Working on GIS Materials

[nickeubank]

@kylebradbury I'm starting to work on pulling together GIS materials for next semester.

If you're interested, I'd love to approach putting them together with a Coursera GIS in Python course in mind that the two of us could put together.

If that sounds of interest, let me know!

The instructional material I have in hand you can find starting Jan 21st: https://www.practicaldatascience.org/ids541_specific/class_schedule_541.html

They're from a couple years ago and need review, expansions, and the addition of exercises, but I think the structure feels reasonable.

I don't have any raster readings yet, as my raster days were from back when I was an ArcGIS guy, but I was thinking you could add materials there pretty easily?

[kylebradbury]

Hey @nickeubank - would definitely be interested in collaborating on this and thanks for sharing what you have on the topics. Perhaps we can chat about this to nail down the right coverage? There are a few topics that may be useful to touch on w.r.t. rasters. Happy to discuss what level we'd want to target, although this would be my "wish list" for working with raster data that I currently would like to see in the world:

• Common raster data types and their peculiarities
• Introduction to rasterio (and xarray)
• Plotting raster data
• Plotting both raster and vector data
• Common types of analysis
  • Band algebra
  • Reprojection, resampling, and interpolation
  • Cropping and masking
  • Combining raster and vector data: zonal aggregation (summary statistics)
• Raster machine learning / computer vision (this could be a potentially large topic)
  • Supervised learning (image classification, semantic segmentation / pixel-wise classification, object detection)
  • Unsupervised learning (unsupervised segmentation)
• Exporting data (formats, compression, data precision, etc.)

[kylebradbury]

I've come across a few resources that do a great job at the intersection of GIS and Python, in particular Geocomputation with Python which largely covers topics I've found myself touching on when mentoring students. Another good resource is Geographic Data Science with Python.

Overall, I think there's some exceptional content in those resources. There are a few things that I've found a bit lacking in each of these that I've also encountered in trying to help students get ramped up in GIS that I have some ideas around and I've listed some of those challenges in priority order:

1. GIS basics are typically not explained in detail; map projections and transformations are inherently confusing. The basic concept students seem to get, but connecting it to the tools quickly highlights gaps in understanding. In most of these books, they assume students have some GIS knowledge. Not only are the concepts new, but so is the terminology used in the documentation and often students aren't prepared for it. I think this is one of the biggest gaps, in my opinion. I have some ideas for a series of figures to help make that a bit clearer, I hope!
2. Understanding the Python ecosystem (packages, file formats, etc.) for GIS. Geocomputation with Python does the best job of the options I've seen yet, but I think it takes a bit more to understand all the pieces (Geopandas geometries are shapely objects manipulated by several projection libraries all working together). The same is true on the raster side, and is further complicated with raster/vector interactions.
3. Making presentation-quality and/or interactive maps in Python. Most Python maps that are produced are lackluster and customizing them brings in all of the peculiarities of matplotlib. If you want interactive maps, it introduces new libraries that are often opaque in their documentation. Most resources cover the very basics but no more.

[nickeubank]

Those are great! Probably good to muse a little more on whether we have enough added value to do our own thing then, no? I take your point some of these things aren't covered as well as you'd like, but is that sufficient to be worth the investment in a coursera course?

[kylebradbury]

Glad you find them helpful! We could pull together what we can to meet any immediate instructional needs (courses, capstones, mentoring on projects) between existing resources and content we create and if there's enough new material, we could package it for Coursera or a book, etc.?

[nickeubank]

Yup, sounds good. Let's also get together in person soon to discuss MIDS stuff more broadly