Years to Seconds

Every Rate Is Published Per Year and Consumed Per Second

Almost nothing that matters is measured at the rate it is quoted. A corrosion allowance arrives as millimetres per year; the probe that checks it samples in seconds. A capacity plan starts from petabytes per year; the link it has to survive is rated in megabytes per second. A finance team reports transactions annually; the queue that carries them is sized per second. In every case the same physical quantity is being described, and the only thing standing between the two statements is the number of seconds in a year.

Get that divisor wrong, or round it early, and the error rides through the whole model — usually in the direction that makes the design look comfortable.

Conversion factor: 1 yr = 31 556 952 s here, the Gregorian mean year of 365.2425 d. To restate an annual rate per second, divide by that figure — or multiply by its reciprocal, 3.168 873 85 × 10⁻⁸. A wall-loss rate of 0.1 mm/yr becomes 3.17 × 10⁻⁹ mm/s, about 0.0032 nm every second.

What the Divisor Actually Decides

Annual figures are reporting artefacts

Nothing corrodes once a year or arrives in a single yearly burst. The annual number exists because that is the interval somebody audits, budgets or publishes on — the underlying process runs continuously and has to be modelled that way.

Three year lengths, three answers

A plain 365-day year is 31 536 000 s, the Julian year 31 557 600 s, the Gregorian mean 31 556 952 s. They spread across roughly 0.066 %, invisible in a capacity estimate and very visible in a remaining-life calculation carried to four figures.

A per-second mean is not a peak

Dividing a yearly total by 31.6 million gives the flat average the year would have had if demand never moved. Real traffic concentrates into business hours and campaign days, so the mean is a floor for sizing, never the target.

Instruments push the other way

A thickness probe, a flow meter or a metrics agent emits per-second samples that have to be annualised before anyone will accept them beside a published figure. Same factor, used in reverse, with the same rounding discipline.

Restating an Annual Rate for a Model

The converter supplies the divisor; the judgement is entirely in what you divide and what you do with the result afterwards.

Get the second count for your span

Type the number of years in the left field and the second total appears on the same keystroke. One year is 31 556 952 s; a 25-year design life is 788 923 800 s. Decimal fractions work too, so a 0.5-year inspection interval resolves without a separate sum.

Divide your annual quantity by it

Carry the second count into your spreadsheet and divide the yearly figure by it: 500 TB a year over 31 556 952 s is 15.84 MB/s sustained. Keep the full second count in the formula rather than a rounded 31.6 million.

Flip it to annualise a measurement

Press swap (↔) and the pair runs s → yr, which is what you want when a logger has handed you an elapsed second count and the report expects years. Either field accepts input directly, so a second total typed on the right returns the year figure without touching the swap control.

Copy the raw number into the sheet

The copy button hands over digits only — no unit, no thousands grouping — so a formula cell treats it as a number instead of text it refuses to multiply. Ctrl + C inside the field does the same thing.

An averaged per-second rate sizes nothing on its own. A billion API calls a year is 31.69 req/s averaged flat, but if two thirds of it lands inside eight business hours the busy-second rate is several times that. Convert first, then apply a peak-to-mean factor you can defend.

Published Annual Rates, Restated Per Second

Representative annual figures from corrosion engineering, geophysics and platform capacity work, each divided by 31 556 952 s. The point of the third column is not that anyone quotes these numbers out loud — it is that this is the form a rate has to be in before it can enter a simulation, a sensor threshold or a throughput budget.

Quantity	Annual rate	Per-second rate	Where the per-second figure is used
Atmospheric corrosion, carbon steel	0.1 mm/yr	3.17 × 10⁻⁹ mm/s (0.0032 nm/s)	Electrochemical rigs that log mass loss continuously
Pipeline internal wall-loss allowance	0.5 mm/yr	1.58 × 10⁻⁸ mm/s (0.0158 nm/s)	Inhibitor dosing models and remaining-life projections
Global mean sea-level rise	3.4 mm/yr	1.08 × 10⁻⁷ mm/s (0.108 nm/s)	Altimetry drift budgets and tide-gauge trend fitting
Mid-ocean ridge spreading	25 mm/yr	7.92 × 10⁻⁷ mm/s (0.79 nm/s)	Plate-motion models fed by continuous GNSS baselines
Data lake growth	500 TB/yr	15.84 MB/s (127 Mbit/s)	Replication link sizing and ingest pipeline throughput
Central log ingest	2 PB/yr	63.38 MB/s (507 Mbit/s)	Broker partition counts and retention disk planning
Public API traffic	1 billion calls/yr	31.69 req/s averaged	Baseline for autoscaling floors and rate-limit budgets
Card transaction volume	12 million/yr	0.38 tx/s averaged	Queue depth and settlement batch window estimates

The two halves of the table behave very differently. Physical rates collapse into numbers so small they only make sense in scientific notation, which is precisely why corrosion work keeps its published unit and converts only at the point a sensor or a solver demands it. Volume rates land in a comfortable range and gain real meaning on the way: 500 TB a year sounds like a storage question until it becomes 127 Mbit/s and turns into a network one.

What This Pair Does While a Model Is Being Built

The exact divisor, not a remembered one

31 556 952 appears the moment you type 1, so the figure that goes into the formula is a stated Gregorian basis rather than the 31.5 million somebody half-recalls.

Tiny results stay in exponent form

Anything below about a millionth switches to scientific notation automatically, which is the only readable way to show the seconds side of a sub-nanometre-per-second rate.

Annualising a logged interval

Both boxes take input, so an elapsed-second count from a data logger converts back to years in place — useful when a test ran for 4 730 000 s and the report wants that stated in years.

Longer horizons without a second sum

Decades and centuries sit in the same dropdowns, so a 40-year asset life or a 200-year exposure scenario converts straight to seconds instead of being multiplied up by hand.

Questions About Rates That Change Timescale

Which year length should I divide an annual rate by?

Three are in circulation. The common year of 365 d gives 31 536 000 s; the Julian year of 365.25 d gives 31 557 600 s and is the one astronomy standardised on; the Gregorian mean of 365.2425 d gives 31 556 952 s, which is what this page uses throughout. The widest gap between them is about 0.066 %, so for capacity sizing the choice is noise. It stops being noise when a rate is integrated over decades, or when two teams compare results to four significant figures — then write the basis beside the number, because a reader cannot infer it.

My corrosion allowance is quoted in mm/yr — is a per-second value ever useful?

Not for reporting, but constantly for instrumentation. A 0.5 mm/yr allowance is 1.58 × 10⁻⁸ mm/s, and that is the form needed to set the expected drift on a continuously logging thickness probe, or to feed a transient model that steps in seconds. It also gives a sense of scale the annual figure hides: at that rate the surface loses roughly sixteen picometres a second, so any single reading that appears to show a jump is instrument noise rather than metal. Convert down for the sensor, convert back up for the integrity report.

Can I provision capacity from a yearly call volume divided by seconds?

Only as the starting line. A billion calls a year is 31.69 req/s if demand were perfectly flat, and it never is. A business-hours service typically concentrates most of its volume into about a fifth of the week, which alone lifts the busy-hour rate several times above the mean, before any campaign spike or retry storm. The usable method is to convert the annual total to the mean, then multiply by a peak-to-mean ratio measured from your own traffic rather than assumed. Keep the mean anyway — it is the right number for cost forecasting, where the peak would badly overstate the bill.

How do I turn a yearly data volume into a link speed?

Divide by the second count, then convert bytes to bits. Five hundred terabytes a year over 31 556 952 s is 15.84 MB/s, and multiplying by eight puts it near 127 Mbit/s of sustained payload. Two things then have to be added before that meets a purchase order: protocol and retransmission overhead, and the fact that the transfer almost certainly is not spread evenly — a nightly replication window compresses the same volume into a fraction of the day and multiplies the required rate accordingly. The annualised figure is an honest floor, not a specification.

Where does rounding actually bite in this conversion?

In the reciprocal, far more than in the divisor. Dividing by 31 556 952 and multiplying by a rounded 3.2 × 10⁻⁸ are not the same operation: the shortened constant runs 0.98 % high, roughly fifteen times the spread between the three year definitions people argue about. The habit that avoids it is to keep the second count as a full integer in the spreadsheet, divide rather than multiply by a hand-typed reciprocal, and round only in the cell that is displayed. Where a result is genuinely tiny, leave it in scientific notation instead of trimming it to two decimals and losing the value entirely.