Advances in remote sensing techniques for monitoring and predicting plant physiology and biochemistry

This thesis investigates novel remote sensing approaches to monitor and predict plant physiology and biochemistry in response to environmental stressors and seasonal changes. Divided into two chapters, each explores a distinct remote sensing technique and its application in understanding the connection between plant physiology and remote sensing.Chapter 1 introduces a novel nighttime low-cost photodiode method, which was tested during a drought response experiment of LED-induced canopy-level chlorophyll a fluorescence (LEDIF) in Polygala myrtifolia. Far-red LEDIF (720 nm - 740 nm) was retrieved using low-cost photodiodes (LEDIFphotodiode) and a hyperspectral instrument (LEDIFhyperspectral). To link the LEDIF signal with physiological drought response, we tracked stomatal conductance (gsw) using a porometer as an indicator of plant water status, two leaf-level vegetation indices — photochemical reflectance index, PRI; normalized difference vegetation index, NDVI— to represent chlorophylls and xanthophyll pigment dynamics, respectively, and a pulse-amplitude modulation (PAM) device to measure photochemical and non-photochemical dynamics of photochemistry. Our results demonstrate a similar performance between the photodiode and hyperspectral retrievals of LEDIF (R2=0.77, P < 0.01). Furthermore, LEDIFphotodiode closely tracked drought responses with photochemical quenching (PQ, R2=0.69, P < 0.001), Fv/Fm (R2=0.59, P < 0.001), and leaf-level PRI (R2= 0.59, P<0.05). The results demonstrate the potential of this cost-effective method to accurately track changes in photosynthetic status and overall plant health, offering valuable insights into the relationships between the physiological mechanisms of photosynthesis and chlorophyll fluorescence.Chapter 2 employs hyperspectral reflectance data to predict an array of chlorophyll and carotenoid pigment concentrations in Pinus palustris (Longleaf Pine) using a partial least squares regression modeling approach. This study took place in north-central Florida, at the Ordway Swisher Biological Station (OSBS), and more specifically, the National Ecological Observation Network (NEON) flux tower site within it. The research site is dominated by mature Longleaf Pines and low-lying perennial grasses. From six Longleaf Pine trees, branches were harvested, and needles were either 1) immediately stored for later pigment extraction and 2) made into needle mats to retrieve reflectance measurements with our hyperspectral spectroradiometer’s leaf clip. Needle mats were assembled by laying individual needles flat and continuously side by side until they were approximately 5-6 cm wide, and were held together by two pieces of tape, arranged at the top and bottom of the needles. Using a PLSR modeling approach, our prediction task for each PLSR model was to predict the average pigment content of a tree, using a hyperspectral reflectance measurement of a single needle mat (for each pigment/pool) per tree. Our results reveal the potential of hyperspectral remote sensing to estimate plant pigments accurately and efficiently; our PLSR models successfully predicted the concentrations with R² values > 50% for eight of the ten pigments/pools. We were able to best predict lutein and neoxanthin, as well as chlorophyll b and a (R2 = 0.91, RMSE = 0.01; R2 = 0.77, RMSE = 0.0; R2 = 0.68, RMSE = 0.02; R2 = 0.65, RMSE = 0.05, respectively). This research demonstrates the value of leaf-level remote sensing in advancing our understanding of the physiological status of evergreen species and their underlying pigments.Collectively, these chapters showcase the power of remote sensing for monitoring and predicting plant responses to environmental stressors and seasonal changes. They offer valuable insights into the relationships between plant physiology and remote sensing, paving the way for improved strategies to assess plant health and resilience in the face of changing environmental conditions