Evaluating remotely piloted aircraft estimates of crop height and LAI against satellite and crop model outputs
Loading...
Date
2021-02-17
Authors
Dzinic, Omar
Journal Title
Journal ISSN
Volume Title
Publisher
University of Waterloo
Abstract
Crop simulation models (CSM) have been a method for decision makers to study the effects of crop management activities for predicting, planning, and improving crop growth for the past several decades. While the applicability and robustness of CSMs had been rapidly evolving, the methods of gathering input and validation data for CSMs has remained predominantly the same. However, the application of remote sensing technologies including remotely piloted aircraft systems (RPAS) and satellites for agricultural purposes has demonstrated the potential for automated rapid and high detail CSM validation data. This study evaluated the accuracy of validation data acquired using RPAS and satellite technologies when compared to CSM outputs and observed crop measurements. Imagery of an agricultural field was acquired throughout a growing season with the use of a multi-sensor RPAS and existing satellite missions. Field work was performed alongside the RPAS imagery acquisitions to collect input data for crop modelling and accuracy assessments. Using the acquired imagery, the crop height and leaf area index (LAI) values of crops in the field were estimated for multiple dates. The LAI was estimated using 1) a regression-based method and 2) a function of the fractional vegetation cover and the leaf angle distribution method. A CSM was run alongside the remote sensing to simulate crop height and LAI values. When the estimated values were compared to observed measurements, showing the RPAS-derived crop height values were significantly more accurate (RMSE=193.6 cm, RMSE=161.3 cm) than the satellite-derived crop heights values (RMSE=223.4 m, RMSE=117.1 m respectively) yet less accurate than the CSM crop heights values. The RPAS-derived LAI value accuracies (RMSE=0.42, RMSE=0.66) and satellite-derived LAI value accuracies (RMSE=0.56, RMSE=0.56) were similar but the RPAS was found to, on average, estimate LAI more accurately than the CSM. Overall, the RPAS methods showed moderate accuracy across both crop height and LAI estimations and was found to perform better than the CSM in some situations. Future work may include additional imagery acquisitions throughout a growing season to further test the accuracies of RPAS-derived estimates as well as integrating estimates directly into CSMs for validation purposes.
Description
Keywords
rpas, uav, drone, crop modeling, leaf area index, lai, remote sensing, geomatics, crop height, gis, satellite, sentinel