Urban Green Cover & Heat Risk Assessment: Singapore-region

Yaragarla, Ramkumar

Observation Metadata

Satellite Platform	Sentinel-2C (Sentinel 2)
Observation Date	2025-12-05
Cloud Cover	10.657708%
Spectral Bands Used	B04 (Red), B08 (NIR), B11 (SWIR1), B12 (SWIR2)
Processing Date	2026-01-03

Sentinel-2 satellite preview of Singapore-region, captured on 2025-12-05 from an altitude of approximately 786 km.

Statement of Problem

Rapid urbanization and climate change are placing unprecedented stress on cities worldwide. As built-up areas expand and green spaces decline, urban heat islands intensify, air quality deteriorates, and residents face increased health risks. Understanding the spatial distribution of vegetation and development is essential for evidence-based urban planning, climate adaptation, and public health protection. This analysis quantifies the current balance between green cover and built-up surfaces in Singapore-region, identifies thermal vulnerability hotspots, and provides actionable recommendations for enhancing urban forestry and green infrastructure.

1. Urban Forestry Assessment: Current Vegetation & Built-Up Status

1.1 Vegetation Distribution (NDVI Analysis)

The Normalized Difference Vegetation Index (NDVI) analysis for the Singapore-region, as of the observation date of 2025-12-05, reveals a mean NDVI value of 0.20, indicating a moderate level of vegetation cover. The median NDVI value of 0.15 suggests that half of the region exhibits vegetation cover below this threshold. A standard deviation of 0.22 points to a considerable variability in vegetation density across the region. Spatially, areas with higher NDVI values, as shown in the NDVI color and greyscale maps, correspond to regions with denser green cover, primarily concentrated in parks, nature reserves, and residential areas with substantial landscaping. Conversely, lower NDVI values are observed in urban and industrial zones, highlighting the challenges in maintaining green cover in highly developed areas.

1.2 Built-Up Intensity (NDBI Analysis)

The Normalized Difference Built-Up Index (NDBI) analysis for Singapore-region indicates a mean NDBI value of -0.04, suggesting a relatively low level of built-up intensity across the region. However, the maximum NDBI value of 0.67 identifies specific zones with high built-up density, likely corresponding to central business districts and industrial areas. The standard deviation of 0.13 reflects variability in urban development across the region. High NDBI values, as depicted in the NDBI maps, are concentrated in areas with extensive infrastructure and minimal green spaces, indicating zones under significant development pressure. This pattern underscores the need for balanced urban planning to mitigate the environmental impacts of urbanization.

NDVI Color Visualization

NDVI Greyscale (Index Values)

NDBI Color Visualization

NDBI Greyscale (Index Values)

Additional Vegetation & Water Indices

Beyond NDVI and NDBI, three complementary indices provide deeper insights into vegetation health and water presence:

EVI (Enhanced Vegetation Index)

EVI improves on NDVI by correcting for atmospheric conditions and canopy background noise. It is more sensitive in areas with dense vegetation, making it useful for monitoring forest health and identifying vegetation stress that NDVI might miss.

EVI Color Visualization

EVI Greyscale (Index Values)

MNDWI (Modified Normalized Difference Water Index)

MNDWI enhances water body detection by using green and shortwave infrared bands. It is more effective than NDWI at separating water features from built-up areas, making it ideal for mapping urban water bodies like lakes, rivers, and reservoirs.

MNDWI Color Visualization

MNDWI Greyscale (Index Values)

NDMI (Normalized Difference Moisture Index)

NDMI measures vegetation water content by comparing near-infrared and shortwave infrared reflectance. Higher values indicate well-hydrated vegetation, while lower values suggest drought stress. This index is valuable for assessing irrigation effectiveness and identifying areas at risk of vegetation die-off.

NDMI Color Visualization

NDMI Greyscale (Index Values)

2. Vegetation-Development Balance Analysis

2.1 NDVI-NDBI Difference Analysis

The NDVI-NDBI difference analysis provides insights into the balance between vegetation and built-up areas in Singapore-region. The difference bins data shows that 51.73% of the region falls into the bin ranging from -0.27 to 0.21, indicating areas where vegetation and built-up structures are relatively balanced. Positive values, where vegetation dominates, account for 26.55% (bin 0.21 to 0.69) and 12.85% (bin 0.69 to 1.17) of the region, primarily in parks and green spaces. Negative values, indicating built-up dominance, cover 0.52% (bin -0.75 to -0.27) and are mainly found in the central business district and industrial areas. The difference map and legend illustrate these patterns, with colors ranging from green (high vegetation) to red (high built-up), providing a visual representation of the vegetation versus built-up balance across the region.

2.2 Overlay Interpretation

The combined overlay image of NDVI and NDBI for Singapore-region reveals distinct patterns when both indices are viewed together. Areas with high NDVI and low NDBI values appear in green, indicating zones with abundant vegetation and minimal built-up structures, such as parks and nature reserves. Conversely, areas with low NDVI and high NDBI values are shown in red, representing highly urbanized zones with dense infrastructure and limited green spaces. This overlay highlights the spatial distribution of green versus built-up areas, aiding in the identification of zones requiring attention for urban planning and environmental management.

NDVI-NDBI Difference Map (Green = vegetation-dominated; Red = built-up-dominated)

Combined NDVI-NDBI Overlay

3. Urban Heat Island Risk & Thermal Vulnerability

3.1 Heat Risk Distribution

The Urban Heat Index (UHI) analysis for Singapore-region identifies zones at varying levels of heat risk. The heat index bins data shows that 51.73% of the region falls into the moderate-risk zone (bin -0.21 to 0.27), where NDVI and NDBI values are relatively balanced. High-risk zones, where NDBI exceeds NDVI, cover 12.85% (bin -0.69 to -0.21) and 0.02% (bin -1.65 to -1.17) of the region, primarily in densely built-up areas with minimal vegetation. Low-risk zones, where vegetation dominates, account for 26.55% (bin 0.21 to 0.69) and 8.85% (bin 0.27 to 0.75) of the region, mainly in green spaces and parks. The heat index map and legend illustrate these zones, with colors ranging from blue (low risk) to red (high risk), indicating the spatial distribution of heat risk across the region. The implications for the urban heat island effect and public health are significant, as high-risk zones are more susceptible to heat-related illnesses and require targeted interventions to enhance green cover and improve urban microclimates.

3.2 Key Findings & Implications

Total green cover percentage: 33.57%
Healthy vegetation percentage: 14.31%
Vegetation health score: 0.47
Green infrastructure score: 0.474

Urban Heat Index Map (Red = high risk zones)

4. Strategic Recommendations for Green Infrastructure

4.1 What's Working Well

The region maintains a significant green cover percentage of 33.57%, indicating successful efforts in preserving and promoting vegetation across urban and natural areas.
Areas with healthy vegetation, comprising 14.31% of the region, suggest effective management practices in maintaining the vitality of green spaces.

4.2 Critical Challenges

The vegetation health score of 0.47 indicates room for improvement in the overall health and resilience of the region's vegetation, particularly in urban areas.
High-risk heat zones, particularly in densely built-up areas, highlight the need for targeted interventions to mitigate the urban heat island effect and protect public health.

4.3 Evidence-Based Recommendations

Enhance green infrastructure in urban areas by increasing the planting of trees and shrubs, particularly in high-risk heat zones, to improve air quality, reduce urban heat, and provide recreational spaces for residents.
Implement targeted vegetation management practices to improve the health score of existing green spaces, including regular monitoring, maintenance, and the introduction of native species to enhance biodiversity and resilience.
Develop and enforce urban planning policies that promote a balanced distribution of green and built-up areas, ensuring that new developments include adequate green spaces to mitigate the environmental impacts of urbanization.

Analysis Coverage Area

The interactive map below shows the exact geographical bounds of this satellite analysis. The colored overlay represents the NDVI coverage area overlaid on OpenStreetMap. You can zoom and pan to explore how the analysis boundaries align with streets, neighborhoods, and landmarks in Singapore-region.

Note: The analysis boundaries may extend beyond administrative city limits as they represent the satellite image crop captured on 2025-12-05. This ensures complete coverage of the urban area and surrounding regions for comprehensive vegetation and heat risk assessment.

Why Urban Green Cover Matters

Urban forests and green spaces are critical infrastructure for healthy, livable cities. Trees and vegetation reduce air pollution, lower urban temperatures through shade and evapotranspiration, manage stormwater, support biodiversity, and improve mental health and well-being. As cities grow denser and climate change intensifies heat events, monitoring and protecting urban green cover becomes essential for public health and environmental resilience.

Satellite remote sensing provides objective, repeatable measurements of vegetation health and urban development patterns across entire metropolitan areas. The Normalized Difference Vegetation Index (NDVI) quantifies photosynthetic activity and green cover, while the Normalized Difference Built-up Index (NDBI) identifies impervious surfaces and development intensity. Together, these indices reveal the changing balance between nature and urban growth—and highlight where intervention is most needed.

Understanding NDVI and NDBI

NDVI (Normalized Difference Vegetation Index) measures the difference between near-infrared light (strongly reflected by healthy vegetation) and red light (absorbed by chlorophyll). Values range from -1 to +1, with higher values indicating denser, healthier vegetation. Typical ranges: 0.6-0.9 = dense forest; 0.3-0.6 = moderate vegetation; 0.1-0.3 = sparse vegetation; <0.1 = bare soil or built-up areas.

NDBI (Normalized Difference Built-up Index) uses shortwave infrared and near-infrared bands to identify constructed surfaces. Positive values indicate built-up areas (roads, buildings, concrete), while negative values suggest natural or vegetated land. When NDBI exceeds NDVI in an area, it signals high development intensity and elevated urban heat island risk.

README Note

Our Mission: We want this research dataset brief to be Simple, Authentic, and Repeatable.

1. Title of the Dataset

Urban Green Cover and Heat Risk Assessment

2. What This Dataset Is About

This dataset was created to help anyone understand how vegetation and green cover in this region is changing over time. It brings together processed satellite data, simple calculations, and a few observations that give the bigger picture.

3. Why This Dataset Matters

At I Hug Trees, we believe that Geospatial Satellite imagery and processed data should be accessible to everyone. While grounded in scientific principles, outputs are presented in accessible formats so that technical imagery and calculations resonate with ordinary people and communities. Because only when it is relatable, can it tell clear stories about our greenery and urban life: shaping how we live, how we breathe, and how we cope with rising heat.

This dataset tries to make that easier. Whether you are a researcher, policy maker, student or just curious about the environment, these numbers and images help you see trends that are not obvious at first glance.

4. Source of the Data

Satellite: Sentinel-2
Provider: Microsoft Planetary Computer
Acquisition Window: Past 60 days (filtered by cloud cover)
Cloud Cover Threshold: < 30%
Initial Tile Discovery: Copernicus Data Space Ecosystem browser

5. How the Data Was Processed

Basic Preprocessing:

It is important to get the right tile from the Sentinel-2 database for imagery processing. The browser feature from the Copernicus Data Space Ecosystem helped us identify the correct tile, its bounds, and the subset coordinates needed for extraction. It is always essential to double-check the preview image to verify that the fetched tile truly corresponds to the target region.

Next, we used an AWS Lambda environment for the bounds-discovery phase. This step involved fetching band data from the Microsoft Planetary Computer for dates where cloud cover was below 30% within the past 60 days. Once the acquisition date met this condition, we moved to AWS EC2 server scripts written in Python to download raw band data and process them into COGs (Cloud Optimised GeoTIFFs) along with additional indices.

These processed COGs and index outputs are then used for image displays, NDVI interpretation, and HTML digest features published on our platform. All indices raw value outputs are in JSON files for easy repeatable processing. We currently run this workflow on a quarterly schedule for each identified region.

Cloud Masking:

This dataset uses Sentinel-2's Scene Classification Layer (SCL) to remove pixels affected by clouds, shadows, haze, and saturation. Only surface-clear classes are kept, ensuring that the vegetation indices are calculated from clean, reliable pixels. The masking is applied at the pixel level, meaning every index (NDVI, EVI, etc.) is computed only on valid areas after stripping away noisy regions. This results in more trustworthy COGs, cleaner previews, and more meaningful temporal comparisons.

Calculation Formulas Used:

NDVI = (NIR − Red) ÷ (NIR + Red)
EVI = 2.5 × (NIR − Red) ÷ (NIR + 6×Red − 7.5×Blue + 1)
NDWI = (Green − NIR) ÷ (Green + NIR)
NDBI = (SWIR1 − NIR) ÷ (SWIR1 + NIR)

Tools Used:

Data Source: Microsoft Planetary Computer (Sentinel-2 L2A), rasterio, GDAL
Computation: Python 3, NumPy, SciPy, Pillow, Rasterio, rio-cogeo
AWS Pipeline: Lambda (triggers), EC2 (processing), S3 (storage), Bedrock (AI summaries)
Mapping: Leaflet.js, tile layers served from S3
Automation: Python boto3, Cron (EC2 quarterly jobs)

6. File Contents

The dataset includes metadata.json with satellite tile information, various index outputs (NDVI, EVI, NDWI, NDBI, MNDWI, NDMI), and statistical summaries. Please find the detailed list of files available for download in the Download Data & Maps section below.

7. How to Use This Dataset

You can explore the NDVI trends, plug the json file into your favourite tool, build visualisations, or compare it with earlier datasets. It's created to be flexible.

8. Leveraging AI

AI helped speed up some parts of the work, like spotting unusual patterns, creating brief insights, and checking for inconsistencies. All metadata and bin-statistics JSON files were loaded and parsed into structured dictionaries, ensuring the AI receives clean, context-rich inputs for stable summarisation and interpretation.

9. Limitations & Things to Keep in Mind

Cloud cover may affect accuracy
NDVI has known limitations
Spatial resolution is 10 meters
Some patterns may need ground truth validation

10. License / Permissions

Please refer to the How to Cite This Analysis section below for citation guidelines. This dataset is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).

11. Contact

For any questions, collaborations, or clarifications, feel free to reach out at: nature@ihugtrees.org

Data & Methods

Data Sources

Satellite: Sentinel-2C Level-2A (atmospherically corrected)
Provider: Microsoft Planetary Computer
Observation Date: 2025-12-05
Cloud Cover: 10.657708%
Spatial Resolution: 10 meters (NDVI), 20 meters (NDBI, resampled to 10m)

Index Calculations

NDVI = (NIR - Red) / (NIR + Red) using Bands 8 and 4
NDBI = (SWIR1 - NIR) / (SWIR1 + NIR) using Bands 11 and 8
Difference Index = NDVI - NDBI (positive = vegetation-dominated; negative = built-up-dominated)
Urban Heat Index = Composite metric flagging areas where NDBI > NDVI (high heat risk)

Processing Workflow

Images were processed using Python with the pystac-client and rasterio libraries. Cloud masking was applied using the QA60 scene classification layer. Statistics were computed using rasterio zonal statistics and exported as JSON for analysis. All geospatial outputs are provided as Cloud-Optimized GeoTIFFs (COGs) for efficient web access and GIS integration.

Limitations

Analysis represents a single-day snapshot; seasonal and temporal trends require time-series analysis
Cloud cover and atmospheric conditions affect image quality
10-meter resolution may not capture individual trees or small green spaces
Study area boundaries reflect the satellite image crop, not administrative city limits
Heat risk is inferred from spectral indices; ground validation recommended for mitigation planning

Download Data & Maps

Images & Visualizations

Geospatial Data (Cloud-Optimized GeoTIFFs)

Statistical Data

How to Cite This Analysis

Recommended Citation

Yaragarla, R. (2026). Urban Green Cover & Heat Risk Assessment: Singapore-region. I Hug Trees. Retrieved from https://ihugtrees.org/data-analytics/sentinel-ndvi/Singapore-region/2026/01/02/digest.html

Satellite data: Copernicus Sentinel-2 (ESA), processed via Microsoft Planetary Computer.

BibTeX Entry

@misc{ihugtrees_urban_singapore-region_2026,
  author = {Yaragarla, Ramkumar},
  title = {Urban Green Cover \& Heat Risk Assessment: Singapore-region},
  year = {2026},
  publisher = {I Hug Trees},
  url = {https://ihugtrees.org/data-analytics/sentinel-ndvi/Singapore-region/2026/01/02/digest.html},
  note = {Satellite data: Copernicus Sentinel-2}
}

License

This analysis and associated datasets are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). You are free to share and adapt this work with appropriate attribution.

Planetary Computer Citation

If using Microsoft Planetary Computer data, please cite: microsoft/PlanetaryComputer (2022)

Urban Green Cover & Heat Risk Assessment: Singapore-region A Research Brief on Urban Forestry Data