Skip to content

Comparison of Model Output Statistics and Adaptive Regression based on Kalman Filters for the Lorenz-96 Model.

Notifications You must be signed in to change notification settings

anilzen/adaptive_regression

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

32 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Adaptive Regression and Model Output Statistics

This repository contains Python code, Jupyter notebook, LaTeX source, and documentation for comparing Model Output Statistics (MOS) and Adaptive Regression (AR) for the Lorenz-96 model.

The Jupyter notebook sets up an Observing Systems Simulation Experiment (OSSE) where Nature is given by the Lorenz-96 system. Biased versions of this system are used as models with which forecasts are made. Bias-correction performed by MOS and AR is compared under different scenarios. AR uses Kalman filters based on an idea by Eugenia Kalnay.

Bias can be constant or stochastic. It can vary in space across the grid with a normal distribution or vary in time mimicking seasonal patterns. One can also use blended models by combining forecasts using different types of biases (each specific bias choice with a type and amplitude correponds to a mode).

Statistics are collected for a 5-year nature run and compared to both MOS and AR forecasts. Forecast quality significantly depends on the type of bias. AR is most accurate for time-varying biases, which is the most realistic scenario.

Fore more information on usage, see the online documentation.

About

Comparison of Model Output Statistics and Adaptive Regression based on Kalman Filters for the Lorenz-96 Model.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published