Spectral network analysis illuminates coordinated plant traits across a climate gradient

Understanding how plant populations respond to environmental variation through functional leaf traits remains challenging due to limitations of traditional phenotyping approaches. Hyperspectral reflectance offers a powerful high-throughput solution, simultaneously capturing leaf biochemistry, water content, and structural properties across hundreds of wavelengths. We present a framework combining hyperspectral data, inverse modeling, and network analysis to investigate population-level variation in Streptanthus tortuosus. Using a common garden experiment with four populations, we apply supervised methods (partial least square discriminant analysis; ridge regression) to identify which spectral features differ among populations, and an unsupervised spectral network approach to characterize how wavelength correlations are organizationally structured within each population, where we treat coordination architecture itself as a population-level phenotype that can vary with environment. The framework detects distinct, heritable spectral signatures across populations, population differences in anthocyanins, carotenoids, Chl, water content, and population-specific network architectures. Thermally variable environments were associated with greater spectral modularity, demonstrating that trait coordination architecture varies with climate of origin. This approach addresses the phenotyping bottleneck in evolutionary ecology, providing a scalable, high-throughput tool for characterizing genetically based population differences in both individual traits and their coordination, with broad applications for monitoring plant population responses to climate change.