Temporal and Spatial Autocorrelation as Determinants of Regional AOD-PM2.5 Model Performance in the Middle East
Round 1
Reviewer 1 Report
Studying the important factors that affect the estimation performance of the regional AOD-PM2.5 model is of great significance to monitor PM2.5 based on remote sensing. The research is innovative and suitable for publication in special issues of the Remote Sensing journal. However, the manuscript is not well-developed. Specific comments:
- The manuscript is entitled "Temporal and spatial autocorrelation as determinants of regional AOD-PM2.5 model performance in the Middle East". The manuscript also made a spatio-temporal autocorrelation analysis of AOD, but the results of the analysis cannot support the conclusion that spatio-temporal autocorrelation is a decisive factor in the performance of the regional AOD-PM2.5 model.
- I suggest adding to the analysis in Figure 3 the reason why the accuracy of MISR AOD Raw products is better than that of MISR AOD products.
- In Table 1. Why is the accuracy of using a linear model to estimate PM2.5 based on MAIAC AOD lower than the accuracy of using a linear model to estimate PM2.5 based on MISR AOD, while the accuracy of using a nonlinear model is the opposite? I suggest that statistical index values such as MAE, MSPE and RMSE are given in Table 1.
- In Figure 4. A large number of documents indicate that relative humidity is an important factor affecting the accuracy of PM2.5 estimation. Why does the author's conclusion differ greatly from it? Why does the author use MSE to define the importance of many driving factors instead of PCA, CCA, CA, DCA and other methods commonly used in statistics?
- The discussion section is lengthy, the content is not concise enough, and the literature contribution is not enough. I suggest to supplement and concise it.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper is very well written, it deals with establishing models for PM2.5 over particular areas in complex land-use terrain of the Middle East. Paper represents massive work behind shown results; it provides many technical details, well elaborated. In the end authors, clearly discussed the advantages and disadvantages of use methods. The largest limitation of this work is definitely a small sample size of data; however, as authors clearly described issues at the end of paper, their understanding and explaining of problems somehow overcome this issue. I got only a couple of comments listed below. After minor corrections, the presented manuscript merits publication in the journal Remote Sensing
Minor comments:
Ln 88, 92. When referencing, the year should be noted as well. This is a comment for all references used in the manuscript.
Ln 167. Did authors tried to interpolate ERA5 data to local sites, are authors aware of the potential lowering of the performance of neural networks due to the relatively coarse resolution of grid data? Why not using measurements as well?
Ln 179. The same comment for MERRA – coarse resolution could lower performance?
Ln 386. What can be a reason for this behavior, is the model maybe too rough and meteorology is not playing a crucial role? Over this complex land-use terrain, meteorology should be important.
Ln 463. If model is applied on same type of areas, e.g. only for rural monitoring stations, is it possible to increase model transferability? When dealing with AQ models, rural monitoring stations are usually much more usable for comparisons with modeling data as it is hard to capture trends (and sources) in urban areas.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
I wonder if, in areas where measurements are scarce, the application of dispersion models, considering the different factors affecting air quality (e.g. meteorology, emissions, dispersion, chemistry, and deposition) could provide additional predictors for the adopted ML techniques and also useful information on the spatial distribution of the pollutants. Moreover, I guess that the order of the figures in appendix should be modified: Figure A3 was cited in the text before figures A1 and A2
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The author made a good response and revision to my comments. I suggest accepting the manuscript
