Date of Award

Spring 5-3-2024

Level of Access Assigned by Author

Open-Access Thesis

Degree Name

Master of Science (MS)

Department

Forest Resources

Advisor

Parinaz Rahimzadeh-Bajgiran

Second Committee Member

Isabel Munck

Third Committee Member

José Eduardo (Dudu) Meireles

Abstract

Canopy foliar traits serve as crucial indicators of plant health and productivity, forming a vital link between plant conditions and ecosystem dynamics. While common remote sensing techniques utilize field spectroscopy or imagery to estimate foliar chemistry, the application of remote sensing-derived foliar traits data for plant disease detection remains largely unexplored. Eastern White Pine (Pinus strobus L., EWP), a species of ecological and economic significance in the Northeastern USA, faces a growing threat from the fungal disease complex known as White Pine Needle Damage (WPND). The rising prevalence of WPND over the past two decades poses a serious risk to EWP health and longevity, necessitating a comprehensive understanding of its landscape-level impact. In this study, to understand the impact of WPND on the EWP foliar traits, leaf samples were collected in summer 2022 in Bethel, Maine. Traits such as nitrogen (N), chlorophyll (Chl) and equivalent water thickness (EWT) were measured for 38 asymptomatic and 60 symptomatic trees. Field spectroradiometry was used to collect hyperspectral data from these leaf samples. This study tested the feasibility of using remote sensing variables to model foliar traits and potentially use them for WPND detection and classification. Results indicated that the combination of field-measured traits and remote sensing data can be useful in disease classification into asymptomatic and symptomatic classes as evident by nearly 77% accuracy with Kappa coefficient (K) of 0.46 observed using the Random Forest (RF) model. The RF model based solely on remotely sensed spectral vegetation indices (SVIs) demonstrated even higher accuracy of nearly 87% and K of 0.68. In comparison, the best accuracy achieved by traits only model for the classification of WPND was 70% (K: 0.38). These findings contribute valuable insights and highlight the potential of both field-derived foliar and remote sensing data for effective monitoring and management of WPND in EWP. For modeling foliar traits (Chl, N and EWT) using remote sensing data, several parametric and non-parametric models (machine learning; ML) were tested. Results from linear and polynomial regression of foliar traits revealed notable correlations between SVIs and field-derived foliar traits, with MERIS Terrestrial Chlorophyll Index (MTCI; R2 = 0.62) emerging as the best index for Chl content estimation. The red-edge indices performed consistently better compared to other SVIs in estimating Chl content. However, similar accuracies could not be replicated by these indices when used in ML models. The use of non-correlated SVIs improved the estimation accuracy of foliar traits with best performance in Chl estimation (R2: 0.45; RMSE: 3.73). Among studied traits, the best modeling accuracy was observed for Chl estimation for both parametric and non-parametric methods. We tried developing modified versions of best performing SVIs using hyperspectral bands but they did not show significant improvements in the modeling accuracy of traits. ML models were further used to identify significant spectral regions for each trait. Unsurprisingly, red-edge region was identified significant for Chl. For EWT, important bands were in VNIR and SWIR region and for N; the blue part of the VNIR region was found to be important from ML algorithms. The findings shed light on the complex nature of leaf traits, and need for tailored approaches for different functional groups and ecosystems. This study addresses emerging threats faced by EWP trees and offers valuable tools for WPND detection and foliar trait modeling. With exponential rise in forest pests and pathogens in recent years, remote sensing can prove beneficial for timely and accurate detection of disease and improved forest management practices.

Download

Available for download on Sunday, August 10, 2025

Share