Enclosed, please find my comments.

The manuscript describes a QCL-based photoacoustic aerodynamic gradient system for measuring half-hourly NH3 fluxes under agricultural field conditions. The system's performance was validated through field measurements of NH3 fluxes under predominantly dry soil conditions. The key drivers of NH3 surface-atmosphere exchange during daytime and nighttime were resolved by using a machine learning method. Overall, the combined QCL–photoacoustic gradient system had robust field performance and low instrumental bias and can be used for long-term NH3 flux measurement. The manuscript well fits the scope of AMT. It can be recommended for publication after the following comments are addressed.

Specific comments:

Reliable quantification of NH3 remains challenging due to the inlet interactions. However, the NH3 losses in the inlet were not evaluated or calibrated. Though the NH3 loss in both inlet lines was similar, it would bias the accurate quantification of NH3 fluxes. The uncertainties induced by the inlet interactions should be evaluated. This can be done by using standard gases or a permeation tube of NH3.

Lines 38-40: Given that the system has a high detection limit (~ 3 ppb with 2 min average), it may not be applicable to relatively clean conditions. The authors should emphasize the applicability for “agricultural conditions”.

Lines 161-162: How long is the sampling line? What is the residence time in the sampling line?

Lines 177-178: Please elaborate on how the sampling-related measurement uncertainty was derived.

Lines 307-310: Is the intrinsic system memory induced by the measurement system or soil property? More explanation is needed to help understand the mechanistic presence of system memory.

Lines 336-340: This is an important result and should be given as a figure in the main text.

Lines 477-485: It is better to have figures to show the comparison between predicted and measured NH3 fluxes.

Technical comments:

Line 93: define QCL.