What this is
On 31 July 2024 the Ministry of Environment, Forest and Climate Change published draft notification S.O. 3060(E) (PDF, 4.9 MB, 279 pages), listing 4,402 villages across Goa, Gujarat, Karnataka, Kerala, Maharashtra and Tamil Nadu proposed as an Ecologically Sensitive Area in the Western Ghats. The notification names those villages in a table. It does not publish their boundaries.
This atlas reconstructs them. Every village named in Annexure C was matched to its polygon in the all-India census village boundary dataset, giving 4,331 of 4,402 rows (98.4%) a mapped extent carrying its 2011 Census code and LGD code.
Reading the gazette
The village list and the boundary coordinates are read directly from the text layer of the notification published by MoEFCC, so every name and every coordinate is taken exactly as printed rather than recognised from an image.
Download the notification — PDF, 4.9 MB ↓
279 pages, hosted by MoEFCC at
moef.gov.in/storage/tender/1723024218.pdf. Everything in this atlas is derived
from that file.
Two checks confirm nothing was dropped in the reading. All six state serial runs reconcile exactly and contiguously — Goa 108, Gujarat 64, Karnataka 1,449, Kerala 131, Maharashtra 2,515, Tamil Nadu 135, giving 4,402 rows, which is the count the notification itself states. Annexure B yields 401 boundary points, and all 402 coordinate pairs printed in it are accounted for: the single one excluded is a typo in the gazette, Karnataka point 45, which reads 71.2007, 31.6736 — a location in Punjab, some 2,000 km from the Western Ghats.
Matching names to polygons
Village names were matched to census records hierarchically — district and taluka first, then district, then state-wide — under a strict 1:1 constraint, so no polygon is ever assigned to two gazette rows.
Three things made the difference over a naive fuzzy match:
- Each census village offers three romanised spellings
(
name,censusname,lgd_villagename) as match aliases. - Both sides are additionally compared under a transliteration folding
(
oo→u,ee→i, doubled consonants collapsed, trailing vowels dropped). Without it, Marayur against Marayoor scores only 80 and is rejected; with it, the pair is effectively identical. - Thresholds scale with how much the search was narrowed. District-plus-taluka already cuts the pool to a handful of villages — Devikulam taluka has twelve — so it runs a looser threshold (72) than a district-wide (85) or state-wide (93) search.
A name can match perfectly in the wrong place. The gazette's Khadi in Surat scored a perfect 100 against Khada in Junagadh — 295 km away in Saurashtra, nowhere near the Western Ghats. Raigad's Wandoli matched a village in Gadchiroli, 631 km across Maharashtra.
So a geographic gate is applied: villages matched at district-and-taluka scope define each state's ESA core, and any wider-scope match landing more than 50 km outside that core is rejected and retried against its remaining candidates. Thirty outliers were caught; twenty-three found a correct nearer village instead. No match now sits beyond 50 km of its state core.
Three independent verifications
A village-name match can be confidently wrong, so the result is checked three ways that do not share a failure mode.
1 · Against the gazette's own coordinates
Annexure B lists 401 "prominent points on the outer boundary" of the ESA. These are boundary points, so the correct test is how far each lies from the edge of the reconstructed polygon — not whether it falls inside. A point sitting exactly on a boundary is a coin-flip for a point-in-polygon test, and roughly half of them do fall inside, which is the expected signature.
| State | Points | Median distance | Within 2 km | Within 5 km |
|---|---|---|---|---|
| Goa | 39 | 101 m | 39 | 39 |
| Gujarat | 61 | 244 m | 60 | 61 |
| Tamil Nadu | 109 | 404 m | 94 | 101 |
| Maharashtra | 135 | 495 m | 127 | 135 |
| Karnataka | 57 | 536 m | 48 | 56 |
Goa's median of 101 m is finer than the census village boundaries' own positional accuracy, so the reconstructed edge is effectively the edge the gazette describes.
2 · Against Kerala's own published ESA
Kerala released its ESA as KML — an independent government source for the same ground. The direction of this test matters, so stated precisely:
- 98.1% of the official Kerala area falls inside these village polygons. Only 170 km² of it lands outside — and that residue is spread across 833 fragments of which 820 are smaller than 1 km², i.e. sliver mismatches along shared edges rather than any missing region. 67 of its 98 polygons are more than 99% covered, 86 are more than 95% covered.
- 66.7% of these village polygons fall inside the official area. The other third is the part of each village that was not notified.
Both numbers are correct and they measure opposite things. The first says the official ESA is almost entirely contained by this reconstruction — which is the test that matters. The second is low by design, because Kerala notified portions of villages while this atlas holds whole villages; a whole village is necessarily bigger than the slice of it that was listed.
On the map this looks counter-intuitive at first. Switch on the Kerala official layer and the blue outline sits inside the purple village polygons, with purple spilling beyond it — that spill is the 33%, not a mismatch. If the blue appeared to extend past the purple, that would be the failure. It does not.
3 · Against the notified areas
Annexure A states an area per state, giving a third check that is independent of both geometry sources.
| State | Reconstructed km² | Notified km² | Ratio |
|---|---|---|---|
| Maharashtra | 17,117 | 17,340 | 0.99 |
| Goa | 1,436 | 1,461 | 0.98 |
| Gujarat | 464 | 449 | 1.03 |
| Tamil Nadu | 6,422 | 6,914 | 0.93 |
| Karnataka | 17,785 | 20,668 | 0.86 |
| Kerala | 13,030 | 9,994 | 1.30 |
| Total | 56,254 | 56,826 | 0.99 |
The three checks improved together across successive revisions of the method, which is the reassuring part: a matching rule loose enough to invent false pairings would push polygons to wrong places and degrade checks 1 and 2 even while the match count rose.
Where it is weakest
Bifurcated villages are drawn whole. The gazette marks some villages with an asterisk, meaning only part of the village falls inside the ESA. No partial extent is published anywhere, so those polygons show the entire village and therefore over-state the ESA. Only the State Governments' physical-verification maps can resolve them. The atlas flags every such village when you open it.
Kerala is portion-wise by design. This is why Kerala's ratio is 1.30 rather than 1.00 — it is not an error in the matching. For Kerala, treat the official state KML layer as authoritative and these village polygons as context. You can overlay the official layer directly in the atlas.
Karnataka is the least complete state, at 0.86 of its notified area, with 19 rows unmapped. Its census transliterations diverge from the gazette's more than any other state's.
The 71 unmapped rows are mostly structural. They are composite cells
naming two features at once (Amravathi R.F./Anamalai R.F.), gazette rows repeated
for different survey blocks (Valparai (TP) appears four times, and the 1:1
constraint gives the polygon to the first occurrence), placeholder text
(Area Under Forest), and small Raigad hamlets below census village granularity.
The full list, with a reason for each, is in the atlas under "Not mapped".
This is not an official record. It is an independent reconstruction for reference and research, with no legal standing. It must not be used to determine whether a specific parcel of land falls inside the ESA. For that, consult the notification itself and the relevant State Government.
Why there are no Zone 1/2/3
A common expectation is that the Western Ghats ESA is graded into Zone 1, Zone 2 and Zone 3. It is not — and the distinction matters if you are comparing this atlas against older maps or reports.
The graded scheme comes from the Western Ghats Ecology Expert Panel (the Gadgil report, 2011), which proposed dividing the range into ESZ 1, ESZ 2 and ESZ 3 with progressively lighter restrictions. That scheme was never notified.
What followed was the High Level Working Group (the Kasturirangan report, submitted 15 April 2013), which the 2024 draft notification is built on. It took a different approach: rather than grading the whole range, it identified roughly 37% of the Western Ghats — 59,940 km² of natural landscape — as a single continuous ecologically sensitive band, and left everything outside it unclassified.
So S.O. 3060(E) defines one undivided ESA. Searching the full
279-page notification for zone terminology returns nothing: the only occurrences of "ESZ" are
an email address (esz-mef@nic.in) and the file number
F.No.1/04/2012-ESZ-RE. The only "categories" it names are categories of
projects and activities that are prohibited or regulated, and the CPCB's
Red/Orange/White industry categories — neither of which is spatial.
Consequently the categories in this atlas are not statutory zones. They record how each village was mapped and how much to trust its extent — nothing about differing levels of legal protection. Every village shown is subject to the same single set of ESA restrictions.
What the categories mean
Every village in the atlas carries one of five categories. They are not levels of legal protection — as above, the ESA is single and undivided, and every village shown is subject to the same restrictions. What the category tells you is how that polygon was arrived at, and how much to trust its extent. Two of the five are warnings.
Whole village — 4,131 villages, 40,850 km²
The ordinary case, and about 73% of the mapped area. The gazette names the village without qualification, so the entire revenue village falls inside the ESA and the polygon is simply that village's census boundary. Nothing is being inferred beyond the name match itself, so the extent is as good as the census boundary is.
Bifurcated — 26 villages, 111 km² (treat as an over-estimate)
The gazette marks some village names with an asterisk. That marker means only part of the village lies inside the ESA — the boundary cuts through it.
The notification does not publish where that cut falls. No annexure, no coordinates, no sub-village list. So the honest options were to drop these villages entirely or to draw them whole and label them clearly. This atlas draws them whole, which means every one of these polygons is larger than the ESA actually is. The map flags it when you open one.
The scale is small — these 26 cover 111 km², well under 1% of the mapped area — so the effect on any regional total is minor. For a specific village it is not minor at all. Only the State Governments' physical-verification maps can resolve the true extents.
All 26 are in Goa (9) and Maharashtra (17). The gazette marks 54 villages this way in total; the remaining 27 are mostly small Raigad hamlets that have no census village polygon at all, and appear under Not mapped instead.
Kerala village — 122 villages, 12,360 km² (use the official layer)
Every Kerala row carries this category, because Kerala is a special case. The State defines its ESA portion-wise: not "this village is in", but "this part of this village is in". The gazette lists the village names; Kerala publishes the actual portions separately as KML.
So these 122 polygons are whole villages standing in for what is really a set of partial ones. That is the entire reason Kerala's area ratio is 1.30 rather than 1.00 — it is not a matching error, it is the difference between two definitions.
Kerala's official layer is included in the atlas. Switch on Kerala official ESA under Reference layers to see the real boundary sitting inside these village polygons, and treat that layer as authoritative for Kerala.
Notified forest — 51 areas, 2,263 km²
Not revenue villages at all. These are Reserve Forests and forest blocks that the gazette
lists in the same table as villages — entries like Thadagam R.F.,
Anaikatti North R.F. and Jaganarai Slopes R.F. They are identified by the
R.F. suffix the notification uses.
All 51 are in Tamil Nadu, which lists forest blocks far more than the other five states do. They are mappable here because the census boundary dataset carries Reserve Forests as features in their own right. They range from 1.5 km² to 365 km², so a few of them are large.
Kerala, bifurcated — 1 village, 670 km²
A single village, Chittar, that is both a Kerala row and asterisk-marked in the gazette — so both caveats apply at once. It is worth its own category because it is large: at 670 km² it is by far the biggest polygon carrying a partial-extent warning, and it alone accounts for most of the 781 km² of asterisk-marked area in the dataset.
In the data files these appear in the esa_category field as
whole_village, bifurcated_village, kerala_village,
notified_forest and kerala_bifurcated_village. The separate boolean
bifurcated field is the quickest way to filter out every partial-extent polygon
in one step.
The analysis layer
Everything above is a transcription of a legal document. Everything in this section is derived — the ESA polygons joined to census, land-cover, protection and hydrology data. It is a weaker kind of claim, it inherits the errors of every layer it touches, and it is kept separate for that reason.
How each figure is computed
Geometry and units
Every area is measured in EPSG:7755 (India-wide equal-area), never in degrees. Overlays are polygon intersections against the village boundary, reported both in km² and as a share of that village.
One trap dominated this stage. Several source layers contain self-overlapping polygons — Recorded Forest Area reached 154% of a single village, wildlife corridors 173%. Summing per-feature intersections double-counts wherever a layer overlaps itself. Each layer is therefore unioned into one geometry before measurement, and any result still exceeding the village area is logged as a fault rather than quietly clipped.
The same logic applies across layers: "protected" is the union of national parks, sanctuaries and Parivesh ESZs, not the larger of the two. Taking the maximum understated protection by about 2,000 km², because ESZs sit largely outside park boundaries.
Population and census indicators
Joined on the 2011 Census village code that the reconstruction already carries — 4,330 of 4,331 villages match. This is an exact key join, with no fuzzy matching at this stage.
A few villages straddle a sub-district boundary, and the Census reports each part as a separate row under the same code. Those parts are summed, not de-duplicated: dropping one would silently halve that village's population. Seventeen such cases exist nationally, all in West Bengal, so none affect this dataset — but the pipeline handles them correctly regardless.
Two census directories, not one. The Census publishes rural villages and towns as separate Primary Census Abstracts. Joining only the village directory left every Town Panchayat, Municipal Council and Census Town without a population — which published Valparai, a hill town of 70,859 people, as uninhabited. Both directories are now joined: 4,330 of 4,331 records match, and 33 entries are urban local bodies counted from the town directory. Including them raised the population total by 537,562.
Two traps in that data worth knowing. Census Towns keep village-range codes, so the code number cannot tell a town from a village — only the directory it came from can. And settlements straddling a sub-district boundary are reported as several "(Part)" rows under one code; those are summed, since dropping one would halve the settlement.
"Uninhabited" means a measured zero, never a missing figure. An earlier version of this atlas treated any record without a population as uninhabited, which is how a town of tens of thousands came to be published as having no residents. The two are now distinct: 125 villages are recorded uninhabited by the Census, and exactly one record — Valvane, in Satara — has no population because it carries no census code in the boundary layer at all.
Derived rates: Scheduled Tribe share and literacy are shares of village population; land-dependent workers is (cultivators + agricultural labourers) ÷ total workers; density is population ÷ village area.
Land cover
Categorical zonal statistics over two NRSC rasters at 56 m, counted in the rasters' own Lambert Conformal Conic projection so no pixels are resampled. Each class becomes a percentage of the village.
The Wasteland Atlas is a sparse raster — it holds data only where wasteland exists. Its class counts therefore had to be expressed against village area taken from the land-cover raster, which was verified to cover the villages exactly (ratio 1.000). Left uncorrected, "degraded forest" would have read as a share of wasteland rather than of the village, inflating it roughly fivefold.
Where percentages are aggregated across villages — in the selection panel, or a state total — they are area-weighted. A plain average would let a 0.5 km² village count as much as a 300 km² one.
Highways
Length of the national highway network falling inside each village, in kilometres. Where it feeds a score it is converted to density (km per km²), since raw length simply favours large villages.
Land outside Recorded Forest Area
Village area minus the area under Recorded Forest Area. This is deliberately not called private land — see the caveats below.
The composite scores
Two scores, each the mean of percentile-ranked components. Percentile ranking makes them relative to this set of villages rather than absolute: a score of 80 means "higher than 80% of ESA villages", not "80% of some ideal".
| Conservation value | Human pressure |
|---|---|
| Natural forest % | Population density |
| Recorded Forest Area % | Built-up % |
| — | Highway density (km/km²) |
| — | Land-dependent workers % |
Missing values are not imputed. A village is scored on whatever components
it has, the count is published as conservation_n and conflict_n, and
anything with fewer than two components is left unscored (no village is currently in that state, so the guard has never excluded anyone). Bands are terciles, not a midpoint
cut — a 50/50 split would fall exactly where villages are densest.
Both scores were rebuilt after an audit, and it is worth saying why.
The first version put population density on both axes — inverted on one — and set Recorded Forest Area against its own arithmetic complement. Two of four and two of five components were exact mirrors, which forced 86% of villages onto a single diagonal. The quadrants were an artefact of the formula, not a finding about the Western Ghats.
Removing those mirrors still left a correlation of −0.601. The remaining cause was subtler: land-cover classes sum to 100%, so natural forest and agriculture correlate −0.80 almost by construction. Using forest on one axis and agriculture on the other was one compositional variable counted twice with opposite signs. Pressure now measures people and infrastructure only, and the two scores correlate −0.49 — related, as real conservation value and real pressure should be, but no longer one axis wearing two hats.
Plantation share is still published and can be mapped. It is simply not scored.
Checking the analysis against ground truth
Every derived figure on this site rests on one approximation: villages are drawn whole, even where only part of them was notified. Kerala is the only state that published its own ESA boundary, so it is the one place that approximation can be measured rather than merely disclosed. The same metrics, computed on both geometries:
| Metric | Whole villages | Kerala’s official boundary | Difference |
|---|---|---|---|
| Area | 13,030 km² | 8,863 km² | −4,167 km² |
| Protected / ESZ | 20.3% | 29.1% | +8.8 |
| Corridor outside protection | 5.5% | 6.5% | +1.0 |
| Recorded Forest Area | 75.0% | 92.0% | +17.0 |
| Outside Recorded Forest | 25.0% | 8.0% | −17.0 |
| Natural forest | 77.4% | 83.7% | +6.3 |
| Plantation | 10.6% | 4.0% | −6.6 |
| Built-up | 0.47% | 0.07% | −0.40 |
| National highway | 465 km | 137 km | −328 km |
Every metric moves in the same direction, and that direction is interpretable: the notified portion is the wilder, higher, more protected part of each village, while the part left out is where the estates, roads and settlement are. Plantation more than halves, highway length drops by two thirds, and the share outside Recorded Forest Area falls from a quarter to a twelfth.
What this implies for the other five states. Where no official boundary exists, the whole-village figures on this site almost certainly lean the same way: they overstate development pressure — plantation, built-up, roads, land outside recorded forest — and understate how protected and forested the notified area really is. Kerala suggests the size of that bias, though it is the most portion-wise state and so likely an upper bound on it.
Population is deliberately absent from this comparison. Census counts are per whole village, and apportioning them to a village portion needs an assumption — uniform density, or a dasymetric weighting — that there is no ground truth to validate. An unvalidatable estimate would be worse than none.
Two caveats on the official layer itself: it covers 98 villages where the
gazette lists 131 Kerala rows, so it is not a complete substitute; and at 8,863 km² it is
89% of the 9,993.7 km² Annexure A states for Kerala. Served at
api/v1/kerala-comparison.json.
What every card means
Each figure in the selection panel and on the village card is defined below — what it is, how it is computed, and what it cannot be used for. The same text appears behind the i button on each card, generated from the same file so the two cannot disagree.
Extent
| Card | What it is | How it is computed | Watch out for |
|---|---|---|---|
| Villages count | How many gazette rows are in the current selection. | One row per village named in Annexure C of the notification that was matched to a boundary. 4,331 of 4,402 rows matched overall. | A few gazette rows name the same place more than once for different survey blocks; the 1:1 constraint gives the polygon to the first, so those repeats appear under Not mapped. |
| Total area km² | Ground area of the selected village polygons. | Measured in EPSG:7755, an equal-area projection for India, never in degrees. | Bifurcated and Kerala villages are drawn whole, so this is larger than the area actually notified. Kerala measured against its own official boundary is 8,863 km² rather than 13,030. |
People
| Card | What it is | How it is computed | Watch out for |
|---|---|---|---|
| Population people | Resident population of the selected settlements at the 2011 Census. | Joined on the 2011 census code from both Primary Census Abstracts — the rural village directory and the town/ward directory. 4,330 of 4,331 records match. | This is whole-village population. Where only part of a village was notified, everyone in the village is counted, so treat it as an upper bound. One record, Valvane in Satara, has no census code and no figure. |
| Households households | Number of households recorded by the 2011 Census. | Census field No_HH, summed across the selection. | Same whole-village basis as population. |
| Scheduled Tribe people | Residents recorded as Scheduled Tribe. | Census field P_ST. The percentage is of the selection's total population. | Relevant to Forest Rights Act claims, but the whole-village basis applies here too. |
| Density people / km² | How densely settled the selection is. | Total population divided by total area. | Because the area is whole-village and includes uninhabited forest, density is lower than the lived experience of the settled parts. |
| Land-dependent workers % of workers | Share of working people whose livelihood is directly tied to land. | Cultivators plus agricultural labourers, divided by total workers. Census fields MAIN_CL_P, MAIN_AL_P, TOT_WORK_P. | A measure of who stands to be affected by land-use restrictions, not of poverty. |
Protection & tenure
| Card | What it is | How it is computed | Watch out for |
|---|---|---|---|
| Park, sanctuary or ESZ km² | How much of the selection already sits inside a National Park, Wildlife Sanctuary, or an Eco-Sensitive Zone notified around one. | Union of the India-WRIS national protected-area layer and the Parivesh Eco-Sensitive Zone layer, intersected with the villages. Unioned first, because summing per-feature intersections would double-count overlap. | Replaces the OpenStreetMap-derived layer used until v3.10, which carried unquantified error. Two things to know about the new source. India-WRIS (Ministry of Jal Shakti) REPUBLISHES protected-area boundaries; it is not the notifying authority, which is MoEFCC, so this is an official portal’s copy rather than a notified boundary. And its latest year of notification is 2021, so any park, sanctuary or ESZ notified between 2022 and the 2024 draft gazette is missing. The switch moved protection inside the ESA from 15,759 km² (28.0%) to 15,003 km² (26.7%); earlier downloads will not reconcile with current ones. The WII protected-area shapefile was also compared and rejected: it is a mid-2010s snapshot missing Kodaikanal, Megamalai and Bhimgad, and self-overlaps by 3.5%. |
| Corridor outside protection km² | Corridor land that is NOT already inside a protected area — the ground that connects protected areas rather than forming part of one. | The Parivesh wildlife-corridor layer with all protected area (parks, sanctuaries and Eco-Sensitive Zones) subtracted. Protection is removed rather than overlaid, so this figure and the protection figure are disjoint and may be added. | The subtraction removes 53% of the raw corridor layer as already protected. It also repairs a defect in the source: that layer’s type field files national parks and sanctuaries as "corridor" — Bandipur, Periyar and Parambikulam among them — and 79% of its "tiger reserve" area also sat inside its "corridor" area. Read this as POLICY OVERLAP, not as measured ecological connectivity: it records whether an identified corridor crosses the village, not how much movement the village actually carries. A sole pinch-point and the ragged edge of a wide corridor score the same per unit area; nothing here is least-cost or circuit-modelled. It is zero for 78% of villages, and for that reason it is NOT a component of the conservation score — it is published, mapped and filterable only. |
| Tiger reserve km² | Village area inside a tiger reserve. | Those India-WRIS protected areas whose sub-type field marks them as part of a tiger reserve, intersected with the villages. | This is the protected-area CORE of each reserve, not the full notified reserve. A tiger reserve is core plus buffer, and buffer zones are frequently not themselves notified parks or sanctuaries, so they are absent here — against NTCA’s notified totals these polygons run roughly half. It is a subset of the protection figure, never an addition to it: the source layer is de-overlapped, so every tiger-reserve polygon is also a protected-area polygon. Cross-checked against the WII tiger-reserve layer, which agrees to within 1% inside the ESA (5,471 km² against 5,409 km²). The WII layer used for that cross-check is itself a core-only tiger-reserve layer, so the agreement confirms the core extent, not the full notified reserve. |
| Recorded Forest Area km² | Land already recorded as forest, and so already under forest administration. | The Bharatmaps Recorded Forest Area layer intersected with the villages, unioned first. | Where the ESA overlaps recorded forest, the notification changes relatively little — the state already controls land use there. |
| Outside Recorded Forest km² | The remainder: village area not recorded as forest. | Total area minus Recorded Forest Area. | This is NOT private land. It also holds revenue and common land, water bodies, PSU and defence land, and gaps in the forest layer itself — 346 villages record no forest area despite being more than half forest. Do not quote it as a private-ownership figure. |
| Neither recorded forest nor protected km² | Village area that is outside Recorded Forest Area AND outside every park, sanctuary and Eco-Sensitive Zone — the ground where the ESA notification would be the only instrument in force. | Total area minus the area of the UNION of the recorded-forest and protected layers. It must be measured on the union: most Western Ghats parks are also recorded forest, so subtracting the two figures from each other in turn removes the same ground twice. Until v3.11 it was computed that way and read 18,134 km²; measured correctly it is 20,153 km². | This is NOT private land, and the name is not a description of ownership. It includes water bodies and reservoirs, revenue and common land, PSU and defence land, and every gap in the recorded-forest layer — 346 villages record no forest area despite being more than half forest. A water-body deduction that used to be applied was removed in v3.11 because it double-subtracted; reservoirs are therefore inside this figure. Do not quote it as a private-ownership or an available-land figure. |
Land cover
| Card | What it is | How it is computed | Watch out for |
|---|---|---|---|
| Natural forest % of area | Share under evergreen, semi-evergreen or deciduous forest. | Categorical zonal statistics over the NRSC LULC 250K 2017-18 raster at 56 m, counted in the raster's own projection so no pixels are resampled. Aggregates are area-weighted. | Mapped at 1:250,000 with 112–168 m registration error. For a village under about 25 km² that band is comparable to its whole area, so those figures are indicative only. |
| Plantation % of area | Agricultural plantation — coffee, cardamom, rubber, tea. | LULC class 7, distinct from the forest classes. | This is the estate-versus-forest distinction that drives most local opposition. It is published and mappable but deliberately excluded from the composite scores. |
| Agriculture % of area | Cropland and fallow. | LULC kharif, rabi, zaid, double/triple crop and current fallow classes combined. | Same 1:250,000 scale limit. |
| Degraded forest % of area | Forest land classed as degraded. | NRSC Wasteland Atlas 2015-16, class 6. The wasteland raster is sparse, so shares are expressed against village area from the land-cover raster, not against the wasteland total. | A different epoch from the land-cover layer (2015-16 against 2017-18). |
| Built-up % of area | Settlement and other built surface. | LULC class 1. | At 56 m with 1:250,000 mapping, small hamlets fall below what can be resolved. |
| Reliable at this scale villages | How many villages in the selection are large enough for land cover to be dependable. | Villages of 25 km² or more. Only 416 of 4,331 clear that bar, though they hold 52% of the area. | Below that threshold the satellite registration error is comparable to the village itself. Composition is still shown, but treat it as indicative. |
Pressure & scores
| Card | What it is | How it is computed | Watch out for |
|---|---|---|---|
| National highway km | Length of national highway running inside the selection. | MoRTH highway network from GatiShakti, clipped to the village polygons. | National highways only — state highways and district roads are not included, so this understates road presence. |
| Conservation value of 100 | How much forest a village holds and how much of it is already recorded forest. Read the caveat before calling it ecological value. | Mean of two percentile-ranked components: natural forest % and Recorded Forest Area %. | A percentile rank against the other ESA villages, not an absolute measure. Understand what it does and does not contain. Both components describe forest: natural forest % is measured from imagery, Recorded Forest Area % is tenure status, so this is closer to "forest cover and forest tenure" than to a biodiversity score. It contains no measure of endemism, irreplaceability or spatial configuration, which is what systematic conservation prioritisation normally rests on. A band of "low priority" therefore means low forest cover and non-forest tenure — not low ecological value. Note also that natural forest % counts evergreen and deciduous forest only, so montane grassland (170 km² of the ESA), scrub forest (699 km²) and littoral swamp score zero on it, and the shola–grassland mosaic and Myristica swamps are under-valued despite being among the most endemic-rich habitats in the range; the broader forest_pct field is published for anyone who wants them included. A corridor component was removed in v3.11 after an audit — it was zero for 78% of villages, so percentile-ranking turned it into a flag, and being the complement of protection it pushed protected villages down while recorded forest pushed them up. Of the two survivors, natural forest % is a clean gradient (1.8% zeros, 3,347 distinct values) and Recorded Forest Area % carries a 17.5% tie block at zero. |
| Human pressure of 100 | Where restrictions will be felt most. | Mean of four percentile-ranked components: population density, built-up %, highway density per km², and land-dependent workers %. | Deliberately measures people and infrastructure only. An earlier version shared variables with the conservation score, which forced 86% of villages onto one diagonal; the two now correlate −0.48. |
| Priority bands villages | Where each village falls on the two axes together. | Both scores cut into terciles, giving nine cells, each separately named: protect, protect with some pressure, negotiate, secondary, moderate on both, pressure-led, low priority, contested. | Terciles are relative to this dataset. **The bands changed in v3.11**: removing the corridor component from the conservation score moved 770 of 4,331 villages (17.8%) across a band boundary, so a band cited from an earlier download may not match the current one. An earlier version named only three cells and dumped 59% of villages into an uninformative "mixed". |
On the village card
| Card | What it is | How it is computed | Watch out for |
|---|---|---|---|
| Literacy % | Literate residents as a share of population. | Census field P_LIT divided by total population. | Census definition of literacy, age 7 and above. |
| Gazette entry | How the village appears in the notification itself, before any processing. | State, district, taluka, village name and serial number transcribed from Annexure C, plus the verbatim printed form including the * bifurcation marker. | The gazette's district and taluka names are sometimes older than the census ones — Surat and Tapi, for instance. |
| Name match score 0–100 | How confidently the gazette name was matched to a census village. | RapidFuzz similarity after transliteration folding, with the scope showing how wide the search had to go: district+taluka, district, or state-wide. | A low score, or a state-wide scope, means the pairing deserves checking against the gazette. |
| Setting | Which forest and hydrological units the village sits in. | Largest-overlap assignment against FSI forest divisions and ranges, and India-WRIS river basins and sub-basins. | A location label only. Nothing here measures water yield or forest administration workload. |
Where the analysis is weakest
Whole-village inflation runs through every derived number. Bifurcated and all Kerala villages are whole-village polygons standing in for partial extents. Population is therefore an upper bound — and 39% of the population total is Kerala, where extent is over-stated by roughly 30%. Every percentage-of-area figure moves the other way: 27% protected and 60% recorded forest are under-estimates, because the denominator is inflated while the numerators are real.
Land outside Recorded Forest Area is not private land. It also holds
revenue and common land, water bodies, PSU and defence land, and parks notified on
non-forest land — plus any gap in the forest layer itself. 346 villages record no
forest area despite being more than half forest; those are layer gaps, not tenure
findings, and carry an rfa_gap_suspect flag in the data. Do not quote this as a
private-ownership figure without a cadastral source.
Land cover is unreliable for small villages. NRSC land cover is mapped at
1:250,000 from AWiFS imagery, with registration accuracy of one pixel in plains and two to
three in hills — roughly 112–168 m. For a village under about 25 km² that
uncertainty band is comparable to its entire area. Only 416 of 4,331 villages
clear that threshold, though they hold 52% of the area. The rest carry
lulc_reliable = false and are marked indicative in the interface.
Protected areas come from an official portal, but read the two limits. Until v3.10 this layer was OpenStreetMap-derived via EJAtlas and carried unquantified error. It is now the India-WRIS national protected-area layer, which records each area's type, sub-type, state and year of notification. Inside the ESA it gives 12,968 km² of parks and sanctuaries against the 13,668 km² OpenStreetMap gave, so protection including Eco-Sensitive Zones is now 15,003 km², 26.7% of the ESA, where earlier editions of this page said 28%. Anyone holding an older download will not reconcile with a current one.
First limit: India-WRIS is the Ministry of Jal Shakti's water portal, which republishes these boundaries. The notifying authority is MoEFCC. This is an official copy, not a notified boundary, and it should be quoted that way. Second: its latest year of notification is 2021, so any park, sanctuary or Eco-Sensitive Zone notified between 2022 and the 2024 draft gazette is simply absent from the numerator. The Wildlife Institute of India shapefile was compared and rejected — a mid-2010s snapshot missing Kodaikanal, Megamalai and Bhimgad, and self-overlapping by 3.5%.
Tiger reserve means the protected-area core, not core plus buffer. Reserves are identified by the layer's own sub-type field, so a reserve's buffer zone counts only where the buffer is itself a notified park or sanctuary. Against NTCA's notified totals these polygons run roughly half. The figure is always a subset of the protection figure and must never be added to it — the source layer is de-overlapped, so every tiger-reserve polygon is also a protected-area polygon. Cross-checked against the WII tiger layer, the two agree to within 1% inside the ESA (5,471 km² against 5,409 km²).
"Corridor" is policy overlap, not modelled connectivity. It records
whether an identified corridor polygon crosses the village, after all protected ground is
subtracted — so it is disjoint from the protection figure and the two may be added. It is
not least-cost, circuit or graph-theoretic connectivity: a sole pinch-point and the ragged
edge of a wide corridor score the same per unit area. The source layer's type
field is unreliable — it files Bandipur, Periyar and Parambikulam as "corridor", and 79% of
its "tiger reserve" area also lay inside its "corridor" area — which subtracting protection
largely repairs but does not certify. It is zero for 78% of villages, and for that reason it
is not a component of the conservation score.
Published overlays are display geometry. The GeoJSON overlay files are simplified to about 55 m, drop sub-hectare slivers, and cover the ESA's bounding box rather than the ESA itself. Measuring areas from them will not reproduce the indicator values, which are computed on full-resolution geometry clipped to each village.
"Mining" means derelict land, not active leases. The Wasteland Atlas class is abandoned industrial and mining land as of 2015-16. It finds 15 km² across the entire ESA — far too little to represent active extraction in the Goa–Sindhudurg belt, which is itself evidence the metric does not measure what its name suggests. It is excluded from scoring. Active leases would need Indian Bureau of Mines or state DGM data.
Watershed is a location, not a contribution. Each village is labelled with the basin, sub-basin and watershed it mostly falls in. Nothing computes discharge, upstream area or headwater position, so no claim is made about how much water the ESA yields.
Two raster epochs. Wasteland is 2015-16, land cover 2017-18. Both are 56 m in the same projection so the pixel arithmetic is sound, but a wasteland share is expressed against a denominator two years later.
Sources, credits & licence
The notification
MoEFCC draft notification S.O. 3060(E), 31 July 2024 (PDF, 4.9 MB). Everything in the reconstruction derives from this document.
Base geometry
| Layer | Source | Used for |
|---|---|---|
| All-India village boundaries | Census of India / Survey of India lineage, 661,194 features | The polygon for every gazette-named village, plus 2011 Census and LGD codes |
| Kerala ESA (KML) | Government of Kerala | Independent verification; authoritative for Kerala |
Analysis layers
| Layer | Publisher | Obtained via | Used for |
|---|---|---|---|
| Primary Census Abstract 2011 | Office of the Registrar General & Census Commissioner, India | Direct | Population, households, SC/ST, literacy, worker composition |
| Protected areas (national parks, sanctuaries, tiger reserves) | India-WRIS, Ministry of Jal Shakti (republishing MoEFCC boundaries; latest notification year 2021) | India-WRIS WFS export | Existing statutory protection; tiger reserves are flagged by the layer's own sub-type field, and are the protected-area core only |
| Eco-Sensitive Zones | Parivesh / Bharatmaps, MoEFCC | india-geodata | ESZs already notified around individual parks |
| Wildlife corridors | Parivesh / Bharatmaps, MoEFCC | india-geodata | Connectivity, measured only outside protected areas |
| Recorded Forest Area | Bharatmaps | india-geodata | Land already under forest administration |
| Forest circles, divisions & ranges | Forest Survey of India | india-geodata | Forest administrative setting |
| LULC 250K, 2017-18 | National Remote Sensing Centre / ISRO | Bhuvan | Land-cover composition, 19 classes |
| Wasteland Atlas, 2015-16 | National Remote Sensing Centre / ISRO | Bhuvan | Degraded forest, derelict mining and industrial land |
| River basins, sub-basins & watersheds | India-WRIS, Ministry of Jal Shakti | india-geodata | Hydrological setting |
| National highways | Ministry of Road Transport & Highways, via PM GatiShakti | india-geodata | Infrastructure pressure |
| Basemap | OpenStreetMap contributors, via OpenMapTiles | OpenFreeMap | Geographic context |
How to cite
If you use this atlas, cite it as:
Jambu, N. (2026). Western Ghats ESA Atlas: village boundaries of the 2024 draft notification [Data set]. https://nikunjfriends25-droid.github.io/western-ghats-esa/
Several analysis layers were obtained through India Geodata, a consolidated repository of India's openly-licensed geospatial data. Its own citation file names no individual author, so it is cited as a repository:
India Geodata: Unified Geospatial Data Repository (2026). Released 7 March 2026 under CC BY 4.0. https://github.com/yashveeeeeeer/india-geodata
BibTeX, for reference managers
@misc{india_geodata,
title = {India Geodata: Unified Geospatial Data Repository},
year = {2026},
publisher = {GitHub},
url = {https://github.com/yashveeeeeeer/india-geodata}
}Where a figure rests on a single government dataset, cite that dataset directly rather than this page — the Primary Census Abstract 2011 for population, NRSC land cover for land-cover shares, Parivesh for eco-sensitive zones and wildlife layers, and the notification itself for anything about the ESA's extent. India Geodata in turn aggregates work by ramSeraph (CC0-1.0) and the DataMeet India community (CC BY 4.0) among others; its CITATION.cff carries the full reference list.
Licence
This reconstruction and its derived indicators are released under CC BY 4.0 — attribute Western Ghats ESA Atlas, Nikunj Jambu and link back.
Each underlying dataset retains its own terms. Government of India data is subject to the terms of its publishing ministry; OpenStreetMap-derived layers are ODbL and require attribution to OpenStreetMap contributors; NRSC/Bhuvan products are subject to ISRO's data policy. If you redistribute any individual layer rather than this reconstruction, check the original licence rather than relying on the licence of this page.
For official political maps of India, Survey of India maps are the authoritative reference. Compiled and maintained by Nikunj Jambu.