Ph.D. Dissertation Defense: Temporal Requirements for Future Landsat Systems for Agricultural Monitoring

Ph.D. Dissertation Defense
Temporal Requirements for Future Landsat Systems for Agricultural Monitoring

Emily Myers
Imaging Science Ph.D. Candidate
Chester F. Carlson Center for Imaging Science, RIT
Advisor: John Kerekes

Register Here for Zoom Link

This research aims to determine the necessary overpass frequency for a future Landsat sensor for agricultural growth stage monitoring and yield prediction. Two experiments were conducted to study the effects of imaging frequency on the accuracy of these tasks.

Abstract:
Agricultural monitoring is an important application of earth-observing satellite systems, which may be used for stress and disease detection, growth stage monitoring, and yield prediction in crops at a fraction of the time and cost it would take to survey fields manually. Satellites within the Landsat program are frequently used for agricultural monitoring, but they do not always collect imagery often enough to capture rapid changes in vegetation health. To address this limitation, an increase in revisit rate is being considered for future Landsat systems. This research aims to determine the necessary overpass frequency for a future Landsat sensor for agricultural growth stage monitoring and yield prediction. Two experiments were conducted to study the effects of imaging frequency on the accuracy of these tasks. The first experiment investigated the impact of imaging frequency on growth stage monitoring. Image-derived plot-average Normalized Difference Vegetation Index (NDVI) time-series data collected over a small corn field were used to estimate phenological transition dates. Images were then removed from the original time-series, and dates were recalculated from the resampled data. Using PlanetScope surface reflectance imagery, the average range of estimated dates increased by a day and the average absolute deviation between estimated dates increased by 1/3 of a day for every day of increase in average revisit interval. Using the higher-quality PlanetScope L3H surface reflectance product, these rates of increase were roughly halved. Higher imaging frequency and higher radiometric quality both led to greater precision in estimates. The second experiment investigated the impacts of imaging frequency and time-series end date on yield correlation accuracy. Plot-average Green Normalized Difference Vegetation Index (GNDVI) time-series data collected over a small corn field were resampled to different revisit intervals, gap-filled and smoothed, and correlated with plot-average yield at each day of the growing season. These experiments were then repeated with images removed from the end of the time series. All methods tested performed well on time-series ending 65-72 days or more after green-up, and performed poorly for time-series ending prior to the day of peak GNDVI. Mean R-squared values for GNDVI-yield correlations decreased with increasing revisit intervals. This effect was stronger for time-series ending earlier in the growing season. The findings of this study were used to recommend an optimal overpass frequency of 1-4 days for future yield-monitoring satellite systems.

Intended Audience:
Undergraduates, graduates, and experts. Those with interest in the topic.
To request an interpreter, please visit https://myaccess.rit.edu