Skip to content
This repository has been archived by the owner on Dec 2, 2020. It is now read-only.

cskarby/gridpp

 
 

Repository files navigation

Gridded post-processor

https://travis-ci.org/metno/gridpp.svg?branch=master https://coveralls.io/repos/metno/gridpp/badge.svg?branch=master&service=github

The program post-processes NetCDF files used at MET-Norway by using various downscaling and calibration methods. Post-processed forecasts are placed in a second Netcdf file, which has the desired output grid.

To convert between the two grids, a downscaling method is used. Currently implemented methods are:

  • nearest neighbour
  • elevation gradient (interpolation to new elevations using gradients)
  • smart neighbours (nearest grid points at the same elevation)
  • pressure (interpolation to new elevations using a standard atmosphere)

Calibrators include:

  • spatial smoothing (average within a neighbourhood)
  • quantile-quantile mapping
  • linear regression
  • Kriging of biases at points onto a grid (additive and multiplicative)
  • ensemble calibration using zero-adjusted Gamma distribution (e.g. for precipitation)
  • ensemble calibration using Box-Cox t-distribution (e.g. for windspeed)
  • ensemble calibration using Gaussian distribution (e.g. for temperature)
  • calculation of precipitation phase, using wetbulb temperature
  • calculation of QNH from surface pressure

Installation Instructions

From source

  1. Ensure the following libraries are installed:
    • Boost libraries
    • Netcdf c++ library
    • libgsl0
    • libblas
    • (Optional) Google test library (if developing new code)
  2. Download the source code from a release: https://github.com/metno/gridpp/releases
  3. Edit CXX, CFLAGS_O, IFLAGS, and LFLAGS in makefile
  4. Run 'make'

From debian packages

Follow instructions here: https://launchpad.net/~metno/+archive/ubuntu/gridpp

Running the program

To see program options, run:

./gridpp

To test the program on a fake dataset, run:

./gridpp testing/files/10x10.nc testing/files/10x10_copy.nc\
   -v T -d gradient\
   -v Precip -d smart numSmart=3 searchRadius=3
ncview testing/files/10x10_copy.nc

Running on multiple cores

To run using 8 threads:

export OMP_NUM_THREADS=8
./gridpp ...

Minimizing memory usage

Run the program in sequence for each variable:

./gridpp input output -v T ...
./gridpp input output -v Precip ...
./gridpp input output -v RH ...

Copyright and license

Copyright (C) 2015 MET Norway. Gridded post-processor is licensed under GPL version 2 or (at your option) any later version.

Contact

MET Norway
Postboks 43 Blindern
NO-0313 OSLO

About

Gridded post-processor

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 99.4%
  • Other 0.6%