/sci/ - Science & Math

[None !fATtRiPFAg No.10439123]

This image is 500 pixels by 300 pixels, meaning it is made up of 150,000 pixels.

If you take the average of all 150,000 pixels the resultant color is R:120 G:164 B:127 (a desaturated green).

How could you possibly give me this exact color by only measuring a small subset of pixels? Say you took 50,000 measurements, surely you would still be off by a few hues or shades.

This same principle applies to claims of global climate change. How could you possibly assert a global increase of 1 or 2°C by measuring a few thousand points on a 510.1 million km2 sphere?

[Anonymous No.10439148]

>>10439123
>if you can't guess the value exactly all your measurements and theories are false
like a toddler

[Anonymous No.10439152]

>>10439123

Your example is flawed. Measuring an average is not the same as measuring an average variation. If I know that the variation is constant across all pixels (i.e. all pixels brightness increase by the same amount for a given period of time), then I only need to measure 1 pixel. If I can assume that the variation is mostly constant, then probably a few pixels will give me a fair estimate.

Also, there's a way to estimate the average in your example image, and it's called downsampling. Moreover, you'll lose no information after downsampling if your sampling frequency is still bigger than two times the maximum spatial frequency of the image (Nyquist Theorem).

[Anonymous No.10439153]

These denialist sure are getting colorful with their delusions...

[Anonymous No.10439159]

>>10439123
I can't be bothered actually doing it, but I'm fairly certain I could take 1% of the pixels and reconstruct the rest via interpolation to a satisfying degree.

The rest of your argument is unmitigated horseshit. I wouldn't even know where to start explaining where you went wrong in your thinking. Let's start easy though: have a look at MOD11A1, a land surface temperature / emissivity dataset retrieved from daily satellite measurements of the entire fucking globe.

[Anonymous No.10439162]

>>10439153
>These denialist sure are getting colorful with their delusions...
This is a lesson for all future apes...if you're going to fling poo at us, at least give us a pretty picture. I, for one, salute this fresh dingus.

[Anonymous No.10439170]

>>10439153
He’s not denying science. He’s raising a legitimate point.

[Anonymous No.10439182]

>>10439123
Because the derivative are bounded, we can calculate the color not exactely, but with small error. Its called math fellow.

[None !fATtRiPFAg No.10439219]

>>10439152

What would lead to the assumption that you know variation is consistent across pixels?

>>10439159
What would you consider a satisfying degree? What is a satisfactory margin of error? MOD11A1 is an amazing resource, but only goes back to the year 2000

>>10439182
Again, how small is this error? I'm honestly not trying to troll or act as a denialist or conspiracy tard...genuinely interested in the answers here.

[Anonymous No.10439236]

>>10439219
>What would you consider a satisfying degree?
That depends highly on what number you want to get out of it. a) getting something like the *trend* in global average LST/SST does not require an awful lot of precision obviously (sqrt(N), you know?), and b) even though it 'only' goes back to 2000, it can essentially be used for cross-validation of surface measurement sites so we KNOW how accurate the estimation of a global LST/SST *trend* is if only inferred from in-situ sites.

[Anonymous No.10439254]

Error tends to be minimised when there are a lot of data points which are being monitored constantly.

[Anonymous No.10439271]

>>10439123
>babby's first montecarlo

[Anonymous No.10439272]

>>10439219

>Select 100 random pixels as sample
>Measure those 100 pixels at 100 random time instants
>Save measured values as a 100x100 matrix: each row is a time instant, each column is a pixel
>Create new 99x100 matrix, where each row is equal to the same row from previous matrix minus the next row from previous matrix
>Compute variance for each row
>If the variance is close to zero across all rows, it is safe to say that the variation is consistent

Correct me if I'm wrong, I'm not an expert in statistics (nor in anything).

[Anonymous No.10439288]

>>10439271
Monte Carlo methods require large amounts of random numbers.

What we have here is a very small sample at non-random points.

[Anonymous No.10439320]

>>10439123
Noone is saying climate change is absolute truth (yeah ok some are, but they're wrong), but it's a very empirically tested and probable theory.

[Anonymous No.10439555]

>>10439170
>Let this fellow retard speak. As a retard, I'd like to feel not-retarded for the duration of this thread. As we all know, if someone agrees with my retardation, it's right. Despite all evidence, it's right.
Don't respond. I'm not actually addressing you directly, and only pointing you out as an example.

[Anonymous No.10440876]

>>10439123
By sampling enough points you reduce the error to a negligible amount. What is the issue?

[Anonymous No.10440964]

>>10439170
>He brings up a good point. Statistics is bullshit if you can use it to predict things with 100% accuracy. Therefore, empirical studies are completely useless. Praise Allah.

This is how stupid you sound.