1 of 52

GIS Analysis for Public Health: �Sample Projects and Methods

1737 Cambridge St. CGIS Knafel building Concourse level

Presentation for students in GHP 534: Introduction to Spatial Methods for Public Health

By Jeff Blossom, Center for Geographic Analysis (CGA), Harvard University

May, 2026

2 of 52

Research Question:

What percent of Lesotho’s total population has access to a health facility?

Access defined as < 3 hour walk

3 of 52

What percent of Lesotho’s total population has access to a health facility?

Input data (all publicly available sources)

Elevation – For slope creation to estimate walking speed.

Trails, roads, rivers, bridges - OpenStreetMap.

Land Cover – European Space Agency

Population – WorldPop (U. of Southampton, U.K.)

Health Centers – Ministry of Health, Lesotho.

4 of 52

GIS Analysis Methods

  • ArcGIS Pro raster analysis involving vector to raster conversion, raster reclassification, and integration of horizontal and vertical travel factors in a Path Distance model.
  • Assumed walking speed of 5 kph on flat, smooth surfaces, adjusted accordingly
  • 100m x 100m raster resolution, determining travel time from each cell to the nearest health facility.
  • Overlaid resulting walk times < 3 hours with WorldPop population raster to determine total population with access.

5 of 52

ESA land cover classifications with travel coefficients.

5

6 of 52

GIS Analysis

What percent of total population has access to a health facility?

Access is defined as within a 1 hour walk to a health facility in urban areas; within a 3 hour walk in rural areas.

78% of the population has access

7 of 52

GIS Analysis

What percent of total population has access to a health facility?

Access is defined as within a 1 hour walk to a health facility in urban areas; within a 3 hour walk in rural areas.

By adding 50 new health facility locations based on the GIS analysis to the model, 90% of the population has access.

Link to publication on this work in Global Transitions

8 of 52

9 of 52

10 of 52

11 of 52

PRISM data example – one day, one variable

12 of 52

13 of 52

Conclusions: We found that greater short-term precipitation may trigger airway inflammation in adolescents, particularly among those with asthma.

Link to American Thoracic Journal publication:

https://www.atsjournals.org/doi/abs/10.1513/AnnalsATS.202209-805OC

Link to the code used, ISPRS publication, and FOSS4G presentation: https://gis.harvard.edu/high-performance-computing-address-level-climate-data-extraction

14 of 52

Case 3: Visual Analysis / Cartography

14

15 of 52

Artificial Intelligence (AI) in GIS

15

16 of 52

Language Learning Models (LLM)

16

17 of 52

GeoAI

  • Google Gemini
  • Microsoft co-pilot
  • ChatGPT
  • Adobe Firefly
  • Claude
  • CARTO
  • Cursor
  • …many others

17

GeoAI tools in ArcGIS Pro

18 of 52

ChatGPT can be produced to generate geographic data. If you use ChatGPT for data or informational purposes, make sure to cite it. ��

19 of 52

Leaflet HTML code created from ChatGPT answering “map the last 10 superbowl venues”

19

20 of 52

Jan. 2024�Question to ChatGPT with DALL-E: “Please make a map of Texas with Cities, Rivers, Counties” ��Result 🡪�not great.

20

21 of 52

May, 2025�Question to ChatGPT with DALL-E: “Please make a map of Texas with Cities, Rivers, Counties” ��Result 🡪�pretty close to the real thing.

21

22 of 52

May, 2026�Question to ChatGPT with DALL-E: “Please make a map of Texas with Cities, Rivers, Counties” ��Result 🡪�pretty close to the real thing.

22

23 of 52

May, 2026�Question to Claude, Sonnet 4.6: “Please make a map of Texas with Cities, Rivers, Counties” ��Result 🡪�only cities!

24 of 52

Explore your data with interactive Dashboards in ArcGIS Online

25 of 52

May, 2026�Question to Claude, Sonnet 4.6: “Can you make a line chart with county_name on the horizontal, population on the vertical axis, with all of the 'contains' fields available as filter toggles, and the ability to sort by population with the attached data”

Attached data, in .csv format:

26 of 52

This produced an interactive chart in html format. Download the html at this link and open on your computer to check it out.

May, 2026�Question to Claude, Sonnet 4.6: “Can you make a line chart with county_name on the horizontal, population on the vertical axis, with all of the 'contains' fields available as filter toggles, and the ability to sort by population with the attached data”

27 of 52

A recent project: Mapping riverine villages in Amazonas, Brazil

Challenge: Converting ~150 community location coordinates from a poor quality PDF. Manually entering this into a spreadsheet is tedious and subject to data entry error.

Optical character recognition is needed, but did not work on this PDF in Adobe Acrobat or Illustrator.

Solution: Taking screenshots of the PDF like the one above, and loading these .jpg images into ChatGPT 5 with the prompt “please convert to .csv format with village, lat, long, in decimal degree format”.

Result: A file like this, ready for mapping:

Manually checking a sample of these locations against

modern base maps confirmed their accuracy.

28 of 52

The resulting map focused on showing just 6 of these locations, and was designed with simplicity in mind

29 of 52

AI for Geo

29

Automated feature extraction using CARTO software

https://www.youtube.com/watch?v=lioqmKuJZhA

30 of 52

AI for Geo

30

31 of 52

GeoAI example

  • Neighborhood livability assessment
  • Uses image recognition and interpretation within Google Earth Engine and Gemini on Google Co-lab

31

32 of 52

GeoAI example

  • Neighborhood livability assessment
  • Uses image recognition and interpretation within Google Earth Engine and Gemini on Google Co-lab

32

Give a score from 1 to 10 for the Safety, Greenery, Cleanliness and Walkability, in this image, and provide a short reason for each.

  • Safety Score: 8, Reason: The neighborhood appears to be relatively safe, with a low crime rate and well-maintained streets.
  • Greenery Score: 9, Reason: The neighborhood is lined with trees, providing shade and a pleasant atmosphere.
  • Cleanliness Score: 8, Reason: The streets are clean and well-maintained, indicating a well-kept neighborhood.
  • Walkability Score: 7, Reason: The neighborhood is pedestrian-friendly, with sidewalks and a low volume of traffic.

Current work in development by Lingbo Liu and others, Harvard CGA

33 of 52

AI for Geo �With increasing data collection (both temporally and topically) from Earth observation and human observation sensors, Geo-AI is essential for computation and the metamorphosis of turning data into useful, timely information.��Much of what humans know and do can be loaded into an LLM. How the AI can be best utilized is a more important challenge to solve.

AI is becoming better and better at interactive data visualization, but still struggles with geo-mapping.

33

34 of 52

Training at the CGA

35 of 52

Training at the CGA

36 of 52

One more workshop this spring…

37 of 52

Thank You!

Jeff Blossom

jblossom@cga.harvard.edu

See more CGA projects: https://gis.harvard.edu/research-portfolio

CGA website: http://gis.harvard.edu/

CGA staff email addresses: https://gis.harvard.edu/teams

CGA general contact: http://gis.harvard.edu/contactus

38 of 52

Center for Geographic Analysis (CGA)

  • Founded in 2006

  • A member organization of the Institute for Quantitative Social Science (IQSS)

  • Serves the entire University – anyone with a HUID / HarvardKey

39 of 52

CGA People

  • 4 full-time staff positions
  • 5 full-time or part-time research/teaching appointments
  • Over a dozen visiting/affiliated appointments at any given time
  • Hosted over 100 visiting researchers, fellows, student interns and other affiliates since 2006.

40 of 52

CGA Core Services

  • Consult for research and learning projects
  • Contribute to proposal development
  • Manage geospatial software site licenses
  • Host mapping services
  • Provide GPS mapping equipment

40

41 of 52

  • Credit courses through FAS Government Department
  • Instructor-led workshops
    • Basic Introduction to GIS (2 hours, each semester)
    • GIS for Humanities and Social Sciences (1 day, fall semester)
    • Cartography (1 day, spring semester)
    • GIS Institute (2-weeks, winter and summer)
  • Customized course modules
  • Self-learning tutorials

Instructional Programs

42 of 52

GIS Site Licenses

43 of 52

Web GIS & Hosted Mapping

Fusion Tables

44 of 52

Research Consultation

  • Data issues, e.g.
    • Public and commercial data sources
    • Data collection and management techniques
    • Data conversion, scanning, digitizing, etc.
  • Hardware/software issues, e.g.
    • What GPS to buy – accuracy, compatibility, etc.
    • Which software to use for various tasks
  • Methodology issues, e.g.
    • How to select parameters in spatial interpolation
    • How to build location-allocation models
  • Visualization issues, e.g.
    • How to do 3-D maps or time-series animation
    • How to allow users make interactive maps from a web browser

44

45 of 52

45

Venues of the 2012 London Olympics

CGA Service Projects: Maps and Visualization

Land cover of Gujarat province, India

46 of 52

46

47 of 52

47

The “Out of Eden Walk” project

https://gis.harvard.edu/out-eden-walk

48 of 52

48

49 of 52

Visualizing Paul’s GPS data on a web map

50 of 52

Visualizing Paul’s GPS data in 3D

51 of 52

Geographic storytelling with Story Maps

https://harvard-cga.maps.arcgis.com/apps/MapJournal/index.html?appid=522628408fc5490aa70b0b54fa6612f3

More on the Out of Eden Walk: https://gis.harvard.edu/out-eden-walk

52 of 52

See more CGA projects on our portfolio site