I browse Windy but don't rely on any of those 4 weather forecasting models: I take the median of predicted temperatures and rainfalls instead. Also, I predict rain only if the median exceeds 1mm, and if it's below that threshold but at least 3 models predict (some) rainfall I predict drizzle. Which is the same approach I had at my previous job, using data of doubtful quality to adjust Holt-Winters and Box-Jenkins models in order to forecast drug sales for Big Pharma.

Kaggle by the way began to demand users engaged on modelling competitions to make PDFs explaining their methodologies after learning some cheaters would just combine results from other competitors.

P.S. - Don't average results from different models unless you are really, really sure of what you're doing. Many times the models take turns on which one will output garbage, and you don't want garbage contaminating your average. By switching to median you avoid the crap they sometimes spit altogether - not to mention it's so simple you don't even need to write numbers on paper or use a pocket calculator.

Here goes:

During my dissertation, I was lookig for information on the emissiom of 172nm scintillation light in mixtures of gaseous Xe and CO2 (95:5% - 98:2%), with results being difficult to come by. I found a collaborator who had tested this at lower CO2 concentrations (0-0.5%), but nothing else, no predictions or generalizable applications. Not knowing the optimal search engine terms or what textbook to look in for rules governing gaseous light emission, I ended up looking in fluorescence chemistry papers (my previous field of study) which had something called the Stern-Volmer relation for different concentrations of quenchant in a fluorescent solution. I figured gas scintillation queching was probably similar to liquid fluorescence quenching, but the standard relation didn't quite fit below 10% additive.

I dug around more and found a modification of this relation for diffusion-limited quenching of fluorescent solutions (the same limitation imposed in gas mixtures, quenching due to random Brownian collisions) that employed an exponential term, allowing for a smoother curve down to low additive concentrations. This perfectly matched the available data and allowed me to model the predicted behavior. I discussed this with the one member of my committee who was available, an organic chemist (my PI was on vacation, everyone else was sick, and my dissertation defense was in 2 weeks). He said my reasoning and math for using this formula made sense and gave me a thumbs up to include this analysis. When my PI came back from holiday, he asked me why I didn't use some equation generally used in the field, or even just a generic exponential fit. I was ignorant of his suggestion, but it provided the same general formulation as Stern-Volmer, though Stern-Volmer was more rigorously derived mathematically.

Mixing fields is super cool and can allow a much deeper understanding of the underlying principles, as opposed to limiting yourself to one branch of science. While my PI's recommendation would have given approximately the same answer, understanding and applying Stern-Volmer allowed me to really dig at the principles at play and generate a more accurate and in-depth model, which I managed to write up and defend at the 11th hour.

In vulcano seismology there is this fun little thing called a tremor and its really annoying but also really mysterious as no one knows where it is coming from or what cases it. I've had multiple people try to explain it or I was listenin to talks about it and I have yet to hear solidly overlapping theories. Also not only does the signal look different at every single opportunity (aka every vulcano) that you to look at it, it also hides within a frequency range that is mostly overlapped with random background signal. So to look at it you need to do analzye your seismometers for a directional eigenvalue (not sure if its the correct word or even the only what its just what I'm doing. I'm normally German speaking but what we do is look at the seismometers and whether or nor all or most of the signals are comming from the same direction) so that you can even detect it, meanin just to look at it you already need some statistics.

Not really sure where I'm going with this I just find this concept really really and I'm just once again baffled that we simply do not know about things in science.

maybe this wording works on a certain kind of voter because of the "fuck you I got mine" attitude, they probably think that if they were the scientist they would reap the benefits for themselves

Not a scientist. I have a litany of complex topics that I just can't really talk to anyone about. I'm a big computer networking nerd, and once upon a time, when I didn't know what I didn't know, I was curious what computer networking really entailed... It seemed dead simple, you connect things to a switch, connect that switch to the internet router, not much more.

Then I learned about VLANs, which are cool but it seemed like unnecessary complexity. Then I learned about Routing and L3 switching, and routing protocols and..... Holy shit, how deep is this?

Now-a-days, I want to have conversations about the merits of one routing protocol over another in various contexts, and see/build a spine and leaf network infrastructure that's nearly infinitely scalable.

I want to explore the nuance of IP unnumbered routing. I can't find anyone who will chat about it on a level that's close to my understanding, either someone knows way more than I do, or they know way less.

IP unnumbered routing is a way of connecting devices without setting an IP on the interface that is being routed to/from. The other end uses the routing protocol on top of layer 2, and while the two might have a router ID, often in the form of an IP address, the interface that is connecting the two has no IP. It's basically advanced point to point protocol (PPP) that breaks away from traditional TCP/IP routing in ways that people who have never used anything besides TCP/IP can't really comprehend. The two "IP addresses" (actually router IDs) in play can have nothing in common. Traditional TCP/IP requires that two IPs share a subnet. In routing, this is typically a /30 for IPv4, and the two IPs are adjacent to eachother, eg, 10.254.123.1 and 10.254.123.2 IP unnumbered can have 10.254.123.2 talking directly with 172.30.88.207, with no layer 3 interfaces in-between.

It's really fascinating and interesting and I've been trying to find a good model or guide to help me learn this better, but I keep ending up at dead ends, and I have nobody to talk to about it.

This is why the "secret scientists don't want you to know" always turns out to be some pseudoscience bs that at best is misinformation and at worst is actively harming people. So, yes, they are things scientists don't want you to know.