the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 16 Jul 2025
Advancing Urban Heat Vulnerability Assessment through SAR-Derived Vegetation and Soil Moisture Indicators: A Spatial Modelling Framework for Dhaka, Bangladesh
Abstract. Urban heatwaves are intensifying due to climate change, posing significant risks to public health and infrastructure in densely populated cities. This study develops a spatially explicit framework to assess urban heat vulnerability in the Dhaka Metropolitan Area (DMA), Bangladesh, by integrating vegetation and soil moisture indicators derived from Synthetic Aperture Radar (SAR). Sentinel-1 imagery was used to compute the Radar Vegetation Index (RVI) and estimate surface soil moisture (SSM) through empirical modelling, combining a modified Water Cloud Model (mWCM) with regression calibration against SMAP data. MODIS-derived Land Surface Temperature (LST) was used to characterize thermal variation. A Geographically Weighted Regression (GWR) model, supported by Principal Component Analysis (PCA), quantified local relationships between LST, RVI, and SSM. Spatial autocorrelation analysis using Moran’s I confirmed clustering in both thermal and environmental variables. Results show that areas with higher vegetation and soil moisture correspond to lower LST, highlighting their cooling effects. The model achieved strong performance (R² = 0.8835; RMSE = 0.6126; MAE = 0.4753), demonstrating its robustness and applicability in data-scarce contexts. A Heat Vulnerability Index (HVI) was constructed to spatially map susceptibility to extreme heat. This SAR-based approach supports targeted urban heat adaptation strategies through spatially informed planning.
- Preprint
(1123 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2025-3218', Anonymous Referee #1, 30 Jul 2025
reply
The study uses multiple satellite data and methods to develop a Heat Vulnerability Index (HVI).
The manuscript needs major modification in order to be accepted for publication. First, a better connection between the methods used and their purpose should be explained. Including a methodological diagram could be beneficial. For example, the analysis of spatial autocorrelation and GWR and their goal are not well supported. Further, a clearer explanation of the temporal domain used and its purpose is needed. When should you use annual data, and when is it appropriate to use monthly or daily data? Currently, it is not clear on the document.
Majors
The manuscript needs a study area figure with land cover types.
At the end of the introduction, it is stated that the author will use Moran's I to detect spatial clusters (L75). However, in the results section, Global Moran's I is used instead. For a better identification of the clustering, you should use Moran's I as Local Indicators of Spatial Association (LISA) to achieve this. Furthermore, at least a figure showing the time series of global Moran's I per variable would help the reader. Then in Fig. 4 could be used the variable and year that has the higher Moran's I.
The use of GWR in this study is just a downscaling for LST. But not a predictive model for LST. The absence of in-situ LST measurements explains this.
Minors
L15: RMSE and MAE should have units.
L120: It's unclear why two MOD11 ranges and sources are being used. Please clarify this.
L130 Both datasets are currently available. PerhIt appears that the source (i.e., GEE) used by the author does not contain the complete dataset.ove or modify.
L175: Which resampling method was used? Bilineal? Please, specify.
L220: These extreme heat events were derived from LST or just the BMD & NOAA data. Please, start with a sentence that could clarify this.
L240: Please choose a proper color palette for the figure; currently the colors are very similar, and it is difficult to distinguish between them.
L245: You do not need to indicate "GEE exports". Please delete.
L245: What is the purpose of figure 4? How does it help to understand the manuscript?
L280: In the figure, "coefficients" is stated, but the image can only show one coefficient. I believe it is the slope (B1).
L290: For GWR, you did not use observed data. Here, you make a regression using LST (from MODIS), and it is regressed using PC1.
L295: What are the units °C or Kelvin?
L300: When you run a GWR you get as results spatial indicators for error (Fig. 6) and R2. You should show the spatial R² on the manuscript.
L325: But, in L265, you say something opposite.
Citation: https://doi.org/10.5194/egusphere-2025-3218-RC1 -
AC1: 'Reply on RC1', Aishia Fyruz Aishi, 13 Aug 2025
reply
Dear Reviewer,
On behalf of all co-authors, I would like to sincerely thank you for your constructive and insightful comments on our manuscript. Your suggestions have been invaluable in improving the clarity, methodological rigor, and overall presentation of our work.
Before addressing each comment individually, we would like to highlight the key revisions and clarifications we have made in direct response to your feedback:
1. Clearer linkage between methods and purpose – The manuscript has been revised to explicitly state the rationale for each analytical step. For example, Global Moran’s I quantifies overall spatial clustering, LISA detects localized clusters and outliers, and GWR reveals spatially varying relationships between environmental drivers and LST. These steps together form an integrated approach to assessing urban heat vulnerability. A methodological workflow diagram (new Figure 1) has been added to visually summarize the process from data collection to HVI mapping.
2. Temporal domain clarification – As noted in Line 110 and Line 175 of the preprint, the analysis uses pixel-wise multi-year means (2016-2022) computed from daily data available only on heatwave dates when all datasets were accessible. This choice reduces noise and inter-annual variability while preserving core spatial patterns that are relevant to extreme heat events. Aggregating across only heatwave dates also ensures temporal consistency among datasets with differing native resolutions.
3. Spatial autocorrelation enhancement – In addition to Global Moran’s I, we now present LISA results for LST, RVI, and SSM (new Figure 7) to highlight local clustering patterns and spatial outliers, which are essential for understanding the heterogeneity of heat-related variables. A time series plot of Global Moran’s I values per variable (new Figure 6) has also been added to track changes in spatial clustering over time and to identify the heatwave date with the strongest spatial autocorrelation for further analysis.
4. Study area figure – We have added a detailed land cover map of the Dhaka Metropolitan Area (new Figure 2), showing administrative boundaries, major water bodies, and land cover composition. This provides important context for interpreting spatial patterns in our results.
5. GWR results expansion – We now present spatial R² map (new Figure 9) to better illustrate model performance and spatial variability. We also clarify throughout that the GWR in this study serves as an exploratory spatial analysis tool to quantify local relationships rather than as a predictive model, due to the absence of in-situ LST measurements.
We have responded to your suggestions point-by-point in the attached document, and we hope these revisions will improve clarity, strengthen methodological justification, and remove any ambiguity. Please note that NHESS does not allow us to upload the revised manuscript as part of author response. However, all referenced figures and textual changes are incorporated into the revised manuscript.
Thank you again for your constructive feedback and for the opportunity to improve our work.
Sincerely,
Aishia Fyruz Aishi
aishia.fyruz@du.ac.bd
-
AC1: 'Reply on RC1', Aishia Fyruz Aishi, 13 Aug 2025
reply
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 819 | 42 | 13 | 874 | 17 | 28 |
- HTML: 819
- PDF: 42
- XML: 13
- Total: 874
- BibTeX: 17
- EndNote: 28
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1