Development of a 3D-Scanned Plant Model- and Ray-Tracing Simulation-Based Framework to Reduce 1D?3D Model Estimation Discrepancies in Greenhouse Canopy Light Interception and Photosynthesis = 온실 군락 수광 및 광합성의 1차원?3차원 모델 추정 오차 감소를 위한 3차원 스캔 식물 모델 및 광추적 시뮬레이션 기반 프레임워크 개발

국문 초록 (Abstract)

식물 군락은 작물과 입사 태양복사 사이의 에너지 교환이 이루어지는 공간으로, 군락 내 빛의 흡수와 분포를 결정한다. 이질적이고 복잡한 식물 구조는 변동하는 미기후와 상호작용하며 고차원의 생물물리학적 복잡계를 형성한다. 이러한 복잡성을 단순화한 1차원 군락 광합성 모델은 계산 효율이 높지만, 3차원 구조를 저차원의 매개변수로 나타냄으로써 군락 수광과 광합성을 과대추정한다. 반면, 3차원 모델은 높은 추정 정확도를 가지지만 노동 및 계산 집약적이며, 연속적인 군락 발달을 반영하기 어렵다. 최근 고해상도 3차원 식물 구조 연구를 통해 복잡한 3차원 구조에서 단순화된 패턴을 추출할 수 있는 가능성이 제시되고 있으나, 군락-광 상호작용에서도 이러한 패턴이 안정적으로 추출될 수 있는지, 그리고 생육 단계 전반에 걸쳐 일반화될 수 있는지는 아직 명확하게 밝혀지지 않았다. 따라서, 본 연구는 3차원 스캔 식물 모델과 광추적 시뮬레이션에 기반한 분석을 통해 1차원과 3차원 모델의 군락 수광 및 광합성 추정 오차를 감소시킬 수 있는 프레임워크를 구축하는 것을 목표로 수행되었다. 제1장에서는 온실 파프리카의 3차원 스캔 모델, 광추적 시뮬레이션, Farquhar von Caemmerer Berry (FvCB) 광합성 모델을 통합하여 3차원 군락 광합성 추정 파이프라인을 구축하였다. 온실 규모의 군락 광합성을 엽령에 따른 광합성 능력 변화를 반영한 두 가지 3차원 군락 광합성 모델(3Dleaf와 3Dcons), 1차원 군락 광합성 모델, 온실 CO2 균형법의 네 가지 방법으로 분석하였다. CO2 균형법에 의해 추정된 일중 총 군락 광합성량은 가장 적었고, 3Dleaf 및 3Dcons 모델은 각각 약 2%와 8%의 오차 범위를 보였다. 반면, 1차원 모델은 CO2 균형 추정치 대비 총 군락 광합성을 47–61% 과대평가하여, 3차원 모델과 광추적 시뮬레이션 및 광합성 모델의 결합이 구조적, 생리학적 불확실성을 감소시킴을 입증하였다. 제2장에서는 1차원과 3차원 모델의 군락 수광 분포 오차를 단순하고 재현 가능한 양상으로 나타낼 수 있는지 확인하였다. 온실 파프리카, 토마토, 오이의 고해상도 3차원 스캔 모델과 광추적 시뮬레이션을 이용하여, 1차원 모델 대비 군락 수광 분포의 수직 추정 차이를 정량화하였다. 오차의 정도는 일주기에 따라 변동하였으나, 군락 상단부 광도 기준 정규화를 통해 안정적이고, 종 특이적인 수직 프로파일이 확인되었다. 이를 조정 계수로 적용함으로써 1차원 모델의 과대추정을 크게 줄였으며, 계산 효율성을 유지하면서도 3차원 모델과 유사한 군락 수광 및 광합성량을 얻을 수 있었다. 제3장에서는 이 접근법을 온실 파프리카의 생육 단계 전반으로 확장하였다. 여러 생육 단계에서 구축된 3차원 스캔 식물 모델을 이용하여 일관적인 1차원-3차원 수직 오차 패턴을 확인하였으며, 이를 하나의 통합 조정 함수로 제시하였다. 이 함수를 조정 계수로 적용함으로써 생육 단계에 따른 1차원-3차원 모델 추정 오차를 동적으로 조정할 수 있었다. 종합적으로, 본 연구는 복잡한 3차원 군락 구조에 재현 가능한 저차원적 규칙성이 내재되어 있음을 입증하며, 온실 규모 군락 광합성을 효율적으로 추정할 수 있는 새로운 분석 체계를 제시한다. 이는 향후 온실 환경제어, 재배 관리, 군락 구조 최적화 등 다양한 활용 가능성을 지니며, 시설재배에서의 생산성 향상과 의사결정 지원에 중요한 기여를 할 것으로 기대된다.

다국어 초록 (Multilingual Abstract)

Plant canopies function as the primary interface for energy exchange between crops and incoming solar radiation and govern how light is absorbed and distributed across canopy layers. Highly complex and heterogeneous plant architecture interacts with dynamically fluctuating microclimates, forming high-dimensional biophysical systems. 1D canopy photosynthesis models provide high computational efficiency but tend to overestimate canopy light interception and photosynthesis by reducing complex 3D structures into low-dimensional parameters. In contrast, 3D models can accurately represent canopy structure and within-canopy light distribution, but are labor-demanding, computationally intensive, and difficult to continuously represent canopy development. Although recent advances in high-resolution 3D plant architecture research suggest that simplified patterns may underlie structurally complex canopies, it remains unclear whether such patterns can be stably extracted from canopy–light interactions and generalized across growth stages. Therefore, the objective of this study was to develop a 3D-scanned plant model- and ray-tracing simulation-based framework to reduce 1D–3D model estimation discrepancies of canopy light interception and photosynthesis. Chapter 1 establishes a 3D canopy photosynthesis estimation pipeline by integrating 3D-scanned sweet pepper models, ray-tracing simulation, and the Farquhar von Caemmerer Berry (FvCB) photosynthesis model. Greenhouse-scale canopy photosynthesis was quantified using two 3D models with and without leaf-age-dependent photosynthetic capacity (3Dleaf and 3Dcons), a 1D canopy model, and the greenhouse CO2 balance method. The CO2 balance method estimated the lowest values, while the 3Dleaf and 3Dcons showed estimation differences of approximately 2% and 8%, respectively. In contrast, the 1D model substantially overestimated canopy photosynthesis by 47–61%, demonstrating that 3D plant models combined with ray-tracing and the photosynthesis model reduce structural and physiological uncertainty. Chapter 2 examines whether discrepancies between 1D–3D canopy light models can be summarized into simple, reproducible patterns. 3D-scanned models and ray-tracing simulations of sweet pepper, tomato, and cucumber were used to quantify vertical differences in canopy light distribution compared to the 1D model. Although the magnitude of discrepancies varied diurnally, normalization by above-canopy irradiance revealed stable, species-specific vertical profiles. Applying these profiles as correction factors substantially reduced 1D model overestimation, closely aligning with the 3D model results while maintaining computational efficiency. Chapter 3 extends this approach across developmental stages of greenhouse sweet pepper. Analysis of 3D-scanned canopies at multiple growth stages identified consistent vertical discrepancy patterns between 1D and 3D models, which were expressed as a unified adjustment function applicable throughout the cultivation period. Application of this function enabled dynamic reconciliation of stage-specific 1D–3D discrepancies. Collectively, this study demonstrates that complex 3D canopy architecture contains reproducible low-dimensional patterns and establishes an efficient framework for greenhouse-scale canopy photosynthesis estimation. This proposed framework offers substantial potential for greenhouse environmental control, cultivation management, and canopy structural optimization, contributing to enhanced productivity and decision support in greenhouse horticulture.

목차 (Table of Contents)

• GENERAL INTRODUCTION 1
• LITERATURE REVIEW 9
• Canopy architecture and canopylight interactions 9
• Canopy photosynthesis modelling 11
• Development of 3D plant modelling 14
• Dimension-reduction-based analysis in plant science 16
• LITERATURE CITED 19
• CHAPTER 1 Estimation and validation of the greenhouse-scale canopy photosynthesis of sweet pepper using 3D-scanned plant model and ray-tracing simulation 31
• ABSTRACT 31
• INTRODUCTION 33
• MATERIALS AND METHODS 38
• Experimental workflow 38
• Plant materials and cultivation conditions 38
• Construction of 3D plant and greenhouse structure 41
• Gas exchange measurement 43
• FvCB parameterization 44
• Application of temperature and leaf-age dependence change in photosynthesis calculation 46
• Ray-tracing simulation and 3D canopy photosynthesis estimation 50
• Canopy photosynthesis estimation by the 1D model 51
• Canopy CO2 uptake estimation by greenhouse CO2 balance equation 51
• Model validation procedure 53
• Data analysis 53
• RESULTS 54
• Change in CO2 concentration inside experimental greenhouse 54
• Leaf gas exchange measurement and FvCB parameter estimation 54
• Estimation of 3D canopy photosynthesis by ray-tracing simulation 56
• Canopy photosynthesis calculation by CO2 balance equation 59
• Canopy photosynthesis calculation by 1D model 61
• DISCUSSION 63
• Comparison of canopy photosynthesis by different modelling approaches 63
• Net canopy CO2 uptake estimation by CO2 balance approach 65
• Effect of accurate light distribution in canopy photosynthesis estimation 67
• Application of leaf agedependent change in canopy photosynthesis estimation 68
• Implications and limitations of 3D scanningbased canopy photosynthesis estimation 70
• LITERATURE CITED 74
• CHAPTER 2 Development of dimension-reduction methods for reducing the 1D3D discrepancy in canopy light capture and photosynthesis 87
• ABSTRACT 87
• INTRODUCTION 89
• MATERIALS AND METHODS 93
• Plant materials and cultivation conditions 93
• 3D scanning and structural model construction 95
• Leaf gas exchange measurements and FvCB model parameterization 96
• Comparison of 1D and 3D canopy light interception and photosynthesis 98
• Estimation of canopy light capture and photosynthesis using the 3D model 98
• Estimation of canopy light capture and photosynthesis using the 1D model 99
• Application of leaf age and temperature dependencies to photosynthesis model parameters 102
• Calculation of estimation difference between 1D and 3D canopy light profile 104
• Adjustment of canopy light profile using normalized estimation difference 104
• Data analysis 106
• RESULTS 107
• Construction of 3D-scanned plant structures 107
• Leaf gas exchange traits and FvCB model parameters 107
• Comparison of canopy light capture and photosynthesis from 1D and 3D models 109
• Estimation difference between 1D and 3D canopy light profile 112
• Adjustment of canopy light profile using normalized estimation difference 114
• DISCUSSION 119
• Representation of complex plant architectures via high-resolution 3D scanning 119
• Estimation of canopy light capture and photosynthesis using 1D and 3D models 120
• Adjustment of canopy light profile using normalized estimation difference 125
• Implications and future work 128
• LITERATURE CITED 132
• CHAPTER 3 Development of a unified dimension-reduction-based function for reducing the 1D3D model estimation discrepancy in canopy light capture and photosynthesis across growth stages in greenhouse sweet pepper 146
• ABSTRACT 146
• INTRODUCTION 148
• MATERIALS AND METHODS 151
• Plant materials and cultivation conditions 151
• 3D scanning and structural model construction 151
• Leaf gas exchange measurements and FvCB model parameterization 153
• Comparison of 1D and 3D canopy light interception and photosynthesis 155
• Estimation of canopy light capture and photosynthesis using the 3D model 156
• Estimation of canopy light capture and photosynthesis using the 1D model 157
• Application of leaf age and temperature dependencies to photosynthesis model parameters 158
• Calculation of estimation difference between 1D and 3D canopy light profile 160
• Adjustment of canopy light profile using normalized estimation difference 161
• Construction of a unified 'MasterCurve' function 162
• Scenario analysis for the dynamic applicability of the 'MasterCurve' function 165
• Data analysis 166
• RESULTS 167
• Construction of 3D-scanned plant structures across growth stages 167
• FvCB model parameterization across growth stages 167
• Comparison of 1D and 3D canopy light capture and photosynthesis 170
• Estimation difference between 1D and 3D canopy light profile 173
• Construction of a unified 'MasterCurve' function from normalized discrepancy profiles across growth stages 174
• Adjustment of canopy light capture and photosynthesis using REDL and the 'MasterCurve' function 180
• Scenario analysis for the dynamic applicability of the 'MasterCurve' function 184
• DISCUSSION 187
• 1D-3D model discrepancy across growth stages 187
• Emergence of invariant normalized discrepancy patterns 191
• Correction of canopy light capture and photosynthesis by unified 'MasterCurve' function 193
• Implications and future work 199
• LITERATURE CITED 203
• GENERAL CONCLUSION 213
• ABSTRACT IN KOREAN 215