Skewed Criterion-Weight Distributions for Spatially Regularized Selection of IoT-Enabled Environmental Monitoring Nodes

This work presents a spatially regularized modelling framework for selecting IoT-enabled environmental monitoring nodes within smart-lighting infrastructures. The study does not claim novelty in CRITIC, TOPSIS, mixed-integer programming, or Voronoi tessellation as isolated methods. Instead, its contribution is the problem-specific formulation and validation of a node-selection workflow in which CRITIC-derived criterion weights and TOPSIS suitability scores are used as suitability coefficients in a binary location model, while spatial redundancy is controlled through a distance-threshold penalty. Candidate-level datasets were available for three mining-impacted municipalities in northern Colombia: Albania (70 candidates), Algarrobo (87 candidates), and La Jagua de Ibirico (76 candidates). The implementation uses a minimum separation threshold d0 = 100 m, corresponding to a practical microscale spacing criterion for public-lighting-based monitoring stations, and evaluates sensitivity to the regularization parameter λ. Compared with pure top-k TOPSIS selection, the spatially regularized solutions eliminate close-pair conflicts below 100 m and increase the minimum inter-node distance to approximately 101.2 m in Albania, 105.6 m in Algarrobo, and 100.6 m in La Jagua de Ibirico in the high-regularization MILP runs. Solver runs for the sparse linearized MILP reached zero optimality gap on the reported instances. Voronoi tessellation is then used as an interpretive coverage tool, not as evidence of mathematical novelty. Finally, the analysis of criterion-weight distributions is presented as exploratory and descriptive because of the limited number of main criterion groups.