From Chris:

There is an animation I made back in my early PhD days many years ago that all of us in our group use to try and explain tipping points which hopefully you can see here:


I have made a Quicktime screen recording which looks like this:

And through the BellHouse browser it looks like this:

This is a recording of BellHouse playing the looped version of the first recording five times. The settings on BellHouse are: Sampling interval – 1.5, Sampling threshold – 0.1.

This is a really low sampling threshold as I played it a few times at 5, 2 and 1, but got very little response. So I thought i would try it as low as it would go and got this, which I like. The universal chimes at the beginning and the end reflect the colour change from white to black and back again. You can hear the movement across the tipping point and it might be nice to explore this further to see if there is a way to play particular bells or a better range of bells.

I have also realised that it might play differently if I changed the grid size, which it does:

No time today to playtest the other gif but I will prepare the screen grabs etc at home so I can come in later in the week to play it as soon as I get in.


Playing BellHouse .gif 2

Chris has sent through another version of the same data which he has modified after our discussion in the comments of this post. I have made it into a movie so BellHouse can play it and in it you can see several differences.

Here you can see the 2 graphs from the original animation have changed so that the top element is the same data as the left bottom, but a close up in order to see the movement of the “ball” better. The ball has also been filled with a semi opaque red to help it be seen by the motion capture software and the lines have been made slightly thicker for the same reason. As I understand it, Chris’s intention is that BellHouse plays the variability at the tipping point more clearly as this is the element of the data that is most important to communicate.

In playing it through BellHouse there is a noticeable difference in this section of the animation and how it is played, so I think we are getting towards that point. There are some issues with the software to help to make BellHouse read the whole of the image with all its bells, but we are working on that at the moment.

In the meantime this is how it sounds:

I think you can definitely hear the movement of the “red ball” as the curve begins to flatten out and the consquent change in tone as the ball goes from one dip in the curve to the other.

9 thoughts on “ Chris Boulton: Early Warning Signals of Tipping Points ”

  1. Hi Roop,

    As I said in my email it sounds good! I just wondered if it was worth me making a gif/animation that was more focused on the left hand well that the ball is moving around it. I think I can hear the bells change when this shallows out such that it moves upwards and into a new grid square (hopefully I am not imagining this!) There is a lot of space in the top of the plot of the wells that is wasted for this type of grid approach so zooming in might be useful.

    I think the important thing is the movement from side to side as the ball should ‘slow down’ when the well shallows, and in doing this, it takes longer to return to the bottom of the well and it would be cool if this could be picked up by the bells. Perhaps the cooldown time on the bells would be a problem but I’m happy to have a look if you would like?


    1. Hi Chris,
      As you say, there is a lot of space in the top left that is not being used so it might be useful to try and focus more on what is going on at the bottom of the plot. The other thing to think about is possibly to make the ball solid as then there will be more colour change in the area and that may help to ring bells, esp where it will overlap. It could be solid and a bit smaller if you want more precision. If you want to still see the line underneath it might be useful to make it 50% opacity and a medium red (or whatever colour fits for you) also.
      however one thing also. I think the deepening of the curves as the animation progresses is also making some bells ring. So the other thing to try is to make the line of the curve slightly thicker, maybe by a pixel or 2. What does this signify?
      You can definitely hear the change in the right hand side as the line graph picks up suddenly, though as this is quite near the end it is relatively short. I can loop this movie to enable us to hear the patterns better. I find that when I am working on this I often leave it on whilst I am doing other things (updating the blog etc) and then sometimes can hear the patterns better when i am not concentrating too hard on it (if you know what i mean!). I am going to listen to it again now to hear what you are talking about in the last papragraph and get back t you in a mo.

      1. Hi Chris,
        In relation to the second part of what you have said, I think that the other thing to do would be to slow down the animation as that might help BH to pick up the movement more clearly and therefore respond with more clarity. BH works best if the sampling interval is 1.5 seconds or slower as the beaters have time to reset themselves properly. If it is quicker than this the beaters don’t always strike the pots as well as they should and so sometimes the sounds are weaker and less reliable.
        So in summary, I think we might get a better response if:
        1. The left hand curve were more focused on the area of action
        2. The ball had some fill (50% opacity) of a colour
        3. The animation was slowed down
        4. The curve line was a bit thicker

        one other quick question; What is the time sequence here. I can see on the graph that it goes from 0-1000. what is this in real time? It is probably a daft question, but it would help me to get a sense of it better.
        cheers, roop

        1. Hi Roop,

          I’ve made a new animation which I will email to you in a minute. As I said before I think one of the main things it would be good to hear is the movement across the well so I have added a strip across the top that focuses on either the left or right well depending on where the ball is that is purely picking out it’s left and right movement. I have made the lines thicker but they can be thicker if needed and made the ball a slightly transparent red.

          As for slowing the animation down, I could do this but want to test it a bit more first. There are 2000 slides I have made for this animation and I have stuck together every 5th one. I need to work out whether slowing this down is better than taking say every 10th slide and doubling the amount of time it is on screen. I’d be interested to see how this works as it is to see what to do about this.


          1. Also the time at the bottom of the plot is arbitrary. The system is designed to tip at time 800 but will do slightly before then because it’s being pushed around in the well and it becomes easier to be pushed over the hump in the middle as you approach 800.

  2. Hey the new recording looks great. Is there any chance to have the two (the animation and BellHouse) alongside each other? I think this will inform us on the delay needed more clearly.

    1. Hi Chris,
      We are going to do this in early November when we have Preston Street Filmsin. The plan is to make a playlist of some of the interactions so far. We are going to film a number of recordings and put them through a mixing desk to combine both the input through the sampler (or the animation itself) and BellHouse playing it at the same time. In the meantime, my job is to keep practicing with data to get the best responses. With that in mind, you mentioned previously the data you had about the Pacific. If you wanted to send me that I would be happy to give it a go. I will also look at the Other graph you sent a link to, but that might not be this week as I have to get everything ready for the conference on Thursday. Thanks again though. Did what I wrote make sense?
      Best, Roop

      1. Sure, it might be easier to send the North Pacific data then too as I have to make an animation of it. Now I know a bit more about what BellHouse is looking for I think it’ll inform me more about how to set out the animation.

        The North Pacific is a time series that we assume is in a single well that is shallowing (as in the animation you have but just one well). It might be useful to make this new top panel again. There is also a signal of something increasing to measure that well shallowing which is just a straight time series. This could be represented by a ball moving along a plot maybe so the bells really pick up the change?

        1. Hi Chris,
          What you say about the second element of the data sounds good. I would like to see that. I think it should make a good sound. Thanks again for the intro today. Best wishes, roop

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.