A Three-Level Framework for Spatial Demand Estimation and Latent-Demand Reconstruction in Paid Parking Systems

Estimating paid parking demand is difficult when transaction data are censored by full occupancy and when tariff changes create structural breaks. This study proposes an interpretable three-level framework combining (i) dynamic modelling of the cumulative client base and aggregate parking demand, (ii) recovery of model-implied latent potential demand in capacity-constrained zones using a transferred beta arrival profile and cohort-based survival curves, and (iii) spatial allocation of capacity-adjusted demand using road-network accessibility, land use, tariffs, capacity, and competitive-supply factors. Demand is estimated separately for four parking-duration groups using transaction data from Vladivostok, Russian Federation. Three validation questions are kept distinct. A one-origin, 15-month aggregate revenue scenario forecast after the June 2025 tariff change achieved MAPE 6.5% versus 18.8% for a seasonal-naive benchmark. Leave-one-zone-out spatial cross-validation produced OLS R2 values of 0.49–0.63 across duration groups (0.61 for total sessions), whereas chronological cross-regime transfer of the zone model yielded R2 values near zero or below. The evidence therefore supports spatial demand estimation, latent-demand reconstruction, and transparent scenario analysis; it does not support a claim of robust short-term temporal forecasting for zone-level demand.