Review Reports - High-Resolution Estimation of Monthly Air Temperature from Joint Modeling of In Situ Measurements and Gridded Temperature Data

Round 1

Reviewer 1 Report

The paper describes a method to estimate high-resolution air temperature by combining data from in-situ observation and the Landsat LST. The analysis is done for all seasons, unlike many existing studies that is confined to the summer season. However, some additional explanation of the methodology is desired, especially on the consistency of the analysis for summer and that for all seasons.

[Main comments]
There are a number of studies on high-resolution estimation of air temperature using the Landsat LST data, but they are mainly confined to summer. In this respect, it is valuable that the analysis is done for all seasons in the present study. However, the main part of the analysis and presentation is made for summer, and there are some points where a more detailed explanation is desirable about the consistency of the analysis for summer and that for all seasons.

@ It is better to write in an early part of the paper that the analysis was done for all seasons (so I understand). It is desirable to describe the procedure of the analysis for the entire year, and the meaning of the analysis for summer in this context.
The authors say "we generate --- with less than 15% cloud cover in June, July, and August" (Line 122). Does it mean that the model was constructed using data only for the three summer months, and was applied to other seasons?

@ The analysis is based on fixed coefficients for surface parameters except for impervious surfaces (Line 174). However, conditions of vegetated surfaces should be different according to seasons. Surface conditions will also vary day by day according to the wetness (between dry periods and just after rainfall) and snow cover. It is desired to add a discussion on the possibility of such variations affecting the results. It is at least desired to describe whether there was snow cover in winter.
Additionally, the duration of daytime changes with seasons. Data for 7-8 PM was used in the analysis for summer, but the daytime is shorter in winter. It will be better to add explanation on this point.

[Other comments]
@ There is ambiguity in the expression such as "monthly maximum temperature" and "average maximum temperature". I understand the analysis was made for the monthly average of daily maximum temperature. It will be better to write it in an early part of the paper.

@ Line 17 "Projected temperature increases ---": It is better to add a phrase like "in the world".

@ Line 50 "which estimate four- to eight-fold increases": Please make clear what estimates what.

@ Line 101 "This time range contains --- Landsat 8 LST estimates"; Line 119 "The Landsat LST product has a higher spatial resolution --- between the years of 2014 and 2020.": Isn't it possible to write quantitatively the data amount and/or the number of days that was used for constructing the model?

@ Line 185 "percentage bias (PB)": I understand PB is valid only for a quantity of ratio scale, and is not applicable to temperature.

@ Line 206 "Table A2": A1

@ Line 211 "our estimates show a positive coefficient ---": Where is the result shown?

@ Please write the time of the day for which the analysis in Fig.3 was made.

@ Line 223 "negligible (-0.05)": Do you mean -0.05F?

@ Line 223 "Table A3": A2

@ Please write the season for which the analysis in Fig.4 was made. I understand the analysis was made for July, as written in 2.3. It is right?

@ Caption of Fig.6: "Estimated maximum July temperatures and thermal anomalies ---": I think "and thermal anomalies" should be deleted.

@ Line 257 "(5d-5f)": 6d-6f

Author Response

Thank you for taking the time to review our manuscript. We've addressed your feedback below:

Reviewer 1:

[Main comments]

There are a number of studies on high-resolution estimation of air temperature using the Landsat LST data, but they are mainly confined to summer. In this respect, it is valuable that the analysis is done for all seasons in the present study. However, the main part of the analysis and presentation is made for summer, and there are some points where a more detailed explanation is desirable about the consistency of the analysis for summer and that for all seasons.

Thank you for this feedback. Our original manuscript does use summertime LST data to make predictions for all months, holding many of the effects constant over time. Although the cross validation does not show worse performance for non-summer months (Figure 3), all of the limitations you’ve raised are valid (varying surface conditions, time of day, snow cover, etc.).

We originally investigated using seasonally averaged LST data to capture these effects, but found the composite images were far too cloudy in non-summer months to serve as reliable predictor data. Given our cross validation results, we opted to include all months to show that summertime LST may still be useful for predicting results in other seasons.

However, upon a second round of consideration, we have decided to rerun our model results exclusively for summer months (June, July, August) given our paper’s broader discussions on extreme heat management. We’ve clarified this throughout the manuscript. This ensures that our LST data matches against the time period of study, and the potential limitations you’ve highlighted are much less of an issue. Although we would have liked to include a detailed analysis of the seasonal consistency of results, the transect based model data we used for comparison is only available for July. Without sufficient comparison data, we felt as though restricting our analysis to summer is preferable to having an inadequate analysis of seasonal consistency.

See above context.

The authors say "we generate --- with less than 15% cloud cover in June, July, and August" (Line 122). Does it mean that the model was constructed using data only for the three summer months, and was applied to other seasons?

See above context.

Additionally, the duration of daytime changes with seasons. Data for 7-8 PM was used in the analysis for summer, but the daytime is shorter in winter. It will be better to add explanation on this point.

See above context. We also note that the 7-8pm is for the CAPA model only.

[Other comments]

@ There is ambiguity in the expression such as "monthly maximum temperature" and "average maximum temperature". I understand the analysis was made for the monthly average of daily maximum temperature. It will be better to write it in an early part of the paper.

Thanks for highlighting this. Your interpretation is correct, and we’ve taken a pass through the entire manuscript to clarify this wording.

@ Line 17 "Projected temperature increases ---": It is better to add a phrase like "in the world".

Added “in the U.S” to clarify where this research was done.

@ Line 50 "which estimate four- to eight-fold increases": Please make clear what estimates what.

We have removed this part of the sentence since it was unclear and redundant with previously used statistics.

We added (n=5080) at line 125 to reflect the total number of individual Landsat scenes used in making the composite image for NC. We also added the total number of stations used to train the model (n=191).

@ Line 185 "percentage bias (PB)": I understand PB is valid only for a quantity of ratio scale, and is not applicable to temperature.

Thanks for catching this, we actually report out bias instead of percentage bias, but had the incorrect term listed here. All the values (and Table A2) appropriately use bias, which has degrees Fahrenheit for units.

@ Line 206 "Table A2": A1

Fixed.

@ Line 211 "our estimates show a positive coefficient ---": Where is the result shown?

Table A3

@ Please write the time of the day for which the analysis in Fig.3 was made.

This is observed daily maximum temperature, and doesn’t necessarily correspond to a constant time of day. We’ve tried to clarify this in the Figure labels. We introduce the two time of day windows for the comparison to the CAPA data because those were collected between specific hours.

@ Line 223 "negligible (-0.05)": Do you mean -0.05F?

Yes, these are bias units in degrees Fahrenheit. We’ve added the F for clarity.

@ Line 223 "Table A3": A2

Fixed.

@ Please write the season for which the analysis in Fig.4 was made. I understand the analysis was made for July, as written in 2.3. It is right?

Yes, the CAPA data was collected on July 23, 2021 as mentioned in the text. We’ve added that to the figure caption as well to make sure it is clear.

@ Caption of Fig.6: "Estimated maximum July temperatures and thermal anomalies ---": I think "and thermal anomalies" should be deleted.

We agree, we’ve adjusted the label.

@ Line 257 "(5d-5f)": 6d-6f

Fixed.

Reviewer 2 Report

The present article is focused on high resolution estimation of monthly air temperature from joint modeling of in situ measurements and gridded
temperature data.

The abstract is well developed and includes relevant information.

Introduction is very well structures, it is properly sustained by international references and provides a synthesis of the following sections.

Methods.

Please explain why do you select June as example in your figure.

At a general level, almost all figures must be improved in terms of quality. Please pay attention to this aspect.

In the sub-section 2.2 please check to provide explanations for all terms that you used in the equations.

Results

Figure 3 should benefit of more explanations. What is the accuracy of your model at monthly level?

Figure 6 is not introduced in the text. Please check and correct this aspect.

Overall, the article approaches several actual and important issues related to quantifying extreme heat.

More information should be however provided on the accuracy of your model in relation to other models that approach the same subject.

The authors need to perform a minor revision.

Author Response

Thank you for taking the time to review our manuscript. We have addressed your feedback below:

Reviewer 2:

The present article is focused on high resolution estimation of monthly air temperature from joint modeling of in situ measurements and gridded temperature data.

The abstract is well developed and includes relevant information.

Introduction is very well structures, it is properly sustained by international references and provides a synthesis of the following sections.

Methods.

Please explain why do you select June as example in your figure.

This is just an example to show what the data looks like, corresponding to one of the months of our study. We’ve added “To visualize the coverage of both data sources” to help clarify why we’ve included this figure.

At a general level, almost all figures must be improved in terms of quality. Please pay attention to this aspect.

We’ve checked our figures for readability and clarity to ensure they sufficiently communicate the intended information. If there are more specific aspects of individual plots that you would like us to improve, we would be happy to address these concerns.

In the sub-section 2.2 please check to provide explanations for all terms that you used in the equations.

We’ve checked to make sure that all terms are defined and explained in the text.

Results

Figure 3 should benefit of more explanations. What is the accuracy of your model at monthly level?

This is included in Table A3.

Figure 6 is not introduced in the text. Please check and correct this aspect.

Fixed.

Overall, the article approaches several actual and important issues related to quantifying extreme heat.

More information should be however provided on the accuracy of your model in relation to other models that approach the same subject.

In addition to our detailed comparison against the Durham, NC Heat Watch data, we have made clear that our results have comparable accuracy to other studies using LST to predict air temperature (Line 226).

The authors need to perform a minor revision.

Round 2

Reviewer 1 Report

I appreciate the authors' effort of revision. The article will be acceptable after checking some minor points.

@ The expression "maximum summertime monthly air temperature" (Line 6, 84) may be taken for monthly extreme temperature. It will be better to write in an early part of the paper that it means monthly averaged daily maximum surface air temperature (Line 284).

@ It is better to write the unit of temperature (F) in Figs.1 and 3.

@ Line 120 "Google Earth Engine [44?]" --- There may be a font error.

@ Caption of Fig.5 "nighttime CAPA" --- Do you mean evening?

@ Line 259 "5d-5f" --- 6d-6f

Author Response

We appreciate the second round of reviews. We've addressed the feedback as follows:

Thanks for catching these additional instances. We've updated them.

@ It is better to write the unit of temperature (F) in Figs.1 and 3.

Added.

@ Line 120 "Google Earth Engine [44?]" --- There may be a font error.

Was a missing reference, it is now added.

@ Caption of Fig.5 "nighttime CAPA" --- Do you mean evening?

Yes, adjusted.

@ Line 259 "5d-5f" --- 6d-6f

Corrected.