Combining Spatial and Temporal Data to Create a Fine-Resolution Daily Urban Air Temperature Product from Remote Sensing Land Surface Temperature (LST) Data

Round 1

Reviewer 1 Report

 

Remotely sensed land surface temperature (LST) is often used as a proxy for air temperature in urban heat island studies. However, two pre-dominantly used sensors to observe the LST each have shortcoming that limit their utility for urban application. In this study, the authors optimized a DisTrad method to combine data of several source to provide fine scale LST at four time daily any time of the year. This work is interesting and the authors give an overall clear organization of their manuscript. However, there are a few issues I’d like the authors to clarify before publication, thus I’d give major revision at this moment. As follows:

Major comment

1.     In Fig. 5, while the regression for the intercept is seemingly good fit, the fit for the slope is uncertain due to large uncertainty at around 25°C-30°C, I’d like the authors to provide more justification of their use of logistic functions on this issue.

2.     Does the linear relationship assumption hold throughout the whole data. In Figure 11, it seems that although most data are within the linear line, there are individual stations (1 or 2) well out of the bound of this linear regression, I’d like the author give this a little analysis on this issue.

3.     How is the present study applicable to stations other than used in the present study? Since the framework is designed to apply to a wide range of stations, it would be interesting to discuss this applicability in the conclusion and discussion section.

Minor comments

Line 6, “ JOANNEUM RESEARCH Forschungsgesellsch and nd t mbH”,  I have searched the internet and it seems the one resemble this is “Joanneum Research Forschungsgesellschaft mbH” ? Forgive me if I searched it wrong.

Line 16-23, it seems the authors give too much introduction in this paragraph since these sentences can mostly be shown in Introduction rather than here. The abstract is meant to be concise and straight to the point, and more background info are mostly kept in introduction instead.

Figure 1-4, would it be better if the 4 figures can be combined to show an overall work step of the method rather than separating them into 4?

Line 250-251, following the major comments 1, it may be best to illustrate some more in the non-straightforward or “species-dependent” cases to show why they seem to scatter more about the fitted lines than the others.

Line 333-334, “Top left” , “Top right”, where do these refer to? Since I do not see an upper & lower figures in Figure 9.

 

Author Response

Please see the attached word document

Author Response File: Author Response.docx

Reviewer 2 Report

Review report on the manuscript entitled “Combining spatial and temporal data to create a fine-resolution daily urban air temperature product from remote sensing Land Surface Temperature (LST) data” by David Neil Bird, Ellen Banzhaf, Julius Knopp, Wanben Wu, and Laurence Jones submitted in the journal atmosphere.

I appreciate the attempt by the authors to propose a method for generating finer spatiotemporal resolution air temperature over an urban area using LANDSAT and MODIS data. The flow and representation is good with some interesting points. Although the problem is highly demanding for the benefit of the society, however, I have some issues including a few major related to the present manuscript which are provided below:

1. The authors have used a high pass filter that has a tendency to smooth out the data, particularly highly deviated data. In practical, those peaks may be true and may not be outliers, which means that the proposed method will be less efficient in capturing the temperature extremes. The authors may explain or do some extra analysis to understand that whether the proposed method is suitable/satisfactory to capture the temperature hotspots or not.
2. For the validation over the two stations, the authors may check the Tmax deviations of 1- and 2-standard deviations of long-period and validate with the converted Tmax. What is mean hit score of generated data compared to observations in capturing warmer days over those stations?
3. Line 361. Figure number will be 11 instead of 10.
4. Considering Le Bourget and Orly stations Tmax as shown in Figure 11, it is found that linear relationship may not be a good approach for hot days. In that case, it is suggested to do a bias correction for capturing the warmer days.
5. What is Prob (F) in Table 4? It is not clear what is UTMX (and UTMY) showed in Figure 10.
6. What is the period for generating synthetic LST? Is it sufficient to do statistical test?

Author Response

Please see the attached PDF documents

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I think the author has answered my questions, and I'd like to suggest publication.

One thing to remind during the proofreading stage is that the author can modify typo for the minor mis-reference of figure 11 (quadratic regression) as figure 10 in around line 358-362.