Turmeric sourced from six retailers was processed into a powder and adulterated with metanil yellow at concentrations of 0.0-30% w/w. A handheld near infrared (NIR) spectrometer was used to obtain spectral scans of the samples, which were preprocessed using Savitzky-Golay first derivative (SG1) approximation using 61 smoothing points and second order polynomial. The preprocessed spectra were analyzed using principal component analysis (PCA) followed by classification by soft independent modeling class analogy (SIMCA) and used to group the adulterated turmeric powder samples according to the source (i.e., processor) of adulteration. Results showed the first principal component (PC1) of PCA models was sensitive to adulteration level, but when coupled with SIMCA, un-adulterated and adulterated samples could be classified according to their source despite having high levels of metanil yellow. At 5% level of significance, all of the samples were correctly classed for origin during validation. Some samples were classified under two groups, indicating possible inherent similarities. When PCA model was built using only un-adulterated samples, the PCA-SIMCA model could not classify the adulterated samples, but could classify those with very low levels [ 2% (w/w)] of metanil yellow, allowing for segregation of adulterated samples but not identification of sources.The combination of NIR and PCA-SIMCA modeling is a great tool to detect not only adulterated turmeric powder but, potentially, deter it in the future as the source of adulterated food can be traced back to the source of adulteration.