The manuscript describes the use and evaluation of an online energy-dispersive X-ray fluorescence (EDXRF) detector, the Horiba PX-375, for elemental analysis of ambient particulate matter. The team characterized the performance of the detector, including its limit of detection and measurement uncertainties, and compared the field measurements.

This online EDXRF technique offers advantages for non-destructive, near-real-time, and continues measurements, as well as source apportionment. A comprehensive study and understanding of the detector’s performance is highly desired. I would recommend accepting the manuscript with minor edits.

Line 160-171: Equations (1) and (2) need clarification. In particular, is the standard deviation term the average of three standard deviations? Also, how is the calibration curve derived?  

Line 175-177: Is the standard uncertainty calculated from the mean of a series of observations, or from the standard deviation of the observations, or from several means of several series of observations?

Line 198-199: Are 50 kV and 15 kV the energies of the incident photons radiating the samples? Or are they the incident beams hitting some targets, generating photons that then excite the samples. My understanding from the manuscript (Line 137) is that they are the incident photon energies directed at the sample. However, in that case, I am not sure why 50 kV is used instead of 15 kV for Fe. Fe is excited with 15 kV and has a higher cross section at 15 keV than 50 keV. In fact, many elements listed in the 50 keV section of Table 2 should be excited with 15 keV incident photons. It may be helpful to include more instrument details.

Line 240: Does “self-absorption” here refer only to the signal absorption by the particles themselves, or does it include more general signal absorption, such as by the air path, detector window, etc.? Corrections for the air path and window thickness should be implemented in the data quantification. The impart of particle absorption can be estimated by the size of the particles.

Lastly, I suggest adding a description of the spectrum analysis, including, for example, spectrum fitting and peak identification.

Technical  corrections:

Line 82, should be “several months”