<p>To reduce the environmental risks associated with livestock manure and to facilitate the nutrient recycling facilities, precise characterization of manure nutrients is vital. However, conventional laboratory testing facilities are financially burdensome, time-consuming, and also prone to sample degradation during transportation. To overcome these challenges, this study utilized near-infrared spectroscopy (NIRS) combined with an optimized partial least squares (PLS) framework to rapidly characterize the eight vital nutrients of manure: dry matter (DM), organic matter (OM), ammonium nitrogen (NH₄⁺), total nitrogen (Total-N), phosphorus pentoxide (P₂O₅), calcium oxide (CaO), magnesium oxide (MgO), and potassium oxide (K₂O). The model was trained on a diverse dataset of 490 fresh manure samples across six distinct livestock types. A two-step feature selection technique was introduced to identify the significant wavelengths across (852–2502&#xa0;nm) range in an initial count of 1003 features. After the removal of zero-variance features with a correlation-based filtering approach, variable importance in projection (VIP) analysis was applied to identify the significant NIRS wavelengths. Two distinct modelling pipelines were evaluated: PLS-comprehensive model with identified wavelengths on the full spectral range and PLS-reduced model was solely trained on the peak dominant wavelengths. The PLS-comprehensive model demonstrated superior predictive performance across all the target variables, DM (R² = 0.95, SD = 2.94), OM (R² = 0.92, SD = 29.18), Total-N (R² = 0.91, SD = 2.00), AN (R² = 0.91, SD = 0.81), P₂O₅ (R² = 0.88, SD = 2.22), K₂O (R² = 0.89, SD = 2.60), CaO (R² = 0.84, SD = 2.60) and MgO (R² = 0.92, SD = 0.68). The model was successfully validated and proved its effectiveness to rapidly characterize the concentrations of multiple key manure nutrients.</p> Graphical abstract <p></p>