The Weather Prophet Over the Range: What I Found When I Tested Inigo Jones

On 2 February 1893, a 20 year old farm hand at Crohamhurst, one valley north of us past Woodford, emptied his rain gauge and wrote down 907 mm. In 24 hours. That's still the Australian record, and I reckon most people in this valley have never heard that it happened barely 40 km from Dayboro as the crow flies.

The young bloke with the gauge was Inigo Jones. He spent the next sixty years trying to prove that Queensland seasons could be forecast years ahead from the orbits of Jupiter, Saturn, Uranus and Neptune. He built an observatory at Crohamhurst in 1935 to do it. Farmers and graziers swore by him. The Bureau, then and now, swore at him. And sicne I run a weather station and a forecast model on a hill in the same subtropical hinterland he worked in, I decided this winter to settle it for myself: I pulled his original papers and put his actual method through 130 years of rainfall records.

Who Jones was

Jones trained under Clement Wragge, the government meteorologist who gave cyclones names and Queensland its first real forecasting service. Wragge handed him two ideas in the 1890s: the roughly 11 year sunspot cycle, and a 35 year climate cycle a European researcher called Bruckner had found in old records. Jones took those and went further. Jupiter circles the sun in 11.862 years, which is suspiciously close to the sunspot period. Three Jupiter orbits make 35.6 years, close to Bruckner. From there he derived a family of cycles, and by the 1940s his working set looked like this.

The method itself was simpler than the astronomy suggests. In a 1947 interview he laid it out: to forecast a season, look up what the weather did 35, 59, 71, 84 and 165 years ago, chart those old seasons side by side, and where they agree, that is the forecast. His words: "If, for instance, all periods indicate rain in January, that is our forecast. We just read off the graph." When the periods disagreed, judgement took over, and when a forecast missed he did not drop the cycle, he moved the event to a longer one and set a new date, decades out. His 1870 flood repeat failed its 71 year test in 1941, so he calmly rebooked it for 1954 on the 84 year cycle. You can see the problem. A method that can never be wrong is also a method that can never be right.

In 1923 he called the end of a drought and became famous for it. By 1939 the federal government was funding an inquiry into his methods, and the scientists at the ANZAAS congress that year tore his paper apart. Official reviews in 1939 and again in 1953 both came back negative, the later one scoring his forecasts at about 50 percent. A coin, in other words. He kept publishing anyway, kept his subscribers, and died in 1954 still certain the proof would come once the records grew long enough. His grand cycle needed data until 2006 to complete one full lap. He knew he would never see it tested, and he kept working the method anyway for thirty more years. That takes a kind of conviction I respect even where I think it was misplaced. Well, the records are long enough now.

What I actually tested

This is the part I care about, because his defenders have a fair complaint: the official reviews judged his published forecasts, which mixed cycles with his own judgement and daily sunspot watching. Nobody, as far as I can find, ever took the documented cycle method on its own and ran it against long station records. That is now not so hard to do. The Bureau publishes daily rainfall for Crohamhurst from 1893 to 2003, Brisbane from 1841, and the old Dayboro Post Office gauge from 1931 to 2012. The National Library has digitised his 1944 paper where he explains the method himself, and I went through the old newspapers where he laid out the exact cycle lengths his observatory used.

So the test is his own recipe, nothing added. For every season in the record, look up the same season 35, 59, 71, 84 and 165 years earlier, mark each one wet or dry against its long term median, and when all the available cycles agree, make the call. Score the calls. I ran it twice, once with whole year steps the way his worked examples round off, and once with the observatory's exact lengths, where the analogue drifts through the calendar, his own example maps July 1886 onto November 1945 on the 59 year cycle. I also tested each cycle on its own, and separately checked whether rainfall here varies with Jupiter's position at all, in 60 different seasonal slices.

What the numbers say

I wanted the result to be more interesting than that. The local romantic in me was waiting on a faint signal in the Crohamhurst data, his own gauge, his own valley. It's not there, not yet, not in the 130 years his own valley has given us. The 59 year cycle looked mildly promising at Crohamhurst and then pointed the opposite way at Dayboro, which is exactly what noise does when you slice it enough ways.

Actually, I need to correct something I just implied. I called his method a coin, then quoted a number under 44 percent right after it, which reads like it scored worse than the coin. It did not. A sample of 66 unanimous calls carries a margin wide enough that 43.9 percent and 50 percent are not statistically different, the p value says so plainly. Small samples wobble on both sides of the line, and this one wobbled low. The honest statement is: no skill, in either direction, anywhere I looked, at any sample size I could build.

The pushback, and round two

When the first version of this piece went round, the pushback came quick, and it was fair. I had tested Jones's cycle arithmetic, the years and the lags, but not the engine he believed sat behind it. Jones never claimed the years did anything by themselves. He believed the planets steer the sun and the sun steers us. Where was Saturn's actual position in my test? Where was the sunspot count? I could have left it there and defended the first result, but that would be exactly what I accuse the cycle followers of, defending the theory instead of testing it.

So I built round two. This time with the JPL DE440 ephemeris, the same planetary tables the space agencies fly missions on, so no rounded years anywhere: the real position of Saturn, of Uranus, of Neptune, the actual angle between Jupiter and Saturn as it closes and opens over its 19.86 year lap, and the sunspot record itself, both how high it sits and which way it is heading. Ninety fresh tests across the three stations and every season, on top of the sixty from round one. And I wrote the pass mark down before I ran anything: a result only counts if it survives the statistical correction at one station and then shows up again at a second one. No moving the goalposts afterwards, in either direction.

The count of survivors: zero.

Nine of the ninety flickered at the loose end of the statistics, where chance alone predicts four or five, and each one fell apart when I looked closer. The one that had me leaning forward was the Jupiter and Saturn angle, because a 2025 study of 347 eastern Australian rain stations found a real cycle of about 20.4 years in the rainfall, and 19.86 is temptingly close. But the flicker showed up only at my two short records, and not at Brisbane, which holds 186 years and covers that angle nine times over. A real driver gets stronger where the record is longest. This one got weaker. By now I recognise that pattern on sight, it is what noise does.

I will say this for the method, in fairness. The Bonferroni correction I used across those ninety tests is about the strictest one going, it assumes every test is independent, and mine were not. Rainfall in one season leans on the season before it through the state of the Pacific, so the true number of independent tests sitting under those ninety was smaller than ninety, and the bar I set was higher than it strictly needed to be. A real but modest signal could sit under that bar and I would not see it with this design, that is the honest limit of what I built. And his longest cycle, the 165 year Grand cycle he pinned to Neptune, still cannot be properly tested at any of my three stations, the same wall Jones ran into himself. Nobody alive has watched two full laps of it. That is not evidence for the cycle. It is a genuine gap in what anyone can currently know, and I would rather name it than pretend the ledger is fully closed.

While I was at it I checked the claim you hear from the almanac sellers, that the dying phase of a solar cycle brings drought to our side of the equator. Our three stations say the summers under the steepest solar decline were, if anything, a whisker wetter than average. Not drier. The claim is not just unproven, it points the wrong way.

Where I ended up on the man

Here is what surprised me form the original papers. Jones was more careful than his reputation. His 1944 paper records his failed predictions in plain sight, he committed to testing every event forward as it arrived, and he quotes a colleague telling him the full proof would take three hundred years of data. He was wrong, but he was not a con man. He was a man in love with a beautiful idea, with a filing system for excuses built into it.

And one part of his legacy is solid gold: the observations. His Crohamhurst record is one of the longest high quality rain gauge series in Queensland, and without it the very test that undoes his theory could not exist. The forecasts failed. The measuring, the daily discipline of writing down the number in the gauge for sixty years, that outlasted everything. There is a lesson in that for anyone who runs a weather station, me included. He was chasing the wrong mechanism, I am fairly sure of that now, but he was not chasing nothing. The instinct that the sky was worth watching that closely turned out to be dead right, even where the planets were not the part doing the work.

Well, does the sky drive the weather or not?

Here is where I need to be careful, because the easy conclusion from all this is the wrong one. The numbers do not say astronomy has nothing to do with weather. Astronomy is most of weather. The seasons are nothing else than orbital geometry, the tilt of the planet doing a lap of the sun, and they are the strongest weather signal that exists. Day and night is astronomy. The moon raises a measurable tide in the atmosphere itself, about a tenth of a hectopascal, tiny but real, sitting in the barometer records if you stack enough of them. The solar cycle leaves detectable fingerprints in the upper atmosphere, and on the scale of ice ages the slow wobbles of Earth's orbit run the whole show. Anyone who says the sky does not touch the weather has not thought about it for five minutes.

So the question was never whether the sky matters. The question is narrower and harder: do the positions of the outer planets add anything to a rainfall forecast for this valley beyond what the calendar and the oceans already tell you? That is the only version of the question Jones's theory actually needs, and on 186 years of local records the answer keeps coming back no.

Where that leaves me, and the forecasts

Time to put my own cards on the table. I am a fan of Inigo Jones. I did not run these tests to knock the man down, I ran them because I take him seriously, and testing something properly is the most respect you can pay it. When heart and the data disagree, the data gets the vote. Saying that out loud, after the result went against the side I was barracking for, is not weakness. It is the whole difference between forecasting and fortune telling. Jones published his own failures in his 1944 paper, so in a way I am only doing what he would have done with my dataset.

Science is allowed to change its mind, that is the entire point of it. Wragge tried cycles, Jones built a life on them, the Bureau went to physics and computers, and I sit on a hill with a weather station testing the lot against 186 years of rain. Each step only happened because someone was willing to put a loved idea on the bench and publish what came out. My opinion moved this winter. It moved because I did the work, not because anyone shouted louder, and it can move back the moment the numbers say so.

I wrote in the first version of this piece that our long range outlooks do not use planetary cycles, and I left it there like a full stop. It should have been a comma. The 30 day and the 5 year outlooks are staying, and they run on the things that measurably work here, local climatology and the state of the Pacific and Indian oceans. But the Jones machinery stays too, running beside them as a labelled experiment, and every July the whole test suite runs itself again on the freshest records. That is my standing wager: the day any of his cycles, and the Jupiter and Saturn angle is the one I am watching hardest, earns its way in, it goes into the forecast at exactly the weight it earns, and you will read about it here the same week. His 107 year cycle, the one he mentioned once in 1944 and never explained, I still wonder abotu. Wondering is allowed. Wondering with a test bench attached is even better.

One more thing on the list. I trade commodities using W.D. Gann's methods, and Gann drank from the same planetary well as Jones did, same era, same fascination with orbits and anniversaries. Sooner or later his weather ideas go on the same bench, same rules, pass mark written down before the run. If they fail, you read it here. If they hold, you read it here first. The gauge does not care what Jupiter is doing, but I will keep asking it.

If you want to check any of this yourself, the National Library has his 1944 paper "Long Range Weather Forecasting" free online, Trove has his newspaper columns from the 1920s to the 1950s, and the Bureau's Climate Data Online will give you the same Crohamhurst daily record I used. Control my numbers, I would welcome it.