Application of non-destructive sensors opens a promising opportunity to provide efficient information on nutrient contents based on leaf or canopy reflectance in different crops. In potatoes, nutrient levels are estimated by conducting chemical tests for the petioles. In thinking of deploying sensors for potato nutrient estimation, it is necessary to study the spectrum based on petiole chemical testing rather than leaf chemical testing. Thus, this study aimed to investigate whether there is a correlation between the chemical testing of phosphorus (P) in potato petioles and leaf spectral data. A total of 40 datapoints were collected from outdoor farming following the standard application of nutrients. Additional 20 datapoints were collected from indoor growing with excessive and absence application of P. Both petiole and leaf samples were collected from the fourth leaf from the tip of the shoot following the current tissue testing protocol. Samples were collected biweekly starting from the 45th day after planting. Leaves and petioles were dried before chemical testing and spectral analysis. Chemical testing of petiole was done following the official methods of the Association of Official Analytical Chemists (AOAC). The dried leaves were placed in the NIRS Analyzer to measure the reflectance between 400-2500 nm. A dataset was then developed between P content and leaf spectrum based on a linear relationship. Prediction accuracy was improved by following the shrinking method of Lasso multiple linear regression modelling (Lasso MLR). Performance of the generated model was evaluated using the coefficient of determination (r2), and RPD ratio between the standard error of prediction and standard deviation of actual concentrations. Measurable content of P ranged between 0.07% to 0.7%. The results showed higher concentrations in early stages of the growing season than later stages for the samples taken from the outdoor farms and indoor growing area. The excessive application of P indoor provided datapoints of high-concentration P which enhanced the variation of datapoints across different concentration levels which improved modelling. Lasso MLR showed an equal distribution of estimated concentrations around the fitting line with an r2 value of 0.81 and an RPD value of 2.31 that showed an excellent model performance. Our results showed that the significant wavebands for P estimation were in the visible and very near infrared (400 - 1100 nm). The results of this research show that there is a correlation between the P content of potato petioles and dried leaf spectrum. Further work is planned to validate the significance of the developed model.