Diémoz et al. presents a new PM10 source apportionment framework, RASPBERRY, based on aerosol physical properties derived from OPC and aethalometer measurements. By comparison with chemical PMF results, the authors showed the clear strengths of RASPBERRY. The manuscript is well organized and easy to follow. However, I still have some issues before the manuscript can be accepted.

Line 292, the authors mentioned “heuristic uncertainty” was optimized through trial-error methods or iterative approach. Given the strong influence of uncertainty definition of PMF outcomes, can authors provide more details how to derive this parameterization to ensure the reproductivity?

For traffic emissions between chemical and physical PMF results, their correlation is not that high, r2=0.45. The authors attributed to the detection limit of OPC. Is it possible to improve the physical PMF by introducing other factors, such as NOx?

The authors also mentioned RASPBERRY improves the efficiency of PMF. How much faster is RASPBERRY is compared to a normal PMF? Can we also apply RASPBEERY to the chemical PMF? If so, it would be helpful to discuss whether and how the results would differ from those obtained using a normal chemical PMF approach.

see attached pdf

This study presents RASPBERRY (Real-time Aerosol Source apportionment using Physics-Based Experimental data and multivaRiate factoR analYsis), a novel PM10 source apportionment framework that integrates particle size distributions and spectrally resolved light absorption into a unified Physical PMF (Positive Matrix Factorization). Applied to a five-year hourly dataset (2020–2024) from an urban background site in the Italian Alps, the method successfully identifies six aerosol source factors and demonstrates strong agreement with traditional chemical source apportionment techniques, ground-based remote sensing, and atmospheric modeling tools.

RASPBERRY addresses a critical gap in air quality monitoring: the ability to perform high temporal resolution, real-time, continuous, and cost-effective aerosol source apportionment that is based on measurements of aerosol physical properties.  

The paper is well written and methodologically sound. The 5-year dataset provides excellent statistical robustness and seasonal coverage. The bootstrapping approach for uncertainty estimation reflects best practices in PMF analysis. Validation with chemical PMF, lidar, AERONET AOD, and CAMS reanalysis are thorough.  Overall, this paper represents a meaningful and timely contribution to the aerosol science and air quality communities.

There are several minor issues:

1. In line 309, the impact of the selected coefficient A, α, and C₃ are not mentioned. Also it is mentioned that “no other modeling uncertainty was included”. But other uncertainty sources are not mentioned here but instead was in discussion. So it should be briefly explained to guide the audience to discussion about the rest of the uncertainties.
2. In line 330, the seasonal balanced sample selection is mentioned, if the samples are not daily, is the diurnal sampling also balanced?
3. Line 610, the pink means dust touching the ground in Figs. S24 or in Figure 7? Also is there any lidar to proof that the differences between AERONET and RASPBERRY are due to aloft and surface dust?
4. Figure 8a, the differences between chem-PMFd1 and RASPBERRY is very hard to tell due to similar colors. For Fig. 8b, the liner regression cannot represent the true trend of this data. What causes the large scattering in traffic emission contribution? For Secondary rich factors, high bias can be seen throughout, what is the main cause?
5. Figure 13 e and f, the vertical level where backscattering shows smoke, does not have valid retrieval of depolarization ratio in g and h. (For example, the small chunk of smoke at 5km in 13f, the location is corresponding to noise/no-retrieval for 13h.)