Limitations

This project packages public magnetic anomaly data as an inspectable web layer. It is not a navigation system and makes no accuracy, safety, or GPS-replacement claims.

Data Epoch

The source layer is a compilation from many surveys collected across different years and specifications. This package records the layer as a multi-survey US/Canada compilation with a USGS data release date of 2025. Earth's magnetic field changes over time, and historical surveys may have been corrected using different reference-field models.

Product Type

This v0.1 layer uses the USGS magnetic anomaly grid for the United States and Canada. It must not be blended with RTP, vertical derivative, pseudogravity, EMAG2, WDMAM, or other products as if they measured the exact same quantity.

Resolution

The native source grid has an observed 0.00833 degree pixel size in GDAL inspection. Web tiles are a visualization product and are not a replacement for querying the source raster.

The viewer fades data layers beyond their native tile zoom range to avoid false precision at scales finer than the source grid supports.

Layer color scales are explicit. The default per-layer mode is for reading structure inside the active layer: California uses +/-500 nT, EMAG2v3 uses a percentile-derived +/-300 nT, and Geoscience Australia South Australia uses a percentile-derived +/-700 nT. Shared mode uses a fixed +/-500 nT scale for cross-layer comparison; values beyond +/-500 nT saturate, so high-amplitude features such as Olympic Dam can honestly hit the visual ceiling.

Colors should not be compared directly across layers unless the viewer is in shared scale mode.

Globe view renders the same Web Mercator-cut MAG tiles as map view. Latitudes beyond approximately +/-85.05112878 degrees remain uncovered by design.

Terrain is optional topographic display context from Mapzen/Tilezen via AWS Open Data. It is not a geophysical layer package, not magnetic data, and not package evidence.

NoData

NoData is rendered transparent. Transparent areas should be interpreted as absent source data, not as zero magnetic anomaly.

The optional gap overlay visualizes existing coverage masks of this package only. Hatching means absence of a valid source value in this layer package, not proof that no measurement exists anywhere. EMAG2's dotted overlay marks cells where the value exists but the official uncertainty/source-code availability mask is false; it is not a terrain, weather, or hazard signal.

Observation Levels

Each layer records a human-readable observation_level and a machine-readable observation_level_kind. These fields explain the physical level represented by the package: aerial survey compilations, global upward-continued compilations, and future street/drone packages are separate views of the field. MAG does not infer missing street-level or drone-level coverage from aerial/global layers.

Dateline / Source Bounds

The raw USGS raster extent crosses or wraps around the antimeridian. GDAL inspection of the v0.1 source reports a west bound near -187 degrees longitude, outside the usual [-180, 180] range.

This does not affect the California clip used in v0.1, because the clip bbox is well inside the continental United States. It is a known processing risk for future global layers and must be handled explicitly before global reprojection or tiling.

EMAG2v3 Source Metadata

EMAG2v3 is a 2 arc-minute global magnetic anomaly compilation. It is useful as a low-resolution background layer, not as a high-resolution survey or a navigation-grade map.

The EMAG2v3 UpCont GeoTIFF is geographically gridded, but the downloaded file does not expose an EPSG authority tag that GDAL identifies. The pipeline assigns EPSG:4326 from the official NCEI EMAG2v3 metadata and records that assignment as an explicit processing/provenance step.

The EMAG2v3 GeoTIFF NoData tag is the float32 minimum sentinel -3.4028234663852886e+38. The 99999 NoData value mentioned in some text/CSV-oriented documentation is not the GeoTIFF tag observed in this file and must not be used for raster transparency unless it is actually present in the data.

The EMAG2v3 GeoTIFF bounds are pixel-center registered and inspected as approximately [-0.0166666666675, -89.9833333378325, 359.9833333513325, 89.9833333378325], not the idealized product envelope [0, -90, 360, 90]. The global web layer clips to Web Mercator latitude limits in later processing.

The EMAG2v3 UpCont layer is described as upward continued to 4 km. Source documentation wording varies between geoid and WGS84 ellipsoid phrasing, so the manifest records the observation level with that caveat rather than pretending there is no ambiguity.

The Web Mercator tile layer omits latitudes beyond approximately +/-85.05112878 degrees. Polar source data may exist in the original raster, but it is outside the standard slippy-map projection used by this viewer.

No Cursor Values

The viewer intentionally does not report nT values under the cursor. PNG tile colors are not a reliable numeric data source. Numeric querying uses the processed GeoTIFF through scripts/sample_point.py, not the visual tile layer.

v0.5 observability scores are deterministic field-structure heuristics. They are normalized per layer and are not directly comparable across layers. They do not estimate position, heading, trajectory, localization confidence, or platform fitness.

EMAG2v3 companion rasters add official Error and Code grids as separate artifacts. They were metadata-snapped to the parent value grid without interpolation because their shape matched and the observed geometry delta was below 0.5 pixel. The original and snapped geometries are recorded in the companion manifests.

EMAG2 Error sentinels remain semantically separate from value NoData. In the source grid, value NoData is 1,812,768 cells; Error -999 is 2,099,963 cells; Error -888 is 589,885 cells; and 877,080 cells have a magnetic value while uncertainty/source-code availability is false. This is not hidden or merged: value coverage, uncertainty coverage, and provenance coverage are three independent masks.

EMAG2 Code 888 means ambiguous grid-edge source code and Code 999 means no data per the EMAG2 readme. Source-code decoding beyond these sentinel meanings is deferred; the provenance raster preserves raw codes.

Geoscience Australia TMI 2019 South Australia

The ga_tmi_au_sa layer is a South Australia WCS subset of the Geoscience Australia Total Magnetic Intensity Grid of Australia 2019, seventh edition, 80 m cell size. It is a re-levelled national compilation with mixed contributing survey lineage, not raw survey flight-line data and not RTP/VRTP.

The source CRS is GDA94 (EPSG:4283). GDAL observed the WKT name GEOCENTRIC DATUM of AUSTRALIA without an authority code, so the pipeline records EPSG:4283 from the official metadata. The datum transform to WGS84 / Web Mercator is explicit in provenance. The approximate GDA94 to WGS84 positional shift on the Australian continent is about 1.8 m, far smaller than the approximately 80 m source cell size.

The source WCS subset uses NoData -99999. Phase B source inspect recorded 193,971,016 valid pixels and 40,043,984 NoData pixels before reprojection. After reprojection to Web Mercator, the numeric package records a separate value coverage mask; uncertainty and provenance refs remain null because no official companion error/provenance grids were onboarded for this product.

The layer uses a per-layer visual scale of +/-700 nT from the documented formula S = ceil(max(abs(P2), P98) / 50) * 50, with Phase B source percentiles P2=-534.7974 nT and P98=667.7047 nT. Its colors are not directly comparable to the California or EMAG2 layers without reading each layer legend.

Not Navigation Grade

This package is for research, education, and data inspection. It does not validate the source layer for real-world navigation, robot localization, aerial localization, marine localization, underground localization, or GPS-denied positioning.