Know your data!
It’s the first rule of any engineering project you undertake. Mistakes often happen at the base level of data collection, errors in the early stages propagate through a project once the data is accepted, the effects of which are often misunderstood or unquantified. Nobody is immune to this of course, it’s just a point I want to make. From an engineering standpoint, it means heavy verification, often multiple times until you’re sure that your answer is reasonable. In my world if I’m wrong, things don’t work and big money is flushed down the toilet. In climatology if they are wrong and it is pointed out, there is little recourse or concession of the mistake. It would just be another messed up science if the politicians and polyscienticians weren’t forcing ridiculous policies down our throats, but that’s another story.
Jeff C found a site on the NSIDC where they provide AVHRR satellite data from 1982 to 2004. This data is monthly data which means daily cloud masking and error correction have already occurred. The masking is different than Steig used in his Antarctic paper and it may have had a substantial impact but at least we have a point of reference. There are several questions to answer, are 3 pc’s enough to represent the data, are the trends similar to the presented trends, are there other problems in the data.
Honestly, I’ve spent many hours now looking at the data and while I don’t have the answers to all these questions, there will be some interesting things over the next week or two. I really need to send Christmas cards to the team for all the entertainment they provide.
Jeff C deserves enormous credit for accessing and processing the data, we did work on it together but there were big hours spent in analysis and verification. Anyway, instead of getting too far ahead of myself and into some of the incomplete but still interesting finds I’ve decided to catch my readers up.
Below is a movie of the skin surface temperature of the Antarctic. Watch the video a couple of times and you may see some interesting features.
As a second confirmation of the movie above, I plotted the standard deviation of each gridcell so everyone can see cells that change the most and I think more importantly, the least!
The edges of the antarctic satellite data are surrounded by pixels which don’t change skin temp. You can see in the movie they stay red despite the rest of the continent turning blue. These pixels appear to represent contamination of land data with ocean surface temperatures. This data is nearly exactly the same grid points as Steig used yet clearly from the video and the SD plot, ocean pixels are included as surface data. How does Steig expect the ocean to change temperature with monthly air?
The peninsula is one of the worst areas after watching the movie, would you use that portion of the grid to establish continental trends? Would you allow the high density of sites comprising that data to account for over 25% of the reconstruction? Or would you realize, these stations are nothing more than ocean temps and work instead with the inland data?
There are a few months without data and a couple more where the collection resulted in a helical star pattern, the removal of which is at this point unknown. Those who have been keeping up will realize the mountain range in the autocorrelated PC data is nowhere near the influence it appeared to be. The PC trend simply locked onto the peninsula as the dividing point, the apparent coincidence with the antarctic mountains was therefore likely just by chance.
As far as the sea data contaminating the trends, I’ve already masked and removed the data simply by the fact that these pixels rarely drop below 10C, even in the coldest months. My point being that it wasn’t very hard to find ocean pixels and subsequently mask them out of the reconstruction.
Rather than me going on about how crazy it is to allow ocean pixels in a land reconstruction, I’m guessing some of you can help me out.
14 thoughts on “Evidence of Sea Contamination In Antarctic Sat Data”
And the Steig paper is what we can expect from peer-reviewed science? Will someone with the proper expertise please re-create this paper with accurate results and send it through peer-review in one of the “proper” climate related publications? Surely the truth will win out.
Thanks for the work here Jeff’s. Besides the obvious ocean pixels, there is an interesting symetry to the warming and cooling patterns as you move through the seasons. West Antarctica seems to just extend naturally from the geographic core of Antarctica rather than mountains as boundary. As you move further out on the peninsula, the temp becomes more disconnected from inland and essentially ocean as you observe.
I imagine it is just a matter of time before you show us a movie of anomalies rather than temp and compare to the 3 PC movie.
Keep pluggin Jeff.
Steig’s paper really is starting to look like a pigs ear, with them getting the results they expected (or wanted, if you are cynical), so they never bothered to scrutinise the data or algorithms for problems.
Sad that the chances of Nature withdrawing that (front cover) paper are about zero…
I should say that if Nature does not with draw it, then it’s only going to confirm my (already strong) suspicion that climate science has become a Cargo Cult Science:
I would urge the haters out here to not jump overboard. While I agree that sea should not be included (even if the trim cuts a little land), the amount of “mis-trim” seems very tiny.
For instance, comments that the paper should be withdrawn, can not be supported based on a few pixels.
I agree with TCO, the pixels are a sign of sloppiness but may not have a huge affect on the reconstruction. Time will tell.
Chris is right though that we’ve found more weaknesses than I would have expected.
I’m just surprised that they didn’t clean these pixels up before continuing their calculations. It’s not like you have to be Sherlock Holmes to find the problem.
I would have to assume that the darker shade of blue in the SD plot would be at or near 0.0, while deeper shades of red are “larger” SD numbers. It is interesting to note, that if I am correct with my assumption above, and if I remember correctly that most of the AWS are relatively close to shore, then couldn’t one draw the simple conclusion that the Steig reconstruction is based more on the land/sea interface rather than actual land mass temperature fluctuations/trends?
It seems to me that the peer review process is simply check for grammatical errors and sign off. I am not trying to devalue the Jeff’s work, but the above is not that complex, tedious and painful, yes, but not complex. It is this work, and the work of others that scare the be-jesus out of the Steig’s et al. And they say this is too complex for the laymen to understand.
Another thing I would like to point out to TCO is that unless you analyze the distance between the land/sea interface and location of the land based AWS, there is no way to tell what type of impact it has on the final outcome. Like I said above, if I remember the pictorial of AWS locations correctly, the majority (66%) fall within the darker blue shaded area. If this is the case then, the impact will be significant.
TCO, I don’t think anyone is criticising the paper based on a single Air Vent post – more like a dozen posts (not that I counted them)…
And even then, I only said “it’s start to look like”. Of course, it would be very easy to tell *for sure* whether Steig got his calculations mostly right or not – he could release the full source code. (I do actually expect him to do so eventually, but most likely not any time soon.)
#8 Why Not
Not sure if you are confusing or not, AWS was a separate reconstruction used for validation of the sat reconstruction. Your point of the effect is even more pronounced with the sat reconstruction however, because the surface station data is used in this. Almost all of the surface stations are on the peninsula or the coast.
Thank you for correcting my own thoughts, as I was thinking about the sat reconstruction and the resultant trends.
Can you tell us if Eric and/or Micheal are taking the valid information generated here and elsewhere about their paper to heart?
#12 I doubt that TCO is allowed in such a liberal blog! LOL
Hey good work J&J!!!!!!
(I hope I had been some of the poking on cloud masking, IE RH, blah blah blah.)
what we ur uhm you need is a sub routine to remove the water temp noise from the data… like if temp = temp over time pic5 then trash. this would also remove some of the peninsula temp noise as well.
don’t get concerned for missing this data…. we want to know the Antarctic temps, not Antarctic ocean temps LOL…..
It would be so very rewarding for you and team,
to verify double dipping into cloud masking,
I hope and pray that it works out for you and us(readers), you work it out and I’ll E-mail around the truth !!!!!!!! go go team!!!!!!!
new name lol