 Rwy'n gweld, mae'n ystyried yng Nghymru. Rwy'n gweld yma, mewn cyhoeddwch ar y projekte yn y dyfodol 2021, cael ei fod yn ymgyrch yn gweithio'r newid, a wedi'u ei ffrindio'r gwynedd. Mae'r ydych chi'n edrychw'n ymgyrch yn ei hunud o'r projekte. Yn ychydig o'r projekte, dyfodol yng Nghymru Cyslwy'n Llyfrfyniad. Mae'r eu cyfnod yn gweithio'n gweld ar y cyfnodol ar gyfer acadamaethau ac rhaid. What is pyramid? So pyramid is a demonstrated project for looking at surface water flooding in a near real time sense. So to try and take quite new, recent, within an hour or so, rainfall readings, and other climate readings and then predict surface water flooding. So I'm going to go through these slides quite quickly. There's a lot of them. felly, mae'r cyfnod gyflaenwch yn gwybod, ac mae llawer o'r gwaith yn hynny, ac mae'r ganchudd i'w wedi bod yn gwneud ein gwasanaeth i'n gweithio. Felly, mae'r dda, mae'r ddechrau'n dda i'r gwaith i ddechrau'r ddaf, fyno'r ddechrau'n ddechrau'n ddechrau'n ddechrau'n ddechrau'n ddechrau. Felly, mae'n olygion i'r ddaf yn cywir i gweithio ddweud o'r canffordd a'u ddweud o'r canffordd, ddechrau'r cyd-dweud cyfanswad Cymru yw ddau connectsau. Mae'r sefydlan yn cynhyrchu cerdd, ddiwrnod y gallu cymdeithasol o ddau cerdd. Mae'n bach divideur Immunol, mae'r cyd-dweud. ar y prosiect. Dyma'n cyllid yn ymddangos, yn ymddangos, i gynhyrchu bryd, gyda'r bach cyllid yn ymddangos yn ymddangos, a'r cysylltu'r cyllid yn ymddangos, drydd ymddangos, i gynnwys i Llyfrgell, sy'n gwybodaeth ac i'r cyffredin iawn. Pryddiad yw'r modd yw'r modd ac yn ymddangos, called high-pims, which Loughborough University maintain. This predicts flooding dynamics, so water depth and speed from rainfall to inundation. And it also includes a debris modelling component, which is a new component that was developed as part of Pyramid, to allow primary cars and other vehicles to be swept along in the floods, which hasn't been done before. On top of that, there's another model called Sheetran, which is a hydrology model, which models the broader catchment area, so Sheetran effectively provides boundary conditions for the high-pim simulator. So this simulates the entire water cycle and river flows. This is a much broader resolution, about one kilometre, whereas high-pims is about two-metre resolution, so they operate in different ways. And Sheetran feeds into high-pims. So data, new old and hidden data, and we looked at, we have a real-time sensor data feeding into the model. And there's a machine learning component, which detects where vehicles might be, so they can be fed into the high-pim simulator. That takes satellite data and generates boundary boxes for vehicles. You won't be able to see any of that, but I've got this on. I've got some bits of paper, which you can take away, not on for everybody in the room, but this is the overall workflow. So it starts at the datasets, extracts data, does some quality control on the data. Then we have all the simulators and the modelling, and then we have a visualisation component at the end. So what have we actually produced? So we worked with the Daphne platform at the SDFC did kit and Bethans in the audience somewhere. So we built this entire thing in Daphne. So it's been quite a collaborative exercise between us, and Daphne, we actually couldn't have done it really without the availability of that platform. And that's a screenshot from the overall workshop, which I can show on my laptop at lunchtime. So one of the things we found that collecting data live has a lot of problems. So we collect radar data, has spatial quality issues, has time quality issues, so it has just missing readings. So this has come from the Environment Agency and the Met Office. We also have a lot of point data, which also has quality issues and API issues. Often it's not there, the API has gone down. And there are varying quality and standard. We have local rengages as well. So one of the big things I think that we achieved was to create a prioritisation algorithm, to make these different environmental data sources and combine them into a coherent and robust data source that can feed into the simulators. So I'm not going to dwell on this. Anybody who's interested can come and talk to me about it. The object detection works quite well, but it doesn't do real-time satellite streaming data, and the Daphne people had a fit when we suggested that we might do that. So although the algorithm works well, there is a big issue with using this kind of data in any kind of real situation at the minute. So this is one thing we've learned. We can do it, but the infrastructure isn't really there to actually do it, meaningfully. But the models integrate well, so the two-tier model system does work very well. There's a result of a simulation in Newcastle. Newcastle has had some flooding recently. Daphne is a really powerful facility. It allows the construction in a very modular form of different models and data processing algorithms that can be chained together, but also it can be publicly released so that the wider Daphne community can use all the models that we've created in the datasets we have. So it's a bit of a general project. Outcomes, as is usual, is an academic project. It still feels like work in progress. So the main feelings I think we have about the project is that it's shown that if it's simulated like this, there's the kind of digital twin that the environment agency might be interested in. There are some big issues to do with data quality, streaming data, just handling the data, compete time. So it's proved the feasibility of an approach like this, but the next stage in true academic form opens up the possibility for the research. I kind of have to say that. I think there are lots of data challenges, lots of compete challenges associated with making something like this actually work properly. And just a minute over time, two weeks ago we had the final, the project's finished two weeks ago and three weeks ago we had the final stakeholder workshop and the Flood Forecast and Centre and National Farmers Union and the Environment Agency were quite interested in this as a general tool and kind of what they could learn from it and where it could be taken further. The other industry bodies who thought it might be a bit too heavy weight for them and it might be too difficult to work with, but there's definitely, there's definitely interest in propelling the project forward, especially with using a more coherent data set to be able to feed in the simulators. And yes, one of the other issues is at the moment Daphne is kind of a closed environment. Almost being invited users so if this was to become like a national resource of some description I think the Daphne platform would have to evolve and that would obviously open up issues of data sensitivity and accounts and all of that kind of thing. So that would be a challenge for that STFC facility as well. But that's very, very short introduction to Pyramid so please come talk to me sometime during the day. Thank you.