Thirty-three Julys in Huron County show no rising storm signal

The number, and what kind of number it is

Start with the question as asked: how much lightning hits Huron County in July? The honest answer begins with an admission. Nobody publishes that figure. Environment and Climate Change Canada operates the Canadian Lightning Detection Network and does publish lightning density data, but the open feed is a real-time product — ten-minute rasters, retained for about a day. There is no public archive you can query for "Huron County, July."

So the figure below is derived, and it is worth being precise about what that means. We take published flash-density climatology for this part of southwestern Ontario, multiply by the county's land area of — km², and apply July's share of the year's activity. That yields roughly — cloud-to-ground flashes over the county in an average July — on the order of — a day — inside a plausible range of — to —.

Why so much lightning here? Lake Huron. Through summer the lake breeze sets up a convergence zone inland of the shoreline, and storms crossing the lake organise as they come ashore. It is the same mechanism that produced the county's costliest single weather event: the F3 tornado that struck Goderich on 21 August 2011, which began as a waterspout over the lake, came ashore at the harbour and tracked through the town — one death, 37 injuries, and roughly $75 million in insured damage against about $130 million in total property loss. That was also the year the Forest record was set, 75 kilometres down the same shoreline.

Then the harder question

An insurer does not really want the July flash count. It wants to know whether the number is going up. That question cannot be answered from lightning data at all, because the archive does not exist — so we went at it from underneath, through the conditions that produce the lightning.

Thunderstorms need moisture, instability and lift. The most reliable single gauge of the first two at a surface station is dew point, which measures how much water the air actually holds. Relative humidity will not do: it moves with temperature, so in a warming record it will happily show you a drying trend that is really just a warming one. Dew point has no such problem.

There is exactly one long-running station inside Huron County — —, climate ID —, on the shoreline. We pulled every hourly and daily observation it has recorded for July from — to — — — Julys with a complete hourly record, and — of those with a complete daily rainfall record as well.

The scatter is the first finding. A single hot, humid July tells you nothing here: the record contains Julys with almost no humid hours at all and Julys with several hundred, sometimes back to back. Any analysis that samples three or four years from this record and draws a line between them will produce a confident trend in whichever direction the sampling happened to favour. We know, because our own first pass at this did exactly that.

Across the full record the verdict is —. The same holds at the more extreme 22 °C threshold: —, moving from — hours per July in the first decade to — in the last.

Warmer air, no extra water

Put temperature and dew point on the same axis — they share a unit, so this is a fair comparison rather than a trick of scaling — and the shape of the change becomes visible.

Between the first and last decade of the record, July mean temperature moved from — °C in — to — °C in —, and hours at or above 30 °C went from — to — per July. Mean dew point over the same windows went from — to — °C.

Both of those are decade averages, and it needs saying plainly that neither passes a significance test on this record: temperature shows —, and dew point shows —. A warming direction here is consistent with what is happening regionally, but this single station over this length of record cannot establish it. What the station can say is narrower and more useful: whatever warming is under way is not being matched by added moisture.

That matters because moisture, not heat, is the binding constraint on convection. Warmer air can hold more water, and the standard expectation is that it will. Here it has not — and the one measure in this entire analysis that does clear the significance bar points the same way. The peak July dew point, the single most humid hour of each month, shows — at —. The most convectively favourable hour of a Huron County July is, on this record, slightly less favourable than it was thirty years ago. Read the limits below before leaning on that one: it is also the result most likely to be an instrument artefact.

Rainfall tells a compatible story, and a more operationally useful one.

The pattern in the decade means is fewer wet days, an unchanged number of heavy ones, and less rain overall — precipitation concentrating, with the drizzle thinning out while the downpours hold. It is a coherent story and it is the one an underwriter would want to hear about. It is also, on the significance test, — for wet days and — for heavy ones. — Julys of rainfall counts are simply not enough observations to resolve a change of this size.

What survives is the negative result, and for the heavy-rain tier it is the one that matters: the frequency of claim-generating rainfall days in a Huron County July has — over three decades. It has not fallen either. It is, as far as this record can tell, the same as it was.

What this means for a P&C book

The temptation, given a rising loss environment, is to attribute it to locally intensifying weather. In Huron County in July, the local record does not support that. Losses are rising across Ontario — the Insurance Bureau of Canada has been clear that annual insured severe-weather losses now routinely exceed a billion dollars where two decades ago they seldom passed half that — but the driver visible in this county's own observations is not more storm days. Three things follow.

• Do not price a local convective trend you cannot measure. Thirty-three years of in-county data show no detectable increase in convective fuel. A rate assumption that bakes in rising local storm frequency for this geography is, on this evidence, unsupported. It may still be right — see the limits below — but it is not demonstrated, and the difference matters when you have to defend a filing.
• The exposure story is doing the work. Rising losses in a flat-hazard geography point at what is insured rather than at what the sky is doing: replacement costs, building values, the scale and specialisation of agricultural structures, and the density of high-value equipment on farms that were far less capital-intensive in 1994. Huron County is among Ontario's most productive agricultural counties, and the value at risk per hectare has moved a great deal further than the dew point has.
• Tail risk is not the same as trend risk. A geography can have a flat hazard trend and still be badly exposed to a single event. The county's benchmark loss came from one storm on one afternoon, and nothing in this analysis reduces the probability of the next one. Frequency assumptions and severity assumptions should be argued separately here; the evidence bears on the first and is silent on the second.

The lake is the reason for both halves. It suppresses nothing about the tail — the shoreline convergence that makes this corridor Canada's lightning capital is still there every summer — while the county's thermodynamic record simply has not moved much in a single consistent direction over the period we can observe.

What this cannot tell you

Six limits, because they bound every claim above.

• The flash figure is an estimate, not a count. It is arithmetic on a published climatology, not a tally of detected strikes. It carries no information about any particular July, and it should never be quoted without its range.
• No detectable trend is not the same as no trend. Thirty-three years is a short record for climate detection, and these series are noisy enough that a real trend of modest size would not clear the significance bar. This analysis rules out a large, obvious local increase in convective fuel. It does not rule out a small one, and it is not evidence against regional or global projections built on far longer records and physical modelling.
• One station is standing in for a county. Huron County covers — km². The only long-running station in it sits on the lakeshore, where the lake moderates temperature and humidity. Inland townships very likely behave differently, and the lake-breeze convergence that drives convection here is a spatial feature that a single point cannot resolve.
• Instrument homogeneity is unverified. A declining peak-dew-point trend at an automated station is exactly what a humidity-sensor change or drift would also look like. We have not been able to rule that out from the published record, and it is the single result here most likely to be an artefact. Treat the one statistically significant finding in this paper as the least trustworthy one.
• Dew point is a proxy, not a storm count. It captures moisture, not lift or shear. A July can deliver a damaging storm without a notably humid month, and the county's worst event did not require a record dew point to happen.
• Nothing here is actuarial, financial or insurance advice. It is a description of a public observational record and an explicitly derived estimate, produced for discussion. It is not a rate indication and it does not speak to any individual risk.