Soil-informed multivariate decision support to assess urban ecosystem service potential

Abstract

Metropolitan expansion across Africa requires robust soil information to support ecosystem services and urban resilience. However, digital soil mapping often prioritises predictive accuracy over identifying where information is reliable for decision making. This study evaluated numerical and South African Soil Taxonomy within a decision framework to assess where ecosystem service inferences can be made with confidence in Gauteng, South Africa. Although both approaches achieved comparable subsoil accuracy (71% vs. 80%), their information spaces differed markedly. The numerical classification captured high-frequency variation but generated pedodiversity through class fragmentation, producing weakly constrained spatial patterns (synergistic state = 21%). In contrast, the taxonomic system preserved pedological coherence through process-based transitions, maintaining pedodiversity with higher predictive confidence (synergistic state = 63%). These results show that classification choice, not accuracy alone, determines where reliable soil information exists. We recommend integrating classification, probability, and uncertainty into a pedologically grounded decision-support framework for rapidly urbanising landscapes. Current digital soil mapping approaches prioritize predictive accuracy without explicitly identifying where soil information is reliable for decision making. This study demonstrates that classification structures govern the translation of pedodiversity into actionable knowledge, showing that higher accuracy does not necessarily yield more reliable ecosystem service inference. By integrating classification, probability, and uncertainty, this work introduces a framework for identifying decision-relevant soil information in rapidly urbanising landscapes.