The potential of near infrared spectroscopy (NIRS) for determining the total arsenic (As) content in the prostrate amaranth (Amaranthus blitoides S. Watson) was assessed. Seventy-four samples belonging to this species, were harvested at different maturity stages along the polluted area, and then were scanned by NIRS. Their As reference values were obtained by atomic absorption spectrometry and they were regressed against different spectral transformations using modified partial least square (MPLS) regression. First derivative transformation equation of the raw optical data, previously standardized by standard normal variate (SNV) and De-trending (DT) transformations, resulted in a coefficient of determination (r(2)) in the external validation of 0.63, indicative of equations that can be used for a correct separation of the samples into low, medium and high groups. The standard deviation to standard error of prediction ratio (RPD) and range to standard error of prediction ratio (RER) for the first derivative equation were similar to those obtained for other trace metal calibrations reported in NIRS reflectance. Major cell components such as chlorophyll, lipids, starch and proteins were used by MPLS for modeling the equations. The use of NIRS for the determination of the As content in A. blitoides plants offers an important saving of time and cost of analysis.