Battery Autonomy Days Map: How Many Days of Storage Your Location Needs

A battery bank is not sized by how much sun a place gets. It is sized by how long the sun goes away. Those turn out to be almost unrelated questions: across the 90 states and countries below, the link between darkest-month peak sun hours and the days of storage a site needs is effectively nil. An Arizona array and a German array differ by a factor of five. Their batteries differ by a day and a half.

What One Day of Autonomy Buys

A day of autonomy is one day of the household’s normal consumption, held in the battery as usable energy, with nothing coming in from the array. Three days means three days of ordinary living with the panels producing nothing at all.

Usable carries the weight. A bank rated 30 kWh and run to 80% depth of discharge holds 24 kWh you may spend, and days of autonomy count the 24.

The chain runs: days × daily load = usable kWh, then usable kWh ÷ depth of discharge = the nameplate you buy. The off-grid battery bank calculator runs it with your own load list and chemistry; this page supplies its first number.

The Formula, in Three Lines

One. Size the array so an average day in the site’s darkest month produces 25% more than the daily load. That margin covers battery round-trip loss, controller and inverter loss, soiling and panel aging.

Two. Run the site’s real daily irradiance record through that array for 19 winters, letting the battery fill and drain. Record the deepest drawdown of each winter. The 90th percentile of those 19 numbers — the one-in-ten winter — is the design figure, in days of load.

Three. Round up to the next half day, with a floor of one day. That is the “design days” column in both tables.

A one-in-ten winter is used rather than the worst on record, which in several places roughly doubles the bank for capacity that then sits idle. The worst of the 19 is in the CSV.

Days of Autonomy, Every US State

Sorted by design days, largest first. Any column heading re-sorts; a state name opens its off-grid page. Peak sun hours are the darkest-month daily average on a south-facing plane at latitude plus 15 degrees.

Oregon at 5.5 days is the highest figure in the country, set by the Willamette valley winter, where low cloud settles into the basin and stays. Its worst of the 19 winters drained 7.7 days of load.

Texas at 4.5 is the surprise. It gets more than twice Oregon’s December sun and still buys the same battery, because the array is sized small to match that bright December average and Gulf stratus then arrives in multi-day sheets.

Days of Autonomy, Every European Country

Same method and columns, one reference point per country at the capital city.

Romania at 6.5 days is the extreme of the dataset, and it is set by the tail rather than the norm. A typical Romanian winter drains 2.6 days of load. The one-in-ten winter drains 6.4. The worst of the 19 reached 9.4.

All three sit in the CSV, because the gap between them is the decision: a bank built for 9.4 spends most of a decade holding energy it never needs.

The Maps

Both maps are schematic. Every state and country is one square of the same size, shaded by its recommended days, so no place reads as louder than another for being larger on a globe.

The pattern is not the one a sunshine map draws. The heavy shading sits on the mid-continental United States, the Pacific Northwest, and a belt through central and southeastern Europe. Scandinavia, Iceland and Alaska, the darkest places on the list, sit at or below the median.

Why Sunshine Does Not Predict Storage

Across all 90 places, the correlation between darkest-month peak sun hours and the one-in-ten-winter storage figure is 0.05. Peak sun hours explain essentially none of the variation in battery size.

The reason sits in step one. The array is already sized to the local darkest month, so absolute darkness is priced into the panels first. What is left for the battery is the shape of the departures: how far and how long a site drops below its own winter average.

Finland has 0.45 peak sun hours in December and needs 2.0 days of storage. Wyoming has seven times the December sun and needs 1.5. Both are consistent places. A Finnish December is reliably dark and a Wyoming December is reliably bright, and consistency is what a small battery requires.

The expensive climates are the changeable ones. Continental interiors with dry cold air clear within a day or two of a front. Marine layers, river basins and mountain valleys trap low cloud for a week or more, and that persistence is what the battery rides out.

Panels or Batteries

Days of autonomy are not a fixed property of a place. They depend on array size, and the two are directly substitutable. The chart below sweeps array size from break-even to double at five sites.

The curves are steep at the left and flat at the right, and that shape is the lesson. Germany needs 7.2 days of storage with an array sized exactly to the darkest-month average, 3.2 days at the 25% margin used in the tables, and 1.3 days once the array is half again bigger.

Across all 90 places, 79 fall to 2.5 days or less once the array is oversized by 50%. In medians: 6.7 days at a break-even array, 2.9 days at the 25% margin used in the tables, 1.9 days at 50%.

Read the last step on its own. Going from a 25% margin to a 50% margin is a fifth more panel, and it takes out a median third of the storage.

Where it stops paying is visible at the far right of the chart. Germany sheds 4.6 days of storage between a break-even array and a 30% margin, and 0.3 days between the 180% and 200% arrays. What is left out there is a run of days with almost no sun, which no extra panel reaches.

Set the array with the solar sizing calculator, then bring the figure back to the battery bank calculator.

Where More Battery Is the Wrong Answer

The first is the far north. Alaska, Finland, Sweden, Norway and Estonia all report 1.5 to 2.0 days, which is true and misleading at once. Their December peak sun hours run 0.45 to 0.72, against 5.28 in Arizona — an array seven to twelve times the size for the same house. Nobody builds it. Those households let a generator carry midwinter and size the battery for the shoulder seasons.

The second is the deep tail. Where the one-in-ten figure passes about five days, the last two days of storage cover an event that turns up once or twice in twenty years, and a generator carrying those days is usually the smaller commitment. The backup generator guide sets out that comparison.

Wind changes the arithmetic without removing it. Ohlendorf and Schill, in Environmental Research Letters in 2020, ran 40 years of MERRA-2 reanalysis over German onshore wind, 1980 to 2019. They found a period of around five consecutive days below a 10% capacity factor — under a tenth of what the turbines are rated for — in an average year, close to eight days once a decade, and just under ten days in the longest event of the record, in March 1985.

Those are the same weather patterns that produce the solar drawdowns above, so a hybrid site should size its storage against the combined shortfall rather than against either resource on its own. The European event record sits in the outage and blackout atlas.

Method, Data and Sources

Irradiance comes from NASA POWER, which serves CERES SYN1deg Edition 4.2 satellite-derived surface shortwave flux on a one-degree grid from January 2001 onward. Its earlier years run on a different source, NASA/GEWEX SRB Release 4-IP, so the window here stops at 2001 and one product covers the whole record.

Two products are used: the 2001 to 2020 monthly climatology for the darkest month and its peak sun hours, and daily all-sky irradiance over the same period for the persistence simulation.

US reference points are each state’s 2020 Census center of population, from the Census Bureau’s CenPop2020 state file. European reference points are capital cities. Two pairs of capitals fall inside the same one-degree grid cell — Vienna with Bratislava, Skopje with Pristina — so Austria and Slovakia carry identical figures, as do North Macedonia and Kosovo.

The persistence simulation runs on horizontal irradiance, because the drawdown depends only on the ratio between a cloudy day and that month’s average, which tilt barely changes. A drawdown year runs 1 July to 30 June so a winter is never split, and the two partial years at the ends of the record are discarded, leaving 19 complete winters.

Peak sun hours were cross-checked against the European Commission’s PVGIS at eight European points (PVGIS-SARAH2 and ERA5, 2005 to 2020). Horizontal figures agreed within about 12%; on the tilted plane NASA POWER ran 3% to 25% lower, the widest gaps at Berlin and Madrid. Treat that column as a regional indicator and run your own coordinates through PVGIS or PVWatts before ordering panels. The days column is unaffected, being built from ratios.

Three limits. A one-degree cell is roughly 111 km on a side, so a mountain valley or a coastal fog pocket can differ sharply from its cell average, and one reference point cannot represent Texas or Norway. The simulation also assumes a flat daily load, where a real house uses more in the same cold, dark weather that starves the array.

Verified September 2026 against the NASA POWER API as served, the CenPop2020 state file as published, and the PVGIS v5.2 API. Refresh: annual, as NASA POWER extends the daily record. The full table, including the worst single winter and both other array sizes, is data.csv. The underlying data is public; this compilation is free to reuse with credit to Grid CEO.

Common Questions

Is three days of autonomy still the standard answer?

Three days is the median of the 51 US entries here and half a day short of the European median, so it is a fair place to start. It is also wrong by a factor of two in both directions at the ends of the table: 1.5 days in Wyoming, 6.5 in Romania.

Why is the number lower in Finland than in Texas?

Because the array has already absorbed the darkness. A Finnish array sized to a 0.45 peak-sun-hour December is enormous, and Finnish winter is consistent enough that it rarely stays far below its own average. Texas gets more December sun but loses it to multi-day Gulf stratus, and that gap is what the battery pays for.

Does this include losing the panels to snow?

No. The simulation models cloud cover only, so a site that spends days under snow needs either the discipline to clear the array or an extra day on top. Steeper tilts shed snow faster, a second reason for the latitude-plus-15 winter tilt used here.

Can I use these numbers for a weekend cabin?

Only loosely. The figures assume a flat load all year. Intermittent occupancy usually works in your favor, because the bank reaches a dark spell already full. Run the real pattern through the battery bank calculator rather than scaling this figure down.

Sources: NASA Prediction Of Worldwide Energy Resources (POWER), daily and climatology point APIs, all-sky and tilted-surface products, CERES SYN1deg Edition 4.2, retrieved September 2026, with its energy-fluxes methodology documentation; U.S. Census Bureau, Centers of Population by State, 2020 Census (CenPop2020_Mean_ST); European Commission Joint Research Center, PVGIS v5.2, PVGIS-SARAH2 and ERA-5, 2005 to 2020; Ohlendorf and Schill, “Frequency and duration of low-wind-power events in Germany”, Environmental Research Letters 15(8) 084045, 2020.