pyresample.ewa._fornav module

Resampling swath data to uniform grid using an Elliptical Weighted Averaging (EWA) algorithm.

pyresample.ewa._fornav.fornav_weights_and_sums_wrapper(cols_array, rows_array, input_array, grid_weights, grid_accums, input_fill, output_fill, rows_per_scan, weight_count=10000, weight_min=0.01, weight_distance_max=1.0, weight_delta_max=10.0, weight_sum_min=-1.0, maximum_weight_mode=False)

Python wrapper around the C interface to fornav weights and sums steps.

The main difficulty is that the C code can operate on multiple input arrays, but the python->C cython interface doesn’t automatically know the data type of the numpy arrays inside a tuple. This function casts the array pointers to the corresponding C type and then calls the templated cython and C++ functions.

This algorithm works under the assumption that the data is observed one scan line at a time. However, good results can still be achieved for non-scan based data is provided if rows_per_scan is set to the number of rows in the entire swath.

Parameters:

  • cols_array – numpy array of grid column coordinates for each input swath pixel

  • rows_array – numpy array of grid row coordinates for each input swath pixel

  • input_array – numpy array with the same geolocation and same dtype

  • grid_weights – numpy array to be filled with the sum of weights data (must be writeable)

  • grid_accums – numpy array to be filled with sum of weight * image pixel data (must be writeable)

  • input_fill – fill value in input data arrays representing invalid/bad values

  • output_fill – fill value written to output data arrays representing invalid/bad or out of swath values

  • rows_per_scan – number of input swath rows making up one scan line or all of the rows in the swath

  • weight_count – number of elements to create in the gaussian weight table. Default is 10000. Must be at least 2

  • weight_min – the minimum value to store in the last position of the weight table. Default is 0.01, which, with a weight_distance_max of 1.0 produces a weight of 0.01 at a grid cell distance of 1.0. Must be greater than 0.

  • weight_distance_max – distance in grid cell units at which to apply a weight of weight_min. Default is 1.0. Must be greater than 0.

  • weight_delta_max – maximum distance in grid cells in each grid dimension over which to distribute a single swath cell. Default is 10.0.

  • weight_sum_min – minimum weight sum value. Cells whose weight sums are less than weight_sum_min are set to the grid fill value. Default is EPSILON.

  • maximum_weight_mode – If -m is not present, a weighted average of all swath cells that map to a particular grid cell is used. If -m is present, the swath cell having the maximum weight of all swath cells that map to a particular grid cell is used. The -m option should be used for coded data, i.e. snow cover.

Returns:

boolean if any input data was used on a any output grid cell

pyresample.ewa._fornav.fornav_wrapper(cols_array, rows_array, input_arrays, output_arrays, input_fill, output_fill, rows_per_scan, weight_count=10000, weight_min=0.01, weight_distance_max=1.0, weight_delta_max=10.0, weight_sum_min=-1.0, maximum_weight_mode=False)

Python wrapper around the C interface to fornav.

The main difficulty is that the C code can operate on multiple input arrays, but the python->C cython interface doesn’t automatically know the data type of the numpy arrays inside a tuple. This function casts the array pointers to the corresponding C type and then calls the templated cython and C++ functions.

This algorithm works under the assumption that the data is observed one scan line at a time. However, good results can still be achieved for non-scan based data is provided if rows_per_scan is set to the number of rows in the entire swath.

Parameters:

  • cols_array – numpy array of grid column coordinates for each input swath pixel

  • rows_array – numpy array of grid row coordinates for each input swath pixel

  • input_arrays – tuple of numpy arrays with the same geolocation and same dtype

  • output_arrays – tuple of empty numpy arrays to be filled with gridded data (must be writeable)

  • input_fill – fill value in input data arrays representing invalid/bad values

  • output_fill – fill value written to output data arrays representing invalid/bad or out of swath values

  • rows_per_scan – number of input swath rows making up one scan line or all of the rows in the swath

  • weight_count – number of elements to create in the gaussian weight table. Default is 10000. Must be at least 2

  • weight_min – the minimum value to store in the last position of the weight table. Default is 0.01, which, with a weight_distance_max of 1.0 produces a weight of 0.01 at a grid cell distance of 1.0. Must be greater than 0.

  • weight_distance_max – distance in grid cell units at which to apply a weight of weight_min. Default is 1.0. Must be greater than 0.

  • weight_delta_max – maximum distance in grid cells in each grid dimension over which to distribute a single swath cell. Default is 10.0.

  • weight_sum_min – minimum weight sum value. Cells whose weight sums are less than weight_sum_min are set to the grid fill value. Default is EPSILON.

  • maximum_weight_mode – If -m is not present, a weighted average of all swath cells that map to a particular grid cell is used. If -m is present, the swath cell having the maximum weight of all swath cells that map to a particular grid cell is used. The -m option should be used for coded data, i.e. snow cover.

Returns:

tuple of valid grid points written for each output array

pyresample.ewa._fornav.write_grid_image_single(output_array, grid_weights, grid_accums, output_fill, weight_sum_min=-1.0, maximum_weight_mode=False)