Spatiotemporal Variation and Influencing Factors of Ecological Quality in the Guangdong-Hong Kong-Macao Greater Bay Area Based on the Unified Remote Sensing Ecological Index over the Past 30 Years

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Taking Guangdong, Hong Kong and Macao Greater Bay Area (GBA) as the study area, this paper proposes an improved Unified Remote Sensing Ecological Index (URSEI) to address the limitations of the traditional Remote Sensing Ecological Index (RSEI) in long-term ecological monitoring in cloudy and rainy climatic zones, and applies spatial autocorrelation and geodetic detector to systematically analyze the spatio-temporal evolution of the ecological quality and the driving mechanism from 1990 to 2020, which is useful for regional ecological protection and It is of reference value for regional ecological protection and policy making. In order to make the article more perfect, the reviewers put forward the following comments for the authors' reference.

1. This paper proposes URSEI based on RSEI, and discusses that the latter has better results, but the paper does not clarify the differences between the two. It is suggested that the specific differences between URSEI and the traditional RSEI in terms of indicator selection, normalization method, and synthesis strategy should be clarified in the introduction or methodology.
2. In the methodology section, it is mentioned that the construction of the “invariant area layer” is based on the land cover data from 1990 to 2020, but the identification of the invariant area is rather vague, and it is suggested to add the quantitative criteria for the screening of invariant area in order to enhance the scientific rigor.
3. Results:

(1) In 3.2.Spatiotemporal Changes in Ecological Quality, “Over the past 30 years, the proportion of different URSEI levels has undergone varying degrees of change, with the overall mean URSEI showing a fluctuating downward trend.”, according to the picture, it is not a fluctuating downward trend, but rather a first decline and then an increase.

(2) 3.4. Ecological Quality Spatial Autocorrelation Analysis, “this study adopts a 5 km × 5 km grid to sample the URSEI values of each period ”, is the 5 km sampling grid here of low resolution, making it difficult to identify local differences and increasing the error.

(3) In 3.5.3 Risk Zone Detection Analysis, “Among natural factors, temperature also exhibited a negative correlation with URSEI, with the highest URSEI mean value occurring at Temperature Level 1 (15.52-19.74°C),suggesting that areas with lower temperatures tend to have In terms of elevation and slope classifications, the highest URSEI mean values appeared in Level 4 and Level 5, respectively, indicating that regions with high elevations and slope classifications have better ecological quality. In terms of elevation and slope classifications, the highest URSEI mean values appeared in Level 4 and Level 5, respectively, indicating that regions with high elevations and steep slopes tend to have better ecological quality, possibly due to lower human disturbance. “The basis for the “positive and negative correlations” involved is not stated in the text, so it is recommended that this be checked and supplemented.

4. Discussion part, “with 2010 serving as a turning point, dividing the evolution into two stages”, 2010 as the turning point of ecological quality from decline to rise, it is suggested to combine with regional policies, industrial transformation It is suggested to enhance the causal explanation with regional policies, industrial transformation, etc. Generally speaking, this part should have a part that discusses the limitations of the study, which is not mentioned in the text, so it is suggested to add it.
5. In terms of presentation and charts, in Table 9 The result of single detection from 1990 to 2020, the expression of the column heading “2000-2020” is doubtful, whether it needs to be amended to “1990-2020” to avoid ambiguity.

Author Response

Taking Guangdong, Hong Kong and Macao Greater Bay Area (GBA) as the study area, this paper proposes an improved Unified Remote Sensing Ecological Index (URSEI) to address the limitations of the traditional Remote Sensing Ecological Index (RSEI) in long-term ecological monitoring in cloudy and rainy climatic zones, and applies spatial autocorrelation and geodetic detector to systematically analyze the spatio-temporal evolution of the ecological quality and the driving mechanism from 1990 to 2020, which is useful for regional ecological protection and It is of reference value for regional ecological protection and policy making. In order to make the article more perfect, the reviewers put forward the following comments for the authors' reference.

 

Comments 1: This paper proposes URSEI based on RSEI, and discusses that the latter has better results, but the paper does not clarify the differences between the two. It is suggested that the specific differences between URSEI and the traditional RSEI in terms of indicator selection, normalization method, and synthesis strategy should be clarified in the introduction or methodology.

 

Response 1: Thanks for your comment.

The differences in data selection and normalization methods have been incorporated into Sections 2.2.1 and 2.3.3, respectively. As for the synthesis strategy, the methodological distinctions were originally presented in Section 2.3.4. The corresponding paragraph contents are as follows:

(1)The difference in data selection

“Given the long acquisition cycle of Landsat data and the frequent cloud cover in the GBA, this study modified the conventional approach of selecting a single cloud-free or minimally-clouded image from the target year. Instead, images were selected from the target year and the adjacent years within the vegetation growth period (September, October) and the non-growth period (December, January), ensuring cloud coverage below 30%.”

(2) The difference in normalization methods

“The RSEI model employs range normalization to eliminate dimensional differ-ences among indicators. In existing studies, normalization is typically performed sep-arately for each time period by applying the minimum and maximum values from in-dividual datasets to normalize corresponding indicators. However, this method is sus-ceptible to extreme values, which may introduce biases in the results. Over the past 30 years, the Greater Bay Area has undergone significant urban expansion and ecological changes, leading to substantial differences in the statistical distribution of ecological indicators. While conventional probability-based normalization methods (such as Z-score or Gaussian normalization) can adjust data distribution, they may reduce time-series consistency, affecting the spatial-temporal comparability of indicators and the accuracy of evolution pattern analysis.

To address the issue of indicator normalization, this study proposes an invari-ant-region-based normalization method.”

(3) The difference in synthesis strategy

“The RSEI model uses Principal Component Analysis (PCA) to construct a compo-site ecological index for each monitoring period and standardizes the RSEI values to a range of 0-1 using range normalization, effectively reflecting the relative ecological condition of a specific spatial region at a given time. However, in areas with signifi-cant ecological changes, the RSEI value only represents the relative condition within a specific period, making it difficult to accurately reflect the long-term ecological evolu-tion trend across different years.

In this study, ecological indicator factors from four monitoring periods are fused using a unified principal component analysis to construct a composite ecological index. First, the normalized indicator factors are used as variables, and they are input into PCA synchronously and with equal weight to ensure that the data are processed under the same standard. Then, the contribution rate of the first principal component (PC1) is assessed. If it exceeds 70%, PC1 can be used to create the composite ecological index. Unlike the traditional RSEI model, this study does not re-normalize the composite eco-logical index. Instead, it normalizes the feature vector of PC1 and applies it uniformly to all monitoring periods to generate the Unified Remote Sensing Ecological Index (URSEI), ensuring the comparability of ecological quality changes across different time periods.”

 

Comments 2: In the methodology section, it is mentioned that the construction of the “invariant area layer” is based on the land cover data from 1990 to 2020, but the identification of the invariant area is rather vague, and it is suggested to add the quantitative criteria for the screening of invariant area in order to enhance the scientific rigor.

 

Response 2: We sincerely appreciate your constructive suggestion regarding the identification of invariant areas. Following your advice, we have revised Section 2.3.3 to include quantitative screening criteria and validation steps:” First, using land cover classification data of the Greater Bay Area from 1990 to 2020 [49], a pixel-by-pixel comparison method was applied to identify areas with unchanged land cover types. To accurately delineate long-term stable land cover regions, three decadal intervals (1990–2000, 2000–2010, and 2010–2020) were established to generate stable land cover layers for each period. Finally, spatial intersection analysis was performed to extract areas where land cover types remained unchanged over the entire 30-year period. The formula is as follows:”

 

Comments 3: In 3.2.Spatiotemporal Changes in Ecological Quality, “Over the past 30 years, the proportion of different URSEI levels has undergone varying degrees of change, with the overall mean URSEI showing a fluctuating downward trend.”, according to the picture, it is not a fluctuating downward trend, but rather a first decline and then an increase.

 

Response 3: This mistake has been addressed and corrected as:

Over the past 30 years, the proportion of different URSEI levels has undergone varying degrees of change, with the overall mean URSEI showing a “decline-then-rise” trend.

 

Comments 4: 3.4. Ecological Quality Spatial Autocorrelation Analysis, “this study adopts a 5 km × 5 km grid to sample the URSEI values of each period ”, is the 5 km sampling grid here of low resolution, making it difficult to identify local differences and increasing the error.

 

Response 4: Thank you for your valuable comments regarding the sampling grid resolution. The selection of a 5 km grid in this study was primarily based on the following considerations:

1. Scale Compatibility: Given the vast spatial extent of the Guangdong-Hong Kong-Macao Greater Bay Area (total area: 56,000 km²), a 5 km grid effectively balances the accuracy of global spatial pattern characterization with computational efficiency, while avoiding the introduction of local noise that might obscure broader trends.
2. Methodological Consistency: Similar regional-scale ecological studies (e.g., Wu et al., 2022; Liu et al., 2021) have employed 5 km grids for spatial autocorrelation analysis, with methodologies and conclusions widely validated in the academic community.

We fully acknowledge the potential of higher-resolution grids (e.g., 1 km) to reveal finer local variations, which will be a key focus of future research. However, for the objectives of this study, the current grid design remains robust in supporting the core conclusion regarding spatial clustering patterns of regional ecological quality.

 

Comments 5: In 3.5.3 Risk Zone Detection Analysis, “Among natural factors, temperature also exhibited a negative correlation with URSEI, with the highest URSEI mean value occurring at Temperature Level 1 (15.52-19.74°C),suggesting that areas with lower temperatures tend to have In terms of elevation and slope classifications, the highest URSEI mean values appeared in Level 4 and Level 5, respectively, indicating that regions with high elevations and slope classifications have better ecological quality. In terms of elevation and slope classifications, the highest URSEI mean values appeared in Level 4 and Level 5, respectively, indicating that regions with high elevations and steep slopes tend to have better ecological quality, possibly due to lower human disturbance. “The basis for the “positive and negative correlations” involved is not stated in the text, so it is recommended that this be checked and supplemented.

 

Response 5: In the revision, the basis for the ”positive and negative correlations” involved has been added as:

”Among natural factors, each variable was similarly classified into five levels(Level 1:lowest, Level 5: highest). temperature also exhibited a negative correlation with URSEI, with the highest URSEI mean value occurring at Temperature Level 1 (15.52–19.74℃), suggesting that areas with lower temperatures tend to have better ecological quality. In terms of elevation and slope classifications, the highest URSEI mean values appeared in Level 4 and Level 5, respectively, indicating that regions with high elevations and steep slopes tend to have better ecological quality, possibly due to lower human disturbance.”

 

Comments 6: Discussion part, “with 2010 serving as a turning point, dividing the evolution into two stages”, 2010 as the turning point of ecological quality from decline to rise, it is suggested to combine with regional policies, industrial transformation It is suggested to enhance the causal explanation with regional policies, industrial transformation, etc. Generally speaking, this part should have a part that discusses the limitations of the study, which is not mentioned in the text, so it is suggested to add it.

 

Response 6: We sincerely appreciate the reviewer's valuable suggestions. We have carefully addressed each point in our revised manuscript as follows:

“In the second stage (2010–2020), ecological quality in the GBA showed an im-proving trend. This shift can be attributed to several key factors. First, the pace of ur-banization slowed down, and the expansion of industrial land significantly decreased. Second, after 2010, both national and local governments introduced a series of ecolog-ical and environmental protection policies, such as the Outline of the Plan for the Re-form and Development of the Pearl River Delta Region (2008-2020) and the Guang-dong-Hong Kong-Macao Greater Bay Area Development Plan, which explicitly set ecological conservation and green development as key objectives, thereby promoting the implementation of ecological restoration projects. Lastly, the regional industrial system progressively transitioned toward high-tech industries, while traditional high-pollution, high-energy-consumption industries were either phased out or up-graded, effectively reducing industrial pressure on the ecological environment. For in-stance, cities like Shenzhen vigorously developed high-tech industries and modern service sectors, leading to decreased industrial pollution emissions.”

“4.4 Limitations and Future Works

This study systematically analyzed the spatiotemporal evolution characteristics and driving mechanisms of ecological quality in the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) from 1990 to 2020 using the Unified Remote Sensing Ecologi-cal Index (URSEI) and the geographical detector model. However, several limitations need to be acknowledged, and future improvements are proposed as follows:

Regarding data limitations, due to constraints in historical data availability, the nighttime light and GDP data for 1990 were substituted with data from 1992. While such substitution is common in long-term ecological studies, it may introduce certain biases in the analysis of early-stage driving factors of ecological quality. Future re-search could employ data interpolation or modeling approaches to optimize the com-pleteness of early-stage data and reduce potential biases caused by direct substitution.

In terms of analytical dimensions, this study primarily relied on remote sensing data and large-scale driving factors for assessment, with relatively insufficient consid-eration of micro-level socioeconomic elements (e.g., environmental protection invest-ments, specific policy implementation effects, corporate pollution control measures). Subsequent studies should incorporate field surveys or socioeconomic statistical data to further refine the analysis of driving mechanisms.”

 

Comments 7: In terms of presentation and charts, in Table 9 The result of single detection from 1990 to 2020, the expression of the column heading “2000-2020” is doubtful, whether it needs to be amended to “1990-2020” to avoid ambiguity.

 

Response 7: Thanks for your comments. The mistake has been corrected.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This is a very interesting article which details a thorough procedure for improving RSEI. It is design appropriate and novel enough to deserve publication. My recommendation is for minor revision provided that the following topics are addressed. 

1) In Figure 1, clarify better where major cities and transportation networks are. These are very relevant for URSEI, since you determine human factors have become increasingly important in showing where it got degraded.

2) In the discussion, you must address the fact two variables were not based on 1990 data (but 1992). 

3) Figure 2 needs improvements: the maps are too small and lack proper contrast. I also suggest adding road networks and location of larger cities such as in Figure 1.

4) Same applies to Figure 5.

5) Discussion section doesn't address quantitatively how URSEI is an improvement based on RSEI. Could you add some more details on this?

6) Also, I understand the discussion should address how the URSEI indices may be used in more concrete scenarios.

7) Finally, all points you raise in the discussion (for example, ecological corridors, urban clusters, etc) should be supported by maps that show their exact location vis-à-vis URSEI and clusters. This means the discussion needs to be enhanced considerably so visitors not acquainted with the Guangdong-Hong Kong-Macao may understand how it relates spatially.

Author Response

This is a very interesting article which details a thorough procedure for improving RSEI. It is design appropriate and novel enough to deserve publication. My recommendation is for minor revision provided that the following topics are addressed.

 

Comments 1: In Figure 1, clarify better where major cities and transportation networks are. These are very relevant for URSEI, since you determine human factors have become increasingly important in showing where it got degraded.

 

Response 1: Thank you for your careful review and valuable suggestions! Regarding your recommendation to include the locations of major cities and transportation networks in Figure 1 , we have carefully considered this and made adjustments. The original figure uses DEM to visually represent the topographic characteristics of the Greater Bay Area, which directly aligns with the URSEI assessment findings on "spatial differentiation of ecological quality" (e.g., better ecological conditions in high-altitude areas and degradation in low-altitude urban zones). For the purposes of subsequent results and discussion, the locations of the urban core area of cities in the Greater Bay Area are also shown in Figure 1. However, overlaying transportation networks on top of this may lead to excessive visual elements and color complexity（as shown in the figure below）, potentially obscuring key geographic information. Therefore, we recommend retaining the current adjustments to ensure map readability and scientific clarity.

 

Comments 2: In the discussion, you must address the fact two variables were not based on 1990 data (but 1992).

 

Response 2: In the revision, we have clearly stated in Section 4.4 ("Limitations and Future Work") that two variables (nighttime light and GDP data) were based on 1992 data rather than 1990 data due to availability constraints.

 

Comments 3: Figure 2 needs improvements: the maps are too small and lack proper contrast. I also suggest adding road networks and location of larger cities such as in Figure 1.

 

Response 3: Thank you for your thorough review and valuable suggestions. Regarding the improvements to Figure 2 , we have implemented the following revisions:

1. Optimization of Map Size and Contrast

We have enlarged the figure size and enhanced color contrast to improve the clarity of ecological quality levels (Excellent, Good, Moderate, Poor, and Very Poor).

2. Rationale for Not Overlaying Additional Elements

The primary purpose of the URSEI map is to display the continuous spatial pattern of ecological quality. Overlaying road networks or urban locations may overcrowd the figure and obscure the core data. Thus, we opted to maintain the current design for optimal clarity.

 

Comments 4: Same applies to Figure 5.

 

Response 4: Thank you for your thorough review and valuable suggestions. Regarding the improvements to Figure 2 , we have implemented the following revisions:

1. Optimization of Map Size and Contrast

We have enlarged the figure size.

2. Rationale for Not Overlaying Additional Elements

The primary purpose of the URSEI map is to display the continuous spatial pattern of ecological quality. Overlaying road networks or urban locations may overcrowd the figure and obscure the core data. Thus, we opted to maintain the current design for optimal clarity.

 

Comments 5: Discussion section doesn't address quantitatively how URSEI is an improvement based on RSEI. Could you add some more details on this?

 

Response 5: In the revision, we have significantly expanded the quantitative comparison between URSEI and RSEI in Section 4.1:

“This study proposed multiple improvements to the Remote Sensing Ecological Index (RSEI) to address the challenge of maintaining temporal consistency in ecological quality assessment for cloudy and rainy regions. Given the persistent cloud cover in the Greater Bay Area, our approach moved beyond reliance on single images by se-lecting imagery from both growing and non-growing seasons within a three-year win-dow (the target year plus one year before and after). Annual indicator factors were generated through median synthesis followed by averaging, which effectively miti-gated the impacts of extreme weather and solar altitude variations while enhancing data stability and representativeness. To address thermal data gaps caused by cloud cover, we employed a random forest model using land cover type, NDVI, and elevation as predictors to reconstruct and correct LST data, ensuring spatial consistency.

In terms of indicator normalization, traditional RSEI normalizes each temporal dataset independently, making it vulnerable to extreme values and temporal incom-parability. Our improved method established a land cover invariant reference layer (1990-2020) and extracted indicator values from these stable areas as global normali-zation thresholds. This approach simultaneously eliminated extreme value effects and maintained consistent normalization baselines across different time periods. For prin-cipal component analysis, conventional RSEI performs PCA separately for each tem-poral dataset, resulting in inconsistent weights across periods. Our enhanced method merged indicator factors from all four time periods for unified PCA, generating global eigenvectors. The application of fixed weights across all periods ensured the temporal comparability of URSEI.

The average correlation test revealed that URSEI's correlation with individual in-dicators was significantly stronger than any single indicator, confirming its superior comprehensive representation capability. Furthermore, the URSEI classification maps of the Greater Bay Area (Figure.2) showed no apparent edge-matching or mosaic arti-facts, with highly consistent spatial patterns of ecological quality across time periods, ensuring result coherence and integrity while improving visualization effectiveness. Similarly, the ecological quality change detection maps (Figure.5) exhibited no anom-alous regions, with detected changes closely matching actual conditions, demonstrat-ing URSEI's effectiveness in maintaining temporal comparability for long-term se-quence analysis.”

 

Comments 6: Also, I understand the discussion should address how the URSEI indices may be used in more concrete scenarios.

 

Response 6: In the revision, the potential concrete applications of the URSEI index have been added as:

“In future applications, the URSEI model could be extended to other regions exhibiting similar climatic characteristics and ecological dynamics. Particularly suitable candi-dates include rapidly urbanizing areas such as the Yangtze River Delta urban agglom-eration and Chengdu-Chongqing economic zone, which share with the Greater Bay Area both intensive anthropogenic pressures and frequent cloudy/rainy conditions. The model would also prove valuable for monitoring ecological changes in other southeastern coastal regions where persistent cloud cover and precipitation similarly challenge conventional remote sensing assessments of fast-evolving environments.”

 

Comments 7: Finally, all points you raise in the discussion (for example, ecological corridors, urban clusters, etc) should be supported by maps that show their exact location vis-à-vis URSEI and clusters. This means the discussion needs to be enhanced considerably so visitors not acquainted with the Guangdong-Hong Kong-Macao may understand how it relates spatially.

 

Response 7: Thanks for your comment.

Regarding ecological corridors, the period of this study (1990-2020) did not cover the ecological corridor policies implemented in the Greater Bay Area after 2020, and thus its impact was not directly analyzed. However, the ecological barrier function of the high-high concentration areas (such as Zhaoqing and Huizhou) mentioned in the text may provide a basis for future corridor planning.

“The central urban cluster” in the text refers to cities such as Guangzhou, Foshan, Dongguan, Shenzhen and Zhongshan, which are geographically located in the central part of the Greater Bay Area. The distribution of each city has been marked in Figure 1.

 

 

 

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Accept after minor revisions