Lesson 0.3Lesson 0.3 · Ground Rules — Thinking Spatially
Coordinate Systems, Projections & Datums
Why two correct maps refuse to line up
Your roads sit in the Arabian Sea. Nothing is broken. Your CRS is.
You load a plot boundary and a road layer that you know are from the same city — and they land a thousand kilometres apart, or one sits in the ocean. No file is corrupt. The layers simply describe their positions in two different coordinate reference systems, and until you tell the GIS which is which, it cannot reconcile them. The CRS is the most-skipped idea in GIS and the one that quietly ruins the most work. This lesson makes sure it never ruins yours.
North is a decision, not a fact — always check your CRS before you believe a distance.
Datum: the model of the Earth's shape
The Earth is not a sphere; it is a lumpy, slightly-squashed spheroid. A datum is an agreed mathematical model of that shape, plus where it is anchored, so that a latitude and longitude actually point to a real spot on the ground. The world's default is WGS 84 — the datum your phone's GPS uses.
Datums matter because the same latitude and longitude on two different datums can be tens or hundreds of metres apart on the ground. India's historical surveys used the Everest 1830 ellipsoid (the Kalianpur datums) — beautifully fitted to the subcontinent in its day, but not the same as WGS 84. An old Survey of India sheet and a fresh GPS track can disagree by a couple of hundred metres for exactly this reason. Mixing datums without a transform is a classic, invisible error.
Same lat/long, different datum, different ground point. Always ask: on which Earth?
Projection: flattening the globe (and the price you pay)
A projection is the recipe for squashing the curved Earth onto a flat plane — a map, a screen. And here is the iron law, proven by Gauss: you cannot flatten a sphere without distorting something. Every projection preserves some properties by sacrificing others. It can keep shape (conformal), or area (equal-area), or distance along certain lines — never all at once.
This is not academic. Web Mercator (the projection of Google Maps and most web maps) keeps shape but wildly inflates area toward the poles — which is why Greenland looks the size of Africa when it is a fraction of it. If you measure areas or draw distance buffers in a distorting projection, your numbers are simply wrong. Measure in a projection built for measuring, near your area of interest.
Geographic vs projected: degrees or metres
Put datum and projection together and you get two families of coordinate system.
A geographic CRS gives positions in degrees of latitude and longitude on the datum — e.g. EPSG:4326 (WGS 84). It is perfect for storing and sharing global position, but a degree is not a fixed distance (a degree of longitude shrinks toward the poles), so you should not measure lengths or areas in it.
A projected CRS gives positions in metres on a flat grid — e.g. a UTM zone. Now distance and area are honest, because the grid is in real metres. The working rule: store and share in geographic (4326); analyse and measure in a projected metre-based CRS suited to your region.
The right CRS for India
India spans several projected zones, and you must pick the one over your site:
• UTM (WGS 84): India falls across UTM zones 42N–46N — EPSG:32642 (zone 42N) through EPSG:32646 (zone 46N). Most of peninsular India sits in EPSG:32643 (43N) and EPSG:32644 (44N). Excellent, metre-accurate, and globally standard for local work.
• India NSF (National Spatial Framework): a Lambert Conformal Conic set on WGS 84 for seamless national mapping — EPSG:7755 is the whole-of-India national LCC, with per-state zones 7756–7764. Use it when you need one consistent CRS across state or national extents.
• Legacy Everest/Kalianpur: old Survey of India work used Kalianpur 1975 zones — EPSG:24378 (zone I), 24379 (IIa), 24380 (IIb). You will meet these when georeferencing historical sheets.
⚠️ One trap to memorise: EPSG:4240 “Indian 1975” is _not_ an India CRS — despite the name it is used for Thailand. For historic Indian survey data, reach for the Kalianpur 1975 / India zone codes, not 4240.
Peninsular India: think EPSG:32643/32644. National seamless: EPSG:7755. Never 4240.
On-the-fly reprojection: convenience, not a cure
Modern GIS shows layers with different CRSs together by reprojecting them on the fly for display — so your mismatched layers will often line up on screen anyway. This is a convenience and a trap. The data on disk is still in its own CRS, and any measurement or analysis uses the project's CRS. If your project CRS is geographic (degrees), your buffers and areas will be wrong even though everything looks fine.
The professional habit: set a sensible project CRS (a metre-based projected system over your area), know the CRS of every layer you load (check the .prj), and reproject layers permanently to a common metre-based CRS before serious analysis. Looks-right is not is-right.
EPSG:4326
WGS 84 geographic (degrees) — the storage/sharing default
GPS/web default. Store and exchange in this; do NOT measure area/length in it.
EPSG:3857
Web Mercator (metres) — web base maps
Shape-preserving but badly area-distorting; fine for display, wrong for area analysis.
EPSG:32643 / 32644
WGS 84 / UTM zones 43N & 44N — most of India
Metre-based projected CRS for local measurement across peninsular India.
EPSG:7755
WGS 84 / India NSF LCC (national)
Seamless national metre grid (Lambert Conformal Conic); per-state zones 7756–7764.
EPSG:24378–24380
Kalianpur 1975 / India zones I, IIa, IIb (Everest 1830)
Legacy Survey of India datum; needed when georeferencing historic sheets. (NOT EPSG:4240.)
Workshop — break it, then fix it
The fastest way to understand CRS is to reproduce the classic bug and repair it. You will make two layers refuse to line up, then set the right CRS and measure an honest area.
QGIS 3.44 or ArcGIS Pro; any Indian polygon layer; an internet base map/OSM layer.
Goal: reproject a layer to the correct India UTM zone and measure a plot's true area in m² Data: any Indian plot/ward polygon (or draw one) + an OSM road layer Time: ~35 minutes
- 1Load an Indian polygon layer and note its CRS. In QGIS: bottom-right shows the project CRS; layer CRS is in Layer Properties → Information. In ArcGIS Pro: Map Properties → Coordinate Systems; layer Source → Spatial Reference.
- 2Set the project CRS to EPSG:4326 and use the measure tool to read the polygon's area. Note the figure.
- 3Now reproject the layer to the correct UTM zone (EPSG:32643 or 32644 for peninsular India). QGIS: right-click → Export → Save Features As → choose the CRS. ArcGIS Pro: Project (Data Management) tool.
- 4Set the project CRS to that UTM zone and measure the area again. Compare — the degree-based figure is meaningless; the metre-based one is real.
- 5Add an OSM road layer in a different CRS and confirm on-the-fly reprojection still lines them up on screen — proving that “looks right” and “measures right” are different things.
You’ll walk away with
One correctly-reprojected layer, a true area in square metres, and permanent muscle memory that the project CRS must be metre-based before any measurement.
Three altitudes on the same idea
Read the band that fits you — or all three.
Do your site measurement in metres, not degrees. Set your project to the local UTM zone (e.g. EPSG:32643 for much of western/central India) before you compute plot area, setbacks or distances, or your numbers will be subtly wrong. When a surveyor hands you a total-station file, ask which datum and zone it is in before you overlay anything.
Consistency across the plan area is everything. For a city or region that crosses a UTM zone boundary, the India NSF LCC (EPSG:7755) gives one seamless metre grid. Statutory area figures — land-use totals, FAR yields — must be computed in a projected CRS; a plan whose areas were measured in EPSG:4326 is quoting fiction.
Walkable-distance work is meaningless in degrees. A 400 m catchment, a 15-minute isochrone, a block perimeter — all must be measured in a metre-based CRS. Fix the project CRS first; then your “five-minute walk” is a real 400 m, not a distorted guess that changes with latitude.
“The layers line up on screen, so the coordinate system is fine.”
Do it yourself
Fix the CRS in your head before you touch the software.
- 1Your site is in Bengaluru. Which UTM (WGS 84) EPSG code should you measure in — 32643 or 32644? (Bengaluru is near 77.6°E.)
- 2Explain in one sentence why you should not compute plot area in EPSG:4326.
- 3A colleague reaches for EPSG:4240 for a historic Indian sheet. What do you tell them?
- 4Name the datum your phone's GPS reports positions in.
The rule that saves your data
Peer-reviewed journals & authoritative standards
- 01Longley, P.A., Goodchild, M.F., Maguire, D.J. & Rhind, D.W. — Geographic Information Science and Systems, 4th ed. — Wiley, 2015.
- 02de Smith, M.J., Goodchild, M.F. & Longley, P.A. — Geospatial Analysis: A Comprehensive Guide, 7th ed. — Winchelsea Press, 2025.
- 03Survey of India — Dept. of Science & Technology, Govt. of India, ongoing.
Now that every point knows where it truly is, we need to know how it is packaged on disk — shapefile, GeoJSON, GeoPackage, GeoTIFF — and why the wrong container costs you a `.prj` file and a week. Formats are next.
The author
Amogh N P
Architect, interior designer, and creative polymath. Studio Matrx began in his notebooks — his vision of design made honest, useful, and open to everyone. Its Academy is written and taught in his memory, and free, forever.
More about Amogh →