, I’m biologically required to endure the identical loop of small discuss yearly: “It’s boiling, isn’t it? Means hotter than 2020,” or the basic, “Again in my day, we really had 4 seasons, not simply ‘Pre-Oven’ and ‘Deep Fryer.’”
Truthfully, I’m tempted to nod alongside and complain too, however I’ve the reminiscence of a goldfish and a mind that calls for chilly, onerous details earlier than becoming a member of a rant. Since I can’t keep in mind if final July was “sweaty” or “molten,” I’d like to have some precise information to again up my grumbling.
I work at icCube. It’s mainly an expert sin for me to get right into a data-driven argument with out bringing enterprise-level tooling to a back-of-the-napkin debate.
On the subsequent apéro, when somebody begins reminiscing about how “1976 was the true scorcher,” I shouldn’t simply be nodding politely whereas nursing my pastis. I needs to be whipping out a high-performance, pixel-perfect dashboard that visualizes their nostalgia proper into oblivion. If I can’t use multi-dimensional evaluation to show that our sweat glands are working more durable than they did within the seventies, then what am I even doing with my life?
Whereas this journey started as a quest to settle an area argument within the South of France, this submit goes past the local weather debate. It serves as a blueprint for a basic information problem: the best way to architect a high-performance analytical system able to making sense of many years of historic information relevant to any area requiring historic vs. present benchmarking.
The Battle Plan
Right here is the plan mapping out our tactical strike in opposition to obscure nostalgia and anecdotal proof:
- Scouting the Intel: Searching down the uncooked numbers as a result of “it feels sizzling” isn’t a metric, and we’d like the high-octane stuff.
- Constructing the Conflict Room: Architecting a construction sturdy sufficient to carry many years of heatwaves with out breaking a sweat.
- The Analytical Sledgehammer: Deploying the heavy-duty logic required to show uncooked information into simple, nostalgia-incinerating proof.
- The Visible “I Instructed You So”: Designing the pixel-perfect dashboard to finish any apéro argument in three seconds flat.
- Put up-Victory Lap: Now that we’ve conquered the local weather debate, what different home myths we could incinerate with information?
Scouting the Intel
Information is central to our mission. Subsequently, we have to safe correct, high-fidelity historic temperature data from France.
Méteo-France, the nationwide meteorological and climatological service, is a public institution of the State. It makes accessible to all customers the information produced as a part of its public service missions in its public information portal: datagouv.fr. God bless public information portals. Whereas half the world’s information is locked behind paywalls and registration varieties that ask on your blood sort, France simply… fingers it over. Liberté, égalité, température.
The info used on this submit is made accessible underneath the Open License 2.0.
The Observations
Climatological (every day/hourly) information from all metropolitan and abroad climate stations since their opening, for all accessible parameters. The info have undergone climatological management: www.
The Climate Stations
Traits of meteorological climate stations in metropolitan France and abroad territories in operation: www.
Early Evaluation & Transformations
Being like Saint-Thomas, I wish to see and overview a bit on my own the precise information to get first an excellent understanding and carry out a little bit of sanity checks earlier than drawing any conclusions afterward.
To maintain issues clear, I’ve been extracting uncooked temperature information from the pile of observations we have now. Being an unrepentant Java geek, I’ve constructed a set of courses for this mission and tossed them right into a Github mission. Be happy to tear by the code, re-use it as a lot as you want.
I’m not going to bore you with a dry lecture on the information proper now. That may be like serving a lukewarm rosé, completely legal, presumably unlawful in sure Provençal villages.
I’ll be diving into the gritty particulars when wanted.
Constructing the Conflict Room
If we’re going to settle these terrace debates as soon as and for all, we are able to’t simply flip up with a spreadsheet and a dream. We want an OLAP schema; a construction so sturdy it makes the native historic stone masonry look flimsy. We’re protecting it lean for this particular combat, however belief me, it’s constructed to scale when the subsequent “mildest winter ever” argument inevitably breaks out.
Let’s break down the structure.
The Dimensions
- Stations: It lets us pinpoint the precise climate station within the France map as a result of saying “someplace within the South” received’t minimize it. We want coordinates, names, the works.
- Time/Calendar: The same old suspects: years, months, days. Boring? Certain. Important for proving your neighbor’s reminiscence is rubbish? Completely. We’re tossing in Months and Days of Month to gasoline a calendar widget that can let me level at any particular date and say: “See? July 1st, 2025 was an absolute hellscape”. Precision is vital if you’re ruining somebody’s nostalgic buzz.
The Info (aka., Measures)
- Temperatures: The “Holy Trinity” of information factors—Common, Most, and Minimal. That is the first enter for our “Deep Fryer” versus “Pre-Oven” evaluation.
The complete schema definition is parked over within the GitHub mission with the supply code, prepared for if you’re feeling notably vengeful.
The Dice
The ultimate end result? A loaded schema containing greater than 500 million rows of French temperature information stretching again to 1780. Is it absolute overkill for an informal chat over olives? In fact it’s. That’s the purpose.
It provides us a playground to hack into different metrics afterward. However let’s save these for after we actually wish to make folks remorse mentioning the climate within the first place.
The Analytical Sledgehammer
Time to construct the question that can shut down the subsequent apéro debate in three seconds flat.
To chop by the noise, I’m utilizing the MDX language: a question language particularly designed for this type of multi-dimensional heavy lifting. To show that we’re certainly residing in a “Deep Fryer”, I’m going to match every day’s temperature in opposition to a historic reference interval.
Should you don’t communicate MDX, skip to the beautiful image. The question mainly tells the information engine to seek out the typical “regular” for this particular day over 30 years and subtract it from right now’s temperature.
First, the reference interval (aka., our regular baseline) is outlined as a static set utilizing the vary operator (e.g., 1991 – 2000):
with
static set [Period] as {
[Time].[Time].[Year].[1991] : [Time].[Time].[Year].[2020]
}“Why 30 years?” As a result of that’s what climatologists and the World Meteorological Group determined counts as “regular” earlier than the planet began experimenting with new thermostat settings. It’s the gold normal for a “climatological regular”; lengthy sufficient to easy out the bizarre years, quick sufficient to nonetheless keep in mind what “regular” used to really feel like.”
The every day common temperature is outlined as the typical of the utmost and minimal temperatures of the day. I’ve experimented with hourly averages; the outcomes are practically equivalent. So let’s keep on with this easy and effectively accepted definition:
with
[T_Avg_Daily] as
( [Measures].[Temperature (max.)] + [Measures].[Temperature (min.)] ) / 2
, FORMAT_STRING=".#"Now, we have to know what the temperature ought to be. We calculate the typical of these every day temperatures aggregated over our reference interval:
with
[T_Avg_Period] as
avg( [Period], [T_Avg_Daily] )
, FORMAT_STRING=".#"Lastly, we calculate the distinction, measuring precisely how a lot hotter (or colder) it’s right now in comparison with my previous years. This delta worth places a exact quantity on our collective sweat:
with
[T_Avg_Diff] as
IIF( isEmpty( [T_Avg_Daily] ), null, [T_Avg_Daily] - [T_Avg_Period] )Placing all collectively, right here is MDX question that compares the 2025 every day temperatures in Uzès in opposition to the report:
with
static set [Period] as {
[Time].[Time].[Year].[1991] : [Time].[Time].[Year].[2020]
}
[T_Avg_Daily] as
( [Measures].[Temperature (max.)] + [Measures].[Temperature (min.)] ) / 2
, FORMAT_STRING=".#"
[T_Avg_Period] as
avg( [Period], [T_Avg_Daily] )
, FORMAT_STRING=".#"
[T_Avg_Diff] as
IIF( isEmpty( [T_Avg_Daily] ), null, [T_Avg_Daily] - [T_Avg_Period] )
choose
[Time].[Months].[Months] on 0
[Time].[Days of Months].[Days of Months] on 1
from [Observations]
the place [T_Avg_Diff]
filterby [Time].[Time].[Year].&[2025-01-01T00:00:00.000]
filterby [Station].[Station].[Name].&[30189001] -- Nîmes CourbessacThe attentive reader will discover I’ve swapped the native Uzès station for the Nîmes-Courbessac station. Why? As a result of I would like that candy, candy historic information to gasoline my “again in my day” comparisons, and Nîmes merely has an extended reminiscence. It’s proper subsequent door, so the temperatures are just about equivalent although, if I’m being trustworthy, Nîmes normally runs a bit hotter.
Within the subsequent part, I’ll present you the best way to splash some shade on these values so you may spot the heatwaves at a look.
The Visible “I Instructed You So”
So it’s time to cease gazing uncooked code and truly construct a visible for that MDX end result. My plan? Cram your complete yr right into a single 2D grid, as a result of taking a look at a scrollable record of 365 dates is a one-way ticket to a migraine.
The setup is easy: months throughout the horizontal axis, days of the month on the vertical. Every cell represents the temperature delta, that’s, the (Celsius levels) distinction between 2025 and our reference interval. To make it “idiot-proof” for the subsequent time I’m three pastis deep, I’ve utilized a warmth map: the warmer the day was in comparison with the previous, the redder the cell; the colder, the bluer.
Full disclosure: I’m not a “visible man.” My aesthetic desire normally begins and ends with “does the question return in underneath 50 milliseconds?” However even with my lack of creative aptitude, the information speaks for itself.
One look at this grid and it’s painfully clear: 2025 isn’t simply “a bit delicate.” It’s a sea of offended crimson that proves our reference interval belongs to a world that was considerably much less “pre-oven.” If this doesn’t shut down the “again in my day” crowd on the subsequent apéro, nothing will.
My Nostalgia Previous Years (1980-2000)
I’m recalibrating the baseline to match the years of my youth. By shifting the reference interval to these “glory days,” it seems my mind wasn’t exaggerating; the information confirms a transparent shift from the manageable summers of the previous to this new depth.
No surprise the lavender is pressured.
#Days > 35
I began getting curious; was it simply my creativeness, or is the “oven” setting on this planet really rushing up? I made a decision on a fast train: counting what number of days per yr the thermometer hits or cruises previous the 35°C mark.
To the shock of completely no person, the information confirms the “pre-oven” section is shrinking, and the “deep fryer” period is formally taking on.
2003: When Summer time Grew to become a Tragedy
There, within the information, a stark peak that towers above all others. The summer season of 2003. Fifteen thousand folks didn’t survive these relentless days above 35°C. In France alone. A nation that hadn’t understood how lethal warmth could possibly be. The chart doesn’t seize the empty chairs at dinner tables that autumn, the households endlessly modified, the belief that got here too late.
These charts don’t show world local weather change by itself; they merely show native lived actuality with rigor.
Put up-Victory Lap
And that’s the way you flip an informal sundown drink right into a data-driven interrogation.
We’ve formally unleashed the information and MDX to show that “it was cooler” isn’t only a senior citizen grumbling after one too many Ricards; it’s a verifiable truth. Is bringing a multi-dimensional heatmap to a social gathering the quickest strategy to lose buddies and cease getting invited to apéros? Most likely. However is the silence that follows a wonderfully executed “I informed you so” price it? Each single time.
Information received’t cease the warmth however it’ll hopefully cease the unhealthy arguments about it.
The “Mistral Insanity” Index
Now that the warmth is settled, I’m setting my sights on the legendary Mistral. In each village sq. from Valence to Marseille, there’s a sacred “Rule of three” that claims as soon as the Mistral begins, it should blow for 3, 6, or 9 days. It’s the type of native numerology that individuals defend with their lives.
I’m already prepping a brand new “Wind-Chill” schema to cross-reference hourly gust speeds with this calendar delusion. I wish to see if the wind really cares about multiples of three, or if it’s simply our brains looking for patterns within the chaos whereas our shutters are rattling.
Should you’ve loved watching me over-engineer an answer to an informal dialog, comply with my descent into analytical insanity over on Medium. We’re simply getting began.
