Google Summer of Code 2025: Cloudcasting UI

[peterdudfield]

This space is for you to ask any questions you have about this project. We're here to provide clarifications and help you understand the project's goals, scope, and requirements. Feel free to ask about anything that interests you!

Please note that this discussion is for questions and clarifications, not for formal applications.

Project Description

We've been working on forecasting clouds using satellite imagery, and we would now like to take the next step and make an exciting visual representation of these cloud forecasts. This is an exciting and innovative project in the weather forecasting space.

We expect this to be implemented and deployed as a new, separate application to our existing frontend apps, but ideally would be written in a tech stack, methodology and code style that is as compatible as possible with our other frontend repos, particularly the two apps in the quartz-frontend monorepo.

Prospective tech stack, in brief:

• React (potentially Next.js as a full-stack solution as there will be some data handling and fetching, but also open to either another similar, slimline(ish), unopinionated React framework for the right candidate , or a vanilla React app + backend server – the main driver is to get this off the ground within the timeframe in a considered way)
• Typescript (non-negotiable, sorry not sorry!)
• Mapbox (again, other library ideas are welcome and encouraged to pitch in regard to specific needs/requirements, especially from research in how to visualise the output xarr data, but this is our current default.
• Depending on the frontend chosen, there could be a need for a backend server, for which I would suggest Node.js or similar.
• Vercel for deployment, either of full-stack app or just the frontend; we also use Render as a backend PaaS currently.

For clarity, always starting with the latest versions of existing packages using up-to-date documentation to set up necessary dependencies is expected, so you should be comfortable tweaking configs to plumb together the new application, but otherwise this should be possible with minimal custom implementations.

This is important to us here at OCF to try to make our codebase easily accessible to future contributors and/or open source users.

Expected Outcome

A UI that clearly shows the output from our cloud forecasting service.

As mentioned, this is brand new territory; the exact criteria are expected to develop during the initial stage, so there will be an element of end-to-end design and development in this project.

This is overall a proof of concept, and just happens to be a perfect time for a GSoC candidate to take a brand new resource and show it off to the world.

Other Key Information

• Expected Size: 175hrs
• Skills: Plotting and visualising in a Javascript framework, understanding of user experience, good problem solving skills.
• Difficulty level: Medium
• Related Reading: https://openclimatefix.org/projects/cloud-forecasting
• Potential mentors: @braddf
• Example .zarr data download link: https://drive.google.com/file/d/1eoQgBBkZ14F1nlqiSKHI_czGyJshE4dZ/view

Timeline

Details of the GSoC process can be found here for convenience: https://developers.google.com/open-source/gsoc/timeline

As of Mon 24th March, we are now in the "GSoC contributor application period begins" stage, so head over there to check out how to apply, and let us know any questions you might have on this thread to fill in any gaps from the above description!

N.B. if you find it useful to put down ideas on Github to share with us to clarify your questions or check your early thoughts, that's a-okay – but there is absolutely no expectation for you to draft a load of code in anticipation of applying this project.

GL;HF!

[suvanbanerjee]

hey @peterdudfield i am interested in this project and would love to contribute for the same

A bit about me, My name is Suvan Banerjee, currently a 3rd year college student from Bengaluru, India, i am very active in open source space from past 1 year and have got a few merged and open pr in open climate fix as well, i am a maintainer of open voice os main website and have sone work in webspace, I am also a full stack intern at a startup so added bonus for that, I will try to come up with a draft pr and will share with you either here or via email, please do let me know what you prefer

[suvanbanerjee]

Hey, @braddf thanks for the clarification, coincidently 😉 I have chosen a similar stack as you mentioned and will be done with my draft proposal by this weekend :D
some doubts,

• Do I need to think about user auth like login and emails (like quartz solar does now)
• Is it going to be like a proof of concept rather than a product that the end user will directly use ?
• Should it also show the predicted solar output like quartz solar does now, or only cloudcasting ?

[suvanbanerjee]

@braddf openclimatefix/cloudcasting-app#18 can you check this issue/question,

TLDR - do we have a sample zarr file that model produces so we can know what we have to deal with

[braddf]

Hey Suvan, thanks for the Q's – auth will probably come down to what we can fit into this project and the direction it takes, how much work, etc. and similar for the product overall. It is definitely a proof of concept, but it should be treated as if it will be used by, and be useful to, users at the end of the project. As to what else we show on this UI, that's to be discussed as there are a lot of cool angles we could take this in with the right candidate.

Re the file, I've updated the description with a download link, so check that out whenever you like! ✌️

[suvanbanerjee]

@braddf how about building a custom tiling server, cause looking at the requirements (zarr ie 70mb) we need a solid pre-processing in backend with caching and actions or cron jobs, what are your thoughts on this idea. as its a medium project implementing this will be on a tighter schedule, I'm wondering if I might be overthinking it or making it unnecessarily complex. Let me know your thoughts!

[braddf]

Hey, this may or may not be overthinking – I'm not the best person to judge that myself! 😂 But to me, there could definitely be an MVP version of something like this, as the data processing aspect of this project will be one of the most important aspects to making this work well for the end user.

[MAYANK12SHARMA]

Hi @peterdudfield,

I wanted to get some clarity on the tech stack for building the UI. Are there specific frameworks or libraries you’re considering for the frontend?

I’ve been exploring a few options and was wondering if we might consider building the system entirely in Python. For example, we could use Dash or Streamlit for the frontend, paired with Flask or FastAPI to serve the data. For visualization, libraries like Plotly, Seaborn, Matplotlib, or Cartopy could handle cloud forecasts, including time-series animations.

Alternatively, are you thinking of using React for the component structure, with libraries like D3.js or Plotly for rendering the forecasts? And for handling satellite data (like NetCDF), would you prefer processing it server-side with xarray in Python or directly in the browser using something like netcdfjs?

[braddf]

Hey Mayank, thanks for the questions! Apologies for any confusion surrounding requirements/expectations for this project – this will be a full-stack project in line with our existing technologies, so Javascript over Python. As for libraries, I've added some further clarification (hopefully) in the description above, so have a look if it's still something you'd be interested in, and let us know if you have any further questions. Thanks! 👍

[MAYANK12SHARMA]

Hi @braddf,
Thanks for the clarification! I completely understand the project direction now—moving forward with a full-stack JavaScript approach using React, TypeScript, Mapbox, and any necessary backend (like Node.js) aligns well with my expertise. I’m excited about the prospect of building a new UI that not only integrates with the quartz-frontend monorepo but also provides a clear, engaging visualization of the cloud forecasts.

I do have a follow-up question:

• Regarding data handling for the satellite imagery and forecasts, should I expect an existing API, or would building a dedicated data-fetching layer be part of the scope?

[braddf]

Hey Mayank, this is still TBC – we'll need some kind of datalayer on the server side as the data is pretty chunky, but a few ways to do this with different pros and cons. For now, the data is in an S3 Bucket and needs retrieving in some way, shape or form to be used and displayed, whether that's a dedicated (small) API, a backend service within this new app, or otherwise. 😊

[yuvraajnarula]

Hi @peterdudfield and @braddf,

The cloud forecasting visualization project immediately caught my interest, as I’ve had experience working with meteorological visualization systems.

A few considerations that might be worth exploring:

• Data Representation Challenges: Cloud forecasts inherently carry visual complexity—how are you approaching the representation of uncertainty? Techniques like probabilistic cloud edge rendering using WebGL fragment shaders can visually encode forecast confidence through opacity gradients.

• Temporal Resolution Tradeoffs: For satellite-derived cloud forecasts, the interpolation between prediction steps plays a crucial role in user perception. At what intervals will forecasts be available, and have you considered morphing techniques for smoother transitions?

• Performance Optimization: Rendering high-resolution global cloud forecasts efficiently often requires tile-based rendering with level-of-detail management. Is the data already structured for multi-resolution access, or would that pipeline need development?

From experience, WebGL-accelerated approaches—such as volumetric cloud rendering with custom shaders—have been effective in maintaining interactivity, enabling smooth time scrubbing through forecast sequences, and allowing comparisons between model predictions and observed satellite imagery.

Would be curious to hear if you’re considering incorporating additional meteorological data layers beyond cloud coverage in this tool.

[braddf]

Hey Yuvraaj, glad to hear your enthusiasm – should be an interesting one project for sure!

Re your questions:

• Data Representation Challenges: This is a great idea and good to be aware of from day one, nice. Our plan for the MVP for the Cloudcasting forecast won't include any probabilistic values or ensembles to start off with, but this is definitely something that we would be interested in for the future; my initial thoughts would be to see what standard mapping libraries like Mapbox can offer (probably with a small layer of data manipulation server-side) and then reach for something a bit more powerful as needed.
• Temporal Resolution Tradeoffs:For temporal resolution, our current use case is linked to our solar forecasts, as a means of anticipating and demonstrating the ebbs and flows of solar generation due to clouds; as both the solar forecast and the new cloud forecast are split into 30-minute summative chunks of generation and end-state cloud predictions in the same window respectively, I think this sharing this time-step is actually suitable to orient the user to the data shown. This could be a consideration for future applications though, so would be up for chatting more.
• Performance Optimization: this shouldn't be too strenuous at first for the resolution of the forecasts, we are currently just treating as one tile across the predicted area at the previously mentioned 30-minute time-steps.

I'd definitely be interested in hearing options for optimised visualisations as part of this project, particularly surrounding smoothly sliding through forecast data, as that's something we haven't had the resource to focus on in other places yet.

And as for extra layers of meteorological data, absolutely there could be some crossover here, we have a rolling archive of NWP data that would be super interesting to visualise for instance...

Thanks, hope that helps but let me know if you have any further questions!

[yuvraajnarula]

Hey @braddf ,

Appreciate the insights! It’s great to align on the MVP scope.

For data representation, using Mapbox with server-side manipulation makes sense for now, and a modular approach could help integrate probabilistic overlays in the future. If there’s an existing structure for this, we could explore its adaptability.

The 30-minute temporal resolution seems well-suited for the use case. If future applications demand finer granularity, we might consider efficient sub-interval handling without adding complexity. Understanding user interaction patterns could also refine this approach.

For performance, treating forecasts as a single tile should keep things efficient. Optimized visualizations—progressive loading, smoother transitions, or client-side caching—could enhance usability, especially given the resource constraints in other areas.

Lastly, the NWP archive sounds like a strong candidate for future meteorological data layers. A lightweight integration approach could keep things scalable without performance trade-offs.

Looking forward to refining these aspects as we move ahead!

Thanks,
Yuvraaj

[mrgear111]

@peterdudfield
Hey, I'm interested in working on the Cloudcasting UI project. Could you share the repository where the current work is being done or where we should contribute?

[braddf]

Hey Daksh, the Cloudcasting App (forecasting app, that is) is linked in the description in "Other Key Information" – other than that, this is a brand new project for the chosen GSoC candidate to visualise the results of these new forecasts.

I have also updated the description yesterday/this morning to give a little more detail, so hopefully that helps.

Any other Q's, let us know! Thanks 👍

[prashanth316]

Hi everyone! I'm really excited about the Cloudcasting UI project and would love to contribute. The idea of visualizing cloud forecasts using satellite imagery is fascinating, and I’m eager to help build an intuitive and engaging UI. I have experience in front-end development, data visualization, and user experience design. Could someone guide me on where to start and how I can get involved? Looking forward to collaborating with the community! ☁️✨

Where can i get to the codebase and start contributing?

[braddf]

Hey @prashanth316, great to hear! See my comment above to Daksh, and also the updated description for this project at the top of this page.

Hopefully there should be a bit more on there to think about for now until applications open in a couple of weeks. Any other questions or clarification, do let us know. Thanks!

[Devangap]

I'm Devanga, a third-year undergraduate student majoring in Data Science. I am new to open-source contributions but have experience in full-stack development. Currently, I’m doing a Data and AI internship.I am really excited about this project because of its intersection between weather forecasting, satellite imagery, and UI visualization. I’d love to contribute and learn through this experience.
What kind of cloud forecasting data will we be working with? Is it raw satellite imagery, processed datasets, or API-based data?

[braddf]

Hey Devanga 👋 Sorry for the even more delayed response – great to hear from you! I was actually just trying to pin down an answer about our data here, or better yet an example file, which I have just about done so I'll try and link something to the bottom of this thread in a minute! In general though, we have zarr files as the output of the cloudcasting-app, which should be very similar to the "real" satellite data in format, and beyond that we'll be working to find a nice way of surfacing this data to the front end. Currently, they are just in an S3 bucket, so we'll need some way of retrieving them whether that's with a separate API or some server-side business logic where we can authenticate with AWS.

Let me know if you have any further questions and stay tuned for the sample .zarr!

[chaitra-samant]

Hi @peterdudfield
I'm Chaitra Samant a second year student from VJTI, Mumbai. I find this project quite exciting. I have experience in similar React based applications. I would love to know how are we receiving the data we want to display on the Cloudcasting UI, is there an existing API?

I would also like to know more about the selection process, are there any contributions that need to be made or tasks which need to be completed? Can you provide some insights on how to get started?

Thanks,
Chaitra
Chaitra_Samant_resume.pdf

[braddf]

Hey @chaitra-samant, there is no existing API as yet, so the data fetching element of this project would be planned out with the successful candidate in the initial stage. For more info, check out/re-read the updated description and hopefully that should make clear how this all works. 👍

[lambaaryan011]

Hi @braddf, Do you have a plan to take one or more mentees in this project?

[braddf]

Hey Aryan, to my knowledge it's just one mentee per project. ☝️

[braddf]

Hey all! 👋

Further to the above, hopefully you'll get a notification and I'll add this link to the description too, but here's a link to an example .zarr file which will be the output of the cloudcasting-app and thus the starting point for the data for this project.
As previously mentioned, this is also the same format we have the actual satellite data in, so should be compatible across historic and future imagery. It's fairly substantial, at 70MB for this example for one timestep, so bear that in mind!

Download sample data here: 👉 https://drive.google.com/file/d/1eoQgBBkZ14F1nlqiSKHI_czGyJshE4dZ/view

Any issues with the download or data, please let me know!

[Averagenormaljoe]

Hello,

The sample data link is no longer available. @braddf

[braddf]

Thanks @Averagenormaljoe – new link here and in description 😊

[lambaaryan011]

Hi @braddf, for Cloudcasting UI, are we following the same design as Quartz-Frontend, or are there any specific changes?

[braddf]

Hey Aryan, I would default to something similar so that a) it matches, and b) it's quick to get off the ground without huge design thoughts required; that being said, it can be a much simpler version and should start from a clean, new project, only including elements and features required for this task 👍

[jiagill]

Hi @braddf!

My name is Jia, and I'm currently studying CS + Statistics with a minor in Geography and GIS at UIUC. This project sounds like something that's right up my alley, and I'm really excited to take a stab at it in my project proposal!

A question I had was who the target audience for this product is. Would it be used mostly by researchers, industry professionals, or educators? I can imagine each group would have varying needs—for example, researchers may want low-level information provided on the UI, while laypeople might desire high-level, easy-to-understand visualizations. I'd love to know anything that would help me better understand the intended audience for this tool.

[braddf]

Hey Jia!

Sorry for the delay, but great to hear the enthusiasm!

Great question. Unhelpfully though, I don't have a firm answer for you 😅 in order for the chosen candidate to have maximum impact with their project, we intentionally want to keep this brief/implementation as open as possible in the planning phase, such that the mentee has the space to explore and make a case for what they feel is the best fit. So for target audience, this is a perfect example of this – as you've mentioned, there are many different potential users for a product like this, each with slightly different prior knowledge and requirements, and we can't (and definitely shouldn't) try to accommodate all needs and ideas right off the bat.

That being said, I can give a small nudge on this to say that the model itself is in early stages, so the data won't necessarily have the depth and detail that the hypothetical researchers you've mentioned would perhaps be looking for, so perhaps that pushes us more towards the higher-level, approachable data viz, but still ideally with an "industry flavour". At the end of the day, the primary job of this UI right now will be to showcase the new cloud forecast, and demonstrate the potential value of the underlying data.

Hope that helps!

[lakshaypunia]

Hi @braddf

My name is Lakshay. This project aligns really well with my interests, and I’m excited to put together a strong proposal for it!

I have strong frontend skills in Next.js and backend experience with Express.js. I recently started working on this project in a completely new repo and am looking forward to contributing!

[braddf]

Hey @lakshaypunia!

Love to hear that, let me know if you need any clarification on anything 👍

[Averagenormaljoe]

Hello @braddf

Question about the development process: will we be expected to write unit tests with a library such as Jest when creating the website?

Another question: will it be expected to be responsive and accessible across multiple devices, such as mobile and different desktop screens?

[braddf]

Hey @Averagenormaljoe!

1. Jest (or similar) would be great. Not forcing TDD, but we would really like to see unit tests planned in from the beginning for sure. I'm personally a big fan of Cypress testing, but it can get a bit complicated with data viz UIs, especially in short timeframes!
2. Responsive would be great; I would probably see this as a desktop-first UI, but there's no reason it shouldn't also be viewable and work nicely on smaller devices, especially as its primary role is to showcase – and the best device is the one you have with you!

Hope that helps!

[Nasibodula]

Hi @braddf
I'm interested in the Cloudcasting UI project but couldn't access the example .zarr dataset. Could you provide an alternative link or suggest where I can find similar sample data for testing?

[braddf]

Hey Nasibo!

Welcome to the discussion – and sorry about that! I've updated the download link(s) on this page now, but do let me know if you have any further issues 👍

[Nasibodula]

Hello @braddf
I was able to go through the datasets and i have an important question to ask.
What tools or libraries are recommended for reading and interpreting .zarr files in JavaScript . Most .zarr handling is commonly done in Python with libraries like xarray.

[braddf]

Hey Nasibo, sorry I missed this reply – Github notifications are not my friend!

This is true, lots of folks love and use Python for .zarr files – the main driver here is trying to produce a streamlined, simple solution to show what we can do with these forecasts, so full-stack Javascript would fit this, and various zarr processing libraries are available that I have seen and seem to have decent uptake, though I have not used any of them myself just yet.

Hopefully you managed to get an application in before the deadline yesterday!

[Nasibodula]

Hi @braddf, thanks for getting back to me! No worries about the delayed response. Yes, I did manage to submit my application before yesterday's deadline. I was able to access the datasets after your update, so that was very helpful. I'm looking forward to the possibility of working with you on this project and exploring JavaScript libraries for .zarr file processing. Thanks again for your support!

[Vulcanric]

Hi @braddf,
I'm very interested in the Cloudcasting UI project. Is it okay if I don't know everything about the project at the moment but can catch up with it?

[braddf]

Hey John – sorry, I completely missed your message here, hopefully you managed to get in an application before the deadline yesterday! Either way, look forward to maybe seeing some of your contributions around here soon...

[rishn]

Hi @braddf, this project looks super exciting and I’m really interested in contributing to it through GSoC. I’ve been going through the Cloudcasting documentation and the discussions here, and it’s great to see the thoughtful planning around the MVP scope and visualisation ideas.

I had a quick question about the .zarr data mentioned—are there any preferences or constraints around how it’s served to the frontend? Just trying to better understand how the data retrieval might be expected to work in the initial setup.

Looking forward to learning more and brainstorming with you all!

[braddf]

Hey Rishaan, sorry I missed this!
Hopefully you managed to get an application in yesterday before the deadline; good question though, this will be something I will probably work through a plan with the successful candidate to ensure they have suitable ownership and freedom to solve the problem themself, without it being totally overwhelming of course!
Thanks and good luck 👍

[emlweb]

Google Summer of Code 2025 applications are now closed.

We are currently reviewing all applications. Contributors will be announced 8 May 2025. Thank you!

[peterdudfield]

I'm closing this discussing now as GSOC 2025 is nearly over. Thank you for everyones input and help.

We hope to take part next year and we'll be posting info here