Robust Non-Destructive Prediction of Jackfruit (Artocarpus heterophyllus cv. Tekam Yellow) Colour at Different Maturity Stages Using Visible–Near Infrared Spectroscopy: Influence of Rind and Flesh

Abstract

Abstract The Malaysian jackfruit ( Artocarpus heterophyllus ) industry is increasingly challenged by a physiological disorder known as jackfruit bronzing , attributed to Pantoea stewartii subsp. stewartii . This disorder manifests as a yellowish-orange to reddish discolouration of the pulp while leaving the rind visually unaffected, leading to substantial postharvest quality and economic losses. The Tekam Yellow cultivar, in particular, has demonstrated high susceptibility to this condition. This study aimed to evaluate the potential of visible near-infrared spectroscopy (Vis–NIRS) as a non-destructive analytical tool for the early detection of internal bronzing through the estimation of rind and flesh colour parameters (L*, a*, b*, C*, ΔE, and h°). Spectral data were collected from the rind surface of intact jackfruit samples at 10, 12, and 14 weeks after anthesis (WAA) across the 500–950 nm wavelength range. Partial Least Squares Regression (PLSR) models were developed to establish the relationship between spectral reflectance and reference colour metrics. Various spectral pre-processing techniques—including Savitzky–Golay smoothing, Standard Normal Variate (SNV), and Multiplicative Scatter Correction (MSC)—were applied to enhance signal quality and minimise scattering effects. The optimised models demonstrated high predictive accuracy, with coefficients of determination for calibration ( R c²) and prediction ( R p²) reaching up to 0.98. Correspondingly, root mean square errors of calibration (RMSEC) and prediction (RMSEP) were as low as 0.29. Overall, calibration performance remained consistently strong across traits and maturities ( R c² ≥ 0.62), indicating that the model structures effectively captured the spectral–colour relationships within the calibration dataset. In contrast, the predictive performance of independent validation models varied substantially between normal and bronzing conditions. Under normal conditions, several models achieved excellent predictive accuracy—for instance, rind colour L* at 10 WAA ( R p² = 0.95, RPD = 16.19) and flesh colour L* at 10 WAA ( R p² = 0.98, RPD = 10.88). Conversely, models developed under bronzing conditions frequently exhibited lower predictive coefficients ( R p²) and residual predictive deviation (RPD) values ( p ≤ 0.05 to confirm the significance of observed differences. Collectively, the results demonstrate that Vis–NIRS applied through the rind offers a promising non-invasive approach for the rapid and accurate detection of internal bronzing in Tekam Yellow jackfruit, facilitating improved quality monitoring and early disease detection at both harvest and postharvest stages. This work highlights the potential of Vis–NIRS as a practical, high-throughput phenotyping and quality assurance tool to support sustainable value-chain management in the Malaysian jackfruit industry.

Plant phenotyping relevance

Vis–NIRSとPLSRを用いてジャックフルーツの内部ブロンズ症および果肉・果皮色を非破壊推定する手法を開発・検証しており、植物状態の取得方法が研究の中心である。

Code and data availability

The paper reports Vis–NIR spectral data (1,620 spectra) and PLSR models for jackfruit colour prediction, but no public repository, deposit, or authors' URL for the spectral dataset, models, or analysis code is provided in the supplied blocks. The Data Availability Declaration states the datasets are available but gives

No evidence-backed public reproduction asset is currently recorded.