Climadjust is a web based service which allows users such in a wide variety of sectors (agriculture, planning, infrastructure, etc) to use information from climate models with more specificity for local conditions (topography, land use, etc.) to more gather more accurate data to forecast the possible effects on a particular area as a result of climate change.

As part of the Climateurope webstival 2020, BellHouse has been playing some data from the Climadjust service to explore how this data sounds.

Juan Jose from Predicita sent me six sets of data to play and is for an area in or near Barcelona in Spain. This data is presented in the form of a gif which shows 2 series of line graphs, which show temperature variations over time. The first; a blue field of lines, denotes data generated by a series of climate models (global in nature) with a thick blue line which denotes the average of these models. The space between the highest measurements and lowest is shaded in blue. The second field (grey) denotes the data when bias adjusted by the software.

The data looks at 2 areas? and explores the historical and predicted temperature over time with 2 different prediction data sets. In this post I will set out the recordings for each area.

In the initial discussion about the data, held during the workshop in the 3rd Climateurope Webstival on the 15th October 2020. We played some of the initial sets, in which the transitions in the data are every 0.5 seconds. This means that BellHouse can only play 1 of every 4 transitions, or more accurately, it plays an aggregation of those four transitions as it records all the colour value change between one sampling interval and the next. I asked Juan Jose if it would be possible to re-issue the data with a change every 2 seconds so that BellHouse could play each change in the data. I have recorded all of the second submission of data and 2 of the first stage so the difference can be seen. These are for the first and 5th sets of data and these are ordered to be next to each other so that they can be heard together.

Data Set 1

BellHouse playing a double loop of a graph from the Climadjust software @Bias Adjustment for tamsin historical GSOD (Annual Mean). Sampling interval 2. Sampling threshold 2. In this animation the transitions are at 2 seconds.

In this recording the animation is timed to advance every 2 seconds.

BellHouse playing a double loop of a graph from the Climadjust software @Bias Adjustment for tamsin historical GSOD (Annual Mean). Sampling interval 2. Sampling threshold 2. In this animation the transitions are at 0.5 seconds.

Here is the quicktime recording of the .gif (note- edited to capture just the moving data to allow as many bells as possible to be sampling the movement). The first is a 2 second intervals and the second at 0.5 second intervals.

Bias adjustment for tamsin historical GSOD (annual mean) (as a quicktime movie) with 2 second intervals in the animation.
Bias adjustment for tamsin historical GSOD (annual mean) (as a quicktime movie) with 0.5 second intervals in the animation.

In the 1st of these playbacks of the first animation, BellHouse is responding more clearly to each increase and decrease in the graphs as the animation progresses. You can hear at this speed how BellHouse responds to the colour changes in each area of the grid as they get larger or smaller, much more clearly. In the second playback and the movie that corresponds< Bellhouse is playing every 4 transitions (or the agglomeration of these) and so plays less detail, but much more quickly.

Data Set 2

Bellhouse playing ” Bias adjustment for tasmin historical IBERIA01 (Annual Mean) 2 seconds gif”. sampling interval 2, sampling threshold 2. 2 loops played
Bias_Adjustment_for_tasmin_historical_IBERIA01_(annual_mean)_2secs.gif Actions.mov

In this second animation the two graphs are much close together and most of the time they overlap. The blue graph takes up most of the available space and one assumes this is because there is a wide range of variability in the models that make up this part of the graph. In this payback you can hear that the blue graph is playing up to 4 bells at any one time.

Data Set 3

Bellhouse playing “Bias_Adjustment_for_tasmin_rcp45_GSOD_(annual_mean) 2 secs .gif” Sampling interval 2, sampling threshold 1.
3 Bias_Adjustment_for_tasmin_rcp45_GSOD_(annual_mean)_2secs.gif

In this playback (3), the shaded area is much smaller than previously. I have lowered the sampling threshold to one as initially it was not registering enough to play the bells properly, meaning that were significant gaps. In this adjusted recording the BellHouse is normally playing 2 bells at a time.

Data Set 4

BellHouse playing 4 “Bias_Adjustment_for_tasmin_rcp45_IBERIA01_(annual_mean)_2secs.gif Actions.mov” A model from the Climadjust software by Predicitia. Sampling Interval 2, Sampling Threshold 1
4 Bias_Adjustment_for_tasmin_rcp45_IBERIA01_(annual_mean)_2secs.gif Actions

Like the previous one, this animation has a much smaller range of temperature difference and this can be heard in the bells, but it follows a similar pathway to the last one. In this one, each animation takes 5 mins approx to play through so looped twice this makes over 10 minutes.

At this point I had a couple of questions for Jose (written in my notes) with possible answers:

  1. Why are all the graphs following an upward trend ( Do they refelect overall changes in temperature upwards over time?)
  2. Why is the bias adjusted line always higher?
  3. Why in this one is there much more information?

Data Set 5

BellHouse playing “5 Bias_Adjustment_for_tasmin_rcp85_GSOD_(annual_mean)_2secs.gif Actions” A model from the Climadjust software by Predictia. Sampling Interval 2, Sampling Threshold 1
5 Bias_Adjustment_for_tasmin_rcp85_GSOD_(annual_mean)_2secs.gif Actions

(*From notebook) This one looks to have a much steeper rate of ascent across the timescale so it should play some of the bells to the right. At about 1.10 min into the playback there is a gap, an internet lag I think which then causes the bells to play slightly irregularly as they catch up. This may also be because the blue line at this point in the animation is quite smooth and so may not trigger the responses from the BellHouse. This also occurs in the beginning of the black graph line for about 10 seconds.

  • These are good visualisations for seeing how BellHouse responds to data. You can really hear how each bell corresponds to the movement on the sampling grid as it is playing so slowly. I t might sound a bit monotonous but it is clear.
  • Here each grid is taking 14 strikes to progress through. It means you have time to listen horizontally (as it were).
  • You can also hear/see the amount of variation in a grid to cause a response (about a quarter of the grid needs to change colour to cause a response). So although the main change might be happening across one grid, if there is enough in the grid above or below, then this will cause another bell to strike, which is useful as there is significant variation occasionally from transition to transition in the animations.
  • There is a lot more data transitions in these later graphs, so there is a lot more to hear.

I have also recorded this with the original animations in 0.5 second transitions. Here it is:

BellHouse playing 5 Bias_Adjustment_for_tasmin_rcp85_GSOD_(annual_mean)_0.5 secs.gif Actions.mov. Sampling interval 2, Sampling threshold 1

It is much shorter, and you can hear the difference caused by the transition changes. What do you think?

Data Set 6

BellHouse playing a model from the Climadjust software by Predicitia. Sampling Interval 2, Sampling Threshold 1
6 Bias_Adjustment_for_tasmin_rcp85_IBERIA01_(annual_mean)_2secs.gif Actions

The graph in this final animation seems more varied across the models and slightly wider, therefore likely to play up to three bells but mostly 2. Again the graph rises steeply across the time series from the bottom left to the top right.

Again there is a lot of data in this animation so the recording is over 10 mins.

Summary:

There has been a lot of information to record for this play test, but there are a couple of summary points that might be useful to raise. I have already raised a few questions in the passage of this post already, but these are some more general points.

  • This kind of animation plays the BellHouse in a particular way because of the left to right way that the colour changes across the sampler through time. So it makes a particular pattern of play, especially over several animations which set a pattern that can be listened for.
  • Because of the pattern of the blue and grey graphs, (blue below the grey), you can hear the blue graph often in the larger, lower toned bells, which are represented in the lower part of the sampling grid. The grey graph, because it normally is higher than the blue, generally plays smaller, higher toned bells.

3 thoughts on “ PREDICTIA – CLIMADJUST Playtesting ”

  1. Hi there Roop!
    Regarding the question you asked me about which data would be more useful to record for us, we think that the dataset 5, with the 0.5sec GIF would be the one to go (“BellHouse playing 5 Bias_Adjustment_for_tasmin_rcp85_GSOD_(annual_mean)_0.5 secs.gif Actions.mov. Sampling interval 2, Sampling threshold 1”)
    Just to put a little bit of context on the data, what we are showing are the minimum temperatures throughout the year in Barcelona, projected from now up until the end of the century. We run an ensemble of climate models for that period of time, and compute the minimum temperatures for each day of the year, and then average it for the year. We see that it has an upward trend, as the climate change effects in Barcelona has the effect of creating a warmer climate than the city already has. In short, Barcelona will get warmer in the future.
    The relevant question is how much warmer Barcelona will get. And that’s the key question we’re trying to answer with these data. We have selected a “business as usual” climate scenario: a scenario where we do not mitigate climate change at all, so trends in emissions continue going upwards throughout the XXI century. This is what climate scientists label as “RCP scenario 8.5”. RCP stands for “Representative Concentration Pathway”, or the future concentrations of greenhouse gas in our atmosphere. 8.5 refers to the amount of heat that the Earth retains due to the green house effect. Basically this scenario is usually taken by researchers as a worst-case scenario. For this, climate models provide the output we see on blue in the images.
    However, since Barcelona is a rather small location, these results have to be adjusted: the climate models outputs are valid for larger areas, but for specific locations they are usually biased. Meaning, they give systematically hotter or colder temperatures. So we adjusted the results using Climadjust, our tool, and the results are in black in the graphics. The minimum temperatures are hotter than previously thought. And so, the situation for Barcelona is worse than previously thought.
    I’ll keep thinking on how to better present this information to people outside the climate community, as it would be good to have some context on it!
    Best,
    Juanjo

    1. Hi Juanjo,
      Thank you so much for your comments, they help to explain what is going on in the graphs much more clearly to a non-scientist such as me (especially the abbreviations!). If it is ok. I might Copy & paste some of what you have said into the main body of the post to make your responses to the questions I asked, easier to see for those who may look through this quickly.
      Also one last question: When you mention the 8.5 figure, am I to understand this as a measurement in centigrade or a kind of abstract proportional measurement? If it is the former, that would be pretty scary, no?
      Best, Roop

      1. Hi Roop!

        Yep, you can copy&paste the comments as you wish 🙂

        As for the last question… Luckily 8.5 has nothing to do with the temperature increasing 8.5 C. This is a common misunderstanding (I personally think it was a mistake naming it like that, cause it calls misunderstanding). 8.5 refers to a parameter called “radiative forcing”, that measures the difference between the ammount of sunlight the atmosphere absorbes and the ammount of energy it releases back to space (which is a mix between the sunlight that is reflected, the energy produced by us and the Earth… all of it modulated by the green house effect in the atmosphere).

        In a scenario RCP 2.5, the atmosphere retains 2.6 watts per metre squared – W/m2, a measure of energy – more than pre-industrial conditions. This parameter gives the researchers an idea of the level of emissions we have. A parameter 8.5 encompasses nearly all the models in which we continue business as usual.

        I hope this clarified! Do not hesitate to ask if you have more questions!

        Best wishes (and good look with the filming),
        Juanjo

Leave a Reply to Juanjo Cancel 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.