The hashrate is yours to recompute

Every block records its committed targetThe number a header's hash has to come in under for the block to be valid; a lower target means more work. and its Cardano-attested time. Those two numbers are what a hashrate estimate is made of, and both are on a public ledger. So the work behind the chain can be estimated by anyone, from public data, with no reliance on pool self-reporting.

That matters because the usual hashrate figure for a small chain comes from the pools themselves, which means it comes from the parties with the strongest reason to overstate it.

whitepaper §5, §9

What each block leaves behind

The committed target
Carried in the header's nBits fieldThe four bytes of a header that state the target the miner is committing to; the easiest value this chain allows is 0x7fffff · 2²³², a full-width mantissa at the loosest exponent that stays inside a 256-bit target. and checked by the validator against what the rule allowed at that moment. It is how much work the block was expected to cost.
The Cardano-attested time
From the transaction's validity intervalThe window of chain time within which a Cardano transaction may be accepted, which is where WHAT's clock comes from. rather than from a timestamp the miner reported about itself. See who enforces the validity interval.
The height and the tip hash
In the data attached to the chain's tip output, so the sequence of blocks is reconstructable in order with no gaps.

Line up a run of targets and times and the expected work follows: a target implies how many hashes a block takes on average, and the interval implies how long that took. Anyone with a Cardano node can do it, and two people working from the same blocks with the same estimator should get the same answer.

The estimator is fixed, so two people get the same number rather than two defensible ones. A block's work is its difficulty times 232 hashes; an estimate over a run of n blocks is that work summed and divided by the attested seconds the run spans. Because blocks turn up at random, the naive ratio reads n/(n−1) high and wobbles by about 1/√(n−2), so the published figure corrects the first and prints the second beside itself.

Two runs are published, because no single one does both jobs. Six blocks is about an hour and wobbles by about half, which is what shows an arrival while it is still happening. A hundred and forty-four blocks is about a day and wobbles by about eight per cent, and that is the figure worth quoting. Bitcoin's hashrate sits beside the slower one, taken from mempool.space over that site's own 24-hour window — a different estimator, which the page says rather than implies.

Both runs are computed and drawn on the chain status page, from the blocks the chain is sitting on, re-read while you watch it. It is the same estimator described here and the same code the validator's own checks run under, which is why a figure there can be reproduced rather than only read.

whitepaper §9

Which leaves exactly one thing you have to take from somewhere else

A price. It is nowhere on the chain, it cannot be derived from anything that is, and no amount of care with the rest of the hashprice calculation fixes that. Wherever a price appears on this site it is marked as external, with its source named.

So in the hashprice calculation, the bottom of the division is reproducible and the top depends on an outside observation. It is worth knowing which half is which before quoting either.

There is a second limit, and it is about where the hashrate came from rather than how big it is.