Extracting Urban Footprint for 2 Indian cities using Remote Sensing
Updated: 5 days ago
It's been a while since I've analyzed satellite imagery (Mauritius Oil Spill, August 2020). In this article, I have mapped the spatial extent of built-up areas i.e. the Urban Footprint, of a couple of Indian cities.
To share the context, not only does India have 3 cities in the Top 10 world's fastest growing cities list currently, but also it is predicted to occupy all the 10 spots in the world's fastest growing cities list between now and 2035!
While economic activity, productivity & growth is fuelled by urban centers, unchecked urbanization is risky too, for obvious ecological reasons among others.
So how does imagery from earth observation satellites help us to monitor urbanization?
As you may already know, satellites capture the energy reflected from the earth's surface, the source of illumination being either an external light source such as the Sun or from the radio waves emitted by the sensor transmitter onboard the satellite itself. The former method captures visible light in the form of Optical imagery while the latter captures energy not visible to human eye (microwaves, radiowaves etc.) in the form of Radar imagery. Now, Radar has several advantages over optical imagery which makes it much better placed to observe and identify the earth's surface for the requisite phenomena.
For mapping Urban Footprint in particular, Radar-based remote sensing is effective as it has properties which make it suitable to distinguish impervious, man-made structures which reflect a strong energy signal / back-scatter as well as have high coherence values - from natural landscapes (agriculture, vegetation, water bodies etc.) which have comparatively less backscatter and / or coherence.
Below, I've analyzed Synthetic Radar Aperture (SAR) imagery over New Delhi to map its urban footprint. Let's have a look.
(The sliders below are best viewed on a PC.)
The image on the left of the slider is a RGB composite model of radar imagery bands over a cross-section of New Delhi in October 2020 (Slider of the complete study area can be seen here). Yellow color represents urban built-up areas whereas green represents non-urban areas (such as natural vegetation, agriculture and barren lands). To derive the RGB model, I laid two recent radar imagery on top of each other and studied the properties of the overlay. Pixels of urban areas would remain the same in both the imagery datasets (due to strong backscatter as well as strong coherence) whereas corresponding pixels of non-built up areas would have a small, yet significant, change in the coherence (eg., the presence of wind would cause distortions in the orientation of vegetation thereby resulting is less backscatter and coherence) which enables us to distinguish it from urban areas and attribute a different color to it.
(Note: Each pixel has a spatial resolution of 5 m by 20 m. i.e. 100 sq. m.)
In the slider above, we have extracted only the yellow pixels i.e. only the urban built-up areas and attributed white color to it (binary: 1) whereas all the remaining pixels were attributed black color (binary: 0). Again, the slider above is just a subset of the study area to aid in visual representation.
As you can imagine, we can also compare urban footprint over multiple periods of time to understand how the urban landscape has changed i.e. perform Temporal analysis. The above image slider compares the entire study area over two time periods nearly five years apart - Oct 2020 and January 2016.
Can you spot some deviations in the yellow pixels, in particular? It can be difficult to observe as the spatial extent comprises the entire city's extent and beyond, however, you would be still able to spot minor deviations. Notice that a yellow spot emerges slightly north-west of Ghaziabad in the 2020 image on the right. Let's inspect it closely...
Upon observing that precise spot on Google Earth's basemap, I have been able to ascertain that the cluster of new yellow pixels point to the newly built WUPPTCL Ataur Power Sub-station.
The accuracy of satellite imagery processing output is remarkable!
Want to test other points for yourself? Download the KMZ output files here and open it using Google Earth on your PC.
Malappuram was recently ranked as the fastest growing city in the world (2015 -2020) in terms of population (44% increase). Let's see what Urban Footprint Mapping of this region has to suggest.
You can download the RGB and Urban Footprint files from here.
The increase in white pixels (urban areas) in the image on the right is visually discernible. Let's find out by how much -
2016 has close to 17,000 white pixels
2020 has nearly 20,000 white pixels. An increase of 3000 pixels multiplied by 100 square metres per pixel gives us 300,000 sq. m. or 0.3 sq. km increase in Urban built-up area. This isn't a low figure. It is equivalent to addition of 21+ Eden Gardens cricket stadiums (14,200 square metres each) in terms of built-up area in a space of just 4 years.
Therefore, we can roughly conclude that a 44% growth in population (5 years) is met by an increase of 18% increase in built-up / urban area (4 years).
It may not be completely evident from the imagery output above - these images are not raw images captured by satellites, rather, they have been computed using the various bands within the radar imagery product. The methodology of calculation is technical in nature and backed by years of research. Modern technology makes it much easier to perform the calculations, however, processing the imagery datasets still requires considerable computing resources.
The underlying principle is relatively easy to grasp, though. Each pixel within the imagery dataset gives us plenty of information about the properties of the surface captured (amplitude, phase, orientation, height etc.). However, the true beauty lies in how we play around with these available data points to unearth new insights. For example, if we were to amplify certain properties and repress other properties, that could give us an interesting insight; if we overlay one pixel of the same area captured in a particular time period onto the corresponding pixel of the same area in another time period, the comparison may unearth another useful insight, and so on.
All these insights from processing just a tiny pixel! Isn't this fascinating?
You can read more of my satellite imagery analysis from the links here - 1) Forest Fires in Uttarakhand, 2) Flooding from Amphan Cyclone, 3) Land Subsidence in Kolkata, 4) Urban Heat Index (UHI) in Mumbai, 5) Ship Detection in Gulf of Kutch, 6) Air Pollution in India 7) Volcano Deformation at 3 locations, 8) Snow Cover in Himachal Pradesh, 9) Mapping Water Bodies, 10) Assessing Damage from Beirut Explosion 10) Sargassum Invasion in Caribbean Sea 11) Drought Monitoring at Indirasagar Reservoir, & 12) Estimating Actual Evotranspiration in Punjab
Feel free share your comments, queries or suggestions.
Intelloc Mapping Services | Mapmyops is engaged in providing mapping solutions to organizations which facilitate operations improvement, planning & monitoring workflows. These include but are not limited to Supply Chain Consulting, Drone Services, Location Analytics & Applications, Site Characterization, Satellite Imagery Analytics & Polluted Water Remediation. Projects can be conducted pan-India and overseas.
Several other mapping and operations workflows are documented on this Geo-blog. Reach out to us via email - email@example.com or book a paid consultation (video meeting) from the link placed at the footer of this site's landing page.
Much Thanks to EO College for the training material. Analysis has been done on ESA's SNAP platform using Sentinel 1 SAR imagery