Support the Arctic Sea Ice Forum and Blog

Author Topic: Home brew AMSR2 extent & area calculation  (Read 1974250 times)

etienne

  • Nilas ice
  • Posts: 2045
    • View Profile
    • About energy
  • Liked: 309
  • Likes Given: 23
Re: Home brew AMSR2 extent & area calculation
« Reply #2850 on: January 25, 2017, 07:39:42 PM »
Interesting, thank you!
Did you try to separate the trending in your graph in 2007?

Just curious.  :)
Hello,

I used 2005 based on the Worldbank CO2 emission graph.
http://data.worldbank.org/indicator/EN.ATM.CO2E.KT

I don't think 2007 would change much the picture because the trendline seems quite ok excepted for the 2016 data which could be the next step if the theory is ok. I don't have much time the next days to try it.

Thank you crandles for the links. Now that I think about it, the step story I remember concerns more temperature going up than sea ice extent going down. Here is a link to such a study.
https://www.skepticalscience.com/its-a-climate-shift-step-function-caused-by-natural-cycles.htm

Bye,

Etienne

Peter Ellis

  • Grease ice
  • Posts: 619
    • View Profile
  • Liked: 33
  • Likes Given: 14
Re: Home brew AMSR2 extent & area calculation
« Reply #2851 on: January 25, 2017, 10:07:36 PM »
Tamino has done analysis to consider different rates (changepoint analysis)

for example
https://tamino.wordpress.com/2016/09/11/climate-deniers-embarrass-themselves-about-arctic-sea-ice/#more-8711

He seems to prefer change in rate with no step for Arctic extent:

https://tamino.wordpress.com/2015/10/01/arctic-sea-ice-2/

Of course.  Change in rate is physically plausible.  A step change in extent is physically much less plausible - and indeed breaks the laws of thermodynamics if you interpret it literally as an instantaneous change of extent.

etienne

  • Nilas ice
  • Posts: 2045
    • View Profile
    • About energy
  • Liked: 309
  • Likes Given: 23
Re: Home brew AMSR2 extent & area calculation
« Reply #2852 on: January 25, 2017, 11:05:46 PM »
Tamino has done analysis to consider different rates (changepoint analysis)

for example
https://tamino.wordpress.com/2016/09/11/climate-deniers-embarrass-themselves-about-arctic-sea-ice/#more-8711

He seems to prefer change in rate with no step for Arctic extent:

https://tamino.wordpress.com/2015/10/01/arctic-sea-ice-2/
Hello,

From what I understood, Tamino says that deniers use data of one step taken out of context to say that the situation isn't as bad as what science says. So he says that if you use data on a longer period, you can’t hide what's happening.

My idea when I separated my trend in two steps was to check if the slope had changed (is it melting faster or slower...), and to see if a logarithmic or exponantial trend would be a better fit, if the trend would cross... My surprise was that the slope was very much similar (108 km per year before 2005 and 98 km per year after 2005 - I see 10% as non significant in such a context), but that the trend was around 0.5 million square km lower.

The thing that was not a surprise is that 2016 is out of scope for all trends which could suggest that we would be getting ready to go the next step down.

Well, if data keeps falling down like it does right now, trend for after 2005 will have a steeper slope, trends will cross and an exponential or logarithmic global trend would be ok.

I don’t have the knowledge, time and experience to say anything about climate change with these graphs, but I use the technique to analyse energy and water consumption and I found it very efficient. I use running averages (daily, weekly, yearly...) in order to reduce noise and to have easy access to comparable data, average has to be done on a full cycle of similar lenght (for consumption, a month is not good, 30 days is ok but 28 days would be better in order to have always the same quantity of Sundays in the average - using a year like for my artic sea ice extent graph is good because the error becomes very small and weather has a limited impact on the average). Of course, if somebody want's to compare morning and afternoon, average will have to be done on maximum 4 hours.

Bye,

Etienne
« Last Edit: January 25, 2017, 11:33:54 PM by etienne »

FishOutofWater

  • Nilas ice
  • Posts: 1088
    • View Profile
  • Liked: 696
  • Likes Given: 332
Re: Home brew AMSR2 extent & area calculation
« Reply #2853 on: January 26, 2017, 08:37:50 PM »
Hello,

I updated my Sea Ice Average Extent graph with the data that were available yesterday (up to January 22).
As I wasn't so happy with the global trend (green line), I decided to separate it in two periods, before (red line) and after (purpel line) January 1, 2005. I was surprized by the result, but it is in phase with other things I have seen and read, looks like there are steps on the way down.



Best regards,

Etienne

The record is too short for you to break that line in two and get any idea of the slope of the second line even if there was a step. Tamino has warned against using statistics that way. You don't have sufficient data. Arguing that other people have additional data sets does not make the slopes of your two lines any better defined.

etienne

  • Nilas ice
  • Posts: 2045
    • View Profile
    • About energy
  • Liked: 309
  • Likes Given: 23
Re: Home brew AMSR2 extent & area calculation
« Reply #2854 on: January 26, 2017, 10:35:28 PM »
The record is too short for you to break that line in two and get any idea of the slope of the second line even if there was a step. Tamino has warned against using statistics that way. You don't have sufficient data. Arguing that other people have additional data sets does not make the slopes of your two lines any better defined.

I agree that the record is too short, but it is all what I found. When I look at graphs like CO2 emissions, I think that maybe I should have done the cut a little bit earlier, like 2003. In that case, slope difference between the red and the purple line would have been even smaller. The choice of 2005 was done according to petrol price, which has nothing to do with climate change, excepted that high petrol cost created a switch toward coal.

Well, what that step story might say is that maybe we don't have just a few major tipping points (like the blue ocean event..), but many small ones, that the 0.5 million sqr kilometers lost between the two trend lines really is something to worry about because for example something could have been broken in the way ice freezes during the winter...

I am not a climate specialist so I don't know what to say in this context, but I achieved that type of event twice with the electrical consumption curve, and it was not by putting LED bulbs but with a major optimization effort. Just adopting more efficient technics as they come does not change the consumption trend because efficiency improves regularely and is balanced by new consumptions (electrical cars, more screens per person...).

Of course, maybe this step is only noise and doesn't deserve the time I spent on it. Maybe it makes no sense to define if the trend is linear, logarithmic... because it moves according to external parameters (aerosols, CO2, methane, sea currents...) that move in all kind of directions.

Well, if there really are steps, we are in big trouble because you can't know :
- when the next step will be
- how big the next step will be
- what is needed to have the next step going up instead of down.



« Last Edit: January 26, 2017, 10:52:00 PM by etienne »

DrTskoul

  • Nilas ice
  • Posts: 1455
    • View Profile
  • Liked: 210
  • Likes Given: 60
Re: Home brew AMSR2 extent & area calculation
« Reply #2855 on: January 26, 2017, 11:11:23 PM »
Discontinuity of a continuous variable is a huge no no!!!

etienne

  • Nilas ice
  • Posts: 2045
    • View Profile
    • About energy
  • Liked: 309
  • Likes Given: 23
Re: Home brew AMSR2 extent & area calculation
« Reply #2856 on: January 27, 2017, 08:20:06 AM »
Discontinuity of a continuous variable is a huge no no!!!
Hello,

I'm sorry, I don't understand, could you be more specific. If you talk about physical values, there is a continuity, but if you talk about data points, the line inbetween is just a creation of our imagination in order to be as close as possible of the reality.

If you talk about trends, there is no need for continuity. You see that in nature everywhere. For example :
- if you make a long term trend on the water level of a river, it will be flat, but it is not going to help you to predict how long ships can go under the bridge when water level is going up.
- the hockey stick temperature graph is a discontinuity in the earth temperature trend.
- if ou make a long term (millions of years) linear trend on earth temperature, the hockey stick will be just normal in the picture, but it’s not going to help you very much. If you make different trends for the warmer, colder and changing periods, you get interesting information.
- sea level also doesn’t have a stable linear behavior.https://www.giss.nasa.gov/research/briefs/gornitz_09/
- even for sea ice extent, I believe that it was more or less stable until it started to decrease dramatically.

It is always possible to create a mathematical formula where data would fit, it’s probably what is done when climate models are created. This is a very efficient method if you are in a stable context with predictable events where many things can be neglected because it is only a short time event and because you have a limited possibility of controlling the process, for example to calculate the trajectory of a rocket in order to send satellites around earth, but climate models just give general information which doesn't reduce their value.

I don’t have a problem if you think that my calculation doesn’t make sense, I agree that having such a short period of data is a good argument in that direction and I believe that a result that can’t be explained has a high probability of being wrong, but I don’t agree that I can’t do it. I don’t believe that my calculation gives more arguments to climate deniers because the step theory could be a very strong argument against climate denial (just wait, it will move soon, it’s not linear). The thing I don’t like in this step theory is that you can scare people with some kind of apocalyptic speech (the sky is going to fall on our heads for the people who read Asterix and Obelix).

I believe that I said more or less what I could on this subject and I feel that I am a little bit out of topic. If you have comments to my points, please do them, but if I don't answer, it's because I don't have any thing specific to add.

Thank you all,

Etienne
« Last Edit: January 27, 2017, 08:38:41 AM by etienne »

Peter Ellis

  • Grease ice
  • Posts: 619
    • View Profile
  • Liked: 33
  • Likes Given: 14
Re: Home brew AMSR2 extent & area calculation
« Reply #2857 on: January 27, 2017, 09:49:38 AM »
I'm sorry, I don't understand, could you be more specific. If you talk about physical values, there is a continuity, but if you talk about data points, the line inbetween is just a creation of our imagination in order to be as close as possible of the reality.

If you talk about trends, there is no need for continuity.

Correct, but you are not modelling a discontinuous trend.  You are modelling the DATA as discontinous, i.e. a sudden jump from one absolute value to another.  That is non-physical. 


You see that in nature everywhere. For example :
- if you make a long term trend on the water level of a river, it will be flat, but it is not going to help you to predict how long ships can go under the bridge when water level is going up.
Correct, but you don't see a discontinuity in the data.  The river does not go from 1 foot to 5 feet without passing through 2/3/4 feet.

- the hockey stick temperature graph is a discontinuity in the earth temperature trend.
- if ou make a long term (millions of years) linear trend on earth temperature, the hockey stick will be just normal in the picture, but it’s not going to help you very much. If you make different trends for the warmer, colder and changing periods, you get interesting information.
- sea level also doesn’t have a stable linear behavior.https://www.giss.nasa.gov/research/briefs/gornitz_09/
- even for sea ice extent, I believe that it was more or less stable until it started to decrease dramatically.

All correct, but again you are talking about discontinuity in the trend, not discontinuity in the data.

etienne

  • Nilas ice
  • Posts: 2045
    • View Profile
    • About energy
  • Liked: 309
  • Likes Given: 23
Re: Home brew AMSR2 extent & area calculation
« Reply #2858 on: January 27, 2017, 01:40:42 PM »
All correct, but again you are talking about discontinuity in the trend, not discontinuity in the data.

Who talked of a discontinuity of data ? I think that there is a misunderstanding. I talked of a change of trend where the slope stays the same, but the second trend is 0,5 million sqr km lower than the first one. I talked of steps in trends, not in data.

I calculated a trend for the few years where data was going from the upper trend to the lower trend, but didn't publish it because I though that calculating a trend on such a short period didn't make any sense. To be complete, upper and lower trend had a slope of 100 sqr km per year, and transition trend had a slope of 500 km per year. That 500 value doesn't make sense because changing a little bit the starting and ending points would have brought too much difference in the value.

Best regards,

Etienne
« Last Edit: January 27, 2017, 02:21:25 PM by etienne »

FishOutofWater

  • Nilas ice
  • Posts: 1088
    • View Profile
  • Liked: 696
  • Likes Given: 332
Re: Home brew AMSR2 extent & area calculation
« Reply #2859 on: January 27, 2017, 02:16:05 PM »
This is what you don't want to do.

Gray-Wolf

  • Grease ice
  • Posts: 948
    • View Profile
  • Liked: 131
  • Likes Given: 458
Re: Home brew AMSR2 extent & area calculation
« Reply #2860 on: January 27, 2017, 02:20:49 PM »
This is what you don't want to do.

Indeed! If it were to highlight the large volume drops, or 'steps', then that might be informative?
KOYAANISQATSI

ko.yaa.nis.katsi (from the Hopi language), n. 1. crazy life. 2. life in turmoil. 3. life disintegrating. 4. life out of balance. 5. a state of life that calls for another way of living.
 
VIRESCIT VULNERE VIRTUS

magnamentis

  • Guest
Re: Home brew AMSR2 extent & area calculation
« Reply #2861 on: January 27, 2017, 02:25:20 PM »
All correct, but again you are talking about discontinuity in the trend, not discontinuity in the data.

Who talked of a discontinuity of data ? I think that there is a misunderstanding. I talked of a change of trend where the slope stays the same, but the second trend is 0,5 million sqr km lower than the first one. I talked of steps in trends, not in data.

I calculated a trend for the few years where data was going from the upper trend to the lower trend, but didn't publish it because I though that calculating a trend on such a short period didn't make any sense. To be complete, upper and lower trend had a slope of 100 sqr km per year, and transition trend had a slope of 500 km per year. That 500 value doesn't make sense because changing a little bit the starting and ending points would have brought too much difference in the value.

Best regards,

Etienne

some nice guys who know what they're talking about are trying to give you friendly hints about what's a no go in this context, just reconsider and take on the advice, don't get stuck in defending your point (yourself) for various reason the attempt seems inappropriate while i for my part am open to good arguments, don't hesitate to take this to PM if you think that you were not or misunderstood :-)

Peter Ellis

  • Grease ice
  • Posts: 619
    • View Profile
  • Liked: 33
  • Likes Given: 14
Re: Home brew AMSR2 extent & area calculation
« Reply #2862 on: January 27, 2017, 03:03:40 PM »
Who talked of a discontinuity of data ? I think that there is a misunderstanding. I talked of a change of trend where the slope stays the same, but the second trend is 0,5 million sqr km lower than the first one. I talked of steps in trends, not in data.

The slope is the trend. The "0.5 sqr km lower" is a discontinuity in the data, which is unphysical.

I calculated a trend for the few years where data was going from the upper trend to the lower trend, but didn't publish it because I though that calculating a trend on such a short period didn't make any sense. To be complete, upper and lower trend had a slope of 100 sqr km per year, and transition trend had a slope of 500 km per year. That 500 value doesn't make sense because changing a little bit the starting and ending points would have brought too much difference in the value.

And this is precisely the problem.  That 0.5 million sqr km jump makes no sense.  If it's instantaneous, then it literally breaks the laws of physics. If instead you add a small continuous line joining one period to the next, then you have a very short, very steep slope for a few years - which is also nonsensical as it would imply the Arctic got incredibly warm for a short period and then almost immediately cooled back down again.

When you're doing this kind of change point analysis (as Tamino does, properly, because he's a real statistician), you have to make sure the end points of your trendlines meet each other, without sudden step changes.  The slope (i.e. the trend) can change abruptly, but not the actual extent.

Peter Ellis

  • Grease ice
  • Posts: 619
    • View Profile
  • Liked: 33
  • Likes Given: 14

charles_oil

  • Frazil ice
  • Posts: 337
    • View Profile
  • Liked: 38
  • Likes Given: 74
Re: Home brew AMSR2 extent & area calculation
« Reply #2864 on: January 27, 2017, 03:16:26 PM »
I imagine the justification for a "new" or discontinuous trend line would be where a significant event has occurred that causes a shift so there is the trend before - and after.

Normal seasonal variations though wouldn't be seen as justifying this (but presumably a major meteorite strike would).  Perhaps change in methodology (visual vs satellite observations?) might also qualify.

mikkel

  • New ice
  • Posts: 4
    • View Profile
  • Liked: 0
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2865 on: January 27, 2017, 08:51:04 PM »
Hi guys. I sympathize with etienne here and think it's pretty obvious what the question is. I've seen it in my own work!

From a systems perspective, etienne is merely speculating that there was a rapid phase shift to another attractor, but that the new attractor has the same dynamics as the old attractor of linear decline. How this translates to physical reality is pretty easy to conjecture: the ice decreases until at some time there is a tipping point in a particular basin of the Arctic where the dynamics quickly change, while the rest of the Arctic continues its steady progress down. This isn't exactly hypothetical, it appears that something like this may have occurred around Svalbard -- time will tell.

That said, you guys are completely correct about the statistical treatment for the hypothesis. Etiene's technique doesn't properly test the hypothesis for the reasons you state, however it is a novice attempt at trying to prove the intuitive intuition. I can relate, because I did the same thing early on in my career and it led to this exact same conversation with someone who taught me the proper statistics!

If there is anything I've learned it's that complex systems are very hard to communicate because there are so many different mental models, and analysis relies on making assumptions about which to pick. I've had countless discussions about whether to model things as continuous distributions (say logistic regression) vs multiple linear attractors (with change point analysis) or even as stochastic processes with high dimensionality.

In this example I agree that it looks like it won't pass a change point analysis. However, the residuals are obviously correlated at long tau, so a linear fit isn't correct either. So when it comes to analysis, we have to decide whether to try to fit a continuous distribution or make the assumption that there was a jump to a new attractor but a linear fit is still appropriate for each part.

Both of them rely on assumptions -- which continuous distribution would you pick? -- and neither is going to have statistical validity because there isn't enough data. In this case, like so many, we can be strong in saying that it looks there there is non-linearity, but the type of non-linearity is up for debate.

Unfortunately by the time we know, it will no longer be important. Therefore, at best we can do is use projections in order to try to better grasp what is going on and come to mutual understanding of how to react.

I've learned that quantitative tools have value for expressing qualitative intuition. They can demonstrate what we are thinking even if what we are thinking is not yet provable and therefore there is no way to be "correct." In that view I find Etienne's thoughts very interesting! I personally wouldn't come to the same conclusion, but it gives food for thought.

magnamentis

  • Guest
Re: Home brew AMSR2 extent & area calculation
« Reply #2866 on: January 27, 2017, 09:17:58 PM »
Hi guys. I sympathize with etienne here and think it's pretty obvious what the question is. I've seen it in my own work!

From a systems perspective, etienne is merely speculating that there was a rapid phase shift to another attractor, but that the new attractor has the same dynamics as the old attractor of linear decline. How this translates to physical reality is pretty easy to conjecture: the ice decreases until at some time there is a tipping point in a particular basin of the Arctic where the dynamics quickly change, while the rest of the Arctic continues its steady progress down. This isn't exactly hypothetical, it appears that something like this may have occurred around Svalbard -- time will tell.

That said, you guys are completely correct about the statistical treatment for the hypothesis. Etiene's technique doesn't properly test the hypothesis for the reasons you state, however it is a novice attempt at trying to prove the intuitive intuition. I can relate, because I did the same thing early on in my career and it led to this exact same conversation with someone who taught me the proper statistics!

If there is anything I've learned it's that complex systems are very hard to communicate because there are so many different mental models, and analysis relies on making assumptions about which to pick. I've had countless discussions about whether to model things as continuous distributions (say logistic regression) vs multiple linear attractors (with change point analysis) or even as stochastic processes with high dimensionality.

In this example I agree that it looks like it won't pass a change point analysis. However, the residuals are obviously correlated at long tau, so a linear fit isn't correct either. So when it comes to analysis, we have to decide whether to try to fit a continuous distribution or make the assumption that there was a jump to a new attractor but a linear fit is still appropriate for each part.

Both of them rely on assumptions -- which continuous distribution would you pick? -- and neither is going to have statistical validity because there isn't enough data. In this case, like so many, we can be strong in saying that it looks there there is non-linearity, but the type of non-linearity is up for debate.

Unfortunately by the time we know, it will no longer be important. Therefore, at best we can do is use projections in order to try to better grasp what is going on and come to mutual understanding of how to react.

I've learned that quantitative tools have value for expressing qualitative intuition. They can demonstrate what we are thinking even if what we are thinking is not yet provable and therefore there is no way to be "correct." In that view I find Etienne's thoughts very interesting! I personally wouldn't come to the same conclusion, but it gives food for thought.

kudos for the impressive way of covering both ends, somewhat comprehensive even for laymen like me. certainly among the best reads today and i read many hours each day :-) thanks  :)

etienne

  • Nilas ice
  • Posts: 2045
    • View Profile
    • About energy
  • Liked: 309
  • Likes Given: 23
Re: Home brew AMSR2 extent & area calculation
« Reply #2867 on: January 27, 2017, 10:53:01 PM »
Thank you very much for all your comments. I learned a lot today. Last time I did statistics was 25 years ago in University. Thanks for the refresh

Well, in the world I live, discontinuity is normal, because when I replace a 5 kW motor by another that is 20% more efficient, I really have a discontinuity, and when I replace a lightning system by another one that is LED based, I have a small discontinuity. Small discontinuities give linear graphs since it is a continuous process, and optimization events will create big discontinuities creating changes in the consumption curve similar to what I pointed out with my disconected trends. Maintenance and changes in production systens also bring discontinuities, just like week-ends, turning of the lights in the evening... I don't work with physical values like water going up and down in a riverbed.

An important thing in this forum is that we have all different backgrounds, we need understanding and clarification when methodology is not respected. You can't just say that something is not ok, it has to be explained because maybe it is ok in other contexts.The links toward Tamino's pages (one with a link to the Chow test on Wikipedia) where very helpful for me.

A funny thing I just realized is that here many people work with anomalies instead of absolute value. This is a thing you would never do with electrical consumption because the absolute value is more important than variations. So I "zoom" in the graph, which means than y axis will go for example from 4000kWh/day to 6000kWh/day. Nobody cares where the zero is, but you want to see what is the consumption (absolut value) and in which direction it goes. Being above or below the trend (actually, I use a running average of the last 12 months - so I have a "trend" without discontinuity) is a very important information because it shows an excessive or reduced consumption. Predicting the future doesn't make sense in my professional context because future consuption depends of future projects.

Best regards

Etienne
« Last Edit: January 27, 2017, 11:18:03 PM by etienne »

DrTskoul

  • Nilas ice
  • Posts: 1455
    • View Profile
  • Liked: 210
  • Likes Given: 60
Re: Home brew AMSR2 extent & area calculation
« Reply #2868 on: January 28, 2017, 03:57:32 AM »
Etienne,

You are absolutely right regarding your experience! Especially since you have been dealing with discrete increments ( motor power, number of LEDs etc). I appreciate your true desire to learn and to understand how your experience differs from others.

Unlike pressure and density, temperature for example is a property that is continuous and has a continuous first derivative in time and space except at interfaces under certain conditions (e.g. solid and gas, evaporating droplet etc). On the other hand there can be discontinuities on the level of a river in flow under certain conditions  ( hydraulic jump) where the flow changes regime from fast laminar to slow turbulent flow.  However ocean sea level rise, ice area extent change due to temperature increase etc as explained above cannot and even if the data appear to do so, we have to constrain our models to follow the right physics if we wish to derive the correct conclusions.

Our brains are word to see patterns because there is too much information to process. We have to use our intellect to question what we see..

etienne

  • Nilas ice
  • Posts: 2045
    • View Profile
    • About energy
  • Liked: 309
  • Likes Given: 23
Re: Home brew AMSR2 extent & area calculation
« Reply #2869 on: January 28, 2017, 07:26:23 AM »
Hello,

I'm the engineer here, so I feel I can ask the stupid questions without looking ridiculous. I'm sorry that it is somehow out of topic.

I have no doubt that DrTskoul is right when he says that :

Unlike pressure and density, temperature for example is a property that is continuous and has a continuous first derivative in time and space except at interfaces under certain conditions (e.g. solid and gas, evaporating droplet etc).

Well, I see artic sea ice as a hughe solid and gas interface. Furthermore, when talking about sea ice, there is also a solid-liquid interface that behaves according to totally different rules. So is it so sure that the first derivative of sea ice melting is continuous ?

I agree that if we talk of volume, first derivative has to be continuous, but for area and extent, I'd be surprise if a storm doesn't create discontinuities. I also agree that discontinuities would be minor events and that they couldn't be considered when creating a model to predict the future.

Thank you for all the information.

Etienne
« Last Edit: January 28, 2017, 09:33:08 AM by etienne »

magnamentis

  • Guest
Re: Home brew AMSR2 extent & area calculation
« Reply #2870 on: January 28, 2017, 02:11:41 PM »
go ahead guys, this is a very interesting discussion, unfortunately i can't contribute because i have no clue but that's the fun of it, something really new to learn. i'm truely enjoying this.

cheers  ;)

DrTskoul

  • Nilas ice
  • Posts: 1455
    • View Profile
  • Liked: 210
  • Likes Given: 60
Re: Home brew AMSR2 extent & area calculation
« Reply #2871 on: January 28, 2017, 02:54:59 PM »
When Wip gets back to this thread we will have to stop (really OT). But let's give it one more try. I am a fellow engineer spending too much time with scientists.

Area, extent, volume are all a measure of the mass of ice in the arctic ( 2D projection and 3D). Ice is advected, accumulated {see ridges}, melted or evaporated through the respective processes. The change of ice mass over time regardless the rate, must be a continuous curve. There is no spontaneous mass generation or distruction ( if we avoid nuclear explosions ).  At the same time the various phenomena of mass transfer ( freezing, melting, condensation, evaporation , advection) are all due to heat and momentum transfer from the atmosohere, the ocean, and the sun.

Now due to their nature, the heat and momentum transfer from sun, atmosphere and ocean are smooth in time - again barring nuclear explosions over the arctic. Sun position in the sky changes smoothly with time ( no peekaboo's ). Atmosphere and ocean processes start somewhere else ( cyclones and currents ) and since they don't appear from thin air, their effects are also smooth in time.  Therefore ( unless I have committed a blunder - which I cannot deny that it happens regularly ) the mass and momentum transfer that affects the ice are smooth in time leading to the ice mass, area etc to be smooth in time and not have abrupt kinks or discontinue changes of the rate in time.

Which leads us to fitting trends of area, extent or volume changes. Locally the rate of change has to be at least a series of linear segments that must connect to each other ( continuous ). If one averaged over large periods of time ( e.g what is the linear trend over the last 30 years ) then one represents the mass of ice in time as a line.  Two consecutive 30 year segments will have two descrete and different rates of mass change, constant over those 30 years.  The mass of ice will be represented with lines again, but those lines as physics dictates must come to a single common data point - mass of ice is continuous with time. It is also smooth but for the sake of argument we assumed away that fact when we try to see an average rate.

Which brings us to the start.  Those lines at your original graph. If they represent a linearly  fitted mass, or area,  in time, then they must come to a common data point. Otherwise they are false.

I apologize for being long winded.

crandles

  • Young ice
  • Posts: 3379
    • View Profile
  • Liked: 239
  • Likes Given: 81
Re: Home brew AMSR2 extent & area calculation
« Reply #2872 on: January 28, 2017, 04:19:48 PM »

I agree that if we talk of volume, first derivative has to be continuous, but for area and extent, I'd be surprise if a storm doesn't create discontinuities. I also agree that discontinuities would be minor events and that they couldn't be considered when creating a model to predict the future.

Suppose we consider a block of ice 200m by 10m by 2m thick. This remarkably cracks off a piece 200m by 1m by 2m thick. The crack itself does nothing to area or volume, but then it tips over on its side and suddenly the area goes up to 200m by 11m. If it takes 1 second to tip over and we sample area every 2 seconds which most of the time seems plenty to see changes like melting freezing etc then there seems a discontinuity in area. In reality if we sample area every .001 second we would see that as the slab tips over so the area increases (reaches a maximum and then decreases again) smoothly without a discontinuity.

So I think even area and extent and first differentials are continuous but different processes can have different timescales. Sampling on too slow a timescale can make it look like there is a discontinuity when a faster process kicks in.

Yes this is considering a minor event that couldn't be considered when creating a model to predict the future.
« Last Edit: January 28, 2017, 04:34:01 PM by crandles »

DrTskoul

  • Nilas ice
  • Posts: 1455
    • View Profile
  • Liked: 210
  • Likes Given: 60
Re: Home brew AMSR2 extent & area calculation
« Reply #2873 on: January 28, 2017, 04:27:49 PM »

I agree that if we talk of volume, first derivative has to be continuous, but for area and extent, I'd be surprise if a storm doesn't create discontinuities. I also agree that discontinuities would be minor events and that they couldn't be considered when creating a model to predict the future.

Suppose we consider a block of ice 200m by 10m by 2m thick. This remarkably cracks off a piece 200m by 1m by 2m thick. The crack itself does nothing to area or volume, but then it tips over on its side and suddenly the area goes up to 200m by 11m. If it takes 1 second to tip over and we sample area every 2 seconds which most of the time seems plenty to see changes like melting freezing etc then there seems a discontinuity in area. In reality if we sample area every .001 second we would see that as the slab tips over so the area increases without a discontinuity.

So I think even area and extent and first differentials are continuous but different processes can have different timescales. Sampling on too slow a timescale can make it look like there is a discontinuity when a faster process kicks in.

Yes this is considering a minor event that couldn't be considered when creating a model to predict the future.

True. Sth I missed. That is an instantaneous area generation! So if you only measure the local area around the iceberg there would be a huge area increase with no mass or volume change. And an abrupt area rate change locally.

Thank you for reminding us that.

Tor Bejnar

  • Young ice
  • Posts: 4606
    • View Profile
  • Liked: 879
  • Likes Given: 826
Re: Home brew AMSR2 extent & area calculation
« Reply #2874 on: January 28, 2017, 05:22:31 PM »
Crandles 'published' while I wrote (and made my wife breakfast and talked to a recent immigrant who just moved into a house of their own - how exciting!).

Even a large asteroid strike (or A-bomb - shutter  :'() will change ice area/extent/volume over increments of time, albeit milli- and nano-seconds, not unlike (except for time scale) that caused by CO2e increases. 

When we (attempt to) look at the 'big picture' (e.g., the graph that got this conversation started), it may 'look like'  there was an instantaneous change, but if the early and late rates show unconnected line segments and are 'true representations' of the data, there will be a third line segment to connect the two.  We may not have enough data to authenticate this (including a change point analysis), and without adequate data we'll need an argument for the cause (i.e., circumstantial evidence).  (I've read (above) at least one possible hypotheses that could account for the apparent discontinuity.)

I'm imagining a Kiwi geography student measuring apparent sea level (in relation to a local hill) on the Kaikoura coast and camping on that hill.  Including historical data, the student notes several trends, one associated with the moon, another associated with global warming, etc.  A truly heavy sleeper, this student leaves the tent after a short nap and takes more measurements and discovers the apparent sea level has dropped a couple of meters within an hour.  "Oh my word, an instantaneous change" - only to learn later about the earthquake that changed the local land-sea relationship in seconds.  30 years later, the 'apparent sea level' graph will have curves showing sea level changes before and after the earthquake with a vertical line segment showing the changes that took place during the earthquake.
« Last Edit: January 28, 2017, 07:12:46 PM by Tor Bejnar »
Arctic ice is healthy for children and other living things because "we cannot negotiate with the melting point of ice"

DrTskoul

  • Nilas ice
  • Posts: 1455
    • View Profile
  • Liked: 210
  • Likes Given: 60
Re: Home brew AMSR2 extent & area calculation
« Reply #2875 on: January 28, 2017, 06:58:35 PM »
The point being that sparse data should be constrained when fitted by using model with the correct behavior and physics to be accurate or semi acurrate at multiple times and length scales.

Jim Williams

  • Frazil ice
  • Posts: 398
    • View Profile
  • Liked: 1
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2876 on: January 28, 2017, 08:03:20 PM »

I agree that if we talk of volume, first derivative has to be continuous, but for area and extent, I'd be surprise if a storm doesn't create discontinuities. I also agree that discontinuities would be minor events and that they couldn't be considered when creating a model to predict the future.

Suppose we consider a block of ice 200m by 10m by 2m thick. This remarkably cracks off a piece 200m by 1m by 2m thick. The crack itself does nothing to area or volume, but then it tips over on its side and suddenly the area goes up to 200m by 11m. If it takes 1 second to tip over and we sample area every 2 seconds which most of the time seems plenty to see changes like melting freezing etc then there seems a discontinuity in area. In reality if we sample area every .001 second we would see that as the slab tips over so the area increases (reaches a maximum and then decreases again) smoothly without a discontinuity.

So I think even area and extent and first differentials are continuous but different processes can have different timescales. Sampling on too slow a timescale can make it look like there is a discontinuity when a faster process kicks in.

Yes this is considering a minor event that couldn't be considered when creating a model to predict the future.

I think the technical term is "bergy bits."  To say they were common would be an understatement.

etienne

  • Nilas ice
  • Posts: 2045
    • View Profile
    • About energy
  • Liked: 309
  • Likes Given: 23
Re: Home brew AMSR2 extent & area calculation
« Reply #2877 on: January 28, 2017, 11:01:01 PM »
Hello,

Great discussion. Thanks to all.

Maybe there is another discontinuity related to sea ice extent. On the definition page of NSIDC http://nsidc.org/cryosphere/seaice/data/terminology.html they give this information :

Quote
The Sea Ice Index products have a threshold of 15 percent.

So if you don't have a balance between cells crossing up and down that 15 percent threshold, you could get data that doesn't make sense. If large areas are crossing the threshold in the same direction, it could even look like fast growth or melt, but it could be just a few percent change in these areas (I don't know what is the sensivity). This could also explain oscillations in data.

Don't know if it is relevant in this context because trends cover more years. It should probably be considered as measurement error. Don't know how important error is on extent graphs.

Etienne
« Last Edit: January 28, 2017, 11:12:28 PM by etienne »

jdallen

  • Young ice
  • Posts: 3410
    • View Profile
  • Liked: 650
  • Likes Given: 244
Re: Home brew AMSR2 extent & area calculation
« Reply #2878 on: January 28, 2017, 11:29:10 PM »
In part, etienne, besides sensing, the problem we are dealing with is this... predicting the motions of a compound pendulum while chaotic gnomes are merrily changing the length of the arms on the fly, and whacking them variously to change their energy.

I'm increasingly pessimistic about using sea ice extent *or* area in any sort of predictive fashion. I think both have entered into a state that is far too volatile.

[Image By Catslash - Own work, Public Domain, https://commons.wikimedia.org/w/index.php?curid=10404903]
This space for Rent.

Wipneus

  • Citizen scientist
  • Young ice
  • Posts: 4220
    • View Profile
    • Arctische Pinguin
  • Liked: 1025
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2879 on: February 01, 2017, 09:45:58 AM »
Ice is being blown back through the Bering Strait.

oren

  • Moderator
  • First-year ice
  • Posts: 9805
    • View Profile
  • Liked: 3584
  • Likes Given: 3922
Re: Home brew AMSR2 extent & area calculation
« Reply #2880 on: February 01, 2017, 12:14:15 PM »
Thanks Wipneus. Finally the thread is back on topic...
Interesting that despite the import back into the Chukchi the concentration actually seems to be dropping there, at least initially.

seaicesailor

  • Guest
Re: Home brew AMSR2 extent & area calculation
« Reply #2881 on: February 01, 2017, 12:49:17 PM »
The AMSR2 brightness temperature images I posted in the other thread also show an increase in the last two or three days,

http://forum.arctic-sea-ice.net/index.php/topic,1611.msg101414.html#msg101414

but I think it is because of the direct increase in temperature caused by advection of airmass from the Pacific, may be fooling the concentration calculation. Another (worse) blow is coming in the next few days.
Below the 850 mbar temp anomaly of the last three days (ECMWF), shows very neatly the airmass entering the Arctic



Wipneus

  • Citizen scientist
  • Young ice
  • Posts: 4220
    • View Profile
    • Arctische Pinguin
  • Liked: 1025
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2882 on: February 01, 2017, 12:58:33 PM »
Thanks Wipneus. Finally the thread is back on topic...
Interesting that despite the import back into the Chukchi the concentration actually seems to be dropping there, at least initially.

Indeed, did not want to mention it, but it definitely torch-like.

DrTskoul

  • Nilas ice
  • Posts: 1455
    • View Profile
  • Liked: 210
  • Likes Given: 60
Re: Home brew AMSR2 extent & area calculation
« Reply #2883 on: February 01, 2017, 01:13:24 PM »
Thanks Wipneus. Finally the thread is back on topic...
Interesting that despite the import back into the Chukchi the concentration actually seems to be dropping there, at least initially.

Indeed, did not want to mention it, but it definitely torch-like.

Good old torching....

Wipneus

  • Citizen scientist
  • Young ice
  • Posts: 4220
    • View Profile
    • Arctische Pinguin
  • Liked: 1025
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2884 on: February 02, 2017, 08:55:29 AM »
Notice that Jaxa, that uses bootstrap vs UH's ASI algorithm, does not show this apparent drop of SIC, as such much more believable.

Ninebelowzero

  • New ice
  • Posts: 86
    • View Profile
  • Liked: 0
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2885 on: February 02, 2017, 10:25:53 AM »
......
but I think it is because of the direct increase in temperature caused by advection of airmass from the Pacific, may be fooling the concentration calculation. Another (worse) blow is coming in the next few days.
Below the 850 mbar temp anomaly of the last three days (ECMWF), shows very neatly the airmass entering the Arctic


Jetstream maps show a perfect loop forming over the area and in short order too.

Wipneus

  • Citizen scientist
  • Young ice
  • Posts: 4220
    • View Profile
    • Arctische Pinguin
  • Liked: 1025
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2886 on: February 10, 2017, 04:08:26 PM »
Sudden reverse in the Fram looks like this:

DrTskoul

  • Nilas ice
  • Posts: 1455
    • View Profile
  • Liked: 210
  • Likes Given: 60
Re: Home brew AMSR2 extent & area calculation
« Reply #2887 on: February 10, 2017, 04:22:34 PM »
Sudden reverse in the Fram looks like this:

Wow. Quite the whiplash.  Looks like a bug hitting a moving car..SPLAT!!!

Jim Williams

  • Frazil ice
  • Posts: 398
    • View Profile
  • Liked: 1
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2888 on: February 10, 2017, 05:53:18 PM »
Sudden reverse in the Fram looks like this:

Wow. Quite the whiplash.  Looks like a bug hitting a moving car..SPLAT!!!

More simply:  EEEEKKKKKK!!!!!!

Neven

  • Administrator
  • First-year ice
  • Posts: 9470
    • View Profile
    • Arctic Sea Ice Blog
  • Liked: 1333
  • Likes Given: 617
Re: Home brew AMSR2 extent & area calculation
« Reply #2889 on: February 10, 2017, 06:54:02 PM »
Wow, indeed.
The enemy is within
Don't confuse me with him

E. Smith

Tor Bejnar

  • Young ice
  • Posts: 4606
    • View Profile
  • Liked: 879
  • Likes Given: 826
Re: Home brew AMSR2 extent & area calculation
« Reply #2890 on: February 10, 2017, 07:46:36 PM »
I'm waiting to see if these winds affect (break) the Lincoln Sea ice arch.  Nares Strait is virtually free of ice that originated in the Arctic Ocean (ArO) because this arch last shed ice about February 2nd.  (Little bits of ArO ice have stuck to fast ice here and there, and a little lingers in Kane Basin, I think.)
Arctic ice is healthy for children and other living things because "we cannot negotiate with the melting point of ice"

Wipneus

  • Citizen scientist
  • Young ice
  • Posts: 4220
    • View Profile
    • Arctische Pinguin
  • Liked: 1025
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2891 on: February 12, 2017, 06:34:42 AM »
Continued...

Shared Humanity

  • Guest
Re: Home brew AMSR2 extent & area calculation
« Reply #2892 on: February 12, 2017, 03:13:50 PM »
Wow!

DrTskoul

  • Nilas ice
  • Posts: 1455
    • View Profile
  • Liked: 210
  • Likes Given: 60
Re: Home brew AMSR2 extent & area calculation
« Reply #2893 on: February 12, 2017, 03:47:25 PM »
Who pressed the EJECT button????

Apocalypse4Real

  • Frazil ice
  • Posts: 370
    • View Profile
  • Liked: 3
  • Likes Given: 1
Re: Home brew AMSR2 extent & area calculation
« Reply #2894 on: February 12, 2017, 04:28:05 PM »
Wipneus, I was looking at the Sentinel and Radar Sat individual imagery, through Polarview, but this makes the impact so incredible. Much appreciated. WOW and Yikes!

The power of wind and wave to demolish sea ice is very apparent. Almost all the way to the NP, the ice is rubble.

Gray-Wolf

  • Grease ice
  • Posts: 948
    • View Profile
  • Liked: 131
  • Likes Given: 458
Re: Home brew AMSR2 extent & area calculation
« Reply #2895 on: February 12, 2017, 05:18:30 PM »
http://www.woksat.info/etczb11/zb11-1407-j-apt.html



Another view of the ice to the pole yesterday ( the rubble in the middle right is the corner into Fram with Nares dead centre)

We have just had our first full moon spring tides so lets wait to see if that imparts any stresses into the ice cover?
KOYAANISQATSI

ko.yaa.nis.katsi (from the Hopi language), n. 1. crazy life. 2. life in turmoil. 3. life disintegrating. 4. life out of balance. 5. a state of life that calls for another way of living.
 
VIRESCIT VULNERE VIRTUS

Wipneus

  • Citizen scientist
  • Young ice
  • Posts: 4220
    • View Profile
    • Arctische Pinguin
  • Liked: 1025
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2896 on: March 11, 2017, 08:49:24 AM »
First century drop in UH AMSR2 sea ice concentration:

Update 20170310.

Extent: -104.1 (-107k vs 2016, +25k vs 2015, -78k vs 2014, -861k vs 2013, -777k vs 2012)
Area: -84.5 (-34k vs 2016, +49k vs 2015, +103k vs 2014, -836k vs 2013, -696k vs 2012)
 
You will find the updated graphs in the top post

The details (in 1000 km2):


Extent:
   Central Arctic Basin       East Siberian Sea              Laptev Sea
                    3.3                    -3.7                     1.5
               Kara Sea             Barents Sea           Greenland Sea
                   -0.8                   -29.9                    -9.3
Baffin/Newfoundland Bay            St. Lawrence              Hudson Bay
                  -10.3                   -24.7                    -4.7
   Canadian Archipelago            Beaufort Sea             Chukchi Sea
                   -2.0                     0.0                   -10.7
             Bering Sea          Sea of Okhotsk            Total Extent
                  -14.3                     1.7                  -104.1

Area:
   Central Arctic Basin       East Siberian Sea              Laptev Sea
                    9.8                   -22.1                     2.3
               Kara Sea             Barents Sea           Greenland Sea
                   -0.5                   -22.6                    -6.3
Baffin/Newfoundland Bay            St. Lawrence              Hudson Bay
                  -13.5                   -16.9                    -9.1
   Canadian Archipelago            Beaufort Sea             Chukchi Sea
                   -2.2                     0.9                   -14.1
             Bering Sea          Sea of Okhotsk              Total Area
                    3.4                     6.5                   -84.5



Wipneus

  • Citizen scientist
  • Young ice
  • Posts: 4220
    • View Profile
    • Arctische Pinguin
  • Liked: 1025
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2897 on: March 11, 2017, 08:52:29 AM »
Most of it outside of the Arctic Basin of course. But some of it due to the cracking event in the Chukchi/ESS regions:

Jim Hunt

  • First-year ice
  • Posts: 6268
  • Don't Vote NatC or PopCon, Save Lives!
    • View Profile
    • The Arctic sea ice Great White Con
  • Liked: 893
  • Likes Given: 87
Re: Home brew AMSR2 extent & area calculation
« Reply #2898 on: March 17, 2017, 05:35:14 PM »
A new AMSR2 based product from OSI-SAF/EUMETSAT (OSI-408) is supposed to be operational by now:

http://osisaf.met.no/p/ice/amsr_ice_conc_long_description.html

However the link to the associated maps seems to still end up on the SSMIS based version:

http://osisaf.met.no/p/ice/index.html

I've emailed OSI-SAF about the "issue", but with no response as yet.  Can anybody else find them?
"The most revolutionary thing one can do always is to proclaim loudly what is happening" - Rosa Luxemburg

Wipneus

  • Citizen scientist
  • Young ice
  • Posts: 4220
    • View Profile
    • Arctische Pinguin
  • Liked: 1025
  • Likes Given: 0
Re: Home brew AMSR2 extent & area calculation
« Reply #2899 on: March 19, 2017, 10:50:32 AM »
Perhaps it is fixed now

(following your link I end up in ftp://osisaf.met.no/prod/ice/conc_amsr with lots of  'amsr'-tagged files)