Home
A research project exploring techniques for cleaning Dutch dual-polarisation weather radar polar volumes (pvols) so that biological signal — primarily birds — can be reliably extracted from data dominated by precipitation, electromagnetic (EM) interference and other artefacts.
This site collects the filtering experiments as a sequence of executable notebooks. The production filters distilled from these notebooks live in pipeline.R (full cleaning pipeline) and vispipeline.R (visualisation pipeline) in the project repository.
Notebooks
| Chapter | Topic |
|---|---|
| Banding | Removing the dual-PRF banding artefact in lower-elevation scans using RHOHV and the depolarization ratio (DPR). |
| Simple rain segmentation | A simple rain segmentation built on DPR, median-filter despeckling and a buffered area threshold via imager. |
| Rain stacking | “Stacking” segmented rain across elevation scans into a single composite rain mask. |
| Rain visualisation | Overlaying the rain mask on the bird-only RBC PPI for visual inspection. |
| EM Interference | Detecting and interpolating EM interference beams using per-beam range coverage and linearity tests. |
Data
Polar volumes are referenced as relative paths under data/. On the development machine that directory is a symlink to /mnt/clearskies/ (see .Rprofile). The raw HDF5 polar volumes are not part of the repository and the path is .gitignored.
Dependencies
- R 3.6.3 with package versions pinned via
renv(renv.lock). - Core analysis packages:
bioRad,imager,raster,stars,sf,tidyverse,patchwork,pracma,mclust,ggnewscale. - Documentation built with
bookdownusing thebs4_booktheme.
Restore the environment with renv::restore() and rebuild this site with bookdown::render_book("index.Rmd").