright CIPM 2004-2024