pre-monsoon storm database for northern Thailand 2015-2024

*** pre-monsoon storm database for northern Thailand 2015-2024 ***

We compiled a pre-monsoon (summer) storm database for northern Thailand from publicly available reports. The primary source was the Department of Disaster Prevention and Mitigation (DDPM) website, which posts narrative incident summaries (location, date/time, damages, impacts). Additional official sources (provincial/district offices, utilities, etc.) were consulted when they posted incident information to Facebook or YouTube. To increase coverage, we also consulted credible community/volunteer public pages. Critically, the database stores only the source URLs as provenance. No text, photographs, or video files are downloaded, archived, or retained. All fields are coded by researchers from the public postings; media is referenced solely via links.

Event location (spatial footprint).
The minimum recorded unit is district (amphoe) and province; sub-district (tambon) is coded when reported. When official and non-official sources differ, official sources are prioritized. If only non-official sources exist or footprints conflict, we record the largest corroborated set of affected districts (upper bound), with at least two independent URLs when possible. All entries retain URL provenance, poster type (official/volunteer), and access date.

Damages and impacts.
Counts for affected houses, agriculture, livestock, and assets are taken from official sources when available. When multiple district-level figures exist, we retain the highest reported value as a conservative upper bound and flag uncertainty. We note remaining spatial/reporting uncertainty and plan iterative verification; only URLs are stored.

Hail occurrence and size.
DDPM posts indicating hail presence/absence are treated as primary evidence. Volunteer posts (linked by URL) are used to corroborate and geolocate events. Text-only size descriptions (e.g., fruit/body-part analogies) are coded as low-confidence; Thai coin references are mapped to standard diameters to enable quantitative use. When conflicting sizes appear across sources, we record the largest corroborated size and retain all URLs. We do not store media; we only record whether a post indicates photo/video evidence.

Event timing.
Official timestamps are prioritized. If multiple sources disagree, we define a time window from the earliest reported start to the latest reported end. Vague reports (“afternoon,” “evening”) are coded into coarse bins. Non-contiguous episodes on the same day (e.g., late afternoon and nighttime) are split into separate events after cross-checking URLs, consistent with organized convective line (squall line) passage.

Governance and ethics. All data are from public web pages; we store URLs only, no personal identifiers, and we restrict location to administrative units (district/sub-district). Fields include URL, platform, poster type, post date (as displayed), access date, evidence type (text/photo/video indicated), and confidence flags. This conservative, rules-based approach supports transparency, reproducibility, and future QC while respecting content ownership and platform policies.