Geospatial applications: urban, agriculture, disaster and natural resources

Geoinformatics / Surveying (Geomatics) (B.Tech) · Applications · free study notes for exams, placements and interviews

In short

Geospatial application means using surveying, GNSS, satellite remote sensing and GIS analysis to answer a decision question about a place — where a city is growing, how much wheat was sown, which villages a flood will reach, which watershed to treat first. The same workflow (acquire → georeference → classify/analyse → validate → publish) is re-targeted at four big domains: urban, agriculture, disaster and natural resources. It matters because almost every Indian government mission now runs on a spatial layer: AMRUT GIS master plans, SVAMITVA property cards, PMFBY crop insurance, Jal Jeevan Mission source maps, NDMA hazard atlases. It is examined in GATE Geomatics (GE) under remote sensing and GIS applications, in university "Applications of RS & GIS" papers, and asked in interviews at NRSC/ISRO, Survey of India, state remote sensing centres, ESRI India, Genesys, RMSI, Tata Consulting Engineers and municipal GIS consultancies.

Explained

The common application workflow

Define the decision and the minimum mapping unit (MMU). Scale drives sensor choice: 1:50,000 LULC needs ~23.5 m data, 1:10,000 needs ~5 m, urban utility mapping needs sub-metre.
Data acquisition: satellite imagery, aerial/UAV photography, DGPS/CORS ground survey, total station, LiDAR, plus attribute data (census, revenue records, IoT sensors).
Pre-processing: radiometric correction (DN → radiance → reflectance), atmospheric correction, geometric correction and orthorectification with a DEM, mosaicking, cloud masking.
Analysis: image classification (supervised/unsupervised/object-based), index computation, change detection, overlay, buffer, network and terrain analysis, multi-criteria decision analysis (AHP-weighted overlay).
Validation: field truthing and a confusion matrix (overall accuracy, producer's/user's accuracy, kappa); positional accuracy as RMSE against checkpoints.
Delivery: maps, web GIS/ geoportal services (OGC WMS/WFS), dashboards, decision-support systems — increasingly on Bhuvan, state geoportals and PM Gati Shakti NMP.

Indian data, policy and standards

Indian satellites: Resourcesat-2/2A (LISS-III 23.5 m, LISS-IV 5.8 m, AWiFS 56 m), Cartosat-2S/3 (0.25–0.65 m, stereo for DEM), EOS-04/RISAT (C-band SAR, cloud-penetrating — the workhorse for monsoon floods), Oceansat-3, INSAT-3D/3DR (half-hourly weather), NISAR (NASA-ISRO SAR, launched 2025).
Open foreign data: Landsat 8/9 (30 m, 16-day), Sentinel-2 (10/20 m, 5-day), Sentinel-1 SAR, MODIS (250 m–1 km daily), SRTM/ASTER/CartoDEM/Copernicus DEM.
Policy: the National Geospatial Policy, 2022 (DST) targets a national digital twin, high-resolution topographic survey and 1:500 scale urban mapping; the Guidelines for Acquiring and Producing Geospatial Data, 2021 deregulated collection by Indian entities, removed prior approval for most surveys and mapping, and retained restrictions mainly on finer-than-threshold data access by foreign entities and on negative-list attributes.
Institutions: Survey of India (national mapping agency, CORS network, Open Series Maps), NRSC/ISRO (Bhuvan, NDEM, NICES), SAC, NIC-NSDI, state remote sensing application centres, CGWB, FSI, GSI, NCRMP/NDMA.
Reference systems: modern work uses WGS-84/ITRF with UTM projection; legacy Indian topo sheets used the Everest spheroid with polyconic projection and Indian Grid — never mix the two datums (shifts run into hundreds of metres).
Classification standard: NRSC's hierarchical LULC scheme (Level I/II/III) used for the national 1:50,000 LULC cycles; FSI's forest-cover definition is canopy density classes on ≥1 ha patches (very dense ≥70%, moderately dense 40–70%, open 10–40%), reported biennially in the India State of Forest Report.

Urban applications

Base mapping and master plans: large-scale base maps (1:1,000–1:4,000) from satellite/UAV/total station feed GIS-based master plans — the AMRUT sub-scheme covered ~500 cities; the former NUIS scheme pioneered urban spatial databases.
Property and land records: SVAMITVA drone-based survey of rural abadi areas issues property cards; DILRMP/Bhu-Naksha digitise cadastral maps and link them to record-of-rights for urban/peri-urban parcels.
Urban growth and sprawl: multi-date LULC change detection gives built-up expansion rate; Shannon's entropy quantifies whether growth is compact or dispersed; NDBI and impervious-surface fraction map densification.
Urban heat and environment: land surface temperature from thermal bands (Landsat TIRS) maps urban heat islands and correlates negatively with NDVI; air-quality interpolation over CPCB station networks.
Utility and asset GIS: water supply, sewerage, storm-water, electricity and gas networks as topologically connected networks — used for leak isolation, hydraulic modelling (EPANET/SWMM linkage) and asset registers.
Transport and mobility: GTFS-plus-GIS route rationalisation, accessibility/service-area analysis, traffic shortest-path and emergency response routing, metro alignment and TOD planning, GIS integration with ICCC dashboards in Smart Cities.
Municipal revenue: satellite/UAV-based property-tax surveys detect unassessed and under-assessed buildings — a standard revenue-augmentation project for ULBs.
Site suitability: weighted overlay of slope, LULC, groundwater, road proximity and hazard layers for landfills (per the Solid Waste Management Rules, 2026), STPs and new townships.

Agriculture applications

Crop inventory (acreage): supervised classification of multi-date imagery within the crop calendar; multi-date separates crops by phenology. Operationalised in FASAL for in-season forecasts of major crops and CHAMAN for horticulture.
Crop condition and yield: NDVI/EVI/LAI time series compared with the previous years' profile; yield models are agromet-regression, crop-simulation (DSSAT/InfoCrop) coupled with remote-sensing inputs, or spectral-yield regression. Production = area × yield.
Drought monitoring: NADAMS uses AWiFS/MODIS NDVI anomalies, Vegetation Condition Index and rainfall deviation to grade agricultural drought — an input to the Drought Manual's declaration process.
Crop insurance and surveys: PMFBY uses technology-based yield estimation (YES-TECH) and smartphone-based crop-cutting experiments; the Digital Crop Survey under AgriStack geo-tags every plot's sown crop, and Krishi-DSS brings these layers into one decision platform.
Soil and land resources: NBSS&LUP soil resource maps, soil fertility layers behind Soil Health Cards, soil-salinity and waterlogging mapping, land-capability classification for crop planning.
Irrigation: command-area performance (irrigated vs unirrigated area from imagery), canal network GIS, evapotranspiration-based water accounting (SEBAL/METRIC), tank and reservoir storage from water-spread area and bathymetry.
Horticulture, plantation and forecasting limits: mixed smallholder plots (often <0.5 ha) fall below the pixel size of medium-resolution data, so estimates are given at block/district level with a stratified sampling design, not per field.

Disaster applications (geospatial role)

Hazard zonation (pre-event): flood hazard maps from historic inundation and DEM-based modelling; GSI's National Landslide Susceptibility Mapping at 1:50,000; seismic zoning per IS 1893 (Part 1):2016 (Zones II–V) with microzonation studies for cities; cyclone wind-hazard and storm-surge maps; drought proneness.
Vulnerability and risk: exposure layers (buildings, population grids, lifelines) combined with hazard; risk = hazard × exposure × vulnerability, mapped to prioritise mitigation.
Early warning and preparedness: INSAT-3D/3DR cyclone tracking, Doppler radar nowcasts, CWC flood forecasting, GIS layers of shelters, routes and resources in NRSC's NDEM and Bhuvan disaster services.
Response mapping: SAR imagery (RISAT/EOS-04/Sentinel-1) maps inundation through cloud within hours; change detection and very-high-resolution imagery give damage assessment; emergency routing and helipad/relief-camp siting.
Recovery: pre- and post-event comparison for compensation and reconstruction planning; permanent monitoring of river morphology, glacial lakes and subsidence (InSAR).
Limits: optical data fails under monsoon cloud; SAR confuses flooded vegetation, wet soil and radar shadow; a 30 m DEM is too coarse for urban flood depth.

Natural resources: water, forest, soil, minerals, coast

Groundwater: groundwater prospect maps (1:50,000) from lithology, geomorphology, lineaments and recharge conditions — used to site borewells under Jal Jeevan Mission; CGWB's NAQUIM aquifer mapping.
Watershed management: watershed delineation from DEM, prioritisation by sediment yield/erosion using USLE/RUSLE and runoff by SCS-CN, then siting check dams and treatments — the standard WDC-PMKSY practice.
Surface water: reservoir and tank water-spread mapping (NDWI/MNDWI), the National Wetland Atlas (SAC), river and canal monitoring, snow-cover and glacier mass-balance studies.
Forestry: biennial forest-cover assessment by FSI using LISS-III data with canopy-density classes; ISFR 2023 reported forest cover about 21.76% and forest-plus-tree cover about 25.17% of the geographic area (figures revise each cycle); working-plan GIS, fire alerts (MODIS/VIIRS-based FSI fire alert system), encroachment detection, biomass and carbon estimation (now also with NISAR/LiDAR).
Wasteland and degradation: Wasteland Atlas of India and desertification/land-degradation atlases (SAC) classify saline, ravinous, gullied and barren lands for reclamation.
Minerals and geology: lineament and lithological mapping, hyperspectral mineral targeting, mining-lease boundary and volume monitoring from stereo/UAV DEMs, GSI's National Geoscience Data Repository.
Coastal: shoreline change detection, CRZ mapping and hazard-line delineation under the CRZ Notification, 2019, mangrove and coral mapping, bathymetry.

Accuracy, limitations and comparison of approaches

Resolution trade-off: spatial vs spectral vs temporal vs radiometric — finer spatial resolution usually means a narrower swath, costlier data and longer revisit.
Mixed pixels dominate Indian smallholder and dense-urban scenes; remedies are sub-pixel/spectral unmixing, object-based classification on high-resolution data, or data fusion (pan-sharpening).
Pixel-based vs object-based (OBIA): pixel-based is fast and fine for medium resolution; OBIA uses shape, texture and context and is preferred above ~5 m.
Supervised vs unsupervised vs machine learning: supervised needs training signatures and field knowledge; unsupervised (ISODATA/k-means) needs post-labelling; Random Forest/SVM/CNN now give the best accuracies but need large labelled sets and are harder to defend in audit.
Positional vs thematic accuracy: RMSE in metres versus overall accuracy/kappa in per cent — they are separate and both must be reported. Common acceptance: overall accuracy ≥85%, kappa ≥0.8 for LULC.
Institutional limits: data silos, non-standard attribute schemas, outdated cadastral maps that do not match imagery, and lack of trained staff in ULBs — often the real obstacle, not the technology.

Key points

Geospatial application - converting location data into a decision on land, crop, hazard or resource.
Scale decides sensor - MMU and output map scale fix the resolution needed, not the other way round.
NRSC LULC scheme - the hierarchical Level I/II/III classification used for India's 1:50,000 LULC cycles.
National Geospatial Policy, 2022 - liberalised, mission-mode policy aiming at high-resolution national mapping and a digital twin.
2021 Geospatial Guidelines - removed prior approval for Indian entities collecting and processing geospatial data.
NDVI - normalised vegetation index, −1 to +1; healthy crop/forest typically 0.4–0.9, bare soil ~0.1–0.2, water negative.
NDWI/MNDWI - water-delineation indices; NDBI - built-up index using SWIR and NIR.
Shannon entropy - sprawl measure; value near 1 (relative form) means dispersed, near 0 means compact growth.
SCS-CN method - rainfall-runoff estimation from curve number, used in watershed prioritisation.
USLE/RUSLE - soil-loss estimation (A = R·K·LS·C·P) for erosion mapping and treatment priority.
FASAL / NADAMS / CHAMAN - national space-based crop forecasting, drought monitoring and horticulture assessment programmes.
SVAMITVA - drone survey of rural inhabited land giving legal property cards.
SAR for floods - microwave data images through cloud and at night; indispensable in the Indian monsoon.
NDEM - NRSC's National Database for Emergency Management, the spatial backbone for response.
IS 1893 (Part 1):2016 - seismic zone map (II–V) used as the hazard layer in risk studies.
Kappa coefficient - accuracy corrected for chance agreement; >0.8 is strong.
Mixed pixel - the main error source in Indian smallholder and urban classification.
Datum discipline - mixing Everest/Indian Grid with WGS-84/UTM causes gross positional error.

Formulas and examples

NDVI = (NIR − Red)/(NIR + Red); dimensionless, −1 to +1
NDWI = (Green − NIR)/(Green + NIR); MNDWI = (Green − SWIR)/(Green + SWIR); NDBI = (SWIR − NIR)/(SWIR + NIR)
Area of a class A = N × p²; N = number of pixels, p = pixel size (m) → A in m² (÷10,000 for hectares)
Map scale relation: ground distance = map distance × scale denominator; smallest mappable feature ≈ 0.5 mm × scale denominator
Shannon entropy H = −Σ pᵢ ln pᵢ; relative entropy Hₙ = H / ln n; pᵢ = built-up share of zone i, n = number of zones
SCS-CN: S = (25400/CN) − 254 mm; Iₐ = 0.2S; Q = (P − Iₐ)²/(P − Iₐ + S) mm, for P > Iₐ
USLE: A = R × K × LS × C × P; A in t/ha/yr (R = rainfall erosivity, K = soil erodibility, LS = slope length-steepness, C = cover, P = practice factor)
Accuracy: OA = Σ diagonal / N; Kappa κ = (OA − Pe)/(1 − Pe), Pe = Σ(rowᵢ × colᵢ)/N²
Production = Classified area (ha) × Yield (t/ha); LST from thermal band: Lλ → brightness temperature Tв = K₂ / ln(K₁/Lλ + 1), then LST = Tв / [1 + (λ·Tв/ρ) ln ε]

Example 1 — Wheat acreage, production and value from LISS-III. Classified wheat pixels = 18,450; pixel size 23.5 m.

Pixel area = 23.5² = 552.25 m².
Area = 18,450 × 552.25 = 10,189,012.5 m² = 1,018.90 ha (10.19 km²).
At a district yield of 3.8 t/ha, production = 1,018.90 × 3.8 = 3,871.8 t = 38,718 quintals.
At a wheat MSP of ₹2,425 per quintal (RMS 2025-26; MSP is revised each season), value = 38,718 × 2,425 = ₹9.39 crore.

Example 2 — Indices for one pixel. Reflectances: Green 0.11, Red 0.09, NIR 0.42, SWIR1 0.33.

NDVI = (0.42 − 0.09)/(0.42 + 0.09) = 0.33/0.51 = +0.65 → dense healthy vegetation.
NDWI = (0.11 − 0.42)/(0.11 + 0.42) = −0.31/0.53 = −0.58 → not water.
NDBI = (0.33 − 0.42)/(0.33 + 0.42) = −0.09/0.75 = −0.12 → not built-up. Consistent interpretation: a vegetated agricultural pixel.

Example 3 — Urban sprawl by Shannon entropy. Built-up shares in four peri-urban zones: 0.40, 0.30, 0.20, 0.10.

H = −[0.40 ln0.40 + 0.30 ln0.30 + 0.20 ln0.20 + 0.10 ln0.10] = −[(−0.3665) + (−0.3612) + (−0.3219) + (−0.2303)] = 1.2799.
Hₙ = 1.2799 / ln 4 = 1.2799 / 1.3863 = 0.923 → growth is strongly dispersed (sprawl), not compact; infrastructure cost per household will be high.

Example 4 — Runoff for watershed planning (SCS-CN). Watershed area 12 km², CN = 80 (AMC-II), design rainfall P = 70 mm.

S = 25400/80 − 254 = 317.5 − 254 = 63.5 mm; Iₐ = 0.2 × 63.5 = 12.7 mm.
Q = (70 − 12.7)² / (70 − 12.7 + 63.5) = 57.3²/120.8 = 3,283.29/120.8 = 27.18 mm.
Runoff volume = 0.02718 m × 12 × 10⁶ m² = 3,26,160 m³ (≈3.26 lakh m³) — the storage a series of check dams must handle.

Example 5 — Accuracy assessment of a 3-class LULC map (200 points).

Classified ↓ / Reference →	Built-up	Crop	Water	Row total	User's acc.
Built-up	60	8	2	70	85.7%
Crop	7	85	3	95	89.5%
Water	1	4	30	35	85.7%
Column total	68	97	35	200
Producer's acc.	88.2%	87.6%	85.7%

OA = (60 + 85 + 30)/200 = 175/200 = 87.5%.
Pe = (70×68 + 95×97 + 35×35)/200² = (4,760 + 9,215 + 1,225)/40,000 = 15,200/40,000 = 0.38.
κ = (0.875 − 0.38)/(1 − 0.38) = 0.495/0.62 = 0.80 → acceptable for a 1:50,000 LULC product.

Domain	Typical data	Core analysis	Indian programme
Urban	Cartosat/UAV sub-metre, Sentinel-2	Base mapping, change detection, network, overlay	AMRUT GIS master plans, SVAMITVA, Smart Cities
Agriculture	AWiFS, LISS-III/IV, Sentinel-2, MODIS	Multi-date classification, NDVI time series, yield models	FASAL, NADAMS, CHAMAN, Digital Crop Survey
Disaster	RISAT/EOS-04 SAR, INSAT-3D, CartoDEM	Hazard zonation, inundation mapping, risk overlay	NDEM, Bhuvan disaster services, GSI NLSM
Natural resources	LISS-III, Landsat, hyperspectral, DEM	Prospect mapping, USLE/SCS-CN, cover assessment	NAQUIM, WDC-PMKSY, ISFR, Wetland Atlas

Where it is used

Government and PSU mapping: NRSC/ISRO, Survey of India, state remote sensing centres, GSI, FSI, CGWB, NBSS&LUP — as scientist/engineer, image analyst and project associate roles.
Urban and infrastructure consultancies: RMSI, Genesys, ESRI India, CyberSWIFT, L&T, AECOM, TCE — GIS analyst, utility-GIS surveyor, master-plan and property-tax survey teams.
Agri-tech and insurance: PMFBY implementing insurers, Cropin, Satsure, Agribazaar, NABARD-funded projects — crop acreage/yield analytics and claim verification.
Utilities and energy: power distribution GIS (RDSS consumer indexing), gas pipeline route surveys (PNGRB-regulated city gas networks), solar and wind site suitability.
Disaster and insurance risk: NDMA/SDMA hazard atlases, reinsurance catastrophe models, highway and railway landslide and flood risk assessment.
Platforms and open data: Bhuvan, PM Gati Shakti National Master Plan, state geoportals, Google Earth Engine and QGIS/ArcGIS workflows used daily by analysts.

Interview questions

Q: Which resolution would you choose to map urban plots, and why? A: Sub-metre (Cartosat-3, UAV, or aerial) — plot boundaries and roof detail need an MMU far below a metre; a 23.5 m pixel cannot resolve a 100 m² plot.
Q: Why is multi-date imagery essential for crop classification? A: Different crops have similar spectra on a single date but different phenological curves; multi-date NDVI separates them by sowing, peak and harvest timing.
Q: Why is SAR preferred for monsoon flood mapping in India? A: Microwaves penetrate cloud and work at night, and smooth standing water gives a very low backscatter, so flooded areas appear dark and are easy to delineate.
Q: What do NDVI and NDBI tell you together about a pixel? A: High NDVI with low NDBI means vegetation; low NDVI with high NDBI means impervious built-up — together they help resolve the built-up versus bare-soil confusion.
Q: What is the difference between producer's and user's accuracy? A: Producer's accuracy is based on reference (column) totals — omission error; user's accuracy is based on classified (row) totals — commission error, i.e. the map's reliability for a user.
Q: Why is kappa reported alongside overall accuracy? A: Kappa removes the agreement expected by chance, so it is comparable across maps with different class proportions.
Q: How would you prioritise watersheds for treatment? A: Delineate from a DEM, estimate runoff by SCS-CN and soil loss by USLE/RUSLE, rank sub-watersheds by sediment yield and water scarcity, then validate ranking in the field.
Q: What is the main cause of error when a cadastral map is overlaid on satellite imagery? A: Datum/projection mismatch and the fact that old cadastral maps are non-georeferenced, undigitised and distorted — rubber-sheeting with ground control is needed, and residual error must be reported.
Q: Why do satellite-based yield estimates fail for Indian smallholders? A: Plots are often smaller than the pixel, so mixed pixels dominate; estimates are therefore aggregated to block/district level or need high-resolution data.
Q: What changed for private geospatial work after 2021? A: The Guidelines for Acquiring and Producing Geospatial Data, 2021 removed prior approval and licensing for Indian entities for most collection and mapping, with restrictions retained chiefly for foreign entities and sensitive attributes; the National Geospatial Policy, 2022 then set the national mapping targets.
Q: Which standard gives the seismic hazard layer for a risk map? A: IS 1893 (Part 1):2016 zone map (Zones II–V), refined by city-level microzonation studies where available.
Q: How do you decide between pixel-based and object-based classification? A: Pixel-based for medium-resolution, spectrally distinct classes; object-based for high-resolution imagery where shape, texture and context separate classes better and reduce the salt-and-pepper effect.

Common mistakes

Quoting a class area without the pixel size or the accuracy figure — an area estimate without its confusion matrix is unverifiable.
Mixing datums (Everest/Indian Grid topo sheets with WGS-84 GPS data) and reporting the resulting offset as real change.
Using a single-date image for crop or change studies and misreading seasonal variation as land-use change.
Treating NDVI as a yield value rather than a vigour proxy that needs calibration against crop-cutting data.
Choosing resolution by what is free rather than by the output map scale and MMU, then over-claiming the map's usability.
Comparing two-date classifications made with different schemes or sensors, so change detection reports classification differences, not ground change.
Assuming a 30 m global DEM is adequate for urban drainage or flood depth, where sub-metre vertical accuracy matters.