The promise and challenges of unmanned aerial vehicles for calibrating remote sensing data for agricultural models in the Long-Term Agroecosystem Research network: a case for the southeastern U.S.A.

Authors: Coffin, Alisa; Strickland, Timothy; Smith, Coby; Endale, Dinku; Bosch, David

Submitted to: American Society of Agronomy Meetings
Publication Type: Abstract Only
Publication Acceptance Date: 7/25/2018
Publication Date: 11/4/2018
Citation: Coffin, A.W., Strickland, T.C., Smith, C.M., Endale, D.M., Bosch, D.D. 2018. The promise and challenges of unmanned aerial vehicles for calibrating remote sensing data for agricultural models in the Long-Term Agroecosystem Research network: a case for the southeastern U.S.A.[abstract]. American Society of Agronomy Meeting, November 4-7, 2018, Baltimore, Maryland.

Interpretive Summary: Unmanned aerial vehicles (UAVs), or drones, are now being promoted extensively as a tool to monitor crop conditions in real time and with high resolution. Furthermore, in the U.S.A., the UAV industry has mushroomed in recent years with the introduction of relatively low-cost package solutions that promise high-performance decision tools. While these "agricultural drone packages" may quickly spot within-field problems, such as pest and disease outbreaks, allowing for precision treatments, their use as research instruments requires additional steps to make the data useful for scientific research. However, the protocols for collecting drone data for scientific use are not well-developed, and the promise of these data to serve as “scale bars” for extrapolating from local to regional scales is not well understood. Scientists at the USDA-ARS Long-Term Agroecosystem Research (LTAR) site in Tifton, Georgia, are conducting research to determine the limits of drone data to verify and calibrate global scale satellite data, which are used in models to forecast global crop production. Calibrated multi-spectral data from a UAV-borne sensor were collected over three farms on multiple dates in 2017. Additional hand-held sensors were used to measure vegetation reflectance and leaf area, coincident with UAV collections. Initial reviews of the data indicate mixed results. On one hand, the data provided fine resolution wall-to-wall information across the entire farm, which we are able to compare with the results of hand-held and satellite-borne sensors. On the other hand, the atmospheric conditions produced around solar noon during the summer in southern Georgia--the critical period for collecting multi-spectral UAV data--gave sudden rise to cumulus clouds overshadowing the flight path, creating problems of uneven illumination, affecting the consistency of the data. Future work for 2018 and beyond includes UAV collection of multispectral and thermal data concurrent with overpasses of optical and radar satellites including Landsat, Sentinel and Radarsat.

Technical Abstract: Unmanned aerial vehicles (UAVs) are now being promoted extensively as a tool to monitor crop conditions with fine resolution and low latency. Furthermore, in the U.S.A., the UAV industry has mushroomed in recent years with the introduction of relatively low-cost package solutions that promise high-performance decision tools. While these ensembles may provide adequate data quality to quickly spot within-field problems, such as pest and disease outbreaks, thus allowing for precision treatments, their use as research instruments requires additional steps to calibrate sensors and validate data. However, the protocols for the collection of UAV-based data for scientific use are not well-developed, and the promise of these data to serve as “scale bars” for extrapolating from local to regional scales is not well understood. Scientists at the USDA-ARS Long-Term Agroecosystem Research (LTAR) site in Tifton, Georgia, are conducting research to determine the utility of UAV datasets in quantifying uncertainty associated with satellite-derived data, which serve as inputs to models that characterize and forecast agroecosystem dynamics. Calibrated multi-spectral data from a UAV-borne sensor were collected over three farms on multiple dates in 2017. In some areas, additional proximal sensors were used to measure vegetation reflectance along with leaf area index (LAI), coincident with UAV collections. Initial reviews of the data indicate mixed results. On one hand, the data provided fine resolution wall-to-wall reflectance information across the entire farm, which we are able to compare with both proximal and satellite-based multi-spectral sensors. On the other hand, the data suffered from problems of uneven reflectance due to shadows cast by cumulus clouds rapidly developing during the mid-summer noontime flights in southern Georgia. Future work for 2018 and beyond includes UAV collection of multi-spectral and thermal data concurrent with overpasses of optical and radar satellites including Landsat, Sentinel and Radarsat.