Instructor Notes
Food deserts: overall approach
- The food desert case study has a short reading part (the concept, the USDA method and its limitations) and two notebooks. The Beginner notebook takes about 30 to 45 minutes; the Expert notebook is longer and works best as a guided demonstration or take-home exercise.
- Ask learners to request and activate a Census API
key before the session. The Expert notebook’s
API_KEYis empty by default, and the Census API refuses requests without a key. - The Data Collection module’s Census notebooks (boundaries, the Census API and joining tables to maps) are good preparation.
Food deserts: teaching tips
- Start by asking learners how they would define a food desert. Then show how many choices the USDA has to make: which stores count, how far is “far”, urban or rural, and whether car ownership matters.
- Compare the four LILA maps from the Beginner notebook side by side; the change between measures is the main lesson.
- Mention the July 2026 update: the Atlas is now two maps. The notebooks use the 2019 Large Retailer Access Map data. The new SNAP-Authorized Retailer Access Map counts many more (smaller) stores, so it usually shows less “low access”, and the two are not directly comparable.
- Discuss limitations openly: straight-line distance, the store list, food prices and quality, and the “food swamp” critique.
Grocery access (Parts 1 and 2)
- The two parts share one example: grocery stores in Lafayette and West Lafayette, Indiana. Part 1 is a demonstration of a QField project; Part 2 is hands-on in QGIS. Suggested timing: Part 1 about 45 minutes, Part 2 about 75 minutes.
- Part 1: if time and devices allow, a short walk to map a few campus locations with a prepared QField project makes it memorable.
- Part 2: the most important setting is the project CRS (EPSG:26916). If learners see costs around 0.04, their project is still in EPSG:4326.
- Do not clip the roads to the city limits; the episode shows why. All 17 supermarkets are reachable with the provided road layer.
- With the full layer, routes from the Purdue Memorial Union to the 17 supermarkets average about 5.7 km by road against 5.0 km in a straight line. About 11% of city street length is within 800 m of a supermarket along the network, and about 40% within 1,600 m.
- The grocery store data comes from OpenStreetMap (October 2026) and is not field verified; say so when presenting results.
- The case pairs well with the food desert episode: the USDA uses straight-line distance, and Part 2 shows how much longer real routes are.
- The Data Collection module’s Census episode (notebook 1c) makes the map of households without a car used in Part 2, Step 9. The Data Analysis module’s Network Analysis episode does the same kind of analysis in Python with OSMnx.
Common problems
| Symptom | Likely cause | Fix |
|---|---|---|
| The Atlas cell is slow | The USDA spreadsheet is large | Wait a few minutes; this is normal |
| Census API errors in the Expert notebook |
API_KEY empty, mistyped or not activated |
Paste the key and check its activation email |
| The join produces no matches | Tract IDs stored as numbers | Convert both GEOID and CensusTract to
strings, padded to 11 characters |
| OpenStreetMap request fails or times out | The Overpass server is busy | Wait a minute and rerun the cell |
| Network costs are tiny (about 0.04) | Project CRS is EPSG:4326 | Set the project CRS to EPSG:26916 |
| Some routes have a NULL cost | Network broken, often by clipping | Use the full road layer |
| A web layer loads but nothing appears | Off screen or under another layer | Right-click > Zoom to Layer; reorder layers |