 Okay, welcome back. I thought it would be a good opportunity to try a little bit of a Game water management game, which is a risk-based game using seasonal forecast of hydrological input to a dam So it's quite topical to this workshop as well. It's It will work in a way that you will all have a sheet That you will fill in but don't do it yet. I will explain how to do this as well the information you have Before you start the game is that you are the newly appointed water management manager for Lake Duel and This it's a reservoir That it serves two primary functions. You are you need you need to make sure that you have enough water supply to Swooftown How does some coming that one and you also need to make sure that the safe town is not flooded So you have two basic or two objectives of this this exercise So the What need to remember? I think it's written here as well somewhere that you you You would like to see that the reservoir is full by the August 1st at the end of this rainy season and That means 500 mega cubic meters. That is your game. That's your that is your aim for the exercise So close as possible to 500 That makes means that all the residents can counter water all summer But you also need to keep the releases below 60 mega meters per cubic mega cubic meters each month Otherwise safe town will have damage to their homes So these are two numbers to remember five hundred and sixty This read written here as well that you you're in the first column here You see that your lease can only be between 15 and 60 So you don't have to remember it. It's written here as well The 15 is the lower amount you can release and 60 is the highest amount you can release This is typically the way that man that reservoirs work Anyway, they have restrictions on how much they can release and how they also have a minimum they can release And you also have that that 500 up here as well. So you don't have to remember this these figures So you will be the matter what the management for this for the season from April to August And you will be for each month you will be presented with the probabilistic forecasts of inflows So at each month you have to make a decision on the reservoir releases for the remaining months This means that you only need to make a decision for the actual month But you can also write down if you want to make sort of a plan for the next coming month So you need to make make sure that you have enough water for to reach 500 cubic meters per mega cubic meters But you cannot release more than 60 So if you exceed the level of 500 you're fired Simple as that you lose your job And you have to maintain as I said 15 Minimum of 15 mega cubic meters So we'll have to be played this game as a second to competition as well. So wherever is part Whoever is in the game after all we reach first of August and has not over topped 500 But it's as close as possible to 500 will win a prize So just an example before we start just make it completely clear you need to pencil as well. So I think instead that So this is what the information you will get Now we are in the example box here, which is a gray one. It's already filled in But just to see how it works. So you will be presented with the forecast in this case It would look something like this. So you will get a probabilistic forecast of inflows from March to up to July as you can see here You will have the The max and the minimum which is the 5 and 95 percentiles. These are the numbers here given You will also be given the the median and the 75th and 25th percentiles. You will have a range of possible outcomes So this is your seasonal forecast and you have the starting of 460 So here you are also given that these sort of in numbers in this case your your Forecast this is the median given here in this piece of paper here Or you know, sorry, this is what you write down as your your release schedule So for each month coming off the March, you will write down your release schedule. You can change it at any time You just make sure that It's the actual month for for April. That's the most important one Then you will be given the actual inflow In this case it was five So in this example you you release 15 You received five so your total would have been 460 Plus five minus 15 and that's 450 So you do the math yourselves You just fill in the table and then you transfer for this new number down to the next column here It's already done for you an example, but you have to do it yourself later So as everybody understood this is the clear any questions at all Yes, I will give it after you've written. I will give you the forecast You will write down here in the column. How much you release I Will give you some time to think when everybody's filled For each month step by step, then I will give you what the actual inflow was You do the calculations and see how much of a gained or lost The actual inflow is the actual inflow Yes, I will after you've done your your release schedule I will not tell you beforehand. That's the part of the game you make a prediction Based on this but you will actually get this you can follow that the Prediction they were given as well if you trust it if you trust the seasonal forecast That's that's the whole point. We don't know actually the skill of the seasonal forecast at this point Yes, I would say it's optional It's not necessary that you can do it. You can do you can do it Not why the only important one is actual for the April is only the April value you will use So the rest one is just you can fill it in if you want to But don't care about that if you don't want to just fill in the actual month That's that's what can do as well you can see if what you exactly yes another question All I can say that is you will be only be given the 95th and the fifth percentile So you will not will be basically is covering 90% of the forecast But you have to realize that given if you have a 50 member ensemble for example The observations will be outside of that predicted range two times out of 100 So you you will always have a case when the observed should be outside the forecasted range. That's all I'm saying So don't use the model with care I would also like to have a volunteer to represent the group as well That would be give the play the game a bit differently Any volunteer if I don't get a volunteer, I will pick a volunteer anybody feel up for it a forecaster in the room Okay Okay, we have one here. You don't you don't have to bring your sheet. You can just stand here So this is our volunteer. Okay, so let's start You understood the game. Yeah, so it's April 1st your reservoir is at 450 mega cubic meters And this is the forecast you're given for April 1st. So fill in your weekly schedule If you you can fill it all the way in if you want to or just for April You cannot go below 15 and you cannot go over 60 in the release schedule So I give you some time to think So for the first month for April we have 8 10 and 12. Yes, and then for May June and July We have 36 of what we can expect here that we will get some sort of Larger inflow in in June it could be as much as 130 If you're unlucky No, no wait wait, you will you have it Don't worry, you will be played different kind of game This everybody's ever cheated by the room by the way, yeah Has everybody filled in just for April just for April just the first here. Do not feel further than that So you don't you don't fill in for first of May yet, you will get another forecast the first of May So just fill in this Yeah Okay, ready It's good to take your time to make a proper decision. That's not like that 15 Minimum 15. Yes And you want to close it get as close as possible to 500 Okay, I'm going to go to the next one here So please fill in but you can still see the the forecast here just for a volunteer you will be given three options here option a option B or option C and the evolution of these options are using the medium Which is a dotted line and the minimum and maximum the forecast are the two Solid lines So you can go for a release schedule here option a given that given the forecast Which is kind of intermediate? You have a very careful option here cautious option number B or a more daring option number C So the question to you is which one do you go for you don't have to The initial volume of reservoir will be 460 and the final capacity should not exceed 500 and here It's a high probability of having flood So at the first of the season, I'm going to release as much as possible water to have a conservative Decision on June. Okay There'll be option B as much as water So that's option B then yeah, okay Everybody ready. I'm going off now. You can know the answer for the for the So you go for option B? So the April inflow was 18 So you can fill in the actual inflow for April and then do the calculations depending on how much you released and You will get a new number has anybody over topped everybody is still in the game. Good. Everybody still have a job No, you can't over top yet. It's impossible our volunteer and you released 55 So your volume you started before and 50 The inflow was 18 and your outflow was 55. So you are now 413. So you're quite safe. You still have a job So we go to the next forecast This is first of May The reservoir here now is for our volunteers. You all have your individual reservoirs and Here's the the forecast for first of May So you do the same thing now you fill in how much you want to release from May and you have the May forecast here. You can also do the schedule to in the the rest of the months if you want to just to compare with your previous decisions Okay, everybody ready. Oh, yes, of course They will still fire you if you if you're over top Okay, I'm just gonna go to the next slide You will still see the forecast here, but let's see what the volunteer says You have not three options again option a option B option C You starting from 413 and this is the evolution giving your option a now is the very conservative one very cautious one option B is a bit immediate option C is a bit daring Yeah, this this is giving you can you can use this as your help as well So this is what the range you will reach if you do this one here is your range for this one and here is your range for the It's more likely that I will I will over top but for this case the band shows that I can still have the range of Uncertainties not over top. Okay, you've been very cautious. That's good. So you play for option a Okay, let's see what happens the inflow was 55 and you released so now you do calculations of your of your inflow and release Anybody over topping yet? Everybody still has a job. So what about a volunteer? You still have a job definitely you released 59 and you you received 55 so you have 409 So you still have a job so go to the next let's go to the next forecast everybody ready Okay, here's your forecast for first of June. So now the the Forecast is focusing in on the higher range of the probability So you see now that they have changed the updated if the forecast a bit It's much more certain that you will have a large inflow in June But how large we don't know so Please fill in your schedules. Okay, so let's go to the volunteer here are your three options now You start in the 413 so you can still have a very conservative release or you can Follow the two other schedules if you want to Option B is just that you here's your release 1650 they will get you well below the threshold But you will also be well below 500 given this forecast Option C is a little bit less conservative or cautious. So you will end up somewhere there By options a is 60 and 50 Yes for next month But remember you want to stay as close to 500 as possible as well if you want to please your yeah Yep So which one do you go for so for 409 plus 110 will be 520 for the worst case scenario and If I'm going to release 60 it will be like 460 and In July, I will have the 52 I will over top No, you can you'll be releasing July as well. So you can release the same amount Yeah, so if you release 60 plus 50 is 110 and then you will you will be Still within the threshold. So it's no problem 409 and I will release 45 and The flow will be 110. So This one is little Yeah, I Will go for option B option B again Here we still have some Flood But you see here that you will it's very cautious You will stay but way below the optimum given the forecast you still would be not over dropping here still Okay, but it's your choice you go for option B Okay, so the inflow was 120 actually more than predicted so we're a bit outside the predicted range there So has anybody overtopped Hands up over topping. Okay, so you're you're fired Everybody that still hasn't overtopped or in the game, but you can continue playing anyway So how many people are still in the game? Only two How many have not overtopped Okay, good. So let's go on a volunteer is In a good position you release 60 you had 120 inflows. So you have 469 So let's go to the next book last forecast for first of July Your your reservoir is 469 you have your own reservoirs out there. So here's the here's the forecast so fill in your schedules Oh, sorry, the numbers are 19 for the lower 21 for the median and 28 for the for the top one So given this forecast and you have 469 you have three options. They are still the very cautious one the Not so cautious one and they're going to release As low as water as possible. Okay, which is 25. Yeah Everybody fill in their schedules Okay, so let's go option B So the inflow was 22 and How many are still in the game and our volunteer is still in the game? Well, we don't know yet. You have 466 Okay, so they we have reached the game over now. So now we'll see who is closest to 500 which still hasn't overtopped You got 500. Wow. I don't think it's happened before Well done so you will be Get a bonus from the from the money we promoted very good Also think that our volunteer did an excellent job No, he's fine, he's fine, but here you go a volunteer as well. Thank you very much. Yeah, so that's what's the game. We don't look good community. We have a Organization called HEPEX. I don't know if you've heard of this before It's called the hydrological experiment hydrological ensemble prediction experiment. So we are doing these Sessions at EGU and we try each year to have a different kind of game Which is sort of playful but still has a topic on on the probabilities and seasonal forecast. So I Can actually If you if I collect these afterwards, I would use them for some research as well so you also been part of a research experiment, which is nice to know and That was my game first game I do have another one if you want to do it or if you want to go to a discussion Play a game. Yeah Okay, let's play another game. It's a little bit different, but it's It's also played at One is HEPEX sessions. I have another Schedule for you or another paper for you to use as well. So it's one each This game is more focused on how to interpret ensemble Forecasts, so it's called a peek box game because it's based on research done by Massimiliano Zappa and Catelle Stee in Switzerland where they developed a technique to pick the best Ensemble member out of an ensemble So the way that this game works is that it based on the wisdom of the crowd Which is the process of taking into account? Sort of a collective opinion of a group of individuals rather than a single expert to answer a question And it's been shown that a large groups aggregated answers to questions involving quantity estimation General world knowledge and spatial reasoning has to generally be found to be as good as and Often better as the answer given by any individuals within the group and this is a paper from 1907 by guy come Galton published in Nature where he He asked 787 people to estimate the weight in pounds of a European muscox, which is a part of the European bison and They all gave their estimates in pounds and he collected this information and The median of the crowd wisdom or the estimates were 1,207 Pounds whereas the actual weight was 1198 So that gave an error of the crowd of less than 1% And this has been used in other things as well Maybe you've seen this game called who wants to be millionaire so there's one option there to ask the audience and They are actually correct in 90% of the time and this is quite amazing because they are Questions can be anything The people are just random group of people or happen to spend a weekday in a TV studio. So they do they do have some general common knowledge common wisdom and We're going to try to play this game of wisdom using An ensemble forecast Which is also sort of a collective ideas of what's going to be the evolution in the stream So you will be given an example like this. You will have the runoff the measured runoff here in blue And the yellow one is the runoff used using observations So the presentation temperature observe observations up to the starting date and then you will given the discharge Ensemble discharge So the question is we're going to try to answer this as a crowd of members have more wisdom than a single value forecast So the game is played like this for each time On the sheet you have four different cases you will start with them One by one, so you will fill in how big the observed peak discharge will be and at what time So for each case you will put in an X mark on the sheet where you think that the How big the peak discharge will be and at what time it will occur and then you will be Filling in this these numbers here, which is a coordinate system. So you will fill in G 26 in this case and So you're filling your coordinates for each case So I will give you four examples. You just do one at a time Everybody understood how it works So you do your estimate or where you think that the Magnet the maximum peak discharge will be and what time and you fill it in the paper and afterwards I will give you the answer, but you will be doing all the examples first you can choose freely any points any points, so this is just the the Yeah, the ensemble forecast just to help you to try to estimate where you think that the peak Discharge maximum and the minimum will be or maximum and the timing will be So it doesn't have to be on a certain forecast It can be anywhere and the catchment for your information is a small catchment in Switzerland 186 kilometers is quite flashy forecast catchment and We have the first forecast here Was issued on 25th of April 2013 and this is your your forecast So you're pleased for the first example fill in when you think that you pick discharge But be in terms of a mountain time So I'll give you 60 seconds This is based on a 13 member I think it is Both both of them timing in a month. So you will fill in somewhere Here we use do you think that the Here you have the discharge amount and here you have the timing. Okay, everybody filled in their sheets somebody still thinking No, okay next Forecast is for 26th of April now. You have a new Forecast system there. So please fill in when you think that the peak discharge Will be and how high it will be? Is anybody still thinking? No, there's no penalties for this one. It's just getting as close as possible to the actual Okay, I move to the next one. You have done your papers there as well number three if you turn over your papers Okay, so let's move to the last one. Okay, so everybody ready? Okay so We can use a system or a method called peak box to give an estimation This is what the base this this exercise on so that the way you pick box work is that you you mark the highest Peak discharge of each individual member Then you draw boxes around the quantiles in timing and magnitude So here are the box quantiles for the piece for the 75th and 25th quantile in terms of discharge and also for timing and You also have the the the maximum amount now if you put in the medium for the for the for the discharge peak and the Timing you will end up with the best estimate so the intercept of the median and peak timing a peak magnitude so this is the way that the the peak box works and Is this has been applied to the same situation as you had before as well So for each of these ones you will see you will get an estimate using this peak box peak box method So for the first one, here's a solution. Can I just ask you to shift with your neighbor so you don't correct your own sheet So now you're going to correct that your neighbors sheet and you're going to count the number of boxes you are away I will explain how this work as well. So you correct somebody else is okay No, we're going to compete against the peak box as well, so you will get yes, you will see who one is Closest to the actual observations and we'll see how if you can beat this peak box Method as well. So for the first case the peak box Estimated that the discharge would be on the J 36. That was the estimate The actual solution from observations was at age 35 That was the observed peak discharge So the peak box areas was one box in timing and two boxes in volume It's a total of three boxes. You do the same now for you for your sheets You count how of your how of your neighbor is in terms of timing and Volume and then you add it up So it's not this not a diagonal line. We're just going like you know Vertical and then horizontal Yes from the observed So the observations is how you're comparing with age 35 Sounds like a bingo So age 35 is your observations, so you see how much your neighbors were off from age 35. So anybody still filling in So that was for the first first one. So just do the first one Doesn't the peak boxes is not to your concern. That's just a comparison. Sorry Oh, sorry that so the error in timing in the error in a peak That's an error in the peak box No, the error in timing and the error in peak So the solution was age 35 and you're down about six twenty six Or in total I mean. Oh, that's what for the first one and then it's It's gonna be nine for that one So you sum up the errors in magnitude and peak Somebody's still working on this Is everybody understood somebody who's not filled in yet So you compare your you compare this one age 35. We bought your neighbor put your neighborhood and then you count the number of boxes In each direction, this is this is the peak box estimates, but you would have put somewhere here So you have to count the number of boxes in both directions And then you sum them up peak is in the amplitude exactly so the amplitude error and the timing error, okay So the next solution for next forecast So the peak box most estimate which we're competing against was age 30 the actual observations was g 31 So you compare your neighbors result with the with the g 31 Just for comparison had the peak box error was only two boxes away. So g 31 has everybody filled in Okay, let's go to the next one So for the third forecast the peak box estimate was a deep 26 and the solution from observations was g 27 So in this case it was very very close Okay, somebody still so let's go to the last one then so the the peak box estimate was at one I 24 Where's the actual observation was as age? 23 age 23 is the correct one. So age 23 is the correct answer. So everybody filled in Somebody's still filling in so what you do now you sum up then the neighbor errors in total from all of the four cases So the total error from the peak box estimate was eight boxes Anybody beat eight boxes less than eight boxes error in total for all the four cases now No one Has anyone had less than than 10 errors? 10 nine, okay, we have a winner very good An applaud that was very good. So basically you're as good as an automated system to come So the price was was some Swiss chocolate. I'm not sure this is Swiss chocolate, but So if you want to read more about this and also other blog posts about hydrological Experiments you can visit this website as well and you find both of these games here You can download them and use them if you want to use them for your own purposes for education and so on so and here's also the the reference to how this This peak flow box approaches is described as well how you can use it so But the forecast is still better or almost as good as an automated system is good to know I don't think that anybody has to be beaten the peak box. So it's very very close Good effort. So that's it for me