Add v0.2.0 NEWS entry (omitted from release commit)

Author: NewGraphEnvironment

NEWS.md

@@ -1,3 +1,7 @@
+# cd 0.2.0 (2026-05-04)
+
+* Adds 8 snow-related variables for hydrology departure analysis. Four monthly natives (`swe`, `snowfall`, `snowmelt`, `snow_cover`) ship as 12-band/year COGs alongside the existing 7 climate variables; four annual derived scalars (`swe_max`, `snowfall_fraction`, `snowmelt_doy_50`, `snowmelt_rate_peak`) ship as 1-band/year COGs. New `pct_point_diff` anomaly type in `cd_anomaly()` for `snow_cover` and `snowfall_fraction` (already-percentage variables). New "Snowpack" section in the FWCP Peace vignette with seasonal-curve table, four annual time-series plots, per-ecoregion view, and citation-grounded interpretation; ASWS QA cross-check at 4 BC stations confirms bias is stable over time even where absolute values are biased. Producer pipeline extends with `scripts/backfill_edh_snow.py` (first place ECMWF's accumulation-reset trick lands in code). Also bug-fix in `cd_stac_item()` filename parsing (substring match was mis-routing variables whose names contain other variable names — `swe_max` was being filed under `swe`). Headline finding: regional summer SWE has collapsed by 75% over the record, freshet-timing shift is uniform across all 5 ecoregions at ~1 day/decade earlier melt. ([#55](https://github.com/NewGraphEnvironment/cd/pull/55))
+
 # cd 0.1.7 (2026-05-04)
 
 * Snow-methodology literature review for the upcoming "Snowpack" vignette section. Adds `scripts/rag_build_snow_methodology.R` and `scripts/rag_query_snow_methodology.R`, builds a local ragnar DuckDB from 11 peer-reviewed papers (now in the `NewGraphEnvironment/hydrology` Zotero collection), and ships the methodology synthesis + 15-row "cite this for that" citation map at `planning/archive/2026-05-issue-53-snow-lit-review/findings.md`. Headline finding: `cd_trend()`'s raw Mann-Kendall + Theil-Sen (no prewhitening) is methodologically correct for our 76-year series with strong trends per [Yue & Wang 2002](https://doi.org/10.1029/2001WR000861) — prewhitening would *underestimate* slope when a real trend exists. ([#54](https://github.com/NewGraphEnvironment/cd/pull/54))