Prioritization of Orphaned Oil and Gas Wells Using a Multi-Criteria Scoring Framework and Hierarchical Clustering: A Case Study in Deep Fork National Wildlife Refuge and Caddo Nation Lands, Oklahoma

Abstract

Orphaned oil and gas wells represent a growing environmental and public health liability across the United States, with documented inventories exceeding 140,000 wells and plugging costs that far exceed available funding. In resource-constrained remediation programs, systematic prioritization is essential to direct plugging efforts toward wells that pose the greatest immediate risk. Existing qualitative scoring systems, while useful, are limited by their inability to rank wells within the same priority tier, making it difficult to sequence plugging interventions when multiple wells compete for limited resources, and by the lack of empirical validation of their classification outputs. This study developed and applied a multi-criteria assessment framework for prioritization of 30 orphaned wells across two land-use contexts in Oklahoma: Deep Fork National Wildlife Refuge (Okmulgee County) and Caddo Nation lands (Caddo County). The framework integrates six field-measured risk criteria: methane emission rate, wellhead integrity condition, receptor proximity, site contamination severity, avoided methane release potential, and subsurface integrity and aquifer risk, into a weighted urgency score. An unsupervised hierarchical clustering analysis was applied independently to validate whether the data support the priority groupings produced by the scoring framework, without relying on manually assigned weights. Monte Carlo stability analysis with 1,000 iterations and ±5% Gaussian noise was used to quantify confidence in each well's classification. Before computing stability scores, cluster labels across iterations were aligned to the baseline clustering using the Hungarian algorithm, ensuring that numerically different but structurally equivalent cluster labels were compared consistently across iterations. The scoring framework produced overall weighted scores between 0.41 and 5.00, resulting in ten High priority, seventeen Medium priority, and three Low priority wells. A sensitivity analysis across three weight configurations confirmed that the High priority wells remained stable regardless of weighting assumptions, while boundary wells showed sensitivity to weight changes. Independently, hierarchical clustering identified three well groups that broadly aligned with the scoring framework classifications, with a mean Monte Carlo stability score of 0.997 and all 30 wells receiving High confidence classifications. This data-driven, reproducible, and defensible prioritization approach is underpinned by a scoring framework, independent clustering validation, and Monte Carlo stability analysis that address key limitations of existing ordinal prioritization systems and provide a transferable methodology for orphaned well programs operating under finite remediation resources.