Societal impact assessment
Abundance Potential
Machine Learning Maps Landslide Susceptibility Across eThekwini
Assessment takeaway
Unknown · insufficient evidence
This assessment concerns the regional susceptibility mapping development only. It has plausible planning and preparedness value, but the societal benefits remain conditional because adoption and loss reduction are unreported, and model validation details are missing. Scores are not renormalized; affordability and access remains unknown.
This describes conditional potential, not a probability or measured outcome. Unknown values are not zero, and harms are not subtracted from the total.
Assessment scope
- Development
- Regional landslide susceptibility maps for eThekwini produced using satellite imagery, GIS and machine learning.
- Baseline
- Municipal planning and preparedness without this specific map; its comparative impact is not reported.
- Population
- Municipal planners and residents in eThekwini, including people in formal and informal settlements.
- Geography
- eThekwini municipality, South Africa.
- Target assumptions
- Potential benefits depend on the maps being sufficiently reliable and planners using them alongside local knowledge and other risk assessments. [evidence-1248, evidence-1251]
Five dimensions
- Human wellbeing
- 2/4 · If used to guide development management or preparedness, regional susceptibility information could meaningfully support efforts to reduce landslide harm. The report does not show that the maps have been adopted or reduced losses, so the contribution remains bounded and conditional. [evidence-1251, evidence-1252] Sources: evidence-1251, evidence-1252
- Affordability and access
- Unknown · The evidence does not establish whether the maps improve residents’ access to essentials or useful services, or reduce costs for them.
- Reach and fairness
- 1/4 · The maps cover areas across the municipality, and the report notes that both formal and informal settlements were affected by the 2022 disaster. This creates some potential for broad planning reach, but there is no evidence that map use or resulting protections would be equitable. [evidence-1247, evidence-1252] Sources: evidence-1247, evidence-1252
- Environment and resources
- 1/4 · Maps could help planners consider development in relation to vegetation, land use and slope susceptibility. The report describes these factors but does not show that the maps have changed land management or protected environmental resources. [evidence-1249, evidence-1250, evidence-1251] Sources: evidence-1249, evidence-1250, evidence-1251
- Scientific capability
- 2/4 · The work combines satellite imagery, GIS and two machine learning approaches to produce regional susceptibility maps, a meaningful mapping capability. The reported accuracy lacks validation details, and the evidence does not establish how well the maps generalize. [evidence-1245, evidence-1248] Sources: evidence-1245, evidence-1248
Evidence, confidence and timing
Evidence maturity
Finding: Laboratory. A study produced regional susceptibility maps using satellite data, GIS and machine learning; evidence of operational deployment is not reported. [evidence-1245, evidence-1251]
Societal benefit: Early speculative. The report describes possible planning and preparedness uses but reports neither adoption nor demonstrated reductions in harm. [evidence-1251]
Confidence: Low. The supplied evidence is a report of the study; it gives no details on how the above-98% accuracy was calculated or tested, and reports no real-world use or outcomes. [evidence-1248, evidence-1251]
Conditional time horizon: Near (0–5 years). Because the maps already exist, potential planning use could occur within 0–5 years if municipalities adopt them; the report gives no adoption timeline. [evidence-1251]
Possible benefits
- If incorporated into planning and preparedness, the regional maps could help municipalities identify areas for careful development management and prioritize preparedness efforts. [evidence-1247, evidence-1251] Sources: evidence-1247, evidence-1251
Risks and possible harms
- Because the reported accuracy lacks validation details and the maps provide no event timing or neighborhood-level probabilities, treating them as precise forecasts could misdirect preparedness or create false reassurance. [evidence-1247, evidence-1248] Affected groups: Residents and planners in mapped areas, particularly where regional susceptibility patterns are used for local decisions. Sources: evidence-1247, evidence-1248
Distribution and adoption
Who might be excluded
- Distribution of benefits remains uncertain.
Adoption barriers
- The report does not document municipal adoption; it also provides no neighborhood-level probabilities or event timing, and the accuracy calculation and testing are unspecified. [evidence-1247, evidence-1248, evidence-1251] Sources: evidence-1247, evidence-1248, evidence-1251
Assumptions and conditions that could change the assessment
- The maps are reliable enough for their intended planning scale and are used alongside local knowledge and other risk assessments. [evidence-1248, evidence-1251] Sources: evidence-1248, evidence-1251
- Independent validation and evidence that map-informed planning reduces losses would support higher confidence and more mature benefit ratings. [evidence-1248, evidence-1251] Sources: evidence-1248, evidence-1251
- Evidence on who receives protection from map-informed decisions, including across formal and informal settlements, could change the reach and fairness assessment. [evidence-1252] Sources: evidence-1252
Citations
- evidence-1245 · Using machine learning to build safer cities · N; o; t; ; l; i; s; t; e; d; ; o; n; ; p; u; b; l; i; s; h; e; r; ; p; a; g; e · 10.1038/s41598-026-73316-x · Captured publisher article text · Open source
- evidence-1247 · Using machine learning to build safer cities · N; o; t; ; l; i; s; t; e; d; ; o; n; ; p; u; b; l; i; s; h; e; r; ; p; a; g; e · 10.1038/s41598-026-73316-x · Captured publisher article text · Open source
- evidence-1248 · Using machine learning to build safer cities · N; o; t; ; l; i; s; t; e; d; ; o; n; ; p; u; b; l; i; s; h; e; r; ; p; a; g; e · 10.1038/s41598-026-73316-x · Captured publisher article text · Open source
- evidence-1249 · Using machine learning to build safer cities · N; o; t; ; l; i; s; t; e; d; ; o; n; ; p; u; b; l; i; s; h; e; r; ; p; a; g; e · 10.1038/s41598-026-73316-x · Captured publisher article text · Open source
- evidence-1250 · Using machine learning to build safer cities · N; o; t; ; l; i; s; t; e; d; ; o; n; ; p; u; b; l; i; s; h; e; r; ; p; a; g; e · 10.1038/s41598-026-73316-x · Captured publisher article text · Open source
- evidence-1251 · Using machine learning to build safer cities · N; o; t; ; l; i; s; t; e; d; ; o; n; ; p; u; b; l; i; s; h; e; r; ; p; a; g; e · 10.1038/s41598-026-73316-x · Captured publisher article text · Open source
- evidence-1252 · Using machine learning to build safer cities · N; o; t; ; l; i; s; t; e; d; ; o; n; ; p; u; b; l; i; s; h; e; r; ; p; a; g; e · 10.1038/s41598-026-73316-x · Captured publisher article text · Open source