<Emphasis Type="BoldItalic">Key message</Emphasis> Leveraging affordable red-green-blue (RGB) imaging and neural network algorithms, this study delivers a high-throughput method to quantify seasonal color shifts and genetic variation in Scots pine seedlings. Hue-saturation-brightness (HSB) color analysis, and RGB values can be used for population and seasonal differentiation and hold potential for advancing breeding programs in forestry under changing climatic conditions. Context Scots pine (Pinus sylvestris L.) displays remarkable genetic and phenotypic diversity, with seasonal color changes such as autumn reddening, reflecting population-level responses to local environmental conditions. Advances in imaging and deep learning now enable precise quantification of such phenotypic variation, providing new insights into population-level variation. Aims This study assesses seasonal color variation within and among Scots pine seedling populations, compares the effectiveness of RGB and HSB systems for population and seasonal differentiation, and investigates phenological patterns across progenies of three seed orchards from ecologically distinct populations. Methods One-year-old seedlings from lowland (Plasy, Trebon) and upland (Decin) populations were imaged in a common garden trial in September, October, and January using a handheld camera. Needle-level segmentation was performed via a convolutional neural network. Genetic variability and population differences were analyzed using linear mixed models. Results Population differentiation reached the highest values in the RGB blue channel (QST-blue = 0.64 in September and QST-blue = 0.94 in October) and in HSB (QST-hue = 0.61, QST-saturation = 0.62 in September and QST-saturation = 0.64 in October). Color wheel visualizations revealed converging hue and saturation trajectories, indicating gradual phenological changes in the post-growing season. September values exhibited the highest heritability (h2RGB = 0.12–0.25; h2HSB = 0.12–0.29) among measured optical traits. Conclusion This study demonstrates that RGB and HSB color parameters, extracted from high-throughput image analysis using CNN-based needle segmentation, capture both seasonal and genetic variation in Scots pine seedlings. The highest genetic differentiation and heritability occurred during early autumn, particularly in the blue and saturation parameters. These findings suggest that autumn color traits, quantifiable using simple digital imaging, can serve as cost-effective indicators in tree breeding programs.

Seasonal color shifts and genetic diversity in Scots pine: a generalizable RGB imaging and CNN-based framework for conifer seedlings

期刊论文
出版模式：

开放获取
发表日期：

2025年10月30日

查看全文

Jiří Chuchlík,
Jaroslav Čepl,
Eva Neuwirthová,
Jiří Korecký,
Jan Stejskal

摘要

Key message

Leveraging affordable red-green-blue (RGB) imaging and neural network algorithms, this study delivers a high-throughput method to quantify seasonal color shifts and genetic variation in Scots pine seedlings. Hue-saturation-brightness (HSB) color analysis, and RGB values can be used for population and seasonal differentiation and hold potential for advancing breeding programs in forestry under changing climatic conditions.

Context

Scots pine (Pinus sylvestris L.) displays remarkable genetic and phenotypic diversity, with seasonal color changes such as autumn reddening, reflecting population-level responses to local environmental conditions. Advances in imaging and deep learning now enable precise quantification of such phenotypic variation, providing new insights into population-level variation.

Aims

This study assesses seasonal color variation within and among Scots pine seedling populations, compares the effectiveness of RGB and HSB systems for population and seasonal differentiation, and investigates phenological patterns across progenies of three seed orchards from ecologically distinct populations.

Methods

One-year-old seedlings from lowland (Plasy, Trebon) and upland (Decin) populations were imaged in a common garden trial in September, October, and January using a handheld camera. Needle-level segmentation was performed via a convolutional neural network. Genetic variability and population differences were analyzed using linear mixed models.

Results

Population differentiation reached the highest values in the RGB blue channel (Q_ST-blue = 0.64 in September and Q_ST-blue = 0.94 in October) and in HSB (Q_ST-hue = 0.61, Q_{ST-saturation} = 0.62 in September and Q_{ST-saturation} = 0.64 in October). Color wheel visualizations revealed converging hue and saturation trajectories, indicating gradual phenological changes in the post-growing season. September values exhibited the highest heritability (h²_RGB = 0.12–0.25; h²_HSB = 0.12–0.29) among measured optical traits.

Conclusion

This study demonstrates that RGB and HSB color parameters, extracted from high-throughput image analysis using CNN-based needle segmentation, capture both seasonal and genetic variation in Scots pine seedlings. The highest genetic differentiation and heritability occurred during early autumn, particularly in the blue and saturation parameters. These findings suggest that autumn color traits, quantifiable using simple digital imaging, can serve as cost-effective indicators in tree breeding programs.