Land Use in Selected Indian Cities
Downloads and Resources
-
AhmedabadZIP
The map series for each city is presented within a self-contained QGIS project, which contains the relevant layers corresponding to each year, and displays them in a structured, styled fashion. Sentinel-2 images contain six spectral bands: blue, green, red, near infrared, shortwave infrared 1, shortwave infrared 2. (Images also contain an alpha channel.)
-
BelgaumZIP
-
HindupurZIP
-
HyderabadZIP
-
JaipurZIP
-
KanpurZIP
-
MalegaonZIP
-
ParbhaniZIP
-
PuneZIP
-
SingrauliZIP
-
SitapurZIP
Description
The dataset comprises eleven series of land use/land cover (LULC) maps, each corresponding to one city in India—Ahmedabad, Belgaum, Hindupur, Hyderabad, Jalna, Kanpur, Parbhani, Pune, Singrauli, Sitapur, or Vijayawada—and containing one map per year for 2015–2018. Every map contains several layers: areal LULC (open space, nonresidential, or residential), roadways, a water mask, and contemporaneous satellite imagery.
The maps were generated by bespoke models created with machine learning. A distinct convolutional neural network (CNN) was trained for each city and LULC type (areal, roadways). Training data were constituted from Sentinel-2 imagery and LULC ground-truth from the Atlas of Urban Expansion project.
LULC information has emerged as a key input to decision-making for a host of actors, from national policymakers to urban planners to disaster relief organizations. Strikingly simple at its core—showing what lies where and when—LULC information has a wide and ever-expanding range of applications.
Cautions
These maps represent proof of concept for algorithmic classification of land use/land cover within cities, rather than a fully mature data product. High classification system sensitivity means that substantial noise is manifest. Any conclusions about changing land use based on map changes from year to year should be made with caution. Due to the nature of the training data, performance is strongest in the urban core, and weakest at the far periphery of the urban areas. In particular, road data for some cities are extremely noisy outside of the main built-up areas.
Citation
Kerins, P., E. Nilson, E. Mackres, T. Rashid, B. Guzder-Williams, and S. Brumby. 2020. “Spatial Characterization of Urban Land Use through Machine Learning.” Technical Note. Washington, DC: World Resources Institute.
Access & Use Information
License: Creative Commons Attribution 4.0 International License. Full license text available at Creative Commons Attribution 4.0
Metadata
Project: Urban Land Use
Page Last Updated: June 8, 2020
