Deforestation & Disaster Vulnerability in Sumatra
Mixed-methods triangulation of GFW integrated alerts, NASA precipitation, and 132k social discourse posts across 77 spatial units during the Nov 2025 disasters.
THE QUESTION
Does catchment-scale deforestation statistically modulate flood severity during extreme rainfall, and does digital public discourse align with empirical environmental indicators?
BACKGROUND
Catastrophic flooding and landslides struck northern Sumatra in late November 2025, triggering extensive public discourse attributing the devastation to upland deforestation and palm oil expansion. This project triangulates environmental geospatial observations across 77 spatial units in Aceh, North Sumatra, and West Sumatra with a large-scale computational discourse analysis.
APPROACH OVERVIEW
ENVIRONMENTAL DATA HARMONIZATION
Ingested daily GFW integrated alerts (GLAD/RADD), NASA POWER rainfall, HydroSHEDS network, and BNPB DIBI impact reports.
SPATIAL STRATIFICATION
Harmonized 77 spatial units across 3 provinces (Aceh: 23, North Sumatra: 34, West Sumatra: 20 including major water bodies).
SOCIAL SENSING CORPUS
PRISMA-style deduplication of 132,459 records (23,869 unique tweets and 108,590 unique YouTube comments).
DISCOURSE CODING & LLM VALIDATION
Stratified gold-standard validation () evaluating environmental attribution, blame targets, and physical mechanism references.
TRIANGULATION & AUDITING
Rigorous empirical stress-testing distinguishing hydrological confounding (HydroSHEDS static discharge) from true causal signals.
METHODS
- •Sensors & Environmental APIs: Global Forest Watch Integrated Alerts (GLAD/RADD), NASA POWER daily precipitation.
- •Disaster & Hydrographic Data: BNPB DIBI disaster repository, HydroSHEDS river routing networks.
- •Spatial Scope: 77 spatial units across 3 Sumatran provinces (Aceh, Sumatera Utara, Sumatera Barat).
- •Social Sensing Corpus: 23,869 unique tweets and 108,590 unique YouTube comments (Nov 24 – Dec 9, 2025).
- •Statistical Framework: Spearman rank correlation, Mann–Whitney test, cluster regression, and stratified codebook validation.
KEY DATA SNAPSHOT
RESULTS
Triangulation revealed that while public discourse strongly identifies anthropogenic land degradation (25.3% environmental attribution, 41.3% institutional blame), physical hydrological mechanisms (soil infiltration, retention capacity) are articulated in under 1% of posts. Furthermore, empirical auditing highlighted that static river network properties confound crude attribution models.
DISCUSSION
Public discourse overwhelmingly attributes catastrophic disasters to logging and corporate plantations, yet rarely references catchment hydrological processes.
Auditing the geospatial evidence pipeline demonstrated that static basin attributes (DIS_AV_CMS) can introduce geographic confounding if misattributed as dynamic flood surge responses.
News media accounts exhibit a ~6× volume over-representation relative to their numbers (2.9% accounts driving 17.4% volume), highlighting the necessity of systematic deduplication and actor-type stratification in social sensing studies.
LIMITATIONS
- •Twitter/X dataset lacks user biographical profiles and follower counts, necessitating handle-pattern and behavioral stratification.
- •Hydrological discharge measurements derived from global digital networks require localized river gauge calibration to fully untangle dynamic runoffs from baseline basin scale.
IMPACT & APPLICATION
DATA & REPRODUCIBILITY
Analytical code and specific target coordinates are currently held under institutional review and confidential protocol.
REFERENCES
- 01Nandatama, E. (2026). Deforestation-Driven Disaster Vulnerability: Triangulating Statistical Evidence with Public Sentiment in the November 2025 Sumatra Floods, Indonesia. SSRN Electronic Journal, DOI: 10.2139/ssrn.5965995.
- 02Hansen, M. C. et al. (2013). High-Resolution Global Maps of 21st-Century Forest Cover Change. Science, 342(6160), 850–853.
- 03Lehner, B. et al. (2008). New Global Hydrography Derived From Spaceborne Elevation Data. Eos, Transactions American Geophysical Union, 89(10), 93–94.