vernd

# Vernd — Lykilupplýsingar um umsækjendur um alþjóðlega vernd **Requires:** Tier 2 (browser) — Power BI capture needs Chromium: `uv run playwright install chromium`. Key monthly/weekly stats on **international-protection (asylum) applicants** to Iceland. Power BI dashboard published by **Ríkislögreglustjóri** (National Commissioner of Police) via the Ministry of Justice's mælaborð index at `stjornarradid.is/gogn/maelabord/`. ## Architecture Public `/view?r=<base64>` Power BI embed — same scrape pattern as [landlaeknir](../landlaeknir/SKILL.md). No auth, no token, no Playwright-drivable SPA: just point headless Chromium at the embed URL and intercept the Power BI data responses. | Field | Value | |-------|-------| | Tenant | `509484a8-c0ff-4960-b5da-3bdb75e98460` | | Report key | `a2a07353-0f55-40d2-b761-bab29aba67bc` | | Cluster | `9` | | Embed URL | `https://app.powerbi.com/view?r=<base64(json{k:report,t:tenant,c:9})>` | The cluster number matters — it's embedded in the base64 payload on the official government page. Most `landlaeknir` dashboards use cluster 8; this one uses 9. ## Usage ```bash # Print embed URL + IDs for verification uv run python scripts/vernd.py info # Scrape Power BI query responses (~15s headless browser run) uv run python scripts/vernd.py fetch ``` Output: `data/raw/vernd/responses.json` — list of raw Power BI semantic-query results, one per visual on the page. ## What the dashboard publishes Each Power BI "visual" corresponds to one query response. Captured shape (2026-04): | Visual | Row count | Dimensions | Measure | |--------|-----------|------------|---------| | Yearly total | 5 | `Dags stofnað hælisleit.Ár` | count | | Monthly trend | 57 | `Ár × Mánuður_stutt` | count | | Weekly trend by gender | 142 | `Ár × Vikunúmer × Kyn2` | count | | Age × gender | 9 | `Aldursflokkun2 × Kyn2` | count | | By nationality | 38 | `Hælisleitandi.Ríkisfang` | count | | Nationality × year | 104 | `Ríkisfang × Ár` | count | | KPI cards | 1 each | — | totals/averages | ### Data-model entities | Entity (`Source.Entity`) | Fields used | Meaning | |-------------------------|-------------|---------| | `Nafnaskrá` | `Kyn2`, `Aldursflokkun2` | Applicant demographics | | `Hælisleitandi` | `Ríkisfang` | Citizenship / country of origin | | `Dags stofnað hælisleit` | `Ár`, `Mánuður_stutt`, `Vikunúmer` | Application-opened date | | `Dags beiðni` | `Ár` | Request date | ### Values (sanity check, 2026-04-21 snapshot) Applications-opened per year (from the yearly-trend visual): | Year | Count | |------|-------| | 2022 | 4,493 | | 2023 | 4,159 | | 2024 | 1,932 | | 2025 | 1,712 | | 2026 (YTD) | 421 | The 2022 peak + subsequent decline is consistent with the surge during the Ukraine-war influx and the 2023 policy tightening. ## Decoding the Power BI "DSR" payload Each captured response contains `result.data.dsr.DS[*].PH[*].DM0` — a list of row objects with a `C` array holding values in the order declared by `descriptor.Select`. The first row carries a schema (`S`) that names each slot; subsequent rows omit the schema and reuse the column order. Numeric values are plain ints; categorical values are strings. Minimal parser sketch: ```python import json payload = json.loads(open("data/raw/vernd/responses.json").read_text()) for visual in payload: dm0 = visual["results"][0]["result"]["data"]["dsr"]["DS"][0]["PH"][0]["DM0"] # dm0[0]["S"] holds the column schema; each dm0[i]["C"] is a row tuple for row in dm0: print(row.get("C")) ``` Heads-up: sparse rows use the `R` key to indicate "repeat previous column value" and the `C` key may be absent when all columns repeat. Full decoding requires handling those two shorthands. ## Caveats 1. **No public CSV / API** — the underlying immigration-case database is not published. The dashboard is the only open view. 2. **Definitions.** "Umsókn um alþjóðlega vernd" includes all applicants (Convention refugees, subsidiary protection, humanitarian). The dashboard does not break down by outcome. 3. **Reporting lag.** Latest weeks often under-count — applications logged late in the case-management system filter through over the following weeks. Don't compare the current-week count to prior weeks directly. 4. **Ukraine influx.** The 2022 and early 2023 numbers are dominated by Ukrainian arrivals under the temporary-protection regime. Trend comparisons pre-/post-2022 should flag this. 5. **Cluster number stays at 9.** Don't assume cluster 8 like most `/view?r=` embeds — it's in the payload for a reason. 6. **Encoding.** Power BI responses are UTF-8 throughout (nationality names, month-name strings like "Mánuður_stutt"). Write JSON with `ensure_ascii=False` and read with `encoding="utf-8"`. ## Attribution "Byggir á upplýsingum frá Ríkislögreglustjóra." ## Related - [landlaeknir](../landlaeknir/SKILL.md) — same `/view?r=` scrape pattern, 33 health dashboards - [vinnumalastofnun](../vinnumalastofnun/SKILL.md) — related on labour-market side (work permits) - Policy tracker at `stjornarradid.is/verkefni/utlendingar/upplysingavefur-verndarmala` (`upplysingavefur_verndar` in the mælaborð backlog) — narrative companion, no additional raw data.