pysteps
pysteps copied to clipboard
pySTEPS
Python framework for short-term ensemble prediction systems.
Using the minimum over the entire domain is used instead. Maybe this is more robust anyway and zerovalue can be removed altogether? It is often wrong anyway, see issue #384...
The most generic ones (e.g. odim) should remain in the main repository.
In the extrapolation based nowcasting, the mask (NB: not the radar mask but rather the rain-norain mask in which we allow the generation of new rainfall by the noise process)...
It would be useful if all transformations are fully reversable, but currently `boxcox_transform`, `dB_transform` and `NQ_transform` aren't. In the `boxcox_transform` and `dB_transform` this just because the `metadata["zerovalue"]` does not use...
The nowcast spends a lot of time reading files. This can be made a lot more efficient.
Implementation of an importer for DWD radar data products and a reprojection onto a regular grid. These are necessary to test the implementation of the reduced-space EnKF combination approach (Nerini...
At RMI we are aiming to run pysteps in an operational environment. One thing that could improve copute time is adding MPI optionality. More info will be provided later in...
The current STEPS blending procedure implicitly uses the STEPS nowcasting procedure, but in principle, a similar approach should also be possible with other conventional and machine-learning nowcasting models. We could...
Currently, the STEPS blending procedure determines the weights for the extrapolation, NWP and noise components with generally an increasing weight for the noise component for longer lead times. Although this...