Registry key: export  |  Priority: 9999  |  Category: I/O & Archival

Write climate data to disk in NetCDF, Zarr, or CSV. Handles gridded datasets, multi-point clip results (closest_cell dimension), and collections (lists/tuples/dicts of single-point datasets) with optional location-based filenames and S3 storage for Zarr.

Algorithm

Export runs at priority 9999 (last). It dispatches by export_method first, then by data structure.

flowchart TD
    Init([__init__: parse value dict]) --> Validate["_validate_parameters<br/>(format ∈ {zarr, netcdf, csv}; mode ∈ {local, s3};<br/>s3 requires zarr; export_method valid)"]
    Validate --> Start([execute])
    Start --> MethodCheck{export_method}

    MethodCheck -->|none| ReturnNoop["print message,<br/>return data unchanged"]
    MethodCheck -->|raw / calculate / both + dict result| HandleDict["_handle_dict_result<br/>(raw_data / calc_data keys)"]
    MethodCheck -->|raw / calculate / both + non-dict| HandleSel["_handle_selective_export<br/>(uses _determine_data_type)"]
    MethodCheck -->|data / skip_existing| ExportData["_export_data"]

    ExportData --> MatchResult{match result}
    MatchResult -->|Dataset / DataArray| HasCC{closest_cell dim<br/>and separated?}
    HasCC -->|Yes| SplitCC["_split_and_export_closest_cells"]
    HasCC -->|No| ExportSingle1["export_single"]
    MatchResult -->|dict| LoopDict["For each value:<br/>list/tuple → _export_collection,<br/>else → export_single"]
    MatchResult -->|list / tuple| ExportColl["_export_collection"]
    MatchResult -->|other| RaiseType["raise TypeError"]

    SplitCC --> ExportSingleN["export_single per slice<br/>(filename via _generate_filename)"]
    ExportColl --> ExportSingleN
    LoopDict --> ExportSingleN
    ExportSingle1 --> WriteOne
    ExportSingleN --> WriteOne

    WriteOne["export_single:<br/>1. _generate_filename + extension<br/>2. skip_existing check / unique filename<br/>3. _clean_attrs_for_netcdf"]
    WriteOne --> FormatMatch{match req_format}
    FormatMatch -->|zarr| Zarr["_export_to_zarr<br/>(local or s3)"]
    FormatMatch -->|netcdf| NCDF["_export_to_netcdf"]
    FormatMatch -->|csv| CSV["_export_to_csv (gzip)"]

    Zarr --> Done
    NCDF --> Done
    CSV --> Done
    HandleDict --> Done
    HandleSel --> Done
    Done([Return original data unchanged — chainable])
    ReturnNoop --> Done

    click Init "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L197" "__init__"
    click Validate "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L238" "_validate_parameters"
    click Start "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L326" "execute"
    click ExportData "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L394" "_export_data dispatcher"
    click MatchResult "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L432" "match result"
    click SplitCC "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L820" "_split_and_export_closest_cells"
    click ExportColl "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L457" "_export_collection"
    click HandleDict "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L573" "_handle_dict_result"
    click HandleSel "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L590" "_handle_selective_export"
    click WriteOne "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L1064" "export_single"
    click FormatMatch "https://github.com/cal-adapt/climakitae/blob/main/climakitae/new_core/processors/export.py#L1103" "match req_format"

Data Handling Modes

Mode 1: Gridded Datasets ⚠️ Important

Single xr.Dataset/xr.DataArray with lat and lon coordinate dimensions containing multiple values (e.g., shape (time, lat, lon)).

Key Behavior:

• Options separated and location_based_naming are silently ignored
• Always exports to a single file (cannot split gridded data by location)
• Single filename used with extension based on format

Example Input:

# Dataset from clip with single region or full state
ds.dims  # {'time': 8760, 'lat': 150, 'lon': 100, 'sim': 5}

Parameters Effective for This Mode:

• filename: Used for single output file
• file_format: Determines output format
• export_method: Controls return value
• mode: For Zarr only

Parameters Ignored for This Mode:

• separated (always False, cannot split gridded data)
• location_based_naming (no point coordinates to include)

Mode 2: Multi-Point Clip Results

Single xr.Dataset/xr.DataArray with closest_cell dimension (output from clipping to multiple lat/lon points). The closest_cell dimension represents individual target points.

Key Behavior:

• When separated=True: Data splits along closest_cell dimension
• Each slice becomes separate file with index or location-based name
• When separated=False: All points in single file with closest_cell dimension preserved

Example Input:

# Dataset from clip with multiple points
ds.dims  # {'time': 8760, 'closest_cell': 3, 'sim': 5}
# 3 points: LA, SF, SD

With separated=True + location_based_naming=True:

myfile_34-05N_118-25W.nc  # LA
myfile_37-77N_122-42W.nc  # SF
myfile_32-72N_117-16W.nc  # SD

With separated=True + location_based_naming=False:

myfile_0.nc  # LA
myfile_1.nc  # SF
myfile_2.nc  # SD

With separated=False:

myfile.nc  # All 3 points, closest_cell dimension preserved

Mode 3: Point Collections

A list or dict of xr.Dataset/xr.DataArray objects, where each item represents a single spatial point (scalar lat/lon coordinates with size 1).

Typical Sources:

• Output from batch_select() with return_data_and_metadata=False
• Lists returned by multi-point clipping with separated=True
• Custom collections of point time series

Example Input:

# List of 3 datasets, each a single point
data = [
    ds_la,  # shape (time: 8760, sim: 5)
    ds_sf,  # shape (time: 8760, sim: 5)
    ds_sd   # shape (time: 8760, sim: 5)
]

With separated=True + location_based_naming=True:

• Each dataset exported separately
• Filenames use lat/lon from dataset coordinates

myfile_34-05N_118-25W.nc  # LA
myfile_37-77N_122-42W.nc  # SF
myfile_32-72N_117-16W.nc  # SD

With separated=True + location_based_naming=False:

myfile_0.nc
myfile_1.nc
myfile_2.nc

With separated=False:

• Each dataset exports with base filename
• Duplicate filenames get _1, _2, _3 suffixes

Export Formats

Format	Use Case	Compression	Local/S3	Notes
NetCDF	Climate data standard, archival	netCDF4 / zlib	Local only	Default; preserves CF metadata
Zarr	Cloud / chunked I/O	Internal	Both (S3 requires mode="s3")	Path: s3://bucket/...zarr
CSV	Tabular / spreadsheets	gzip (.csv.gz)	Local only	Slow for large datasets

The Export processor does not support GeoTIFF. For GIS raster output, use rioxarray directly on the returned xr.Dataset after .get().

Format-Specific Notes

NetCDF (_export_to_netcdf)

• Attributes are sanitized via _clean_attrs_for_netcdf before writing.
• Best for archival, ncdump-readable.

Zarr (_export_to_zarr)

• mode="local" (default) writes to current working directory.
• mode="s3" writes to S3 — requires bucket configured outside the processor; only Zarr is allowed when mode="s3".
• Attributes also sanitized prior to write to keep parity with the NetCDF path.

CSV (_export_to_csv)

• gzip-compressed (.csv.gz).
• Practical only for small / point datasets.

Parameters

Parameter	Type	Mode(s)	Description	Constraints
filename	str	All	Base filename (no extension)	Default: "dataexport"
file_format	str	All	Output format	"NetCDF", "Zarr", "CSV" (case-insensitive; aliases inferred)
mode	str	All	Storage destination	"local" or "s3"; "s3" requires file_format="Zarr"
separated	bool	2, 3	Export items to separate files	Ignored for gridded data without closest_cell
location_based_naming	bool	2, 3	Use lat/lon in filenames	Effective when items have scalar lat/lon
filename_template	str / None	All	Custom filename template	Optional
export_method	str	All	Return / handler routing	"data", "skip_existing", "raw", "calculate", "both", "none"
raw_filename	str / None	—	Custom name for raw export	Used by _handle_dict_result / _handle_selective_export
calc_filename	str / None	—	Custom name for calculated export	Used by _handle_dict_result / _handle_selective_export
fail_on_error	bool	All	Raise vs warn on write failure	Default: bool

Code References

Method	Link to Code	Purpose
__init__	View on Github	Parse value dict, set defaults
_validate_parameters	View on Github	Validate format / mode / export_method / types
execute	View on Github	Route by export_method; returns input unchanged
_export_data	View on Github	match result dispatcher (Dataset/DataArray, dict, list/tuple)
_export_collection	View on Github	List/tuple handler honoring separated and location_based_naming
_export_single_from_collection	View on Github	Per-item write inside _export_collection
_handle_dict_result	View on Github	raw_data / calc_data dict (e.g., cava_data)
_handle_selective_export	View on Github	Selective raw/calculate/both for non-dict inputs
_determine_data_type	View on Github	Classify input as raw vs calculated
_export_with_suffix	View on Github	Append suffix (_raw, _calc) and write
update_context	View on Github	Record export metadata in context
_clean_attrs_for_netcdf	View on Github	Sanitize attrs for NetCDF/Zarr serialization
_is_single_point_data	View on Github	Detect scalar lat/lon
_has_closest_cell_dimension	View on Github	Multi-point clip detection
_split_and_export_closest_cells	View on Github	Split along closest_cell and write each slice
_extract_point_coordinates	View on Github	Pull (lat, lon) for filename suffix
_generate_filename	View on Github	Build base filename (location_based or template)
_get_unique_filename	View on Github	Add _1, _2, ... when target exists
export_single	View on Github	match req_format and call format writer

.table_id("day")
.grid_label("d03")
.processes({
    "time_slice": ("2015-01-01", "2015-12-31"),
    "clip": (37.77, -122.42),  # Single point
    "export": {
        "filename": "sf_daily_max_temp",
        "file_format": "CSV"
    }
})
.get())

Writes: sf_daily_max_temp.csv

Columns: time, t2max (one row per day)

## Implementation Details

### Format-Specific Behavior

**NetCDF**: Uses xarray `.to_netcdf()` with CF conventions and compression

```python
data.to_netcdf(filename, engine="netcdf4", encoding={var: {"zlib": True} for var in data.data_vars})

Zarr: Chunked cloud-optimized format (local or S3)

if mode == "s3":
    data.to_zarr(f"s3://bucket/{filename}.zarr")
else:
    data.to_zarr(f"./{filename}.zarr")

CSV: Flattens spatial dims; requires scalar or point data

data.to_csv(filename)  # Works for time series at single points

Skip Existing

With export_method="skip_existing", processor checks if file exists before writing:

if os.path.exists(filename):
    return None  # Don't overwrite
else:
    export_and_return_path(data, filename)

Location-Based Naming

Coordinates are formatted as compass directions:

# (37.7749, -122.4194) → "37-77N_122-42W"
lat_str = f"{abs(lat):.2f}{'NS'[lat < 0]}"
lon_str = f"{abs(lon):.2f}{'EW'[lon < 0]}"
filename = f"{base}_{lat_str}_{lon_str}.{ext}"

Common Patterns

Multi-Format Export

data = (ClimateData()
    .catalog("cadcat")
    .activity_id("WRF")
    .variable("t2max")
    .table_id("day")
    .grid_label("d03")
    .processes({
        "time_slice": ("2015-01-01", "2015-12-31"),
        "clip": "Alameda"
    })
    .get())

# Export to multiple formats
data.to_netcdf("alameda_2015.nc")
data.to_zarr("alameda_2015.zarr")
data.to_csv("alameda_2015.csv")

Batch Export Loop

counties = ["Alameda", "Contra Costa", "Santa Clara"]
for county in counties:
    (ClimateData()
        .catalog("cadcat")
        .activity_id("WRF")
        .variable("t2max")
        .table_id("day")
        .grid_label("d03")
        .processes({
            "clip": county,
            "export": {
                "filename": f"{county.lower()}_2015",
                "file_format": "NetCDF"
            }
        })
        .get())

See Also

• Processor Index
• How-To Guides → Export Data
• Architecture → Data Export
• xarray export docs: .to_netcdf()