Walking distance. Blocks are nodes of a graph; open links join 4-neighbours with length L = 200 m. Shortest path length dij is found with Dijkstra (binary heap). A resident of block i walks to the park edge: d'ij = max(0, dij - 100 m). Walk time t = d' / 80 m/min (typical adult 4.8 km/h). Served means minj tij ≤ T (10 min = 800 m is the common planning standard).
d_ij = Dijkstra shortest path on open street links
t_i = min_j max(0, d_ij - 100) / 80 served_i = [t_i <= T]
served share = sum_i P_i served_i / sum_i P_i
Two-step floating catchment area (2SFCA, Luo & Wang 2003; Gaussian kernel: Dai 2010). Step 1 turns each park's area Sj into a supply-to-demand ratio inside its catchment d0 = 80 T metres; step 2 sums the ratios reachable from each block. The result Ai is accessible park area per person in m², comparable with the WHO-cited guideline of about 9 m² of green space per inhabitant.
R_j = S_j / sum_{k: d_kj<=d0} P_k f(d_kj)
A_i = sum_{j: d_ij<=d0} R_j f(d_ij)
f = 1 (binary) or f = (exp(-d^2/2d0^2) - e^-0.5)/(1 - e^-0.5) (Gaussian)
Conservation: sum_i P_i A_i = sum_j S_j (parks with at least one resident in range)
Inequality of green space. Residents are sorted by A and the Lorenz curve L(p) drawn; Gini G = 1 - 2 ∫ L dp = (1/2μ) ∑∑ wi wj |Ai - Aj|. The income concentration index (Kakwani) uses the fractional income rank r (poorest = 0): C = 2 cov(A, r) / μ = (2/μ) ∑ wi Ai ri - 1. C > 0 means green space is concentrated among the better-off, C < 0 among the poorer.
3-30-300 rule (Konijnendijk 2021). Every home should see at least 3 trees, live in a neighbourhood with 30 % tree canopy, and be within 300 m of a park. Here: visible trees = own street trees + a quarter of the neighbours' + 6 if a park is on or next to the block; canopy = 3x3 block mean of street canopy blended with park canopy (0.5 x park share, block = 4 ha); 300 m = network distance to the park edge.
Siting algorithms. Greedy maximum coverage repeatedly picks the block that newly covers the most weighted unserved residents; the weight is population P (coverage) or P(1 + λ(1 - r)) (equity, favouring poorer ranks). The greedy rule guarantees at least (1 - 1/e) = 63 % of the optimal coverage (Nemhauser et al. 1978) for a fixed budget. Uniform grid ignores people and streets altogether.
Simplifications. Synthetic city (not a real one); one entrance per park at the block edge; every park the same area; people evenly spread inside a block; a park does not remove housing; equal crowding at any park size; no quality, safety, age or disability effects; income and density are generated fields, not census data. 2SFCA is a supply/demand index, not a measured visit rate.