Earth2Studio Weather Data Fetch
Fetch validated weather and climate variables from Earth2Studio sources by time and data type.
The skill is narrowly scoped to data retrieval and discloses remote stores, CDS credentials, and local caching; no malware, credential theft, or destructive default is evident. Seven points are deducted because least-privilege network scope, sensitive-credential protection, pre-download confirmation, rollback, and data-retention controls are not specified.
The workflow, analysis/forecast call signatures, lexicon checks, and troubleshooting table are reasonably consistent, and the supplied evaluation targets cover the main paths. Twelve points are deducted because it depends on live documentation and remote services, lacks tests in the assessed skill directory, and provides no reproducible execution evidence; the benchmark also reports an overall FAIL and findings.
The audience, variable/time/region/data-type inputs, analysis-versus-forecast boundary, and local-file non-fit are fairly clear. Five points are deducted because invocation still depends on agent judgment, Chinese-language interaction guidance is absent, and reachability of remote documentation, GCS, S3, and CDS from mainland China is not assessed.
The materials provide a version, Apache-2.0 license, author, purpose, limitations, troubleshooting, references, and evaluation material, with a generally usable structure. Five points are deducted for no changelog, no explicit maintenance owner or update path, duplicated content between skill documents, and the benchmark's author_format finding.
The skill can select sources around requested variables and times and generate directly modifiable Python scripts; the three evaluation targets specify concrete core outputs. It receives only six points because key APIs and lexicon facts require live rechecking, no actual output is statically verified, and the benchmark reports FAIL with weak discoverability and efficiency results.
Live documentation and lexicon references, a version, evaluation data, target scripts, and benchmark metrics provide some auditability. Six points are deducted because references are not pinned in the skill evidence, no tests exist under the assessed path, there is no corroborating source set, and static review cannot independently reproduce the conclusions.
- Running generated scripts accesses remote services such as GCS, S3, or CDS; confirm network scope, data terms, cost, rate limits, and that credentials are not printed or committed.
- Live documentation and lexicons may diverge from version 0.16.0; verify constructors, call signatures, and variable support in a pinned environment.
- The benchmark's overall verdict is FAIL, with duplicate-content and author-format findings plus low discoverability and efficiency; its metrics should not be treated as passed validation.
- For mainland-China users, validate reachability of the documentation and data sources in advance and provide an accessible or offline fallback.
What does this skill do, and when should you use it?
Earth2Studio Weather Data Fetch is an NVIDIA Earth-2 Team skill, version 0.16.0, for obtaining weather and climate data through Earth2Studio data-source APIs. It identifies analysis or forecast sources, verifies requested variables against source lexicons, and generates Python code that returns an xarray DataArray. It requires Earth2Studio, Python 3.10+, and network access; CDS-based sources additionally require a configured CDS API key. It does not cover installation, model discovery, inference pipelines, or custom data-source creation.
Extracts variables, times, analysis or forecast type, forecast lead times, region, and output format from a request; consults live Earth2Studio data-source documentation; checks source lexicons for variable support; surfaces coverage, resolution, and update-frequency tradeoffs; generates analysis or forecast fetch scripts; shows how to inspect the resulting DataArray and explains local caching.
- A weather researcher needs selected ERA5 or other reanalysis variables for a historical timestamp or range.
- A weather application developer needs recent analysis data from GFS or North American HRRR.
- A forecast workflow developer needs data keyed by initialization time and lead time from GFS_FX, GEFS_FX, or another forecast source.
- A Python user wants remote weather data returned directly as an xarray DataArray.
What are this skill's strengths and limitations?
- Covers source selection across analysis, reanalysis, observational, and forecast data sources.
- Uses live documentation and lexicon checks before recommending sources or writing code.
- Produces Python scripts that return and inspect xarray DataArrays.
- Can be installed individually through the documented skills CLI.
- Requires network access to remote GCS, S3, CDS, or other data stores.
- Variable availability, time coverage, and API parameters can change between Earth2Studio releases.
- CDS sources require API-key configuration and may experience queue-based throttling.
- Does not load local NetCDF/Zarr files directly and cannot mix analysis and forecast source types in one call.
- The supplied material provides no evidence of an independent test suite or fixed platform test matrix.
How do you install this skill?
Use the collection README’s documented command: npx skills add nvidia/skills --skill earth2studio-data-fetch --yes. Omit --yes for interactive selection of the destination. The skill becomes available when the agent next loads skills and encounters a relevant task.
How do you use this skill?
Describe the variables, time range, analysis or forecast type, lead time when applicable, region, and output format. For example: “I need ERA5 500 hPa geopotential height and 2 m temperature for January 1, 2020 at 00Z.” The skill maps these to z500 and t2m, checks candidate lexicons, and generates a fetch script.
How does this skill compare with similar options?
The skill explicitly treats several Earth2Studio sources as alternatives. Its example describes ARCO as free and requiring no API key for ERA5 access, while CDS is described as official and requiring a key. GFS is global operational data; HRRR is North America-only with 3 km resolution. The suitable choice depends on variables, historical coverage, region, resolution, and authentication.