Parallel processing

Parallel processing can be used to speed up the calculation of statistics by splitting the calculation over multiple chunks. Chunking is possible through the chunks parameter in the focal, grouped and zonal statistics functions. The parallel processing is implemented using joblib, and can be used by providing a joblib context manager to the function.

For example, to calculate the focal mean of a raster with 1000x1000 pixels, using 5 threads and chunking the calculation over chunks of 100 pixels, the following code can be used:

a = np.random.rand(1000, 1000)
with parallel_config(backend='threading', n_jobs=5):
    focal_mean(a, window=7, chunks=100)