Next Article in Journal
Socio-Economic Determinants of Climate Change Adaptation Strategies Among Smallholder Farmers in Mbombela: A Binary Logistic Regression Analysis
Previous Article in Journal
Barriers, Opportunities, and Best Practices for Corporate Climate Transition Plans: A Literature Review
Previous Article in Special Issue
An Assessment of the Carbon Budget of the Passively Restored Willow Forests Along the Miho River, Central South Korea
 
 
Article
Peer-Review Record

Spatiotemporal Analysis of Urban Heat Islands in Kisangani City Using MODIS Imagery: Exploring Interactions with Urban–Rural Gradient, Building Volume Density, and Vegetation Effects

Climate 2025, 13(5), 89; https://doi.org/10.3390/cli13050089
by Julien Bwazani Balandi 1,2,*, Trésor Mbavumoja Selemani 2, Jean-Pierre Pitchou Meniko To Hulu 3, Kouagou Raoul Sambieni 1,2, Yannick Useni Sikuzani 4,*, Jean-François Bastin 2, Prisca Tshomba Wola 1, Jacques Elangilangi Molo 1, Joël Mobunda Tiko 1, Bill Mahougnon Agassounon 2,5 and Jan Bogaert 2
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Reviewer 4: Anonymous
Climate 2025, 13(5), 89; https://doi.org/10.3390/cli13050089
Submission received: 7 March 2025 / Revised: 25 April 2025 / Accepted: 28 April 2025 / Published: 29 April 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This work uses Land Surface Temperature (LST) data from 2000 to 2024 to investigate the trends of Urban Heat Islands (UHI) in Kisangani, DRC.   Measuring by the Normalized Difference Vegetation Index (NDVI), the study examines the link between LST, building volume density (BVD), and vegetation density.

The study suggests effective strategies to mitigate the UHI effect in Kisangani, such as integrating green infrastructure, urban planning, and community engagement. But the result and discussion are not mentioned in detail of these strategies; analyze their effectiveness and optimized for the local context.

Are there any discussion about integration of climate change projections to assess how UHI effects might be exacerbated or altered under different climate scenarios. This would help in developing more robust and long-term mitigation and adaptation strategies.

Although this study made use of MODIS data with a 1 km spatial resolution, the studies could not examine UHI trends at a finer scale using higher-resolution data (e.g., Landsat, Sentinel).  This would enable a closer examination of micro-scale fluctuations in UHI inside the metropolis. 

It is clear from the study that GHSL building volume figures can be wrong, especially in cities that change quickly. Even though the authors tried to make sure this data was correct, there may still be some mistakes and incorrect ways to grasp the data. 

Urban-rural gradient zone definition based on morphological factors and population density thresholds could be random in some degree.  Various thresholds or criteria could produce different findings. How the author could explain this situation?

Author Response

Comments and Suggestions for Authors

This work uses Land Surface Temperature (LST) data from 2000 to 2024 to investigate the trends of Urban Heat Islands (UHI) in Kisangani, DRC.   Measuring by the Normalized Difference Vegetation Index (NDVI), the study examines the link between LST, building volume density (BVD), and vegetation density.

Dear reviewer, on behalf of our entire team, we would like to thank you sincerely for your constructive comments, which have helped to improve this manuscript. You will find below our answers to your questions. Thank you very much.

The study suggests effective strategies to mitigate the UHI effect in Kisangani, such as integrating green infrastructure, urban planning, and community engagement. But the result and discussion are not mentioned in detail of these strategies; analyze their effectiveness and optimized for the local context.

R/
We have improved the section of the discussion dealing with these strategies to analyse in detail their effectiveness and contextualise them more in relation to Kisangani. This revision also demonstrates how these strategies are relevant, effective, and adapted to the local context (socio-economic, environmental, climatic, etc.) (lines 479-540).

 

 


Are there any discussion about integration of climate change projections to assess how UHI effects might be exacerbated or altered under different climate scenarios. This would help in developing more robust and long-term mitigation and adaptation strategies.

R/ Thank you very much for your comments. We have improved the discussion section to include climate scenarios in the design of mitigation strategies (lines 485-493).

Although this study made use of MODIS data with a 1 km spatial resolution, the studies could not examine UHI trends at a finer scale using higher-resolution data (e.g., Landsat, Sentinel).  This would enable a closer examination of micro-scale fluctuations in UHI inside the metropolis. 

R/ In our methodology (lines 161-167 ), we acknowledge the limitation highlighted regarding the spatial resolution of MODIS data (1 km), which may not capture the finer-scale heterogeneity of Urban Heat Island (UHI) patterns within the metropolis. However, MODIS was primarily motivated by its high temporal resolution and consistency, which allowed us to analyze long-term UHI dynamics across multiple years and seasons with minimal cloud contamination. Each pixel in MOD11A2 represents the mean value derived from all corresponding MOD11A1 LST pixels recorded over the 8-day interval. This approach reduces the effects of punctual anomalies, such as clouds or sensor errors, by averaging data. It is, therefore, more reliable than products based on a single daily or instantaneous observation. In addition, MODIS LST data (including MOD11A2) have been extensively validated through comparisons with in situ measurements, which have consistently shown high accuracy under clear and stable atmospheric conditions [27-28].

However, we agree that incorporating higher-resolution thermal data from sensors such as Landsat (30 m) or Sentinel-3 SLSTR (1 km but with more spectral detail), or even integrating Sentinel-2 data for urban land cover mapping (10–20 m), would enable a more precise assessment of micro-urban variations in land surface temperature. Our Future work will aim to complement our study by combining MODIS data with higher-resolution imagery, possibly through data fusion techniques or downscaling methods, to better capture intra-urban thermal contrasts and inform local-scale urban planning interventions.


It is clear from the study that GHSL building volume figures can be wrong, especially in cities that change quickly. Even though the authors tried to ensure this data was correct, there may still be some mistakes and incorrect ways to grasp the data.

R/ We appreciate the reviewer’s concern regarding the potential limitations of GHSL building volume data in rapidly urbanizing areas. Recognizing this issue, we undertook a validation procedure tailored to the specific context of Kisangani. In particular, we cross-validated GHSL-derived building volume estimates with independent data sources by computing building volumes from OpenStreetMap building footprints and height values derived from remote sensing. Building heights were estimated by subtracting digital terrain model (DTM) values (SRTM30) from digital surface model (DSM) values (AW3D30), thereby producing an alternative volume dataset.

Subsequent statistical comparisons between GHSL data and our computed volumes showed no significant differences, suggesting that GHSL estimates are reasonably accurate for the study area. This validation step strengthens the reliability of our analysis while acknowledging the general limitations associated with global datasets. (lines 177-183)

Urban-rural gradient zone definition based on morphological factors and population density thresholds could be random in some degree.  Various thresholds or criteria could produce different findings. How could the author explain this situation?

R/
We appreciate the reviewer’s thoughtful comment on the potential arbitrariness in defining urban–rural gradient zones based on morphological and demographic thresholds. To address this concern, we relied on an established and replicable decision tree approach proposed by Marie André et al. [19, 34, 35], which is grounded in morphological criteria (built-up, agricultural, and vegetated areas) and refined using high-resolution satellite imagery available from Google Earth. This method has been recognized for its capacity to closely reflect the spatial reality of urbanization processes [19, 20, 36, 37].

In our implementation, each pixel was classified according to its intensity value (0–255) within a GIS environment, allowing consistent differentiation between impermeable surfaces (built-up: 80–255), transitional/agricultural zones (50–80), and natural vegetation or forest (below 50). These classifications were systematically applied for each year from 2000 to 2024. To enhance robustness, morphological delineation was further cross-validated using population density data, with a threshold of 100 inhabitants/km² used to distinguish zones transitioning to urban character, based on the lowest densities observed around Kisangani’s urban core over the study period.

As with any classification approach, alternate thresholds might slightly shift zone boundaries. However, our method remains rooted in tested frameworks and reflects a consistent, transparent, reproducible approach. Additionally, we aimed to minimize subjectivity by combining spectral, morphological, and demographic dimensions.

Reviewer 2 Report

Comments and Suggestions for Authors

Considering the most important international trend, carbon neutrality and climate change response, the necessity of this study is secured.

Through urbanization, the point of time when the area where the urban heat island phenomenon occurs expands and increases was well analyzed.

The authors also need additional description of the results of the study, the limitations of the study, and future studies.

In addition, if the discusstion or conclusion part also discussed natural disasters and disasters that may occur due to urbanization, it is expected that the completeness of the study would be increased.

Comments on the Quality of English Language

The English could be improved to more clearly express the research.

Author Response

Comments and Suggestions for Authors

Considering the most important international trend, carbon neutrality and climate change response, the necessity of this study is secured.

Through urbanization, the point of time when the area where the urban heat island phenomenon occurs expands and increases was well analyzed.

Dear reviewer, On behalf of our entire team, I would like to thank you sincerely for your constructive comments, which have helped to improve this manuscript. Below are our answers to your questions.

We have improved the quality of the English to make the text easier to understand.

Thank you very much.

The authors also need additional descriptions of the results of the study, the limitations of the study, and future studies.

R/ Thank you very much for your comment. We have improved the Methods section (lines 161-167) to highlight the limitations of the study, while justifying the main reasons behind the choices made.

In addition, if the discusstion or conclusion part also discussed natural disasters and disasters that may occur due to urbanization, it is expected that the completeness of the study would be increased.

R/ Thank you very much for this comment. Lines 531 and 540 of the discussion have been improved to include this aspect of the natural disaster.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors investigated the spatio-temporal patterns of the surface urban heat island (SUHI) in Kisangani and explored the relationships between land surface temperature (LST) and two variables: building volume density (BVD) and the Normalized Difference Vegetation Index (NDVI). However, the manuscript failed to clarify its logic. The reviewer find it challenging to identify the study’s central focus, as each section appears to address topics already extensively covered in prior research.

Specific comments follow:
1.The title should be revised to explicitly articulate the logical connections between analyzing spatio-temporal patterns and the three key themes: "Urban-Rural Gradient, Building Volume Density, and Vegetation Effects." Specifically, the urban-rural gradient represents a spatial dimension, BVD serves as an architectural index, and vegetation effects denote ecological influences. 
2.Lines 44-70 "Increasing global, regional, and local urbanization fosters ... due to such challenges."
In the first paragraph of the introduction, the reviewer considers whether it is too long. The description of the research background should be focused, simple and clear.
3.Lines 72-85 "In this context, Garcia-Herrera et al. ... worsening extreme heat events."
Simply listing existing studies is not an appropriate approach. It is recommended that authors conduct a review to sort out and summarise existing studies.
4.Line 86-94 "Despite extensive studies on the UHI effect ... further highlights spatial variations in UHI effects."
Kisangani faces severe UHI effects illustrating the necessity of selecting the study area. The new study area does not seem to be an innovative point of an article. The authors are advised to explain the differences between this study and other studies in terms of research methodology and elements. Additionally, the authors were advised to choose a core research component rather than studying the spatio-temporal patterns of SUHI and then analysing the relationship between LST and BVD. The authors do not explain the logical relationship that exists between the two.
5. Lines95-109 "We therefore hypothesize that Kisangani's spatial transformations ... while vegetation density's predictive role is anticipated to decline."
Why did the authors of this passage make multiple assumptions. Typically, assumptions are used for future changes and not for past facts.
6.Line137 "Table 1. Spatial characteristics, time scale, and product type of geospatial dataset collected"
The Time scale corresponding to MOD11A2 V6.1 in Table 1 is 01.01 - 31.12, which should be explained by the authors.
7.Lines 139-177 "2.3. Data-Processing and Derivation of the Land Surface Temperature (LST)"
It is recommended that the authors separate the data preprocessing from the inversion of the LST, or place the LST computation in Section 2.4. The two do not fit in one chapter.
8.Line 192 "Figure 2. Methodological flowchart of the study"
The logic diagram should be placed at the top of the methods section to give readers a head start on the specific research path.
9.Lines 370-393 "4.1. Data and Spatial Analysis of the Urban Heat Island"
A discussion of the data or data processing techniques does not seem to be an appropriate content, as these techniques were not created for this study. The discussion should be centred on specific research findings.

Author Response

Comments and Suggestions for Authors

The authors investigated the spatio-temporal patterns of the surface urban heat island (SUHI) in Kisangani and explored the relationships between land surface temperature (LST) and two variables: building volume density (BVD) and the Normalized Difference Vegetation Index (NDVI). However, the manuscript failed to clarify its logic. The reviewer find it challenging to identify the study’s central focus, as each section appears to address topics already extensively covered in prior research.

Dear reviewer, On behalf of our entire team, I would like to thank you sincerely for your constructive comments, which have helped to improve this manuscript. Below are our answers to your questions. Thank you very much.

Following your comments and guidance, we have clarified in the manuscript the particular aspects that constitute the main contributions of this study. The specific details are as follows

Specific comments follow:
1.The title should be revised to explicitly articulate the logical connections between analyzing spatio-temporal patterns and the three key themes: "Urban-Rural Gradient, Building Volume Density, and Vegetation Effects." Specifically, the urban-rural gradient represents a spatial dimension, BVD serves as an architectural index, and vegetation effects denote ecological influences. 

R/ The title has been revised to highlight the logical connection (expected interactions) between the spatial and temporal structure of the UHI, the urban-rural gradient, the building volume, and the vegetation. (Lignes 2-3)


  1. Lines 44-70: "Increasing global, regional, and local urbanization fosters ... due to such challenges." In the first paragraph of the introduction, the reviewer considers whether it is too long. The description of the research background should be focused, simple, and clear.

R/ We have improved this introduction paragraph to make it simple and clear to understand (lines 44-54). In line with the evaluators' recommendations, the following paragraph (lines 56-70)  has been improved to clarify the links with the sustainable development objectives


  1. Lines 72-85: "In this context, Garcia-Herrera et al. ... worsening extreme heat events."
    Simply listing existing studies is not an appropriate approach. It is recommended that authors conduct a review to sort out and summarise existing studies.

R/ Thank you very much for your constructive comment. We have improved this part of the introduction to summarize existing studies and identify general trends (lines 71-82).


  1. Lines 86-94 "Despite extensive studies on the UHI effect ... further highlights spatial variations in UHI effects."
    Kisangani faces severe UHI effects illustrating the necessity of selecting the study area. The new study area does not seem to be an innovative point of an article. The authors are advised to explain the differences between this study and other studies in terms of research methodology and elements. Additionally, the authors were advised to choose a core research component rather than studying the spatio-temporal patterns of SUHI and then analysing the relationship between LST and BVD. The authors do not explain the logical relationship that exists between the two.

R/We have improved this introductory section to better justify the choice of Kisangani and the study's originality compared with previous studies while clarifying the link between LST and urban morphology, as expressed by the BVD (lines 84-100).


  1. Lines95-109 "We therefore hypothesize that Kisangani's spatial transformations ... while vegetation density's predictive role is anticipated to decline."
    Why did the authors of this passage make multiple assumptions. Typically, assumptions are used for future changes and not for past facts.

R/ Thank you very much for this comment. The assumption section has been improved to focus on future changes to be verified/tested. We have also defined these assumptions more clearly (lines 101-110).


6.Line137 "Table 1. Spatial characteristics, time scale, and product type of geospatial dataset collected"
The Time scale corresponding to MOD11A2 V6.1 in Table 1 is 01.01 - 31.12, which should be explained by the authors.

R/We have specified the time interval considered for the MOD 11A2 data, which, like the other data, covers the period from 2000 to 2024 (lines 147).


  1. Lines 139-177 "2.3. Data-Processing and Derivation of the Land Surface Temperature (LST)"
    It is recommended that the authors separate the data preprocessing from the inversion of the LST, or place the LST computation in Section 2.4. The two do not fit in one chapter.

R/ Thank you very much for this guidance, which makes it easier to structure the sections. We have taken this into account while improving the clarity of this section


8.Line 192 "Figure 2. Methodological flowchart of the study"
The logic diagram should be placed at the top of the methods section to give readers a head start on the specific research path.

R/ Thank you for your comment. We have placed the diagram at the top of the methodology section (line 130-135)9.Lines 370-393 "4.1. Data and Spatial Analysis of the Urban Heat Island"
A discussion of the data or data processing techniques does not seem to be an appropriate content, as these techniques were not created for this study. The discussion should be centred on specific research findings.

R/ Thank you very much for your comment. We have focused the discussion directly on the main results of the study.

Reviewer 4 Report

Comments and Suggestions for Authors
  1. The core issue is the classification of UHI levels. The article classifies 0.2<UHI<0.3 as High, and UHI>0.3 as Very High, and uses LST to calculate UHI. The LST difference of 0.3 degrees is actually very weak. It may not be reasonable to use the above thresholds for UHI classification. The author needs to explain or correct this in detail.
  2. Fig.4. The colors on this figure are not consistent with the UHI level, making it difficult to judge the change in UHI level. The author needs to modify the UHI intensity corresponding to each color in Fig.4, for example, dark blue represents Very low UHI (UHI<0). The same problem also appears in Fig.9.
  3. Line 373-384. "Moreover, the influence of Building Volume Density (BVD) as a predictor of Land Surface Temperature (LST) increases over time." and "the influence of vegetation density as a predictor of Land Surface Temperature (LST), as expected, gradually decreases over time." However, as shown in c2, the R2 in c2 is only 0.0554, indicating that the correlation between LST and UHI is not large with the year, and it is not same as "influence of vegetation density as a predictor of Land Surface Temperature (LST), as expected, gradually decreases over time." 
  4. Line 195. "The daily MOD11A1 LST product is generated using pixel-level LST data from individual granules, derived under clear-sky conditions through the generalized splitwindow algorithm". As far as the reviewer knows, in the generalized splitwindow algorithm, NDVI may be used in the calculation of channel emissivities. If this is the case, then the obtained LST is naturally correlated with NDVI. Therefore, the author needs to clarify the specific method of the generalized splitwindow algorithm used here, and if NDVI is used in this method, the author needs to explain and illustrate this point.
  5. The limitations of the article need to be strengthened. For example, NDVI and BVD are correlated to a certain extent, so the correlation between NDVI and BVD needs to be discussed to reduce the interference of their correlation on their respective correlations with LST. In addition, considering only NDVI and BVD is not comprehensive enough, and the influence of more factors on LST needs to be considered in future studies. 

Author Response

Dear Reviewer

On behalf of our entire team, we sincerely thank you for your significant contributions. Below are our answers to your questions and references to corrections made to the text.

 

  1. The core issue is the classification of UHI levels. The article classifies 0.2<UHI<0.3 as High, and UHI>0.3 as Very High, and uses LST to calculate UHI. The LST difference of 0.3 degrees is actually very weak. It may not be reasonable to use the above thresholds for UHI classification. The author needs to explain or correct this in detail.

R/

Thank you very much for this valuable comment. We understand your concern regarding the classification thresholds of UHI intensity based on LST differences. We want to thank you sincerely for your constructive comments. We have clarified this in detail from lines 203-225.

In our study, we used a relative UHI intensity index derived from surface temperature differences, following the normalized form of Equation (1)

Where: UHI refers to the urban heat island intensity (UHI), measured as the relative LST in the area; ΔT is the difference between the i-th pixel LST (Ti ) in °C and the average rural LST (Ts) in °C.

Therefore, a value of 0.2, for instance, implies that the urban temperature is 20% higher than the rural reference, which is substantial in ecological and thermal terms.

We chose this relative formulation because rural LST values vary significantly from year to year due to seasonal and climatic factors. Therefore, using a normalized metric allows us to better track and compare the evolution of UHI intensity over time, independent of absolute temperature fluctuations.

To better illustrate this, consider two different years with distinct rural and urban LST profiles:

 Year 1

  • Urban LST  = 30°C
  • Rural LST (Ts)= 28°C
  • Absolute UHI = ΔT=2°C
  • Relative UHI: 0.071

Year 2

  • Urban LST  = 36°C
  • Rural LST (Ts) = 34°C
  • Absolute UHI = ΔT=2°C
  • Relative UHI: 0.059

Although the absolute UHI remains constant at 2°C, the relative UHI decreases because the rural background temperature is higher in Year 2. The urban heat anomaly is proportionally less significant in Year 2. Such a distinction is only visible with a normalized metric, highlighting its usefulness for interannual analysis.

This method is theoretically justified and widely applied in recent literature. For instance, Huang et al. (2018) [33] used a similar relative index to assess long-term UHI trends across Chinese megacities. Rendan et al. (2023) [31] also employed a normalized LST-based UHI index to evaluate seasonal UHI dynamics in Hulu Langat district, Selangor, Malaysia. Shweta Jain et al. (2019) used the same relative index to assess UHI trends in the fast-growing urban area of Nagpur city in India [32].

Based on this, the normalized UHI approach provides a robust interannual comparison and classification framework, even when the numeric values appear low at first glance.

2. Fig.4. The colors on this figure are not consistent with the UHI level, making it difficult to judge the change in UHI level. The author needs to modify the UHI intensity corresponding to each color in Fig.4, for example, dark blue represents very low UHI (UHI<0). The same problem also appears in Fig.9.

R/ We have improved the representativeness of the UHI levels in terms of their colours, Fig. 4 (line 307) and Fig. 9, line 495

3. Line 373-384. "Moreover, the influence of Building Volume Density (BVD) as a predictor of Land Surface Temperature (LST) increases over time." and "the influence of vegetation density as a predictor of Land Surface Temperature (LST), as expected, gradually decreases over time." However, as shown in c2, the R2 is only 0.0554, indicating that the correlation between LST and UHI is not large with the year, and it is not same as "influence of vegetation density as a predictor of Land Surface Temperature (LST), as expected, gradually decreases over time." 

R/

We appreciate the reviewer’s comment. We acknowledge that the R² value (0.0554) suggests a weak correlation between year and the coefficient of determination (R²) of the NDVI-LST regression. However, our interpretation of a decreasing influence of vegetation density is primarily based on the significant decline in the regression slope values over time (p < 0.05) (Fig.8b2), rather than on R². We have revised the text accordingly to clarify this point and to reflect better the distinction between the predictive power (R²) and the sensitivity (slope) of the NDVI-LST relationship (lines 385-395).

4. Line 195. "The daily MOD11A1 LST product is generated using pixel-level LST data from individual granules, derived under clear-sky conditions through the generalized splitwindow algorithm". As far as the reviewer knows, in the generalized splitwindow algorithm, NDVI may be used in the calculation of channel emissivities. If this is the case, then the obtained LST is naturally correlated with NDVI. Therefore, the author needs to clarify the specific method of the generalized splitwindow algorithm used here, and if NDVI is used in this method, the author needs to explain and illustrate this point.

R/

Thank you for your valuable comment. You are right to note that in some generalized split-window algorithms, NDVI can play a role in determining emissivity. In the case of the MOD11A1 daily product, the algorithm does not directly use NDVI to compute LST. However, the emissivity values used in the algorithm are assigned based on a knowledge base incorporating land cover and vegetation information derived from MODIS products, including NDVI and land cover classifications [26]. Therefore, NDVI may indirectly influence LST values via emissivity assignment, especially in heterogeneous landscapes. We have clarified this point in the revised manuscript (lines 195-202).

5. The limitations of the article need to be strengthened. For example, NDVI and BVD are correlated to a certain extent, so the correlation between NDVI and BVD needs to be discussed to reduce the interference of their correlation on their respective correlations with LST. In addition, considering only NDVI and BVD is not comprehensive enough, and the influence of more factors on LST needs to be considered in future studies. 

R/

Thank you for this valuable comment. We agree that NDVI and BVD may be somewhat correlated, which could influence their relationships with LST. We have now included a discussion of this point (lines 486-493), highlighting the potential interaction effects and suggesting more robust analytical approaches for future work. Furthermore, we acknowledge that limiting the analysis to NDVI and BVD does not capture the full range of factors affecting LST. Accordingly, we have expanded the limitations to suggest integrating additional biophysical and anthropogenic variables, such as albedo, soil moisture, land cover type, and urban structures, in future studies to provide a more comprehensive analysis of LST dynamics (lines 486-493).

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Excellent work; it has already addressed every query that was previously sent. 

Author Response

Dear Reviewer
On behalf of our entire team, we sincerely thank you for your important contributions.

Reviewer 4 Report

Comments and Suggestions for Authors

Accept

Back to TopTop