Indian Case Studies of RS and GIS in Population Analysis

India’s cities are changing faster than any pen-and-paper survey can capture. Between two Census years, neighborhoods rise on the edges of metros, farmlands turn into housing colonies, and informal settlements multiply along highways. To make sense of this constant churn, demographers and urban planners increasingly turn to Remote Sensing (RS) and Geographic Information Systems (GIS). By combining satellite imagery with spatial analysis, researchers can measure how a city has grown, where its population is concentrated, and how it might look a decade from now. The case studies of Kolkata, Delhi, and Gandhinagar illustrate just how powerful these tools have become for population analysis.

Kolkata: Mapping decades of urban sprawl 🔗

Kolkata is one of the most studied Indian cities when it comes to satellite-based urban analysis. Its dense core, surrounding municipalities, and rapidly expanding peri-urban belt make it an ideal laboratory for tracking how built-up land replaces open space over time. Researchers have used multi-temporal Landsat imagery, often spanning four decades, to detect these shifts with surprising precision.

Tracking built-up growth from 1980 onwards 🔗

A landmark study by Basudeb Bhatta applied remote sensing and GIS techniques to identify and analyse Kolkata’s urban growth pattern, applying several statistical methods that revealed an increasingly dispersed pattern of development even as the population growth rate slowed. This finding is critical: it shows that the city is spreading outward faster than its population is increasing, a classic sign of urban sprawl rather than densification.

A more recent analysis covering 1978 to 2017 used supervised Maximum Likelihood Classification to sort satellite pixels into five categories – urban built-up, open land, vegetation, agricultural land, and water bodies. The results showed that urban built-up area progressively increased by about 21.17 per cent, or roughly 239 square kilometres, over the study period, driven by new roads, flyovers, and settlements. Meanwhile, open land, vegetation, agricultural land, and water bodies all gradually declined.

Direction and shape of sprawl 🔗

Beyond the raw numbers, GIS allows researchers to visualise where the city grew. Studies of the Kolkata Metropolitan Area, which spans more than 1,800 square kilometres, found that the physical expansion of the city grew linearly in the south-easterly and south-westerly directions along the eastern bank of the river Hooghly. By overlaying satellite images from different decades, planners can identify which municipalities are absorbing migration pressure and where infrastructure investments should follow.

Delhi: Monitoring change and predicting the future 🔗

If Kolkata’s story is about measuring sprawl, Delhi’s story adds a second layer – forecasting. The capital region is one of the fastest-changing landscapes in South Asia, and researchers have used satellite data not just to record the past but to simulate what Delhi might look like in 2030 and beyond.

Four decades of land-use change 🔗

A widely cited study used Landsat satellite images from 1989, 2000, 2010, and 2020 to classify Delhi’s land use with the supervised maximum likelihood method. The findings are striking: urban and open areas increased by 13.44 per cent and 2.40 per cent respectively, while crop land fell by 10.88 per cent between 1989 and 2020. Forest area, interestingly, grew by 3.48 per cent, largely due to restoration programmes.

Other research focusing specifically on the North-West and South-West districts of Delhi using Landsat and IRS satellite scenes from 1977 to 2014 found a net change of +30.61 per cent in built-up area and −22.75 per cent in cultivated area over those decades. Such precise numbers would be virtually impossible to obtain through ground surveys alone, given the scale of the territory.

Predicting urban growth with simulation models 🔗

This is where Delhi’s research becomes particularly interesting for population studies. Researchers fed past satellite data into the Land Change Modeler (LCM) in TerrSet software to simulate land use in 2030. The simulated map predicted a 24.30 per cent increase in urban area by 2030, with a kappa coefficient of 0.96 indicating very high agreement between simulated and actual maps. A kappa value close to 1 means the model’s predictions match observed reality almost perfectly – a strong validation of using such tools for urban planning.

Another study using integrated cellular automata projected that Delhi’s built-up area will reach nearly 60 per cent of the total area of the city, while cropland and open land will shrink to 19.86 and 0.15 per cent respectively. The largest increases were projected in the northern, western, and south-western sub-districts. For demographers, these projections translate directly into questions about where future residents will live, how services like water and sanitation must scale, and which areas will face the greatest density pressures.

Mass transit and landscape change 🔗

Satellite data has also been used to study how infrastructure transforms cities. A GIS-based study around Delhi Metro stations examined land use changes within a 500-metre buffer zone around high-ridership stations, focusing on built-up areas, green cover, and open space availability. By comparing pre- and post-metro satellite imagery, researchers could quantify how stations like Kashmere Gate and Rajiv Chowk reshaped the surrounding urban fabric. This kind of analysis bridges transport planning, demography, and environmental studies.

Estimating population directly from satellite imagery 🔗

One of the most exciting applications of RS and GIS in population studies is using satellite imagery to estimate the number of people living in an area. This becomes especially valuable in informal settlements, rapidly growing peri-urban zones, and regions where census data is dated or incomplete.

The horizontal and vertical approach 🔗

Traditional population estimation from satellites combines two types of measurements. Horizontal measurements capture the footprint of each building – how much ground area it occupies. Vertical measurements capture height, typically derived from shadow lengths on the imagery, stereo satellite pairs, or Digital Elevation Models. Multiplying footprint by number of floors gives the total residential floor area, which is then converted into an estimated number of dwelling units and, eventually, people.

Of course, this approach requires careful calibration. Researchers must distinguish residential buildings from shops, schools, and offices, and they often use ground truthing – physical site visits – to verify what the imagery suggests. Without this verification, hostels could be mis-counted as family households, or commercial buildings as residences.

A 3D city model approach in Ahmedabad 🔗

An innovative Indian study attempted population estimation using high-resolution data from the indigenous Cartosat-2S satellite over a residential society in Ahmedabad. The researchers used Carto 2S multi-view data to generate a precise Digital Elevation Model (DEM) and Digital Terrain Model (DTM), extracted accurate building heights, validated them against ground data, and built a 3D city model to estimate the number of floors and carpet area of each apartment. This volume-based approach essentially turns the city into a stack of measurable spaces, with population estimated from the total liveable volume.

The Indian Space Research Organisation provides such high-resolution data through its Cartosat series of Earth observation satellites, which has been instrumental for cartographic and urban applications across the country.

Deep learning enters the picture 🔗

The newest generation of research goes further by using artificial intelligence to automate what humans used to do manually. A 2025 study over Gandhinagar combined Convolutional Neural Networks (CNN) to classify buildings as residential or non-residential and Artificial Neural Networks (ANN) to estimate population per building. Using approximately 48,000 building footprints over Gandhinagar’s urban area and high-resolution imagery, the system achieved an F1-score of 0.9936 and estimated the city’s population at 278,954. As the study points out, manual population estimation struggles with today’s denser and more fragmented urban landscapes, especially in cases of vertical expansion and informal housing.

For rural populations, where dwellings are scattered and census enumeration is logistically harder, similar techniques have been explored. Researchers have applied CNN models trained on satellite imagery and nightlight data to estimate population distribution across rural blocks, helping fill in gaps that traditional surveys may miss.

Why these case studies matter for population research 🔗

The Kolkata, Delhi, and Ahmedabad/Gandhinagar studies collectively show three complementary contributions of RS and GIS to population analysis. First, they offer a retrospective view – measuring with high accuracy how cities have changed over decades. Second, they provide a predictive view through models like the Land Change Modeler and cellular automata, allowing planners to anticipate growth. Third, they enable direct population estimation at a resolution that even the decadal Census cannot match.

Census data, while authoritative, captures a snapshot every ten years. Satellite-derived analyses can update that snapshot in near real time and reach into pockets that surveys often miss – slums, transient settlements, and peri-urban frontiers. For a country undergoing massive demographic transitions, this combination of methods is becoming indispensable. Programmes run by organisations such as the National Remote Sensing Centre under ISRO regularly use this approach to support urban planning, disaster response, and resource allocation across India.

For students of population and family health studies, these examples are not just about technology. They are about how we measure human presence in space – and how that measurement shapes everything from health service planning to housing policy.

What do you think? If satellite data can now estimate population with near-Census accuracy, should India consider shorter Census cycles supplemented by remote sensing, or do you see ethical and privacy concerns in relying so heavily on satellite-based monitoring of where people live?

References

1. https://www.tandfonline.com/doi/abs/10.1080/01431160802651967
2. https://link.springer.com/article/10.1007/s12524-020-01177-6
3. https://www.academia.edu/42721874/Mapping_of_Urban_Growth_Dynamics_in_Kolkata_Metropolitan_Area_A_Geospatial_Approach
4. https://pmc.ncbi.nlm.nih.gov/articles/PMC9124063/
5. https://www.researchgate.net/publication/291553507_Monitoring_land_use_change_and_its_drivers_in_Delhi_India_using_multi-temporal_satellite_data
6. https://www.tandfonline.com/doi/full/10.1080/24749508.2022.2132010
7. https://journals.bilpubgroup.com/index.php/re/article/view/9552
8. https://arxiv.org/pdf/2411.04612
9. https://www.isro.gov.in/Cartosat-2_Series_Satellite.html
10. https://arxiv.org/pdf/2509.12926
11. https://www.nrsc.gov.in/