Unsurprisingly, I use science a lot in my daily life.
Science is the discipline of updating our beliefs by actively testing them using systematic observation, controlled experiment, and Bayesian inference.
This process creates knowledge. Nothing else does, although many other things can instill knowledge: reading and listening are particularly useful in this regard. My bias is toward reading, as I come from a literate culture, but non-literate cultures do a very good job of transmitting knowledge from person to person and generation to generation as well. They just use different techniques to do it: poems and stories are often used as mnemonic devices, although there is always a good deal of pure rote repetition too, at least as near as I can tell.
Unlike reading and listening, the processes of systematic observation and controlled experiment both often involve measurement, which is the precise numerical quantification of things. In the cases where no measurement is done, we still quantify, just less precisely. Categorization is a kind of quantification: if I say something is a sailboat, it has a 1 in the "is sailboat" column of my mental spreadsheet, and a 0 in the "is powerboat" column. Those zeros and ones are quantities. Likewise, binary (present/absent) claims are quantitative: "There were no dogs in the room" is a quantitative statement.
So are ordinal claims: "Power struggles within the NDP were a bigger factor than any interest in democratic principles during leadership race in 2022." Bigger, smaller, more, less: these are ordinals, like first, second, third. All of them are quantifiers.
And finally, rough estimates of magnitude are also quantitative: "She was going to beat the band!" is a quantitative statement, albeit one with rather large uncertainty and poor calibration.
In fact, there is no such thing as "qualitative" observation, if we mean by that “something that cannot be turned into a quantitative one”: even saying something like, "The forest air smelled nice" is a quantitative statement. It puts the qualities of the forest air on a specific axis--one that runs in the nasty to nice direction--and places this particular forest air on one side of it. Every quantification has an uncertainty associated with it, and what are normally thought of as "qualitative" statements are nothing but quantitative claims with very high uncertainty.
The difference between systematic or formal observation, on the one hand, and informal or unsystematic observation on the other, is that the latter tell us a great deal about who is doing the observing, and very little about what they are observing, and the former are precisely the opposite: if two people practice the disciplines of systematic observation, they will tend to report the same results, no matter who they are, because they are telling us about the world in a way that is as free as possible from their personal confirmation biases. This does not mean that people with different purposes or categories won’t disagree: it means they are describing different aspects of the world, or the world divided up in different ways.
Controlled experiment gets a lot of focus when talking about knowledge creation, but systematic observation is really where the action is. There are many different aspects to systematic observation, but the one that dominates everything else is the single rule:
Write it down.
If we don't record things fairly soon after observing them, we start relying on reconstruction to recreate them (what is sometimes called "memory", but is in fact nothing like what people mean by "memory".) Any good knowledge creator, from Jane Goodall to Enrico Fermi, wrote things down as they went, relying on their ability to observe and record, and leaving as little as possible to reconstruction.
If I had the power to impose some kind of Newspeak on people, I would use it to replace "remember" with "reconstruct", as the former simply does not accurately represent what is happening when we bring past events back to mind. This process is not one of passive recall of a fixed and more-or-less complete record, but rather is an active reconstruction based on a handful of clues which are easily accessible only because our biases tag them as important, not because they carry the highest amount of information with regard to the event in question.
But we don't have time to measure and write down everything, so we have to be selective about what we do measure. In my case, precipitation is something I measure.
Mrs Wonders and I get our drinking water mostly from rainfall. The downspouts on our gutters run into a water collection system that feeds three cisterns, which hold about 30,000 litres between them, which is almost enough to get us through a dry summer if we go into May with them full, which didn't happen in 2022, although thankfully 2023 is looking better... so far.
Our roof is of such a size that about 2 mm rainfall per day is required to keep us in water, and since the average rainfall year-round is almost 1200 mm we should have plenty of buffer.
In fact, it's a good deal tighter than that, for reasons that only become apparent when we start to measure our daily rainfall, which I did shortly after we moved in, almost five years ago now.
My measurement apparatus is very simple: it's a stainless steel spice shaker of some sort with the lid removed. It has straight sides, a flat bottom, and a sharpish corner between them. All of these are desirable properties: the straight sides mean that the depth of water is a good measure of the amount of rainfall. The flat bottom means that the depth is the same to a half millimeter or so wherever I measure it. And the sharpish angle between them means that after about one millimeter of accumulation there is no variation to speak of in the rate of change of depth with rainfall, and even in the first millimeter the change is negligibly small.
I'm aiming for sub-millimeter accuracy, and with this apparatus that's pretty easy to get. The only problem is that it can be a pain to read: I use a dipstick that consists of a fine threaded bolt, and the water clings to the threads and allows me to use a simple metal ruler to measure the depth to +/- 0.5 mm accuracy. I've done tests with the ruler in the water itself--which is a pain to read unless the cup is very full, which it sometimes is--and confirmed the quality of my regular system.
Every morning at around 8 AM I go out on the porch, dip the cup, hold the bolt up against the sky with the ruler running along it, and measure the water depth. On the rare occasions it snows, I use the ruler direct on the porch itself, and divide the value by ten to get the rainfall equivalent (the division is done in the spreadsheet I keep for the purpose, which has separate rain and snow columns.)
The biggest problem with this system is that I sometimes have to measure in the evening and empty the cup then, to prevent it overflowing if we have particularly heavy rainfall. That and because I'm generally wandering around the house barefoot first thing in the morning it's sometimes cold on my feet when I wander outside, especially if it's been snowing. And if it's still raining, I sometimes get a bit damp. Such is life.
Once I've got the number, I come inside, open up the laptop I keep downstairs, and enter the numbers in the spreadsheet for the current year. The ye olde dayes I would have kept a lab notebook by the door for the same purpose, but there's no point in transcribing if it can be avoided.
The spreadsheet file has a sheet for each year, which lists month, day-of-month, Julian day, rainfall, snowfall, and some columns for processing them into totals for each month. The "Julian day" as I use it is the count of days from the start of the year. Julian day is used in astronomy, and in fact starts 4713 BCE, but I accumulate data yearly and subtract off the starting Julian day for the year to keep the numbers easy to reason about: day 90 or day 180 or day 270 is a lot easier to reason about than day two-million-and-something.
The graph at the top of this post is the rainfall for 2023 (the histogram) compared to the rainfall for 2022 (the dots) and the average at the local airport (the line with points on it).
It's readily apparent that except for the odd month here and there things have been dry. Between June of 2022 and September of 2023, every single month was below average.
But what does this mean?
Well, climate change may be real, eh?
But it's not that simple. Chatting with others on my little island a few months ago, it became apparent that the rainfall variation over the island is large. Like a factor of two from spot to spot. We've had some big rains recently--notice how far above average October 2023 was--and there were days when 50 mm or more fell where I am. In other places it was below 30. In others it was above 70.
I've often said that elsewhere there are micro-climates. Here, every tree has its own weather. This isn't even a joke, really: I made that observation originally while looking at a tree in Desolation Sound that always had a wisp of cloud around it, while the rest of the area remained clear.
So local variability is one factor in why my rainfall measurements may be below the airport averages. I've not tried to compare my data with daily rainfall from the airport, although probably should. I live on the north end of Gabriola Island, which is quite close to Vancouver Island (we're 20 minutes away by ferry, slightly longer by kayak) and just at the point at which the low mountains of southern Vancouver Island become the much higher mountains of the north. The Sunshine Coast, north of Vancouver on the mainland shore, really does get a lot of sunshine because it's in the rain shadow of those mountains, and to some extent so are we.
The airport, on the other hand, is on Vancouver Island almost dead level with the southern end of my island, and therefore should be expected to get a fair bit more rain.
So is it climate change, or is it geography?
This is why long-term records of environmental conditions are vital to understanding our world. Who knows what's changing if we don't have records of it? Fortunately, in many cases we do have longer-term records, but Canada is particularly bad at this. Unfortunately that's starting to matter less as the rate and range of changes increase.
As well as geography, there is the question of: how good are the various instruments being used?
Well-trained, practicing, experimental and observational scientists are a suspicious lot. The motto of the Royal Society is "Nullius in verba", which translates to something like, "Don't trust the other bastard." Nor should we.
If my rain gauge has one flaw is that it's hard to read at low rainfall, and I was in one of our two local hardware stores recently and noticed a fancy conical rain gauge that I thought might help with this. Since the mouth is much larger than the bottom, the depth of water increases rapidly in the first few millimetres, resulting in a more precise reading. The marks on the cone--which is of clear acrylic--are spaced further apart near the bottom, slowly converging on the physical scale near the top.
Unfortunately, "more precise" does not mean "more accurate": the gauge is badly mis-calibrated. Because I'm not incompetent, I know that when you change over from one system of measurement to another--or one system of data processing to another, if anyone from BC Public Health is reading this--it is absolutely necessary to maintain the old system and publish both the old and the new for a sufficiently long period of time that the relation between them can be well-established over a wide range of values. So when I put the new gauge in service by screwing it to my porch railing in late November, I kept the old gauge in place beside it, and now I enter two numbers in the spreadsheet each morning when it rains.
I've only got a month's worth of data consisting of sixteen days of measurable rainfall so far, but the graph below gives a sense of it.
The fancy commercially manufactured rain gauge is off by about 40%, which makes me wonder how much of the variation in rainfall I see reported around the island is due to people using poorly manufactured instruments that suck.
I'm in the measurement business, and NIST-traceable standards take up an implausibly large part of my life, so I appreciate how hard it is to get accurate instrumentation, but there's really no excuse for getting something that's nothing but geometry as wrong as this rain gauge is. A NIST-traceable calibration is one that can be traced back to the National Institute of Standards and Technology in the US, which is the North American Mecca for measurement. Traceabilty happens via physical artifacts that get shipped around.
For example, in thermometry, a manufacturer might ship a standard thermometer to a NIST-certified lab to get it calibrated against a thermometer that comes straight from NIST itself. The manufacturer will then use that device to calibrate it's own thermometers, and it will ship them with a certificate that indicates how their calibration can be traced back to the national standards lab through the chain of instruments used.
That's a much fancier process than anything that has gone into my own measurements, but I will say I've used two different rulers at times to check my rain gauge, because you never measure anything just once, or in just one way, if you care about accuracy. And if you're measuring it all, you should probably care about accuracy.
I will, of course, be rechecking my own gauge to ensure that I've not been reading 40% low all this time, but given the checks I've already performed my prior is pretty heavily biased toward it being correct. Still: when a Bayesian gets new data that conflicts with their prior beliefs, they check all the "ideas that it would contradict if it was accurate" pretty carefully before declaring it bogus.
My plausibility for the idea, "My simple rain gauge is accurate" is somewhat lower now than it was a month ago. But my plausibility for the idea, "This piece of Chinese crap is badly mis-calibated" is much higher.
Trusting instruments made by slave labour in a genocidal dictatorship--not that anyone cares about genocide, or dictatorship--is probably not the best of all imaginable policies. I avoid Chinese goods as much as possible, and gave in to this thing in a moment of weakness, because it's such a clever design. Which should probably be a lesson to me.
If it matters, we should measure it, but what matters to whom is a political and moral question, and how we measure it is, as seen here, a technical problem of some depth. If it's possible to get as much variance as I'm seeing in "How much water fell from the sky in the past twenty-four hours?" just imagine how much variance there could be in "How many people died of covid in Canada last year?" Especially given the various private political corporations, like the BC NDP, that are dedicated to making that number as hard to determine as possible, and ensuring the few estimates we do have are biased to be as low as they can possibly make them.
Access to the data that could reasonably be collected and reasonably be used to guide public policy is a human right.
Access to the data is a human right.
Canada in this regard is pretty much a large inhomogeneous collection of human rights violations.
But data is only useful if you understand how it was collected, and trust the instruments used to measure it. That means governments have to do more than publish the data: they need to include metadata and procedure descriptions that make it clear what is being measured and how. Part of my job as a consulting scientist is to write precisely this sort of documentation, so I know it's not some mysterious and unsolved problem: it's a straightforward application of principles and skills that any undergrad in the sciences should have learned by third or forth year.
In the absence of data collected by the governments who waste our tax dollars on frivolities, we can sometimes collect, and share, our own. There are corporations that do this for weather data, including air quality, and a few open source projects that are engaged in similar initiatives. The latter are, I think, the best place to go looking for an understanding of our world.
Or failing that, with a bit of patience and discipline you can almost always come up with the equivalent of simple cup-and-dipstick device to give you some insight into the world around you.
This is long and talks about many things. Some of them I agree with. Some of them are dead wrong. The idea that "there is no such thing as "qualitative" observation" is just wrong. Qualitative interpretation of the world is how our bodies works. There is a way to quantify this quality, but it took millions of years to figure out how to do this, and still, almost no one is any good at it. That doesn't mean it's not a very powerful way to think about things, and learning to do it has given those who can it the power to do things like pollute the Earth with plastics and order institutions so they can produce the most money. But it's not the way that people's bodies sense the world around them. I will return later with more to say.
Very clear examples. Thanks!