Having worked in data science you just described all of science. It's extremely difficult to get good data so you can build decent models. Weather forecasting abd climate science will always be wrong.
Fuck, this reminded me of a PHD student I tried to help with their thesis project. It was apparently too hard to get good data for their proposed model (lazy & would take time). So, find any old data and jam it into their model and voila, thesis done. I stopped helping before that point.
It’s honestly not his fault either. Science is publish or die these days. Taking your time to craft an excellent paper is rewarded less than the shotgun approach of firing off like 15 that barely meet the most basic requirements
Dr. Mototaka Nakamura (nasa jpl, mit, international pacific research center, etc) wrote a book a year or two ago called “confessions of a climate scientist” that basically explains how climate modeling is complete junk science and that there are very many important aspects to the climate that we barely understand at all and don’t even include in modeling. Most of the book is in Japanese, but he goes over the important stuff in English. Think it costs $1 on kindle, well worth the read.
It seems like it ought to be possible, but Navier-Stokes shows that you simply cannot.
If you don't understand differential equations and how to solve them, and you don't know what the Navier-Stokes equation is, then you won't be able to fully grasp the futility of weather forecasting and climate science.
Smoke predictions in San Francisco have reported with a footnote stating that due to limitations of weather prediction models they cannot accurately predict whether smoke would remain elevated above the city, or come down to ground level.
It makes sense from consideration of model development that only the most important parameter features would be included. Important meaning shows a statistically significant impact on what the weather would be on a given day in the future. Smoke is a rare event thus there is less data available to dial in on. Location and severity of the blaze need to be considered.
A rapid development team could cobble together a model enhancement, but the accuracy would be pretty low improving over time.
In short, shitty models, shitty data in the shitty models, shitty people putting shitty data in shitty models. When it's shitty across the board you don't even get a footnote explaining why predictions may be inaccurate.
Shitty models, shitty data in the shitty models, shitty people putting shitty data in shitty models.
“Reality must take precedence over public relations for nature cannot be fooled.”
Having worked in data science you just described all of science. It's extremely difficult to get good data so you can build decent models. Weather forecasting abd climate science will always be wrong.
Fuck, this reminded me of a PHD student I tried to help with their thesis project. It was apparently too hard to get good data for their proposed model (lazy & would take time). So, find any old data and jam it into their model and voila, thesis done. I stopped helping before that point.
It’s honestly not his fault either. Science is publish or die these days. Taking your time to craft an excellent paper is rewarded less than the shotgun approach of firing off like 15 that barely meet the most basic requirements
Ha yes this is true. The field was computer vision and AI, which is evolving so rapidly you'd see many papers published every week.
Dr. Mototaka Nakamura (nasa jpl, mit, international pacific research center, etc) wrote a book a year or two ago called “confessions of a climate scientist” that basically explains how climate modeling is complete junk science and that there are very many important aspects to the climate that we barely understand at all and don’t even include in modeling. Most of the book is in Japanese, but he goes over the important stuff in English. Think it costs $1 on kindle, well worth the read.
No they won’t, we already know that all things interact with some convergent plane. We know that all changes cause a propagation of interactions.
It ought to be possible to analyze the entire atmosphere as well as the external forces acting on it.
Gravoscope WIP.
It seems like it ought to be possible, but Navier-Stokes shows that you simply cannot.
If you don't understand differential equations and how to solve them, and you don't know what the Navier-Stokes equation is, then you won't be able to fully grasp the futility of weather forecasting and climate science.
So basically energy can’t be predictable enough in any space because from any point it will never be consistent?
My imagination figures that just because you’re inside the fluid doesn’t mean you can’t view the whole container.
I’m saving this comment for later because it captures the idiocy perfectly
Smoke predictions in San Francisco have reported with a footnote stating that due to limitations of weather prediction models they cannot accurately predict whether smoke would remain elevated above the city, or come down to ground level.
It makes sense from consideration of model development that only the most important parameter features would be included. Important meaning shows a statistically significant impact on what the weather would be on a given day in the future. Smoke is a rare event thus there is less data available to dial in on. Location and severity of the blaze need to be considered.
A rapid development team could cobble together a model enhancement, but the accuracy would be pretty low improving over time.
In short, shitty models, shitty data in the shitty models, shitty people putting shitty data in shitty models. When it's shitty across the board you don't even get a footnote explaining why predictions may be inaccurate.
GIGO