easyclimate.core.stats.detrend_spatial

Unified spatial detrending function with multiple computation methods.

Available methods: 1. ‘scipy_reduce’ - Simplified version using scipy.signal.detrend 2. ‘scipy’ - Optimized version using scipy.signal.detrend 3. ‘numpy’ - Manual numpy vectorized version 4. ‘rust’ - High-performance Rust backend version 5. ‘rust_chunked’ - Rust backend with chunked processing 6. ‘rust_flexible’ - Rust backend with flexible dimensions 7. ‘auto’ - Automatically selects the best available method

Functions

calc_detrend_spatial_fast(→ xarray.DataArray)

Remove linear trend along time dimension from spatio-temporal data.

Module Contents

easyclimate.core.stats.detrend_spatial.calc_detrend_spatial_fast(data_input: xarray.DataArray, time_dim: str = 'time', min_valid_fraction: float = 0.5, method: Literal['scipy_reduce', 'scipy', 'numpy', 'rust', 'rust_chunked', 'rust_flexible', 'auto'] = 'auto', **kwargs) → xarray.DataArray

Remove linear trend along time dimension from spatio-temporal data.

Supports multiple computation methods with optional automatic selection.

Parameters

data_inputxr.DataArray

The spatio-temporal data to be detrended.

time_dimstr, default “time”

Name of the time dimension.

min_valid_fractionfloat, default 0.5

Minimum fraction of valid (finite) values required for detrending. Grid points with fewer valid values will be set to NaN.

methodstr, default ‘auto’

Computation method to use: - ‘scipy_reduce’: Simplified version using scipy.signal.detrend - ‘scipy’: Optimized version using scipy.signal.detrend - ‘numpy’: Manual numpy vectorized implementation - ‘rust’: High-performance Rust backend - ‘rust_chunked’: Rust backend with chunked processing (for large datasets) - ‘rust_flexible’: Rust backend with flexible dimension handling - ‘auto’: Automatically selects the best available method

**kwargsdict

Additional arguments passed to specific methods: - chunk_size: int (for ‘rust_chunked’ method) - use_chunked: bool (for ‘rust_chunked’ method)

Returns

xr.DataArray

Detrended data with same shape and coordinates as input.

Raises

TypeError

If data_input is not an xarray.DataArray.

ValueError

If input parameters are invalid.

ImportError

If Rust method is selected but Rust backend is not available.

Examples

>>> import xarray as xr
>>> import numpy as np
>>>
>>> # Create sample data
>>> data = xr.DataArray(
...     np.random.randn(100, 50, 100),
...     dims=['time', 'lat', 'lon']
... )
>>>
>>> # Using scipy method
>>> result1 = calc_detrend_spatial_fast(data, method='scipy')
>>>
>>> # Using numpy method
>>> result2 = calc_detrend_spatial_fast(data, method='numpy')
>>>
>>> # Using Rust method (if available)
>>> try:
>>>     result3 = calc_detrend_spatial_fast(data, method='rust')
>>> except ImportError:
>>>     print("Rust backend not available")
>>>
>>> # Automatic method selection
>>> result4 = calc_detrend_spatial_fast(data, method='auto')

Notes

  • ‘scipy_reduce’: Simplest method but less robust with NaN values

  • ‘scipy’: Optimized scipy version with better special value handling

  • ‘numpy’: Manual vectorized implementation, typically 2-3x faster than scipy

  • ‘rust’: High-performance Rust implementation, typically 10-50x faster than numpy

  • ‘rust_chunked’: Rust chunked version for large memory datasets

  • ‘rust_flexible’: Rust version with flexible dimension ordering

  • ‘auto’: Uses ‘rust’ if available, otherwise ‘numpy’