radolan

radolan

Go package for parsing DWD RADOLAN and RADVOR-RE weather radar composites and working with the DWD polar stereographic grid.

Background

DWD Open Data publishes radar composites in the RADOLAN binary format covering Germany and surrounding areas. This package parses those composites and provides coordinate projection, Z-R conversion, and spatial sampling utilities.

This is a production-hardened fork of jonnyschaefer/radolan (MIT). It has diverged significantly from the original and is not intended to be merged back upstream. Key changes:

• ErrUnknownUnit tolerance in NewComposites: RV composites (5-min precipitation rate) trigger this error on valid data. The upstream library discards such composites; this fork continues and returns them — the data is usable.
• Format ≥ 5 projection fix: RADVOR-RE nowcast composites use Format=5 on the DE1200 national grid. The upstream projection returns NaN for this combination. This fork applies the correct WGS84 polar stereographic math.
• PrecipitationRateAdaptive: Adaptive Z-R conversion that selects between JossWaldvogel70, Aniol80, and MarshallPalmer55 by intensity — closer to DWD operational practice than a single fixed relation.
• NeighbourhoodSample: Spatial averaging over a grid neighbourhood, returning mean mm/h, max mm/h, and rain coverage fraction — useful for point precipitation estimation without single-pixel noise.
• ProjectionFunc: Returns a pre-calibrated closure for repeated projection without per-call composite overhead — important for tight rendering loops.
• Go 1.22+ modernisation: Range-over-int, slices.SortFunc, other idiomatic updates.

Installation

go get github.com/shyrmapp/radolan

Requires Go 1.22 or later.

Usage

Parse a single composite (RADVOR-RE nowcast)

f, err := os.Open("RV2025032615050.bin.gz")
if err != nil { log.Fatal(err) }
defer f.Close()

comp, err := radolan.NewComposite(f)
if err != nil && err != radolan.ErrUnknownUnit {
    log.Fatal(err)
}

// Project a lat/lng to grid coordinates.
x, y := comp.Project(52.52, 13.41) // Berlin
mmh := radolan.PrecipitationRateAdaptive(comp.At(int(x), int(y)))
fmt.Printf("Berlin: %.2f mm/h\n", mmh)

Parse a tar.bz2 archive (RADOLAN RV, 5-min composites)

RV archives contain multiple composites. NewComposites tolerates ErrUnknownUnit and returns all frames sorted by forecast time.

f, err := os.Open("RV_latest.tar.bz2")
if err != nil { log.Fatal(err) }
defer f.Close()

composites, err := radolan.NewComposites(f)
if err != nil { log.Fatal(err) }
fmt.Printf("Loaded %d frames, latest at %s\n",
    len(composites), composites[len(composites)-1].ForecastTime)

Spatial sampling around a point

proj := comp.ProjectionFunc()
x, y := proj(lat, lng)
avgMMH, maxMMH, coverage := comp.NeighbourhoodSample(int(x), int(y), 2)
fmt.Printf("avg %.2f mm/h, max %.2f mm/h, coverage %.0f%%\n",
    avgMMH, maxMMH, coverage*100)

Z-R conversion

// Adaptive (recommended): selects relation by intensity
mmh := radolan.PrecipitationRateAdaptive(dBZ)

// Fixed relation (DWD operational Germany)
mmh = radolan.PrecipitationRate(radolan.Aniol80, dBZ)

// Inverse: mm/h → dBZ
dbz := radolan.Reflectivity(radolan.Aniol80, mmh)

Supported products

RADVOR-RE (RV/YW, Format=5, DE1200 national grid) is fully supported including the WGS84 polar stereographic projection.

Coordinate projection

DWD uses a polar stereographic projection with a spherical Earth model for most grids, and a WGS84 ellipsoid model for DE1200 Format≥5 (RADOLAN RV, RADVOR-RE). This package handles both automatically based on the composite header.

// Grid coordinate from lat/lng
x, y := comp.Project(north, east)

// Pre-calibrated closure for tight loops (avoids repeated header lookups)
project := comp.ProjectionFunc()
x, y = project(north, east)

Coordinates are zero-indexed from the top-left corner of the grid. Non-integer values indicate sub-pixel position; use int(x) for direct array access or bilinear interpolation for smoother results.

Data sources

• RADOLAN RV (5-min observed): https://opendata.dwd.de/weather/radar/composit/rv/
• RADVOR-RE (5-min nowcast, 2h ahead): https://opendata.dwd.de/weather/radar/radvor/rv/
• DWD RADOLAN product documentation: DWD RADOLAN/RADVOR System Description

License

MIT. Derived from jonnyschaefer/radolan (also MIT).