Assessment of nitrogen status and spatial distribution to enhance growth and inflorescence quality in Dendrobium Sonia ‘Earsakul’ using near-infrared hyperspectral imaging

Nitrogen is an essential macronutrient for the growth and quality of cut orchids, and both its deficiency and excess adversely affect plant vigour and inflorescence quality. This study evaluated nitrogen concentration in Dendrobium Sonia ‘Earsakul’ leaves using near-infrared hyperspectral imaging (NIR-HSI). It aimed to select the model from Partial Least Squares Regression (PLSR) and machine learning (ML), and to visualise nitrogen distribution in fresh and dried, powdered leaves. Samples were grouped into non-fertiliser, once-a-week, and twice-a-week fertiliser applications. Twice-a-week fertilisation significantly increased inflorescence length and floret number, which showed parallel increasing trends with leaf nitrogen concentration. A total of 300 samples were used for calibration and 75 for validation. For fresh leaves, second derivative preprocessing with PLSR combined with competitive adaptive reweighted sampling (CARS) provided the best prediction (residual prediction deviation (RPD) = 1.28). However, the model was only suitable for rough screening of leaf nitrogen status. For dried, powdered leaves, standard normal variate (SNV) with PLSR and genetic algorithm (GA) achieved higher accuracy (RPD = 2.00). Nitrogen maps showed clearer differentiation in dried leaves due to reduced water interference. Overall, NIR-HSI provides a rapid, non-destructive approach for nitrogen assessment. Whereas dried, powdered leaves provide a more reliable quantitative prediction.