Source-Dependent Accessibility Discrepancies and Their Effects on Land-Value Models

Accessibility indicators derived from web-map platforms are increasingly used in sustainable spatial planning, service allocation, and land-value modelling, particularly in data-constrained regions. Yet the reliability of such source-dependent measures for decision-making remains insufficiently examined. Using paired parcel-level data from Khyber Pakhtunkhwa, Pakistan, this study conceptualizes accessibility as a spatial measurement process with structured source uncertainty by directly comparing platform-derived (PD) and field-verified (FV) nearest-facility distances across five facility types. Cross-source analysis reveals substantial facility-specific discrepancies in both magnitude and rank ordering, with certain facility types exhibiting near-random or reversed parcel rankings between sources. Spatial diagnostics further demonstrate that discordance events are geographically clustered rather than randomly distributed. An exploratory local amenity-density check further shows that mismatch prevalence varies systematically with nearby POI context, although the relationship is heterogeneous rather than uniformly sparse-driven. Under spatial block cross-validation, land-value models using FV accessibility consistently outperform PD-based models, while PD-based models display fold-level instability. Moreover, coefficient sign orientation and relative importance vary systematically across sources, indicating interpretation sensitivity to measurement choice. Importantly, reducing magnitude error alone does not restore decision reliability when ordering instability persists. These findings show that accessibility source choice can reshape spatial prioritization and inferred price gradients, introducing decision risk into sustainability-oriented planning. We therefore propose a minimum reliability protocol—including discrepancy profiling, ordering diagnostics, spatial discordance mapping, and spatially structured validation—to support transparent and defensible accessibility analytics in data-constrained environments.