서브메뉴
검색
Multidisciplinary Strategies for Grapevine Disease Monitoring in the Age of Digital Viticulture
Multidisciplinary Strategies for Grapevine Disease Monitoring in the Age of Digital Viticulture
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105302
- ISBN
- 9798273309227
- DDC
- 581
- 서명/저자
- Multidisciplinary Strategies for Grapevine Disease Monitoring in the Age of Digital Viticulture
- 발행사항
- [Sl] : Cornell University, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 272 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-07, Section: B.
- 주기사항
- Advisor: Gold, Kaitlin.
- 학위논문주기
- Thesis (Ph.D.)--Cornell University, 2025.
- 초록/해제
- 요약Effective monitoring and management of plant diseases is essential to grape production, particularly in cool climate viticultural areas where concurrent pest and disease pressures are common. This dissertation investigated multi-modal remote sensing tools for detecting and monitoring grapevine downy mildew (GDM) and grapevine leafroll-associated virus 3 (GLRaV-3) in research and commercial vineyards in New York State.The study began with an evaluation of high-resolution commercial satellite imagery for GDM surveillance. Random forest models trained on spectral bands and vegetation indices (VIs) successfully classified areas of high and low GDM incidence and severity, achieving maximum accuracies of 0.85 with SkySat (30cm resolution) and 0.92 with PlanetScope (3m resolution). Significant differences between VIs of high- and low-damage classes were not observed until late July. Cloud cover, image co-registration, and limited spectral resolution were identified as major challenges to operational satellite-based GDM monitoring.Uncrewed aerial systems (UAS) were subsequently examined as an alternate platform for consistent, high-resolution monitoring. Using a pathology research vineyard and a hemp breeding trial as test systems, an image processing pipeline was developed to generate multispectral orthomosaic time series, perform crop segmentation, and extract plant-level spectral data. A convolutional neural network and a vision foundation model produced the most accurate segmentations, with mean intersection-over-union values of 0.85 and 0.95 for vineyard and hemp imagery, respectively. Applying each model to the opposite crop dataset reduced segmentation accuracy, underscoring the need for adaptable, modular workflows for UAS-based analyses in specialty crops.To evaluate hyperspectral UAS imagery for detecting co-occurring diseases, paired UAS-ground surveys were conducted at two commercial Finger Lakes vineyards. Partial least squares regression models trained on visible-to-near infrared imagery detected GLRaV-3 symptoms in the presence of GDM, with F1 scores of 0.69-0.72 for symptomatic vines and highest model performance observed post-veraison. Models differentiating GDM and GLRaV-3 achieved accuracies of 0.55-0.82. Although GDM severity did not correlate with model uncertainty, reduced canopy size significantly increased PLSR prediction variance.Together, these investigations advanced development and implementation of optical remote sensing tools to provide grape growers with efficient, reliable means of vineyard disease monitoring.
- 일반주제명
- Plant pathology
- 일반주제명
- Remote sensing
- 일반주제명
- Climate change
- 기타저자
- Cornell University Plant Pathology and Plant-Microbe Biology
- 기본자료저록
- Dissertations Abstracts International. 87-07B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017360089
■00520260202105302
■006m o d
■007cr#unu||||||||
■020 ▼a9798273309227
■035 ▼a(MiAaPQ)AAI32281892
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a581
■1001 ▼aKanaley, Kathleen.
■24510▼aMultidisciplinary Strategies for Grapevine Disease Monitoring in the Age of Digital Viticulture
■260 ▼a[Sl]▼bCornell University▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a272 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-07, Section: B.
■500 ▼aAdvisor: Gold, Kaitlin.
■5021 ▼aThesis (Ph.D.)--Cornell University, 2025.
■520 ▼aEffective monitoring and management of plant diseases is essential to grape production, particularly in cool climate viticultural areas where concurrent pest and disease pressures are common. This dissertation investigated multi-modal remote sensing tools for detecting and monitoring grapevine downy mildew (GDM) and grapevine leafroll-associated virus 3 (GLRaV-3) in research and commercial vineyards in New York State.The study began with an evaluation of high-resolution commercial satellite imagery for GDM surveillance. Random forest models trained on spectral bands and vegetation indices (VIs) successfully classified areas of high and low GDM incidence and severity, achieving maximum accuracies of 0.85 with SkySat (30cm resolution) and 0.92 with PlanetScope (3m resolution). Significant differences between VIs of high- and low-damage classes were not observed until late July. Cloud cover, image co-registration, and limited spectral resolution were identified as major challenges to operational satellite-based GDM monitoring.Uncrewed aerial systems (UAS) were subsequently examined as an alternate platform for consistent, high-resolution monitoring. Using a pathology research vineyard and a hemp breeding trial as test systems, an image processing pipeline was developed to generate multispectral orthomosaic time series, perform crop segmentation, and extract plant-level spectral data. A convolutional neural network and a vision foundation model produced the most accurate segmentations, with mean intersection-over-union values of 0.85 and 0.95 for vineyard and hemp imagery, respectively. Applying each model to the opposite crop dataset reduced segmentation accuracy, underscoring the need for adaptable, modular workflows for UAS-based analyses in specialty crops.To evaluate hyperspectral UAS imagery for detecting co-occurring diseases, paired UAS-ground surveys were conducted at two commercial Finger Lakes vineyards. Partial least squares regression models trained on visible-to-near infrared imagery detected GLRaV-3 symptoms in the presence of GDM, with F1 scores of 0.69-0.72 for symptomatic vines and highest model performance observed post-veraison. Models differentiating GDM and GLRaV-3 achieved accuracies of 0.55-0.82. Although GDM severity did not correlate with model uncertainty, reduced canopy size significantly increased PLSR prediction variance.Together, these investigations advanced development and implementation of optical remote sensing tools to provide grape growers with efficient, reliable means of vineyard disease monitoring.
■590 ▼aSchool code: 0058.
■650 4▼aPlant pathology
■650 4▼aRemote sensing
■650 4▼aClimate change
■653 ▼aVegetation indices
■653 ▼aGrapevine downy mildew
■653 ▼aUncrewed aerial systems
■653 ▼aClimate viticultural areas
■690 ▼a0480
■690 ▼a0799
■690 ▼a0404
■71020▼aCornell University▼bPlant Pathology and Plant-Microbe Biology.
■7730 ▼tDissertations Abstracts International▼g87-07B.
■790 ▼a0058
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360089▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
