Choosing the right member is a piece of cake
by Massimiliano Zappa, a HEPEX 2015 Guest Columnist
Dear all, I was somehow convinced by the maintainers of this nice blog to make myself available as guest columnist for 2015. The task will be rather challenging since the predecessors in 2014 did a really nice job and compiled a series of interesting contributions. This convinced me to join Twitter and follow all the additional stuff emerging from @hepexorg.
Back to the 2014 columns, it was quite a comprehensive and well balanced mixture of interviews, presentation of current work, opinions and “political” statements. If I have to look back, two posts come first to my mind: the “gender balance post” by Bettina Schaefli, and the “11th advice post” by Anders Persson. You can comment here below about your favorite posts by the 2014 columnists. No matters what you will choose, no post will be superior to the one published by Mc Fools’. I am still waiting for his second post. If he (or she) is reading: “Do you have Twitter? I want to follow you!”
So, a long introduction to come finally to the content of my first guest columnist post. Please read it to the end, even if it is getting lengthy…
The three kings cake problem
Some of us have this problem once a year on January 6th (Three Kings Day). Now, if you can look at the picture here below without getting hungry, you have already a problem less than I do with it (by the way this was my Home-made Three Kings Cake in 2014). The second part of the problem has of course an obvious link with HEPEX, while otherwise it would not make any sense to present this here. Our 2014 cake had 8 small pieces of bread and a big bread in the center. In one of the small breads a figure is hidden (usually a king, but recently also female figures are found, just to come back to the gender balance issue). The person selecting the right small bread will be king (or queen) for the next few days. There is little knowledge about a priori distribution. What is quite sure is that the figure is NOT in the big central bread (only if someone manipulated it). Another fact is that in the last 6 years my daughter won it 4 times and my son 2. There is some evidence that young participants have a bigger chance to win (for sake of peace within the family, you might argue …). How can I prevent this?
[Here a recipe]
| Now, the funny part is coming. I talk now about the second figure here below, and for that I re-arrange the text I just wrote about the first figure (differences are underlined and in italic, some links to past posts are hidden). |
The ensemble flood forecast problem
Some of us have this problem every now and then when the spread of their real-time forecast is high. Now, if you can look at the picture here below without being responsible for an alert, you have already a problem less than I do with it (by the way this was our Home-made forecast on December 15th 2014). Our forecast has 16 members and we plot the median in the center. The users expect that one of the members is THE correct forecast, and he (or she) thinks we should know which one it is. The person taking the right decision will be king (or queen) for the next days (see also a previous post). There is little knowledge about a priori distribution. What is quite sure is that the right decision is NOT in the median (only if someone post-processed it perfectly). Another fact is that in the last 6 years (our operational system runs since 2008) my end-users where quite happy with the system, but still ask me about telling them the RIGHT member. There is some hope that young scientists will join HEPEX and contribute to this fascinating science and come closer to give the users an adequate answer to their expectations (it is maybe utopic, you might argue …). How can we facilitate this?
[Here a recipe]
I wish all of you a nice start in 2015 and that you can choose the RIGHT piece of cake and give the RIGHT support to your end user.
PS: I will put in a comment an update of the statistics within my family. Cross fingers …
Massimiliano Zappa is a HEPEX 2015 Guest Columnist.
January 5, 2015 at 18:50
ready for tomorrow! http://t.co/1xv0uSyaJ6
January 6, 2015 at 00:07
This reminds me a bit of a post I did on modelling versus baking… http://tompagano.blogspot.com.au/2012/10/10-tips-to-being-better-cook-and.html
It was based on a user manual for a forecasting software, written in the form of a cookbook, including 10 commandments
1 Skim the recipe at the start
2 Plan your meals
3 Taste test at every step
4 Clean as you go
5 Use high quality ingredients
6 Be careful when using substitutions
7 Ingredient preparation determines the success of all that follows
8 Know your equipment, have the right tools
9 Master the way things should be done before experimenting
10 Balance taste with presentation
Maybe #11 should be “save the best dish for your most important customer”
January 6, 2015 at 07:58
Dear Tom,
hope Hannah won’t read this, because she will then remind me that I still have to write down a contribution dealing with a cook-book (for operational forecasts). Guess we should have a chat on this, I’m sure the ideas you presented in your blog would perfectly fit in this.
Still have to complete a contribution for IUGG, then I switch to the cook-book
January 6, 2015 at 06:16
ready for saturday? http://t.co/8Vh8DtCJov
January 6, 2015 at 08:11
Dear all,
the decision is taken, my daughter found the right member, again. Guess if she can apply for a forecaster position at EFAS or ECMWF.
Most difficult: she found out the right member even before evaluating all members.
See picture here: https://twitter.com/Hydrology_WSL/status/552360460256030720/photo/1
Good luck to all of you still having a chance for being king or queen today
January 8, 2015 at 23:36
Very nice post! It was a good comparison.
As we are talking about analogies, It inspired me to do something similar. I will post here my version for a recent popular song. Maybe it will contribute:
All About That Spread
Because you know it’s all about the spread
’bout the spread, the ensemble
It’s all ’bout the spread, ’bout the spread, the ensemble
It’s all ’bout the spread, ’bout the spread, the ensemble
It’s all ’bout the spread, ’bout the spread, the ensemble
Yeah it’s pretty clear, it’s not the member one or two
But I can read it, read it like I’m supposed to do
Cause I got that uncertainty that all the operators rate
All the uncertain members in all the uncertain places
I see in the journal of hydrology and in the workshops
We know that that uncertainty is real
We cannot make it stop
If you got an ensemble don’t give it up
Cause every member is imperfect
From the bottom to the top
Yeah, my boss told me to worry about the spread size
He says that I will need to run it again at night
You know I won’t be able to sleep until this rain fall
So, if that’s what’s we’re into
Then join HEPEX and move along
Because you know it’s all about the spread
’bout the spread, the ensemble
It’s all ’bout the spread, ’bout the spread, the ensemble
It’s all ’bout the spread, ’bout the spread, the ensemble
It’s all ’bout the spread, ’bout the spread, the ensemble
Original music: https://www.youtube.com/watch?v=7PCkvCPvDXk
January 9, 2015 at 11:04
Wow, thanks for this. The video you linked has been viewed more than 470 Mio times so far.
Imagine if the HEPEX community would perform your version and upload it …
Maria-Helena, we can surely make something like this in Prague instead of a game!
January 9, 2015 at 13:48
I will do my best to convince the organizers.
Meanwhile, HEPEX members, remember to submit your abstract to our forecasting session before January 31st:
https://hepex.inrae.fr/hydrological-forecasting-and-predictive-uncertainty-advances-and-challenges-of-transferring-science-into-operational-practice-iahs-workshop/
January 14, 2015 at 15:52
On 5 January Massimiliano Zappa made a very nice and witty contribution in his first guest blog. It was followed a sequence of equally witty and nice comments.
But as you may know, in Hell the man responsible for the entertainment is a Swede and being one I have to interrupt all the fun with an austere comment:
Hydrological (and meteorological) forecasting is not a merry TV-Game Show where the participants are supposed to come up with The Right Answer (whether Oslo or Helsinki is the capital of Finland and where the “ensemble median” of Stockholm will not work).
The ambition or purpose of the different ensemble systems is NOT to provide at least one member that may hopefully be “right”. Even if not a single member is “right” a skillful ensemble will still provide useful information in terms of ensemble means or medians and, of course, probabilities.
That is what we have to get used to. To look for the “best” or “right” member is just to put the dear horse in front of the newly purchased automobile.
January 6, 2019 at 11:04
2018 update, I am finally king!
https://twitter.com/Hydrology_WSL/status/1081816260558622720