This project compares food assistance infrastructure with economic need in Washington, D.C.
Research question: Where does food assistance access fail to align with economic need in Washington, D.C.?
Key insightFood presence does not equal accessibility.
Mapping Food Assistance Access Against Poverty in Washington, D.C.
This map compares food assistance infrastructure with tract-level poverty in Washington, D.C. SNAP retailers and farmers markets show where food access points exist, while the poverty layer shows where economic need is higher. The goal is to examine whether food access aligns with the communities most likely to need it.
How to use this map: Toggle poverty, SNAP retailers, farmers markets, and candidate access gaps to compare food assistance locations against areas of higher economic need.
Interactive map using public food access, SNAP retailer, census tract, and poverty context data.
What to Look For
Map question: Where do food assistance locations overlap with higher-poverty census tracts?
Compare SNAP retailer locations against higher-poverty tracts. Areas with higher poverty and fewer nearby access points should be investigated further.
Key findingFood assistance locations are present across Washington, D.C., but their distribution should be interpreted against poverty concentration and nearby SNAP-supported access.
Candidate access gapsSeveral higher-poverty census tracts show limited nearby SNAP retailer coverage. These areas are not final conclusions, but they are useful candidates for deeper food-access review.
Limitation: This analysis uses straight-line distance from tract representative points and does not account for transportation, walkability, food price, store quality, or operating hours.
What this shows: Food access gaps matter most when they overlap with economic vulnerability. Higher-poverty areas with fewer nearby SNAP-supported retailers should be treated as candidate areas for deeper review, not as final conclusions.
Decision value: This screening map can help identify areas for SNAP retailer outreach, food access planning, or deeper network-distance analysis.
Methods
Prepare food location layers, classify SNAP-supported and non-SNAP locations, export GeoJSON, and use web mapping to communicate the difference between availability and practical access.
Tools used
ArcGIS Pro
ArcGIS Online
MapLibre GL JS
GeoJSON
HTML/CSS/JavaScript
USDA SNAP retailer data
USDA Farmers Market Directory
Census ACS
Data sources
USDA SNAP Retailer Location Data
USDA Farmers Market Directory
U.S. Census ACS 2024 5-Year S1701 poverty table
U.S. Census ACS 2024 5-Year B03002 race/Hispanic origin table
U.S. Census TIGER/Line 2024 census tract boundaries
Limitations / next step
The web map is a screening tool for spatial patterns. The next step is to add measured network distance, transit access, or service-area analysis to better represent practical accessibility.
This project shows how food infrastructure should be evaluated against need, not mapped as locations alone.
Changing Patterns of Homelessness in the United States
Geography
United States
Scale
State
Data type
Counts, rates, change, unsheltered share
Output
Static maps + comparison sliders
This project compares state-level homelessness through count, rate, change, and unsheltered share.
Research question: How do state-level homelessness patterns change when measured by total burden, population-adjusted rate, rate change, and unsheltered share?
Important conceptRaw counts show burden. Rates show intensity.
Key insight: The same state can look different depending on whether the question is burden, intensity, change, or shelter condition.
Visual Analysis
The following maps and charts compare homelessness by total count, population-adjusted rate, rate change, and unsheltered share.
Total Count vs Homelessness Rate
Total countRate per population
Use the slider to compare two perspectives. Total counts show where the largest number of people are experiencing homelessness, while rates show where homelessness is most intense relative to population.
Change in homelessness rate can be read in two ways. Absolute change shows how much the rate increased or decreased, while standard deviation change shows which states stand out as unusual compared with the national pattern.
Absolute changeRelative change
Use the slider to compare two ways of reading change. Absolute change shows where homelessness rates increased or decreased the most, while standard deviation highlights which states stand out compared with the national pattern.
Structural Differences: Sheltered vs. Unsheltered Homelessness
Homelessness is not only a question of scale. States with similar homelessness rates can differ sharply in how homelessness is experienced. The unsheltered share shows where people are more likely to be outside the shelter system, while the bubble chart compares rate, unsheltered share, and total count together.
Rate vs. Unsheltered Share, 2024
This chart compares three dimensions at once: homelessness rate, unsheltered share, and total homeless count. California and Oregon show high unsheltered shares, while New York has a high homelessness rate but a much lower unsheltered share, showing a different structure of homelessness.Open full-size chart
Percentage of Homeless People Who Are Unsheltered, 2024
This map shows where a larger share of people experiencing homelessness are unsheltered. Higher unsheltered shares appear across many western and southern states, while some northeastern states show lower unsheltered shares despite high overall homelessness rates.Open full-size map
Structural insightSame scale of problem, different structure. A high homelessness rate does not always mean a high unsheltered share, so policy responses should account for both intensity and shelter conditions.
Decision value: This comparison helps separate where homelessness is largest, where it is most intense, where it is accelerating, and where shelter conditions change the policy response.
Methods
Compile state totals, calculate rates, compare 2020 and 2024 values, and prepare static maps and charts for rate, count-versus-rate, change, and unsheltered percentage.
Tools used
ArcGIS Pro
HUD AHAR / PIT data
Census population estimates
Rate calculations
Choropleth mapping
Static map/chart export
Data sources
U.S. Department of Housing and Urban Development AHAR / Point-in-Time homelessness data
U.S. Census population estimates
Project-derived rate, change, and unsheltered percentage calculations
Limitations / next step
Point-in-time counts are sensitive to local methods, timing, and definitions. Interpretation should account for differences in enumeration practices and local shelter systems.
The next step is to connect these state-level patterns to local housing supply, shelter capacity, and service-system context.
Racial Disparities in Breast Cancer Mortality in the U.S.
Geography
U.S. counties
Scale
County
Data type
Mortality rates, ratios, data suppression
Output
Static disparity maps
This V1 analysis maps county-level Black-to-White breast cancer mortality ratios only where both race-specific values are available.
Research question: Where can Black-to-White breast cancer mortality differences be directly compared at the county level, and what do those comparable counties reveal?
Key insightComparable-county ratios show geographic variation in breast cancer mortality disparities, while suppressed data limits full national interpretation.
Analysis focus: A mortality ratio greater than 1 means Black mortality is higher than White mortality in that county. Dual-data counties are required because suppressed low-count data means some counties cannot be compared directly.
Visual Analysis
Overall mortality context
This map shows county-level female breast cancer mortality rates for all races and ethnicities where data are available. It provides mortality context before comparing race-specific disparities.Open full-size map
Race-specific data coverage
Race-specific county mortality data are incomplete because low-count values are suppressed. Direct Black-to-White comparison is only possible in counties where both Black and White non-Hispanic rates are available.Open full-size map
Comparable-county mortality ratio
This map shows the Black-to-White female breast cancer mortality ratio for comparable counties. A ratio greater than 1 means Black mortality is higher than White mortality in that county.Open full-size map
Key findings
Only 320 counties have both Black and White non-Hispanic mortality rates available for direct comparison.
Among comparable counties, 299 counties have a Black-to-White mortality ratio greater than 1.
The average mortality ratio among comparable counties is 1.44.
The ratio map shows where direct county-level comparison is possible, while the coverage map shows why the analysis cannot be treated as a complete national county map.
Limitation noteBecause many counties have suppressed data, the result should be read as a comparable-county screening map, not a complete national disparity map.
Decision value: This analysis can help identify where comparable county-level disparities are visible and where suppressed data limits interpretation.
Methods
Filter to comparable counties, calculate Black-to-White mortality ratios, map geographic variation, and review where race-specific suppression prevents direct comparison.
Tools used
ArcGIS Pro
U.S. Cancer Statistics mortality data
Census population data
County boundary layers
Mortality ratio calculation
Data suppression review
Static map export
Data sources
U.S. Cancer Statistics county-level mortality data
U.S. Census population data
County boundary layers
Project-derived Black-to-White mortality ratio calculations
Limitations / next step
Suppression and county-level aggregation limit interpretation. The maps should be read as county-level screening tools for spatial pattern recognition, not as evidence of individual risk or causal pathways.
A future next step is to run and document a defensible spatial clustering workflow after the comparable-county dataset is finalized.
This project shows how disparity mapping must account for both visible mortality differences and the limits created by missing county-level values.
Urban Tree Planting Priority Analysis in Windsor, Ontario
Geography
Windsor, Ontario
Scale
City
Data type
Tree density, land surface temperature, public benefit features
Output
Planning-level suitability workflow
This supporting GIS project uses a priority screening workflow to narrow a city-wide urban forestry question into planning-level candidate priority areas for further review.
Research question: Where should Windsor prioritize further review for new tree planting based on heat exposure, low existing tree density, and public/community benefit?
Key insightThe strongest candidate areas were not chosen from heat alone. They emerged where high surface temperature, low existing tree density, and proximity to public/community features overlapped.
Context / why it matters: Urban trees can reduce heat stress, improve pedestrian comfort, support public health, and strengthen green infrastructure. This GIS suitability workflow helps narrow a city-wide problem into planning-level priority areas that can be reviewed through ownership checks, field verification, and community consultation.
Exploded Layer Concept
Layered GIS workflow showing how tree density, land surface temperature, priority layers, public benefit buffers, and candidate priority areas build toward a planning-level screening result.
Workflow summary
Tree data + Landsat surface temperature + public benefit features
Tree density and heat-priority layers
Hot + low-canopy overlap
Public benefit buffer intersection
Candidate priority areas
Areas for further review
How the Layers Narrow the Question
This visual sequence shows how the analysis moves from broad input layers to candidate priority areas. Each step adds one planning filter: heat exposure, existing tree density, public/community benefit, and further review.
↓Keep scrolling to see how each layer narrows the analysis.
1 / 7
High heat + low tree density + public/community benefit proximity = candidate priority areas for further review.
Step 1
Existing Tree Density
Question: Where are there fewer trees?
Right-of-way and park tree datasets were combined to show areas with lower existing tree density.
Step 2
Land Surface Temperature
Question: Where is Windsor hotter?
Landsat surface temperature highlights city-scale heat patterns across Windsor.
Step 3
Priority Heat Areas
Question: Which areas are hot enough to screen further?
The hottest surface temperature classes were reclassified into a binary heat-priority layer.
Step 4
Priority Low-Tree Areas
Question: Where does existing tree density suggest canopy need?
Lower tree-density areas were reclassified as priority areas for canopy improvement.
Step 5
Heat + Low-Tree Overlap
Question: Where do heat and canopy need overlap?
The heat-priority and low-tree-density layers were combined to keep only areas where both conditions were present.
Step 6
Public Benefit Buffers
Question: Which priority areas are near public/community features?
Parks, schools, playgrounds, sidewalks, bus stops, and BIAs were buffered to estimate public/community benefit areas.
Step 7
Candidate Priority Areas
Question: Which areas remain after screening?
Hot, low-tree-density areas near public/community features were identified as candidate priority areas for further review.
Mobile Map Sequence
On smaller screens, the Windsor workflow is shown as a step-by-step map sequence. Each map adds one planning filter and moves the analysis closer to candidate priority areas for further review.
↓Scroll through the maps to follow the screening logic.
1 / 7
Step 1 of 7
Existing Tree Density
Question: Where are there fewer trees?
Right-of-way and park tree datasets were combined to show areas with lower existing tree density.
Step 2 of 7
Land Surface Temperature
Question: Where is Windsor hotter?
Landsat surface temperature highlights city-scale heat patterns across Windsor.
Step 3 of 7
Priority Heat Areas
Question: Which areas are hot enough to screen further?
The hottest surface temperature classes were reclassified into a binary heat-priority layer.
Step 4 of 7
Priority Low-Tree Areas
Question: Where does existing tree density suggest canopy need?
Lower tree-density areas were reclassified as priority areas for canopy improvement.
Step 5 of 7
Heat + Low-Tree Overlap
Question: Where do heat and canopy need overlap?
The heat-priority and low-tree-density layers were combined to keep only areas where both conditions were present.
Step 6 of 7
Public Benefit Buffers
Question: Which priority areas are near public/community features?
Parks, schools, playgrounds, sidewalks, bus stops, and BIAs were buffered to estimate public/community benefit areas.
Step 7 of 7
Candidate Priority Areas
Question: Which areas remain after screening?
Hot, low-tree-density areas near public/community features were identified as candidate priority areas for further review.
What this analysis can and cannot decide
This analysis identifies planning-level priority areas where heat exposure, low existing tree density, and public/community benefit overlap. It does not confirm planting locations for implementation. Final planting decisions would require parcel ownership review, underground utility checks, field verification, municipal planting constraints, and community consultation.
Analysis
I built a GIS workflow to identify planning-level areas where high heat, low existing tree density, and public/community benefit overlap. These areas are candidates for field review, not selected planting locations.
The main output is a set of candidate priority areas for further review, not a planting plan.
Methods
Clipped layers to the Windsor boundary.
Merged right-of-way and park tree datasets.
Created an existing tree density surface.
Converted Landsat surface temperature values to Celsius.
Reclassified heat and tree-density layers into binary priority layers.
Combined priority heat and low-tree-density layers with Raster Calculator.
Buffered public/community benefit features by 300 metres.
Merged and dissolved public benefit buffers.
Intersected heat + low-tree areas with public benefit areas.
Calculated area in hectares.
Reviewed candidate areas against buildings, water, roads, airport proximity, parcels, and imagery.
Tools used
ArcGIS Pro
Landsat Collection 2 Level-2 Surface Temperature
Raster Calculator
Point Density
Buffer
Merge
Dissolve
Intersect
Raster to Polygon
Satellite imagery / basemap review
Limitations
Land surface temperature was based on one Landsat scene, so it represents conditions on that date rather than long-term heat exposure.
Landsat is useful for city-scale patterns, but not detailed enough to evaluate individual planting spots.
Existing tree density depends on available mapped datasets and may miss private trees or recent plantings.
Public benefit was estimated using simple buffers, not actual walking networks or barriers.
Ownership, utilities, soil condition, planting feasibility, and local constraints were not fully confirmed.
The output should be interpreted as a screening result, not a final implementation plan.
Next step
The next step would be parcel ownership review, underground utility checks, field verification, municipal planting constraints, and community consultation before moving from screening results to planting decisions.
Interactive access map
State-level rates and change
Comparable-county ratio map
