Land Use and Land Cover Changes in Kullu District, Himachal Pradesh (2017–2025)

Kullu is a district of river valleys, apple orchards, deodar forests, high pastures and glaciated peaks. Over the last decade it has also seen a lot of building, road widening and tourism growth. But how much has the land actually changed, and where did the change come from? This post works through annual 10-metre land-cover maps for 2017 and 2025 to find out.

At a glance

IndicatorResult
District area analysed5,504.48 km²
Built-up land+70.78% (92.17 → 157.40 km²)
Net tree cover−148.17 km²
Area that changed class19.16%

1. Why Kullu?

The Himalayan landscape runs from dense forested slopes and river valleys to alpine meadows, barren high-altitude terrain and permanent snow. In recent decades, settlement growth, infrastructure, tourism, horticulture and changing land use have put increasing pressure on it. Kullu is a good place to study this. The Beas valley, the Parvati valley and the upper Kullu Himalaya sit side by side, so settlements, forests, farmland, rangelands and snowfields all appear in a single district. This study looks at how much each land-cover class changed and, more importantly, at the transitions between classes.

[Image 1: Location and relief map of Kullu district, showing the Beas and Parvati valleys]

2. Data and Method

2.1 The dataset

I used the Esri Sentinel-2 Land Cover Explorer, which publishes annual land-cover classifications derived from Sentinel-2 imagery at 10 m resolution. The maps have nine classes: water, trees, flooded vegetation, crops, built-up, bare ground, snow/ice, clouds and rangeland.

ParameterDetails
Data sourceEsri Sentinel-2 Land Cover Explorer
Reference years2017 and 2025
Spatial resolution10 m
FormatGeoTIFF, categorical, 1 band, Byte
Coordinate systemWGS 84 / UTM Zone 43N (EPSG:32643)
Raster size8,850 × 11,841 pixels (both years)
NoData value0

2.2 Processing in QGIS

The tiles for each year were combined using GDAL’s Build Virtual Raster tool, clipped to the Kullu district boundary and exported as GeoTIFFs. Because both rasters have identical dimensions, origin, resolution and extent, they can be compared pixel by pixel. Each pixel covers 100 m², so area was calculated as: Area (km²) = pixel count × 0.0001

2.3 The transition trick

To see where each pixel started and where it ended up, I combined the two rasters in the QGIS Raster Calculator: ("2017 clipped kullu@1" * 100) + "2025 clipped kullu@1" Every pixel now carries a unique code for its pair of classes. The pixel count for each code came from the Raster Layer Unique Values Report.

CodeMeaning
202Trees remained trees
207Trees became built-up
507Crops became built-up
811Bare ground became rangeland
809Bare ground became snow/ice

This matters because net totals hide a lot. A class can show a small net change while large areas move in and out of it.

3. The Big Picture: 2017 vs 2025

Land-cover class2017 (km²)2025 (km²)Net change (km²)Relative change
Water10.9313.37+2.44+22.28%
Trees1,713.371,565.20−148.17−8.65%
Flooded vegetation0.040.06+0.02+37.11%
Crops26.6420.26−6.38−23.94%
Built-up92.17157.40+65.23+70.78%
Bare ground917.45610.06−307.39−33.51%
Snow and ice525.38773.78+248.40+47.28%
Clouds1.810.004−1.80−99.80%
Rangeland2,216.692,364.36+147.67+6.66%
Total5,504.485,504.480.00—

Percentage changes are relative to the 2017 area. Minor differences may arise from rounding.

[Image 2: Bar chart of net change by class, from snow/ice (+248 km²) down to bare ground (−307 km²)]

Share of the district

Class2017 share2025 share
Rangeland40.27%42.95%
Trees31.13%28.44%
Bare ground16.67%11.08%
Snow and ice9.54%14.06%
Built-up1.67%2.86%

[Image 3: Side-by-side land-cover maps of Kullu for 2017 and 2025, same legend and extent]

4. Built-up Land: Up About 70%

Built-up area rose from 92.17 km² to 157.40 km², a net gain of 65.23 km². That is the largest proportional increase of any class, and it fits Kullu’s geography, where settlement concentrates along river valleys, road corridors and accessible slopes. The more interesting question is what land became built-up:

Previous classConverted to built-up (km²)
Rangeland45.74
Trees14.03
Crops8.82
Bare ground1.93
Water0.51
Snow/ice0.001
Total gross inflow71.02

About 86.38 km² of the 2017 built-up area stayed built-up, and about 5.79 km² was reassigned to other classes. That is why a gross inflow of 71.02 km² produces a net gain of 65.23 km². The takeaway: growth did not come mainly from farmland. Rangeland contributed the most, and tree cover contributed more than cropland did. Settlement and infrastructure expansion appears to be spreading across several land types at once.

[Image 4: Bar chart of built-up sources by previous class]

[Image 5: Zoomed maps of new built-up pixels around Manali, Kullu–Bhuntar and Naggar]

5. Tree Cover: Down 148 km², but It Is Not That Simple

Tree cover fell from 1,713.37 km² to 1,565.20 km², a net loss of 148.17 km² (−8.65%). Its share of the district dropped from 31.13% to 28.44%.

Where the 2017 trees wentArea (km²)
Stayed trees1,466.32
Rangeland218.41
Built-up14.03
Bare ground6.50
Snow/ice5.44
Water1.67
Crops1.01

At the same time, about 98.88 km² of other classes became trees in 2025. The net loss is the balance between what was lost and what was gained. The trees-to-rangeland shift (218.41 km²) is one of the three largest transitions in the district. It could reflect real vegetation change, degraded or thinned forest, or the way the classifier treats sparse canopy at forest edges and near the treeline. This analysis can’t say which, and it needs checking against high-resolution imagery and field knowledge.

6. Bare Ground, Snow and Rangeland: The Big Shuffle

6.1 Bare ground fell by a third

Bare ground dropped from 917.45 km² to 610.06 km² (−33.51%), the biggest absolute decline in the district.

Bare ground in 2017 became…Area (km²)
Rangeland220.16
Snow/ice188.72
Built-up1.93
Trees1.38
Crops0.03

Flows went the other way too: 46.92 km² of rangeland and 46.34 km² of snow/ice became bare ground in 2025.

6.2 The snow and ice surprise

Mapped snow and ice grew from 525.38 km² to 773.78 km², a 47.28% rise. Its share of the district went from 9.54% to 14.06%.

Transition into snow/iceArea (km²)
Bare ground → Snow/ice188.72
Rangeland → Snow/ice107.25
Trees → Snow/ice5.44
Water → Snow/ice0.65
Crops → Snow/ice0.003
Total gross inflow302.63

Against that inflow, 46.34 km² of 2017 snow/ice became bare ground and 7.63 km² became rangeland. The net result is +248.40 km².

[Image 6: Chart of gross flows into and out of the snow/ice class]

A note of caution: this is not evidence that glaciers are growing. In high mountains, annual land-cover maps are sensitive to when the satellite scenes were taken, how much seasonal snow was on the ground, and how the classifier separates snow, rock and thin alpine vegetation.

6.3 Rangeland gained 148 km²

Rangeland is still the largest class, rising from 2,216.69 km² to 2,364.36 km² (+6.66%). About 457.79 km² of other classes became rangeland, while about 310.12 km² of 2017 rangeland moved elsewhere.

Previous classBecame rangeland (km²)
Bare ground220.16
Trees218.41
Snow/ice7.63
Crops7.59
Built-up2.11
Water0.69

A rise in mapped rangeland shouldn’t be read automatically as more grazing land. It means more pixels were classified that way, which needs independent verification.

6.4 Cropland: a small class, a clear decline

Mapped cropland fell from 26.64 km² to 20.26 km². Only 7.37 km² of 2017 cropland kept its class, and 8.82 km² became built-up. The study can’t say whether the rest was abandoned, shifted into horticulture or converted to something else. At 10 m resolution, small terraced fields and mixed orchards are also hard to map, so this class should be read with care.

7. The Full Transition Matrix

Rows are the 2017 class and columns are the 2025 class. All values are in km².

2017 ↓ / 2025 →WaterTreesFlooded veg.CropsBuilt-upBareSnow/iceCloudsRange
Water7.190.120.000.000.511.770.650.000.69
Trees1.671466.320.001.0114.036.505.440.00218.41
Flooded veg.0.000.000.000.000.000.000.010.000.03
Crops0.041.560.007.378.821.260.000.007.59
Built-up0.260.650.000.3086.382.470.000.002.11
Bare ground0.471.380.000.031.93504.78188.720.00220.16
Snow/ice0.070.190.000.000.0046.34471.150.007.63
Clouds0.000.050.000.000.000.020.560.001.17
Rangeland3.6794.920.0611.5645.7446.92107.250.001906.57

Three transitions stand out:

RankTransitionArea (km²)
1Bare ground → Rangeland220.16
2Trees → Rangeland218.41
3Bare ground → Snow/ice188.72

All three involve classes that are hard to separate in mountain terrain, where bare rock, thin alpine vegetation and snow patches blend into each other.

8. How Much of Kullu Actually Changed?

IndicatorArea (km²)Share
Same class in 2017 and 20254,449.7580.84%
Different class in 20251,054.7319.16%
Total valid area5,504.48100.00%

About four-fifths of the district kept the same mapped class. Trees and rangeland show the highest persistence, though both also exchanged large areas with other classes. This is persistence in mapped classification, not proof that nothing changed on the ground.

9. What This Study Can and Cannot Say

LimitationWhy it matters
Based on a ready-made annual classificationImage availability, season and classifier behaviour can all create apparent change.
No independent accuracy assessmentThe 2017 and 2025 maps need validation before transitions are treated as confirmed change.
Snow/ice class unverifiedThe increase is not evidence of glacier growth or more permanent snow cover.
No elevation-wise analysis yetIn a district running from valley floors to glaciated peaks, change is strongly tied to altitude and slope.
Causes not testedClimate, tourism, hydropower, roads and horticulture are plausible drivers but are not examined here.

Conclusion

Between 2017 and 2025, built-up land in Kullu grew by about 65 km², with gains coming from rangeland, forest and cropland alike. Tree cover fell by about 148 km² in net terms, while the large swings among bare ground, rangeland and snow/ice show how difficult mountain land cover is to map reliably.Net totals alone would have hidden most of this story. Pixel-level transition analysis shows where change came from, and it also shows why verification matters before any conclusions about climate or land use are drawn.

Data and tools

  • Esri. Sentinel-2 Land Cover Explorer, annual land-cover datasets for 2017 and 2025: livingatlas.arcgis.com/landcoverexplorer
  • QGIS Development Team. QGIS Geographic Information System: qgis.org
  • GDAL/OGR Contributors. Geospatial Data Abstraction Library: gdal.org

Editor’s Note: This article currently presents the preliminary GIS spatial data and raster transition matrix for Kullu District (2017–2025). It is a living document. A comprehensive update integrating on-the-ground field observations and socio-economic drivers (including the impact of the Atal Tunnel and highway expansion) is currently in progress and will be published here shortly.