Spatial Data Format of the Week: NetCDF (Network Common Data Form)

The Scientific Multi-Dimensional Array Format

File extension: .nc or .nc4

What it is: A self-describing, machine-independent format for storing array-oriented scientific data—particularly climate, oceanographic, atmospheric, and Earth system model outputs. If you work with time-series gridded data (temperature over time, sea surface height, precipitation), you work with NetCDF.

Why NetCDF Exists

The Multi-Dimensional Problem

Scientific data often has 3+ dimensions:

Climate model output:

Longitude (x): 1440 points
Latitude (y): 720 points
Time: 365 days × 30 years = 10,950 time steps
Variables: Temperature, precipitation, humidity, wind

Total size: 1440 × 720 × 10,950 × 4 variables = ~45 billion data points

GeoTIFF can't handle this (designed for 2D grids + bands).

NetCDF can: Designed for n-dimensional arrays with metadata.

Example Data Structure

dimensions:
    lon = 1440 (0.25° resolution)
    lat = 720  (0.25° resolution)
    time = 10950 (daily, 1990-2020)

variables:
    float temperature(time, lat, lon)
        units: "degrees_celsius"
        long_name: "2-meter air temperature"
        _FillValue: -9999.0

    float precipitation(time, lat, lon)
        units: "mm/day"
        long_name: "daily precipitation"
        _FillValue: -9999.0

NetCDF Structure

1. Dimensions

Define array axes (lon, lat, time, depth, etc.)

2. Variables

Multi-dimensional arrays with metadata:

Coordinate variables: lon, lat, time (dimension metadata)
Data variables: temperature, precipitation, wind_speed

3. Attributes

Metadata about variables and the dataset:

Global attributes: Dataset title, institution, creation date
Variable attributes: Units, missing value codes, scale factors

4. Self-Describing

Everything needed to understand the data is embedded:

Coordinate system (lat/lon, projected, vertical)
Units (Celsius, mm, m/s)
Time encoding (days since 1900-01-01)
Variable descriptions

Common NetCDF Sources

Climate and Weather Data

ERA5 Climate Reanalysis (ECMWF)

Global climate data (1940-present)
Hourly resolution
Temperature, precipitation, wind, humidity
Google Earth Engine: Analysis-ready

PRISM Climate Data (Oregon State)

High-resolution U.S. climate grids
Monthly temperature and precipitation
4km resolution

CMIP6 Climate Models (World Climate Research)

Future climate projections
Multiple models and scenarios
Hundreds of variables

Oceanographic Data

Global Ocean Data Assimilation System (GODAS)

Sea surface temperature, salinity, currents
Ocean heat content
3D ocean state (lon, lat, depth, time)

Copernicus Marine Service

Global ocean monitoring
Satellite + in-situ observations

Satellite Products

MODIS/VIIRS (NASA)

Global vegetation indices
Sea surface temperature
Aerosol optical depth

GPM (Global Precipitation Measurement)

Satellite precipitation estimates
30-minute to monthly accumulations

Working with NetCDF in QGIS

Loading NetCDF

Drag and drop .nc file into QGIS, or:

Layer → Add Layer → Add Raster Layer
Select .nc file
Choose variable from list
Choose time step (if temporal)

QGIS extracts one 2D slice (lon × lat) at a time.

Temporal NetCDF (Time-Series)

NetCDF with time dimension opens as multi-band raster:

Band 1 = Time step 1 (e.g., Jan 2020)
Band 2 = Time step 2 (e.g., Feb 2020)
Band 365 = Time step 365 (e.g., Dec 2020)

View time-series animation:

Layer Properties → Temporal
Enable Dynamic Temporal Control
Use Temporal Controller panel to animate

NetCDF Browser Plugin

Better NetCDF handling:

Plugins → Manage and Install Plugins
Search: NetCDF Browser
Provides variable explorer and metadata viewer

Working with NetCDF in Python

xarray: The NetCDF Swiss Army Knife

import xarray as xr

# Open NetCDF file
ds = xr.open_dataset('temperature_2020.nc')

# View structure
print(ds)

Output:

<xarray.Dataset>
Dimensions:  (lon: 1440, lat: 720, time: 365)
Coordinates:
  * lon      (lon) float32 -180.0 -179.75 ... 179.75
  * lat      (lat) float32 -90.0 -89.75 ... 89.75
  * time     (time) datetime64[ns] 2020-01-01 ... 2020-12-31
Data variables:
    temp     (time, lat, lon) float32 ...

Extract Specific Location

# San Francisco time-series
sf = ds.sel(lon=-122.4, lat=37.8, method='nearest')
sf_temp = sf['temp'].values

Temporal Operations

# Calculate annual mean
annual_mean = ds['temp'].mean(dim='time')

# Seasonal means
winter = ds.sel(time=ds['time.season'] == 'DJF')
winter_mean = winter['temp'].mean(dim='time')

Spatial Subsetting

# Extract California
california = ds.sel(lon=slice(-125, -114), lat=slice(32, 42))

Export to GeoTIFF

import rioxarray

# Open with spatial reference
ds = xr.open_dataset('temp.nc')
ds = ds.rio.write_crs("EPSG:4326")

# Export time step to GeoTIFF
ds['temp'].isel(time=0).rio.to_raster('temp_jan2020.tif')

NetCDF Variants

NetCDF-3 (Classic)

Original format
2GB file size limit
Limited data types

NetCDF-4 (HDF5-based)

Modern standard (use this)
No file size limits
Compression support
Better performance

NetCDF-4 Classic Model

NetCDF-4 features, NetCDF-3 structure
Backward compatibility

NetCDF Advantages

1. Multi-Dimensional Native

Time-series, 3D, 4D+ data
Not "hacked" like multi-band GeoTIFF

2. Self-Describing

CF Conventions (Climate and Forecast)
Embedded metadata
No separate documentation needed

3. Efficient Storage

Chunking: Optimize read patterns
Compression: LZW, DEFLATE (NetCDF-4)
Packed data: Store as int16, scale to float64

4. Scientific Ecosystem

Standard in climate science
Extensive software support (Python, R, MATLAB, IDL)
Analysis-ready formats

5. Parallel I/O (NetCDF-4)

HDF5 parallel access
High-performance computing workflows

NetCDF Limitations for GIS

1. Not GIS-Native

Issue: NetCDF predates GIS conventions

No embedded CRS (like GeoTIFF .prj)
Coordinate system in CF conventions metadata (not always clear)
QGIS struggles with complex NetCDF structures

2. Projection Ambiguity

Common issue: Lat/lon assumed WGS 84, but:

Some use different datums
Projected data (polar stereographic, Lambert) requires careful handling
No .prj file to reference

3. Raster-Only (Mostly)

NetCDF is grid-based:

No native vector support
Point clouds, trajectories possible but uncommon
Use shapefiles/GeoJSON for vector data

4. Learning Curve

Complex structure for GIS users
xarray/nco tools required for advanced use
Not as simple as "drag into QGIS"

When to Use NetCDF

Best for:

Time-series gridded data (climate, weather, oceanography)
Multi-dimensional arrays (3D, 4D+)
Scientific analysis (climate models, forecasts, reanalysis)
Large datasets with compression
CF Convention-compliant data exchange

Avoid when:

2D static data (use GeoTIFF)
Simple web mapping (use COG or tiles)
Vector data (use shapefile/GeoJSON)
Non-temporal rasters (GeoTIFF is simpler)

NetCDF Tools

Command-Line (NCO - NetCDF Operators)

Concatenate files:

ncrcat file1.nc file2.nc output.nc

Calculate mean:

ncwa -a time input.nc output_mean.nc

Extract variable:

ncks -v temperature input.nc output_temp.nc

Python Libraries

xarray: Multi-dimensional arrays (NumPy for NetCDF)

netCDF4: Low-level NetCDF access

rioxarray: Spatial extensions (CRS, GeoTIFF export)

dask: Parallel, out-of-core computation

GUI Tools

Panoply (NASA): NetCDF viewer and plotter

ncview: Quick visualization

ToolsUI (Unidata): NetCDF structure browser

The Bottom Line

NetCDF is the standard format for scientific gridded data—especially climate, weather, and ocean models. If your data has a time dimension, use NetCDF. If it's a single snapshot, GeoTIFF is simpler.

GIS tools handle NetCDF, but it's optimized for scientific workflows, not cartography. Best practice: Use NetCDF for storage/analysis, export time slices to GeoTIFF for mapping.

Key takeaway: NetCDF is GeoTIFF's scientific cousin—same family (gridded data), different priorities (multi-dimensional science vs. 2D maps).