Construction and validation of a multi-dimensional lifecycle evaluation (MD-LEA) system for ecological restoration technologies: Case studies from heavy metal contaminated sites in Nanyang, Hunan, and Xinxiang, China

• A multi-dimensional system (MD-LEA) assesses restoration technology lifecycle. • Adaptive weighting combines Case-Based Reasoning with LSTM-NSGA-II optimization. • System dynamically adjusts decision priorities based on environmental triggers. • Identifies holistically resilient technologies over single-criterion optima. • Validated at three contaminated sites with a two-year field assessment. Selecting optimal ecological restoration technologies requires a comprehensive, long-term, and adaptive assessment, which current methods often lack. To address this gap, we developed and validated a Multi-dimensional Lifecycle Evaluation (MD-LEA) system. This system integrates four key dimensions: restoration efficiency ( η ), lifecycle cost ( C ), ecosystem service value (ESV), and integrated risk ( R ). Its core innovation is a hybrid adaptive weighting mechanism that uses Case-Based Reasoning (CBR) for initial weight setting, combined with a Long Short-Term Memory (LSTM) network and Non-dominated Sorting Genetic Algorithm II (NSGA-II) for dynamic optimization when environmental triggers are activated. The MD-LEA system was applied to three distinct heavy metal-contaminated sites in China, successfully identifying the most suitable technology for each site by balancing competing objectives, yielding comprehensive benefit scores (F-values) ranging from 0.713 to 0.782. Crucially, the system’s adaptive weighting mechanism dynamically adjusted priorities in response to environmental triggers, which in one case led to the selection of a more holistically resilient solution over a technology that was superior in efficiency alone. A two-year field validation confirmed the system’s strong short-term predictive accuracy (e.g., Pb removal showed only a −4.6% deviation from prediction), while larger deviations in long-term indicators such as Ecosystem Service Value (−31.3%) successfully triggered the system’s designed recalibration protocols, underscoring the necessity of its adaptive feedback loop for managing long-term uncertainties. The MD-LEA system provides a scientific and dynamic decision-support tool that, with appropriate regional calibration, can be adapted for diverse ecological restoration challenges globally, promoting the selection of sustainable and resilient solutions.