Resources¶
Preparation before the workshop
We use Google Colab for convenience — nothing to install, everything runs in the browser on Google's infrastructure, with a regular Google account.
For those who want an advanced coding setup (offline, full IDE, AI integration and version control), we recommend VS Code + AI assistants + GitHub — detailed below in the Advanced tutorial: Conda + VS Code, Git & GitHub and Free AI tools for code sections.
What you need¶
- A Google account (your Gmail is enough).
- A modern browser — Chrome, Firefox, Edge.
- An internet connection — Colab runs in the cloud.
That's it. The rest we'll sort out on Friday morning.
Recommended free accounts¶
For working with real data beyond the workshop examples, register in advance for:
| Service | Why it's useful | Link |
|---|---|---|
| Google Earth Engine | Petabyte-scale satellite data, direct Colab integration | signup.earthengine.google.com |
| Copernicus Climate Data Store | API key for ERA5 and other reanalyses | cds.climate.copernicus.eu |
| Copernicus Marine | Black Sea data (SST, waves, currents) | marine.copernicus.eu |
| Microsoft Planetary Computer | STAC catalogue + Sentinel/Landsat without downloads | planetarycomputer.microsoft.com |
Pro tip: register 1-2 days in advance
Copernicus registration takes ~5 min, but API keys can take several hours to arrive. Create your account ahead of time so you don't lose time on Friday morning.
How a Colab notebook works¶
On each module page, you'll see a yellow Open in Colab button. Click → the notebook opens in a new tab, connected to your Google account.
Useful commands:
- Shift+Enter runs the current cell
- Ctrl+M+B adds a new cell below
- File → Save a copy in Drive keeps your changes (the default does not save to the repo!)
Save your first copy immediately
The first time you open a notebook, save a copy to Drive right away. Otherwise, anything you change is lost when you close the tab.
Tutorial: Colab step by step¶
Quick guide for your first session — from zero to a running notebook in 5 minutes.
-
Open the notebook — click the Open in Colab button on the module page, or go directly to colab.research.google.com and open a file (File → Open notebook).
-
Connect the runtime — click the Connect button in the top right (or Runtime → Connect to hosted runtime). You'll see RAM/Disk indicators when it's ready.
-
Run the cells in order — Shift+Enter on each cell, top to bottom. Don't skip the package installation cell (
!pip install ...). -
Install missing packages — if you get
ModuleNotFoundError, add a cell at the top with: -
Save your copy — File → Save a copy in Drive. Do this immediately after opening the notebook.
-
Download generated files — the Files panel (folder icon on the left) → right-click on the file → Download.
-
Share results — the Share button in the top right, just like a Google Doc.
Runtime resets after inactivity
If you leave Colab idle for ~90 minutes, the runtime stops and all variables and files in /content are lost. You'll have to reinstall the packages and re-run the cells. Data saved to Drive stays.
Free GPU
For heavier computations: Runtime → Change runtime type → T4 GPU. Available for free, with daily usage limits.
Useful links
- colab.research.google.com — main interface
- Colab FAQ — frequently asked questions
- Colab keyboard shortcuts — Ctrl+M+H in Colab shows the full list
Advanced tutorial: Conda + VS Code (local)¶
For those who want to work offline, with large datasets or who prefer a full IDE. Same code as in Colab, run on your machine.
Step 1 — Install Miniconda
Download and install Miniconda (Python 3.11+ version). Miniconda is lighter than Anaconda — it includes only conda and Python.
Step 2 — Create the environment
Step 3 — Install the workshop packages
pip install earthengine-api geemap pystac-client planetary-computer \
xarray rioxarray netCDF4 cartopy cdsapi \
numpy pandas matplotlib scipy statsmodels pymannkendall \
jupyterlab ipykernel
Step 4 — Register the kernel in Jupyter
Step 5 — Install VS Code
Download VS Code. After installing, open the Extensions panel (Ctrl+Shift+X) and install:
- Python — Microsoft
- Jupyter — Microsoft
Step 6 — Open the repo
Clone the repo or download the ZIP from GitHub, then in VS Code: File → Open Folder → select the delta-hub-workshop-2026 folder.
Step 7 — Select the conda kernel
Open a .ipynb file. Click the kernel selector (top right) → Select Another Kernel → Python Environments → Python (delta-hub).
Step 8 — Run the notebook
Shift+Enter cell by cell, or Run All from the top bar.
Local GEE authentication
The first time, run in a terminal (with the environment activated):
A browser opens for approval — after that, authentication is saved locally.Local CDS API key
Create the file ~/.cdsapirc (in your home directory, not in the repo) with:
AI-assisted coding in Colab¶
Colab has native integration with Gemini (free with a Google account). What we'll discuss in Module 4 applies to any assistant (Gemini, Claude, GitHub Copilot, ChatGPT).
If you want to use Claude alongside it: claude.ai (free for limited use).
Git & GitHub — version control and collaboration¶
Git is the standard tool for tracking code changes. GitHub is the platform where we store and share repositories.
Installation¶
-
Git (CLI)
The core tool, runs in the terminal.
-
GitHub Desktop
Visual interface — recommended if you're not familiar with the terminal.
-
GitHub account
Free. Required to access and save the workshop repo.
GitHub Education — free for students and teachers
With an institutional email address (@s.unibuc.ro, @unibuc.ro, etc.) and a student ID / certificate uploaded for verification, you get the GitHub Student Developer Pack or GitHub Teacher Toolbox for free:
- GitHub Copilot Pro free (instead of the limited tier) — unlimited code completion and unlimited chat in VS Code.
- Unlimited private repos, GitHub Pro, extended GitHub Codespaces hours.
- Credits and licences for dozens of services (JetBrains, Namecheap, DigitalOcean, MongoDB, etc.).
How to apply: go to github.com/settings/education/benefits, upload your student ID or proof of academic affiliation and wait for verification (usually a few days).
Essential commands (5 are enough for the workshop)¶
# Clone the workshop repo to your computer
git clone https://github.com/DeltaHUB-UB/delta-hub-workshop-2026
# See which files have changed
git status
# Stage files for commit
git add file.py # a single file
git add . # everything that changed
# Save the changes with a message
git commit -m "Add temperature analysis at Sulina"
# Push the changes to GitHub
git push
Clone the repo before the workshop
You'll have all the notebooks and materials offline.GitHub vs Google Drive
Git tracks changes over time (who, what, when). It's not just a storage location — it's a complete project journal. Ideal for code; less useful for large binary files (.nc, .tif > 100 MB).
Free AI tools for code¶
All the tools below have a free tier and can be used without a credit card.
In VS Code¶
| Tool | What it does | Free tier | Link |
|---|---|---|---|
| GitHub Copilot | Code completion + chat in the editor | 2,000 completions/month, 50 chat messages/month | marketplace: GitHub Copilot |
| Codeium | Unlimited code completion | Fully free, no limits | marketplace: Codeium |
| Continue | Chat with any model (Claude, GPT, local Ollama) | Open-source, free | marketplace: Continue |
Quick install (GitHub Copilot):
- Open VS Code → Ctrl+Shift+X → search for
GitHub Copilot→ Install - Sign in with your GitHub account
- Works automatically: inline suggestions as you type, Tab to accept
Activate GitHub Copilot for free: → github.com/features/copilot → Start for free
In the browser¶
-
Claude (Anthropic)
Excellent at Python code and explanations. Free tier available.
-
ChatGPT
GPT-4o mini free. Good for quick debugging.
-
Google Gemini
Natively integrated in Colab. Available from the ✦ button in any notebook.
-
GitHub Copilot Chat
Accessible directly on github.com on any code file or PR.
How to use AI effectively for environmental code
Always specify the context: the library (xarray, geopandas), the data format (NetCDF, GeoTIFF) and the geographic area. A good prompt:
Using xarray, compute the monthly mean temperature from a Dataset
with dimensions (time, lat, lon) and save the result as NetCDF.