Forecasting Urbanisation Along the NH-63 Corridor

Three urban areas on NH-63 at roughly twenty-kilometre intervals: Jagtial (district headquarters), Korutla (cottage-industry hub — the larger settlement at the 2011 Census, so administrative rank and demographic weight are not aligned in this corridor) and Metpally (outer market town nearest Armoor). Each town's analysis extent is a circular buffer on the town centre — radius 5 km for Jagtial, 3 km for Korutla and Metpally — giving fixed extents of 78.54 and 26.71 sq km respectively, invariant across all three epochs, in WGS 84 / UTM Zone 44N.

Sentinel-2A (10 m visible/NIR, 20 m SWIR) and LISS-4 (5 m, green/red/NIR) imagery was acquired for December 2015, 2020 and 2025 — December held constant across years to preserve the post-kharif, pre-rabi window and avoid confounding agricultural and bare-soil classes with crop-cycle timing. LISS-4 drove the primary object-based segmentation; Sentinel-2A's SWIR bands supplied the built-up and bare-soil indices LISS-4 alone cannot.

Object-based analysis — the Orfeo Toolbox Large Scale Mean Shift algorithm via its QGIS plugin, run on the LISS-4 false-colour composite for each town-year — suppresses salt-and-pepper noise and keeps each classified entity a meaningful ground unit. Identical parameters were used across every town-year combination; the resulting polygon layers held between 5,099 and 20,608 objects per town per year.

Water Body, Agriculture (Active), Agriculture (Fallow), Bare Soil/Quarry, Built-up, Vegetation, and a dedicated Wet/Black Soil class — essential because dry Vertisol closely mimics concrete and asphalt reflectance, making its separation the principal control on spectral confusion with built-up. Training samples were assigned by visual inspection of segmentation polygons using the LISS-4 false-colour composite as primary reference, supplemented by Sentinel-2A and Google Earth Pro high-resolution imagery.

Ensemble classification (Breiman 2001) via scikit-learn: n_estimators=500 · max_features=sqrt · bootstrap=True · oob_score=True · class_weight=balanced. Every town and epoch was classified by its own independently trained model — no training data transferred between towns — producing nine independently trained and validated classifications, each split 70/30 by stratified random sampling. Accuracy assessment on the 30% hold-out reports overall accuracy, Cohen's kappa, producer's/user's accuracy, F1, a confusion matrix, five-fold cross-validation, and both Gini and permutation importance across the 27 predictors.

Three administrative data families were compiled from the Telangana Open Data Portal: TGNPDCL electricity records (connections, units, sanctioned load by sector, 2019–2025), TG RTA vehicle registrations by fuel and wheel type (Jagtial & Korutla, 2019–2025), and Beedi & Cigar Workers Act establishment registrations (2023–2025). None share the classifier's error structure — that independence is this study's central methodological contribution, not a supplementary check. Three indicator families — land-cover share, electricity, vehicles — were then extended to 2030/2035/2040 with a two-layer hybrid: a pooled XGBoost regressor supplies a smoothed 2025 anchor (tree ensembles cannot extrapolate past their training years), and a per-series growth-rate layer — logit-space trend for bounded land-cover shares, damped-trend CAGR (Gardner & McKenzie 1985, φ=0.85) for electricity and vehicles — projects the actual trend forward.

Overall accuracy ranges 72.73–90.59%, Cohen's kappa 0.680–0.883; eight of nine models fall in the Landis & Koch (1977) "almost perfect" band, one (Korutla 2015) in "substantial". Cross-validated scores track hold-out accuracy closely — no material overfitting. Built-up is consistently the best-classified category (F1 0.874–1.000); Bare Soil/Quarry the weakest, reflecting dry Vertisol's spectral proximity to impervious surface.

At Jagtial, agriculture fell 67.6% and bare soil 73.7% — cropland and waste ground yielding to built-up, vegetation and Wet/Black Soil. Korutla breaks this pattern: bare soil/quarry collapsed 95.5% (sharpest of the three), yet combined agriculture rose 33.1%, driven by a more than threefold rise in Agriculture (Active) — Korutla's urbanisation drew disproportionately on bare, fallow and quarried ground rather than productive cropland, while fringe land was simultaneously brought under cultivation.

The strongest non-spatial series — seven continuous annual observations across four sectors, wholly independent of the classification. Service connections grew 19.4% (Jagtial), 26.1% (Korutla), 23.4% (Metpally) over six years; load per connection, the closest available proxy for internal densification, rose 45.7%, 58.4% and 74.5% respectively. Commercial connections outgrew domestic in every town (tertiarisation); industrial consumption fell 18.8–32.4%. Streetlight connections — a direct index of municipal road-network extension — grew 7.6% p.a. at Jagtial and 8.6% at Metpally, but only 0.2% p.a. at Korutla, whose built-up CAGR of 4.98% is more than twenty times its streetlight growth.

Cross-sectional comparison for 2023–2025 reveals a structural difference the satellite and electricity data alone could not resolve: Korutla accounts for 54% of the corridor's registered beedi workforce across nine firms, against three at Jagtial and four at Metpally. Beedi rolling is home-based piece work — the registered establishment is a collection centre while production occurs in workers' dwellings — so a town built on this system accumulates dwellings within walking distance of collection points, generates domestic (not industrial) demand, and has little reason to extend the formal street-lighting network as its footprint grows, matching Korutla's measured profile exactly.

Registrations fell 38.9% (Jagtial) and 37.4% (Korutla) 2019–2025 — a compositional transition, not reduced ownership. At Jagtial, four-wheeler registrations rose 51.7% while tractor/agricultural- vehicle registrations fell 44.7%, inverting the car-to-tractor ratio from 0.42 to 1.16, corroborating the agricultural-land loss measured independently from imagery. EV penetration reinforces this: near zero in 2019 to 8.6% (Jagtial) and 11.4% (Korutla) of registrations by 2025.

Built-up, connection and load-intensity growth all move upward in every town, and Korutla's fastest connection and load growth belongs to the town with the second-fastest, not slowest, built-up growth. The one indicator breaking step is the streetlight network.

Pooling all nine town-epoch observations, built-up area correlates strongly with vegetation (r = 0.91, p = 0.001) and water body (r = 0.89, p = 0.001), moderately with Wet/Black Soil (r = 0.70, p = 0.035), and only weakly with agriculture (r = 0.42) or bare soil (r = 0.09) — not significant. This does not mean built-up growth causes vegetation or water-body growth; more plausibly all three rise across the same observations as a shared decadal trend. The more informative result is negative: agriculture and bare soil, which a simple "built-up consumes farmland" model would expect to correlate most strongly and negatively with built-up, in fact show the weakest relationships.

Jagtial's built-up area is projected to reach 16.13 sq km by 2040 (+47.7% over 2025, roughly one-fifth of its mapped extent); Korutla and Metpally grow more slowly in absolute area (+11.5% and +13.1%). Connected load is projected to roughly double at Korutla and Metpally by 2040 (+115.3% and +101.5%) against a smaller but still substantial rise at Jagtial (+74.3%), while service connections grow far more moderately — the densification signature is projected to persist. Vehicle registrations are projected to continue their post-2019 decline in raw count, better read as a compositional shift toward four-wheelers and EVs within a shrinking fleet than a literal headcount forecast.