Abstract
Shrub willow is a perennial crop used to produce biomass for biofuels, bioproducts, and
bioenergy, and to help address environmental issues. The goal of this dissertation was to enhance
willow management by utilizing Sentinel-2 data and unmanned aerial system (UAS) data to
estimate shrub willow leaf chlorophyll concentration (LCC), leaf area index (LAI), and canopy
chlorophyll concentration (CCC) across time, space, and scale and quantify the model uncertainty.
First, we constructed a simple and efficient calibration model for estimating willow LCC using the
SPAD-502 chlorophyll meter. Models including growing degree days (GDD) as a predictor
performed better (R2 = 0.92, RMSE = 4.81 μg/cm2
) than models (R2 = 0.90, RMSE = 5.38 μg/cm2
)
using only chlorophyll meter (CM) readings. Through inclusion of the time and weather-dependent
GDD, models more accurately estimate LCC, which is particularly useful to support seasonal timeseries plant health status monitoring using remote sensing technologies. Second, this dissertation
estimated willow LCC, LAI, and CCC from Sentinel and UAS data across time and space. The
model transferability exploration indicated that neural network (NN) models (for LCC) and
NDVIre-based regression (for LAI and CCC) yielded the best predictive performance when
models were transferred across time and space. Since CCC estimation had smaller normalized
RMSE (NRMSE) values than LCC and LAI estimation for both Sentinel-2 and UAS data, we
recommend using the CCC parameter to support willow health status evaluation. Finally, we
analyzed the impact of different spatial scales—5 m, 10 m, and 20 m—of UAS data on the accuracy
of model predictions for CCC in shrub willow. Models built at 5 m, 10 m, and 20 m could be
applied across time, space, and at different scales, with the 20 m dataset using models built at 5 m
providing the best predictive performance. Compared with RMSE, uncertainty analysis using a
Bayesian approach indicated that model parameter uncertainty increased as pixel size increased
and can guide future experimental design to help save resources. Overall, this dissertation provided
new approaches to enhance willow management and increase the efficiency of practical
applications using remote sensing technology.