Geospatial Data Science

Park Accessibility & Investment Prioritization

A fairer path to park investment.

Which neighborhoods should park investment reach first?

Developed an R and ArcGIS workflow combining demographic, neighborhood, accessibility and park-condition indicators to support investment priorities.

The original project supported a $750,000 federal grant.

Subsequent modernization preserved the historical results while adding reproducible methods, accessibility scenarios and sensitivity analysis.

A neighborhood, measured on footSaved ten-minute walking areas cross census boundaries. Park 13 is highlighted over real Esri aerial imagery. The 2024 origins are archived park points, not verified entrances; the original geometry is preserved.Source: Esri, Vantor, Earthstar Geographics, and the GIS User Community.Full-size PNG SVG

40: Parks rankedSaved 2025 results
49: Source census tractsSeparate 2024 spatial context
10 min: Modeled walking timeHistorical network service areas

Why it matters

Investment priorities need to account for who lives nearby, whether they can reach a park, and the condition of its facilities. Combining these perspectives gives planners a basis for comparing needs, while sensitivity analysis shows how analytical choices can change the order.

Reading the historical context

Population describes estimated residents in the saved walking areas; poverty describes estimated people below the poverty threshold. Income context is an unweighted average of intersecting census-tract medians, not a walking-area median household income. These summaries describe the archived data, not current neighborhood conditions.

My contribution

I developed the original R and ArcGIS analysis, combining census-tract context and summaries within saved ten-minute walking areas with park-condition indicators. I later made the methods reproducible and added comparisons of indicator definitions, walking assumptions, and population allocation while preserving the historical results.

Comparing access assumptions

The newer workflow checks indicator definitions and the populations used as denominators, including the separate poverty-status population. Directed pedestrian routes compare 5-, 10-, and 15-minute coverage at different walking speeds, retaining crossing delays and counting each resident once across overlapping parks. Allocation comparisons hold source totals fixed while distributing people by land area or declared residential weights. These routing and allocation examples use synthetic inputs.

Result

The original project supported a $750,000 federal grant. The modernization preserves the archive and makes analytical choices easier to examine.

Tools

R
ArcGIS
Network Analysis
Spatial Statistics
Sensitivity Analysis

The findings

From local conditions
to investment order.

Park 23 leads the saved 2025 ranking, followed by Park 36 and Park 37. Explore the components behind each park's unchanged score.

The five highest priorities

Select a park to see its score composition. Display values are rounded; the ranking and bar lengths use the original precision.

Score details and indicators

Score components

Percentages show each component's share of this stored score. They are not demographic percentages or revised weights. Rounding can affect the total.

Population need: Demographic and health factors
Neighborhood context: Nearby-park and crime indicators
Park condition: Physical condition assessment

Original condition rating52 / 100

Lower condition ratings contribute more to priority.

Normalized indicators

Poverty0.30

Under 180.77

Older adults0.25

Health0.45

Minority population0.38

Nearby-park index1.00

Crime index0.84

Saved inputs run from 0 to 1. Magnitudes follow each indicator's original definition. Income is not active in this 2025 formula.

Project figures

The evidence behind
the priorities.

The 2024 spatial summaries describe the saved walking zones. They remain separate from the 2025 priority ranking. Select any figure to enlarge it.

An editorial horizontal bar chart shows the original ten highest-priority anonymous parks. Park 23 leads, followed by Park 36 and Park 37. Large rank numbers establish the order; scores are shown to two decimals. — **Where investment rises to the top**The ten leading parks retain their original 2025 rank and score. Values are displayed to two decimals for reading; bar lengths and ranking use unchanged full-precision results.Full-size PNG: Where investment rises to the top SVG: Where investment rises to the top

Three aligned dot distributions show all 40 walking areas for estimated population, people below poverty, and income context. Large rounded medians appear beside each distribution; an interval marks the middle half of observations. — **Forty walking areas. Different local needs.**Across 40 saved walking areas, median population is about 1,573, median below-poverty population is 70, and the median of the area-level income-context values is about $111,000. Income values are unweighted means of intersecting tract medians. These are descriptive summaries of the 2024 GIS snapshot.Full-size PNG: Forty walking areas. Different local needs.SVG: Forty walking areas. Different local needs.

Three unboxed stages connect a real walking-zone polygon over census boundaries, grouped population and neighborhood indicators plus park condition, and the original top-three ranking: Park 23, Park 36 and Park 37. — **From a ten-minute walk to an investment priority**Source attributes are summarized within ten-minute walking zones, normalized into comparable indicators, and combined with park condition to support investment priorities. The diagram uses a real saved 2024 walking polygon and the retained top-three 2025 ranking. It is a method overview, not a new calculation.Full-size PNG: From a ten-minute walk to an investment priority SVG: From a ten-minute walk to an investment priority

Paired-dot chart compares archived and candidate priority ranks for 40 anonymous parks. Twenty-two ranks change, the largest movement is four places, and nine of the original top ten remain. — **How much does the demographic definition matter?**Sensitivity diagnostic: replacing inverse White population count with minority population share, scaled across the 40-park cohort, changed 22 ranks. Maximum movement was four places; nine of the top ten remained. Other contributions were held fixed, and the archived ranking is unchanged.Full-size PNG: How much does the demographic definition matter?SVG: How much does the demographic definition matter?

Decision sensitivity

How definitions change
the priority order.

Replacing the historical minority factor, based on inverse normalized non-Hispanic White population count, with minority population share scaled across the same 40-park cohort changed 22 ranks. The largest movement was four places, and nine of the top ten remained. Other contributions stayed fixed. This is a sensitivity diagnostic; the archived ranking remains unchanged.

The comparison isolates one definition choice. Unresolved source periods and population assumptions prevent treating it as a replacement investment recommendation.

The 2024 spatial context and 2025 rankings are historical snapshots, not current conditions or verified demographic estimate periods. Archived origins were park points or centroids, not verified public entrances. New routing and allocation examples use synthetic inputs. Source recovery remains incomplete; no fully validated replacement ranking has been issued. Anonymous park IDs replace names, but geography remains recognizable.

Reproducibility and verification

Recorded verification includes 1,133 R checks, 18 release tests, and Windows/Linux validation. A preservation check confirmed all 4,521 original files remained unchanged. These checks establish software behavior and preservation; they do not complete source recovery or validate real park entrances and pedestrian routes.