Solar Irradiance Data in India: How to Compare States

Most pages that promise state-wise solar irradiance data for India publish a table with no dataset, no version, and no coordinates. That table cannot be checked, so it cannot be used in a quote or a bank submission. This guide takes the opposite approach. Every number below comes from one named dataset, queried at one named location, on one named date.

The dataset is the Global Solar Atlas, operated by the World Bank Group and ESMAP using Solargis data. We queried a reference site in all 28 states and all 8 union territories on 24 September 2026. The response carried Solargis model version 2.2.68, last updated 1 April 2026, with a satellite record running through 2025.

You will get four things here. First, the difference between GHI, DNI, DHI and plane-of-array irradiation, and which one a fixed-tilt rooftop actually needs. Second, a complete sourced table with annual GHI, daily peak sun hours, optimum tilt, and modelled specific yield. Third, the formula that turns irradiance into expected units, with a worked example. Fourth, the reasons your meter will read lower, and why the sunniest state is rarely the best place to buy.

Short version. Indian solar irradiance ranges from about 1,460 to 2,020 kWh/m² per year, or 4.0 to 5.5 kWh/m²/day, depending on location, per Global Solar Atlas data queried in September 2026. Rajasthan and Gujarat sit at the top, and the north-eastern hill states at the bottom. For a tilted rooftop, use plane-of-array irradiation rather than GHI, then apply a performance ratio near 0.78.

GHI, DNI, DHI and POA: which one your roof needs

Four measures get used interchangeably online. They are not interchangeable. Picking the wrong one is the single most common error on irradiance pages.

A second frequent error is writing GHI = DNI + DHI. The beam term has to be projected first. The correct relation is GHI = DNI × cos(solar zenith angle) + DHI, subject to the measurement convention. Also keep units straight. Irradiance is a power flux in W/m². Irradiation is energy, in kWh/m² per day, month or year.

Peak sun hours: the same number in a different coat

Peak sun hours are not the hours of daylight, and not the hours when the sun is above some brightness threshold. They are the number of hours at exactly 1,000 W/m² that would deliver the same total energy as the real day.

Standard test conditions define 1,000 W/m² as the reference irradiance. So the arithmetic collapses. Divide daily irradiation of 5.5 kWh/m² by 1 kW/m² and you get 5.5 peak sun hours. The number is identical; only the unit changes.

That equivalence is why peak sun hours are useful for mental arithmetic and useless as a separate data source. If a page gives you peak sun hours without saying which plane they sit on, it has told you nothing new. Horizontal peak sun hours and plane-of-array peak sun hours differ by up to 14% in India, as the table below shows.

State-wise solar irradiance data for India

All rows come from the Global Solar Atlas (World Bank and ESMAP, Solargis model v2.2.68, layers updated 1 April 2026, satellite record through 2025), queried at the named city on 24 September 2026. Daily GHI equals annual GHI divided by 365.25. Peak sun hours are given at the local optimum tilt, facing south. Specific yield is the atlas’s modelled PVOUT for crystalline silicon.

Read the last two columns together. Leh ranks sixth on GHI but first on specific yield, ahead of Jodhpur. Shimla ranks twenty-second on GHI but eleventh on specific yield. Cold air is worth more than raw sunlight in a PV model.

Why this is a table of cities, not a table of states

A state average needs a stated averaging method: population weighting, land-area weighting, or a sampling grid. Almost no published Indian irradiance table declares one. That makes the numbers unverifiable.

Point values at named coordinates are reproducible. Anyone can open the atlas, enter the same coordinates, and get the same figure from the same dataset version. That is the whole reason this table names the city in a second column.

Within-state spread is real and worth respecting. In Rajasthan, Jaipur reads 1,882 kWh/m² per year while Jodhpur reads 2,020, a gap of 7.3%. In Maharashtra, Mumbai reads 1,910 and Nagpur 1,876. The Global Solar Atlas publishes its solar resource layers at 9 arcsecond resolution, roughly 250 m, so there is no reason to accept a state-level proxy for a specific roof.

Turning irradiance into expected generation

The governing equation is short. Annual AC energy equals array DC rating, times annual plane-of-array irradiation, times the performance ratio, divided by the 1 kW/m² reference irradiance.

Now the two ways this goes wrong. Substitute GHI, 2,010 kWh/m², and you get 7,789 kWh, 8.4% low. Drop the performance ratio entirely and you get 10,975 kWh, 29% high. Both errors appear in real quotes.

Why your meter reads less than the atlas

Atlas irradiation describes energy arriving at the module plane. Your meter records energy leaving the inverter. Nine loss mechanisms sit in between, and a credible estimate names each one.

Those PVWatts defaults combine multiplicatively, not additively, to 14.08% total system loss. Inverter conversion loss and inverter clipping sit on top of that, and clipping depends on your DC to AC ratio rather than on the weather.

For a defensible performance ratio band, use the atlas against itself. Across all 36 reference sites, the ratio of modelled specific yield to plane-of-array irradiation runs from 0.775 at Ahmedabad to 0.836 at Leh. Annual air temperature at those two sites is 27.0 °C and minus 0.2 °C respectively. Treat 0.78 to 0.84 as the modelled band, then shade it down for a rooftop, because the atlas models a ventilated free-standing array and a roof runs hotter. Our note on summer derating in Indian conditions covers that gap.

Seasonal variation, and why the monsoon is not the whole story

Most pages reduce Indian seasonality to “the monsoon cuts output”. That is only true for part of the country. The worst month varies by region, and in the north it is not a monsoon month at all.

Monthly GHI from the same Global Solar Atlas query, divided by calendar days.

Three regional patterns fall out. In Punjab, Delhi and Ladakh, January is the floor, driven by fog, haze and a low solar altitude. On the west coast and in Karnataka, July is the floor, which is the classic monsoon signature. In Kerala the floor is November, because the north-east monsoon extends the cloudy season past the south-west one.

For context on timing, the India Meteorological Department’s revised normals put south-west monsoon onset over Kerala at 1 June, full country coverage by 8 July, withdrawal from north-west India beginning around 17 September, and complete withdrawal by 15 October. Load matching and any battery decision should follow the local monthly curve, not the annual mean.

Tilt and azimuth: what a flat or east-west roof costs

The old rule says set tilt equal to latitude. Atlas optimum tilt shows that rule works in the south and overshoots in the north.

Thiruvananthapuram sits at latitude 8.5 degrees and its optimum tilt is 11. Mumbai is at 19.1 with an optimum of 22. Bhopal is at 23.3 with an optimum of 25. Then the relationship flattens. New Delhi is at latitude 28.6 with an optimum of 26, and Srinagar is at 34.1 with an optimum of 30. Monsoon cloud and dry-season haze reduce the payoff from a steep winter-biased tilt in the northern plains.

Ladakh breaks the pattern the other way. Leh’s optimum is 35 degrees, above its 34.2 degree latitude, because dry clear air keeps direct beam high through the cold months.

A horizontal array gives up the whole tilt gain. Comparing annual GHI against plane-of-array irradiation at optimum tilt, a flat mount costs about 1.5% at Thiruvananthapuram, 2.5% at Chennai, 8.0% at New Delhi, 8.4% at Ahmedabad, and 14.5% at Leh. The penalty rises with latitude, so a flat terrace in Kerala is nearly free and a flat roof in Ladakh is expensive.

An east-west split roof is a different problem, not a worse one. It flattens the daily curve into two lower, wider peaks, which can suit a daytime load profile. It does need separate maximum power point tracking inputs per orientation. We work through that layout in the east-west roof design guide.

The sunniest state is rarely the best place to buy

This is where irradiance tables mislead buyers. The whole of India fits inside a 38% band on annual GHI, from Itanagar at 1,460 kWh/m² to Jodhpur at 2,020. Exclude the north-eastern hill states and the band narrows to about 20%.

Now compare that with the commercial variables. Retail tariffs are set state by state through each state electricity regulatory commission’s tariff order, and slabs differ sharply between a Mumbai residential consumer and a Punjab agricultural one. Net metering, gross metering and export compensation rules also vary by state and by DISCOM, as our net metering guide sets out.

The central subsidy, meanwhile, is flat. PM Surya Ghar pays ₹30,000 per kW for the first 2 kW and ₹18,000 for the third kW, capped at ₹78,000, according to MNRE for the scheme approved in February 2024. It does not pay more in Assam to compensate for weaker sunlight, and it does not pay less in Rajasthan.

So the ranking that matters is tariff times exported units, under your DISCOM’s rules, minus your installed cost. A 20% irradiance deficit is easily outweighed by a higher tariff slab or a friendlier export rule. Rules change, so verify the current position with your DISCOM before you commit.

Using irradiance data when you size the system and the inverter

Irradiance sets energy, not electrical limits. That distinction decides which tool you reach for.

1. Size the array from consumption. Work out target annual units, divide by the site’s specific yield in kWh/kWp, and you have the DC size in kWp.
2. Constrain it by roof area, shade and your sanctioned DISCOM load. Any one of these can cap the array below the energy-derived size.
3. Choose the DC to AC ratio. Higher irradiance sites and steeper tilts push more hours near the inverter ceiling, which is where clipping enters the economics.
4. Verify the string electrically. String length is governed by module open-circuit voltage at the coldest expected temperature and by current at the hottest, not by an irradiance map.
5. Recheck against the monthly curve. A design that clears the annual test can still clip badly in the peak month.

Step 4 is the one people skip. The Qbits string sizing calculator screens module and inverter voltage and current windows using module temperature assumptions. It does not read an irradiance map and it does not forecast annual kWh, so pair it with the yield method above rather than expecting one tool to do both. Our on-grid inverter range lists the voltage windows those checks need, and the solar yield guide documents the energy side.

The Bottom Line

Solar irradiance data for India is only as good as its provenance. A number without a dataset, a version, a plane and a coordinate belongs in marketing copy, not in a design file. The table above is reproducible because it names all four.

The physics also rewards precision more than geography. Choosing plane-of-array over GHI is worth up to 14%. Applying an honest performance ratio is worth 29%. Moving to a sunnier state is worth 20%.

• Re-query your own coordinates on the Global Solar Atlas and record the dataset version and the date, then use plane-of-array irradiation at your planned tilt.
• Estimate generation with the full formula, state your performance ratio, and cross-check the result against the atlas’s specific yield for the same point.
• Bring your site coordinates, planned tilt and module datasheet to the Qbits technical team and we will help you match the string and inverter windows to that design.