Jupyter notebook examples of data analysis

[batpad]

We need some good examples of jupyter notebooks where we:

• Query the STAC catalog to find datasets that are relevant
• Use eoAPI or cloud-native storage backends to pull in the data we need
• Run some analysis and show some outputs in a notebook.

@geohacker outlined some ideas here: https://github.com/developmentseed/labs/issues/341#issuecomment-1777039503 - but we don't have to restrict ourselves to these. We can also import other / different data-sets and do different demos. The idea is really to demo what's possible in terms of "analysis in the cloud" for large spatial datasets, where one no longer needs to download all the datasets one needs as .zip files or so and then do analysis locally.

This is definitely something to think through further and we should list down humanitarian use-cases that are currently cumbersome, where being able to use an approach like eoAPI really helps with.

@chuckwondo @botanical - is this something that both or either of you would be interested in taking on? Am assigning to both of you for now.

[batpad]

+cc @kylebarron for visibility and inputs.

[botanical]

Thanks for tagging me @batpad! This is something I'm interested in participating in but I don't think I have enough experience to lead this effort nor the capacity to do so for the remaining time in this quarter 😅 I think I will have more availability towards the end of December.

[zacharyDez]

Hey @batpad,

I put some thought into a work plan for the notebooks we will be delivering to IFRC.

As discussed, the notebooks will focus on risk assessment, emergency planning, and impact analysis using the eoAPI and STAC. The objectives of the notebooks are multi-faceted:

• Upskill IFRC analysts in geospatial technologies and notebook usage.
• Demonstrate the capabilities of eoAPI and STAC in practical scenarios.
• Provide a foundation for IFRC analysts to conduct further geospatial analyses.

Some key use cases we discussed:

• Estimating populations affected by emergencies.
• Assessing financial impacts and resource requirements.
• Planning and executing effective emergency and evacuation strategies.

Proposed Notebooks:

• Getting Started with STAC for Risk Assessment: Introduction to STAC, with examples in the context of risk assessment.
• Identifying Damaged Buildings: Using geospatial data to locate and assess damaged structures.
• Locating Nearest Hospitals and Emergency Services: Demonstrating how to find the nearest healthcare facilities or emergency services for each building within administrative boundaries.
• Estimating Populations Affected by Emergencies: Techniques for estimating the number of people affected in emergencies, including population density and distribution analysis.

Type of datasets:

• Administrative boundaries with multiple levels
• Population
• Buildings
• Nighttime lights data
• Emergency services

Timeline:

• ~Dec 1st: Initial meeting with IFRC team for expectations and current skill assessment.
• ~Dec 1st - Mid-Dec: Development phase for the initial set of notebooks.
• ~Mid-Dec: Feedback session with the IFRC team.
• ~Mid Dec - Dec 21st: Refinement and finalization of notebooks based on feedback.
• ~Dec 21st: Final handoff and documentation provision.

@sunu @batpad, it would be good to connect about the datasets as they will quickly become the limiting factors in implementing the notebooks.

[geohacker]

@zacharyDez this sounds great, thank you for taking lead here! The goals and use cases all align with what I've been hoping we'd achieve through this labs effort.

My only suggestion is that if we can prioritise the notebooks in this order:

1. Getting Started with STAC for Risk Assessment
2. Estimating Populations Affected by Emergencies
3. Locating Nearest Hospitals and Emergency Services

I think the nearest hospital identification notebook can be low priority. Getting started with eoAPI for Risk would be a really great notebook to showcase this even outside IFRC.

[zacharyDez]

13 implements the getting started notebook!

[zacharyDez]

I was thinking of some quick hitter notebooks that we could work on to show specific short examples of using eoAPI in practice. The idea is to move from imagery to insight. The outline of these notebooks could follow these general notes:

• Look for signs of impact from imagery (could include digitizing signs of damages or open-air camps)
• Link the impact to the potentially affected population (requires an estimation method and using a spatial join)
• Qualify the population's access to emergency and health services

I started an example for Afghanistan, given that we have not loaded the equivalent datasets for Turkey here.

In #2, I just commented on the issue with the temporal extent of the Afghanistan MAXAR data. I believe the Turkey dataset would provide a better example, but we need to import the country's health services and population hex bins.

As mentioned in #32, the registration of mosaics currently needs to be fixed. We could resume work on this notebook as soon as it is addressed.

[batpad]

The notebooks are ⭐ ⭐ - thanks @zacharyDez !