California lake levels

The answer

Where the state's reservoirs stand

The daily reservoir storage summary carries 48 reservoirs, and every reading below is dated November 4, 2026. Together they hold 13,277,679 AF against 24,193,916 AF of stated capacity, which is 54.9% of capacity. CDEC, managed by the California Department of Water Resources, publishes that Daily Reservoir Storage Summary covering selected reservoirs in Northern and Southern California, and AF in it means acre-feet, the unit the summary reports storage and capacity in.

Measured against their own storage record for this calendar day, 16 reservoirs sit below their median for the date and 24 sit at or above it. Of the rest, 1 has a record too short to rank, and CDEC published no storage figure for 7 more on this date. The fullest by percent of capacity is ENGLEBRIGHT at 98% of capacity, and the emptiest is EASTMAN (BUCHANAN) at 7% of capacity. Furthest below its own normal for the date is PYRAMID, at the 1st percentile of its own record for a November day. Those are two different questions, and the table lets you switch between them.

The table

Every reservoir in the daily summary

48 of 48 reservoirs shown, ranked by percent of capacity

No reservoir matches that filter.

Operators differ across the feed. Oroville Dam is operated and maintained by the California Department of Water Resources, while Cachuma Lake (Bradbury Dam) is operated and maintained by the US Bureau of Reclamation, so the reservoirs carried on the CDEC feed are not all run by the same operator.

The argument

Why the two rankings disagree

Flip the toggle and the order changes. Percent of capacity measures water against concrete. The percentile measures water against that reservoir's own history for the same calendar day. A reservoir can look middling on one and remarkable on the other.

Across the 40 reservoirs that can be ranked both ways, the two orderings carry a Spearman rank correlation of 0.41. Read that as how far the two orderings diverge, not as a p-value, because CDEC prints percent of capacity as a whole number and ties are common.

Take PYRAMID. Among the 40 reservoirs with a long enough record it ranks 8th by percent of capacity, at 88% of capacity, and 40th by where it sits in its own record for the date, at the 1st percentile for a November day.

Both columns divide the same series, reservoir storage in acre-feet, so the disagreement is real and not an artifact of units. Feet would not allow that comparison. USGS keeps one parameter for lake or reservoir water surface elevation defined against the NGVD 1929 datum, reported in feet, and a separate parameter for water surface elevation referenced to NAVD 1988, so surface elevations at different sites can be stated against different vertical datums. Storage is its own parameter measured in acre-feet, distinct from any water surface elevation parameter, and CDEC records reservoir elevation in feet as a distinct sensor from reservoir storage in acre-feet at the same station. So elevation appears in the table as one station's reading with its own date, and never as a subtraction between two reservoirs.

Percent of capacity is the familiar view for a reason. Reclamation describes percent of total capacity as the percentage of total storage capacity used by the water currently stored in the reservoir, and CDEC publishes that column, expressing current storage as a percentage of the reservoir's stated capacity under a heading of % of Capacity. Capacity itself is a defined quantity. Reclamation defines total capacity as the reservoir capacity below the highest of the top of exclusive flood control capacity, the top of joint use capacity, or the top of active conservation capacity, used to express the total quantity of water that can be impounded, and joint use capacity as reservoir capacity assigned to flood control purposes during certain periods of the year and to conservation purposes during other periods of the year.

CDEC also publishes its own comparison against normal, an Average Storage figure for each reservoir with a separate % of Average column computed against it. That average runs over a fixed window, and each reservoir's CDEC information page states the base period used for its published averages as a start year and an end year. A different window gives a different answer, which is why we show a rank over the whole daily record instead. CDEC's average column is not a percentile rank and we do not present it as one. Other operators frame it differently again: Reclamation's percent of daily historical average compares the amount of water currently stored with historical levels on the same date, using a stated multi-year base period rather than a rank within the record.

How we know

Methodology

Every figure here comes out of one series. CDEC publishes reservoir storage at daily resolution, described by the agency as updated daily, and we pull the same sensor data through CDEC's web service, which returns CSV or JSON. Numbers are recomputed in code at build time from the file that pipeline writes. Nothing on this page is typed by hand.

The ladder is what makes the second measure possible. For one reservoir and one numeric day of the year we take every daily storage value in its archive for that day, sort them, and read where the current value falls. USGS defines a percentile as indicating the relative standing of a data value when data are sorted into numerical order from smallest to largest, and calculates its day-of-year statistics from all approved historic observations for a single numeric day of the year. USGS provides those statistics for a specific calendar date at a monitoring location, computed over that location's own period of record. Ours holds 615,712 daily observations across 48 stations, of which 47 have enough history for a ladder. The earliest record starts November 1, 1962 and the longest runs 64 years. Below 10 observations for a calendar day, the row gets no percentile.

We compare against the median rather than the average because USGS notes the mean is less resistant to changes in the magnitudes of outlying observations than the median. The median is the middle value in a data set when the values are arranged in ascending or descending order, so the acre-feet column in the table is the gap between today's storage and that middle value.

Now the limits. USGS calculates its published statistics from approved data only and excludes provisional data from those calculations, while CDEC labels its published station data as provisional and subject to change. Read the percentile here as a rank computed over a provisional archive, not as an approved statistic. Storage at Oroville is derived from a reservoir capacity rating curve rather than measured directly, and CDEC recalculated that curve so that storage values from a stated date forward are calculated using the new rating curve, so a long ladder mixes values computed under different curves and we make no adjustment for it.

I wrote the pipeline behind this table, scripts/fetch-levels.mjs, against keyless government feeds, with a hard timeout and a bounded retry on every call and carry-forward so an upstream outage never blanks a row. The feed label normalizer in src/lib/lakes.ts exists because an uppercase string match on CDEC had been mislabelling every California and Reclamation lake as USGS in the on-page source chain and in the schema. I found that by reading the source chain on a page I had already shipped, and fixed it in code rather than in the copy.

How often these numbers change

CDEC describes its reservoir storage series as updated daily, and the values in its daily reservoir summary are for a period ending at midnight, so the published table is a dated daily summary rather than an instantaneous reading. The date at the top of this page is the date of the readings, not the day you happen to be reading it.

What the reading date means

Readings come from the operating agency's own gage and carry that agency's timestamp. Anything older than seven days is labelled delayed and is never called live. Each row shows its own reading date and its feed, taken from the data rather than assumed, and CDEC warns that its station data are preliminary and subject to revision, so a figure can move after publication.

Why some reservoirs carry no percentile

A ladder needs enough history for one calendar day. Where a station's archive is too thin, the row still shows storage, capacity, percent of capacity and elevation, and says the record is too short to rank instead of guessing. USGS imposes no minimum amount of data for a statistic to be calculated, and reports the number of observations used alongside the statistic. We set a floor instead. Record starts also differ by station, and CDEC's daily reservoir elevation record for Cachuma Lake begins later than the start of the Oroville daily elevation record.

What a percentile does not tell you about a day at the water

A percentile is a position in a sorted list and nothing more. USGS describes day-of-year statistics as representing the expected conditions for a generic day of the year at a location, based on approved historical observations. We do not turn a low rank into a verdict about launching, boating or drought conditions. Not navigation, safety, or engineering advice. Estimates from public gage data. A ramp read is a level-based guide, not a guarantee. Confirm with the lake operator or the managing agency before travelling.

Where to see one lake's whole record

Each reservoir in the table links to its own page where we hold one, and those pages carry the single-lake question. This page does the part no single-lake page can, which is put every reservoir the agency reports on the same ladder for the same day.