Methodology

How to read this site

Every claim carries an evidence label:

• Documented: stated in an official document (clearance letter, committee minutes, EIA, court order).
• Reported: stated by a news outlet or a named researcher.
• Alleged: said by a named person or group and not confirmed by a document. The site always says who.
• Official claim: what a developer or government says about itself.

The ⓘ beside a number opens the exact sentence it comes from, with the document and page. Numbers I computed are marked "derived" and show the formula.

Everything below is a model, not a measurement. Each input is listed with its source, and you can change any of them in model-params.json and rebuild.

1. Forest per unit of storage

For each Western Ghats project that has both a forest-land figure and a storage figure in its clearance papers:

forest ha per GWh = forest land (ha) ÷ storage (GWh)

14 projects have both. Together they need 1,331.2 ha of forest for 139.1 GWh of storage, an average of 9.6 ha per GWh. The range is wide: from 0 at Tarali, where the layout was revised off forest land, to 22.1 at Bhavali.

Pipeline extrapolation (illustrative). Maharashtra has signed MoUs for 78,215 MW of pumped storage. At an assumed 6 hours each, that is 469 GWh. At the average above, that would need about 4,500 ha of forest. This is an illustration of scale, not a forecast: many MoUs will never be built, and siting changes the answer a lot.

Limits.

• Forest ha counts only land legally classed as forest. Lateritic plateaus such as Adoshi (Bhorande's upper reservoir) are often classed as non-forest, so the habitat lost is larger than the forest figure.
• Private and farm land, roughly as large again, is not included.
• Where a project's storage is not stated, I use capacity × 6 h and mark it as an estimate.
• Tree counts come from forest committee minutes, EIAs or site-visit reports and are shown side by side when they disagree.

2. Energy: a pumped storage plant uses more than it gives back

A pumped storage plant pumps water uphill with grid power and releases it later. Some energy is lost each cycle. The share that comes back is the round-trip efficiency, η. I use η = 0.775, the 22.5% declared cycle loss in Tamil Nadu's 2026 tariff order, the best contractual figure in India.

energy in = energy out ÷ η loss = energy in − energy out

Daily full cycle. If the whole pipeline (469 GWh) filled and emptied once a day, it would lose 136 GWh a day, or 49.7 TWh a year.

At a realistic utilisation. Karjat's environmental clearance expects 6,241.5 million units a year from 3,000 MW, a capacity factor of 23.7%. Applied to the pipeline:

output = MW × 8,760 h × capacity factor loss = output × (1/η − 1)

That gives 162 TWh a year delivered and 47 TWh a year lost, about 2.7% of India's net generation in FY2025-26 (1,761.28 TWh).

Sensitivity. Typical plants range from 0.7 to 0.8:

η	Lost per day, full cycle (GWh)	Lost per year at reference capacity factor (TWh)
0.7	201	70
0.775	136	47
0.8	117	41

Limits. The pipeline will not all be built, and real plants do not cycle at a fixed rate. The loss is real only if the charging power would otherwise have been used; power that would have been curtailed is not a loss in practice.

3. Emissions: it depends on what charges it

Storage has no chimney, but it moves emissions in time. Whether it cuts or adds CO₂ depends on which power plants run more when it pumps and which run less when it generates. I use the Central Electricity Authority's CO₂ Baseline Database for FY2025-26: grid weighted average 0.675 t CO₂/MWh and operating margin (the plants that change output first, mostly coal) 0.963 t CO₂/MWh.

Two ways to count.

Attributional intensity: I = EF_charge ÷ η + LCA_infra

Consequential change: ΔCO₂ = EF_charge,marginal ÷ η − EF_displaced,marginal

Scenarios (t CO₂ per MWh delivered).

Scenario	Charged by	Displaces	Result
(a) Coal margin	operating margin	operating margin	+0.28
(a′) Grid average	weighted average	(attributional)	0.87 before infrastructure
(b) Curtailed solar	solar that would be wasted	operating margin	-0.96; -0.88 with infrastructure

In case (a), on 162 TWh a year, the pipeline would add about 45 million tonnes of CO₂ a year. This is an upper bound.

Infrastructure emissions use the NREL life-cycle study of closed-loop pumped storage, base case 86 g CO₂e/kWh (0.086 t/MWh). The study's range is 58–530.

Break-even. If a share s of charging is curtailed solar and the rest is coal margin:

ΔCO₂(s) = (1 − s) × EF_OM ÷ η − EF_OM

This is zero at s = 1 − η. So storage cuts emissions only when more than 22.5% of the charging energy is solar that would otherwise be wasted, or 29.4% once infrastructure is counted.

Limits. Real dispatch is hour by hour; a single margin factor is a simplification. As solar grows, case (b) becomes more common. Developer claims of avoided CO₂ implicitly assume all charging is zero-carbon.

4. Cost: comparing tariffs for batteries and pumped storage

Tariffs are quoted in different units (₹ lakh per MW per month, ₹ crore per MW per year) for plants with different durations. I put them on one basis: ₹ lakh per MWh of storage capacity per year.

₹ lakh/MWh/yr = (₹ lakh per MW per year) ÷ duration (h) (monthly tariffs × 12; ranges use the midpoint)

Recent Indian tenders give 8.6–14.3 for batteries (2–4 h) and 9.7–13.6 for pumped storage (8 h). Every tender used is listed with its source in model-params.json.

Limits. A tariff is a price, not a full cost: viability gap funding, contract length (40 years for SECI's pumped storage tender), degradation and the round-trip loss are not equalised. Batteries and pumped storage of different durations do different jobs.

Data and code

• All data: download page. Every number's source: citations.csv.
• Formulas: src/lib/data/model.ts. Inputs: data/model-params.json. Output: data/generated/stats.json.
• Found a mistake? See the corrections policy.