INTERFACE / PUBLIC COMPONENTS / NAMELIST / FILES / REFERENCES / ERRORS / BUGS / PLANS / PRIVATE COMPONENTS

MODULE reg_factor

Contact: Jeff Anderson
Reviewers:  
Revision: $Revision: 1.2 $
Change Date: $Date: 2004/04/26 20:23:02 $
Change history: see CVS log

OVERVIEW

Computes a weighting factor to reduce the impact of observations on state variables using information from groups of ensembles. Can be run using groups or using archived summary information available from previous group filter experiments.




OTHER MODULES USED

types_mod
utilities_mod



PUBLIC INTERFACE

use reg_factor_mod, only : comp_reg_factor



PUBLIC COMPONENTS


var = comp_reg_factor(num_groups, regress, time_index, obs_index, state_index, [obs_state_ind], [obs_state_max])
integer, intent(in)                           ::  num_groups 
real(r8), intent(in), dimension(num_groups)   ::  regress 
integer, intent(in)                           ::  time_index 
integer, intent(in)                           ::  obs_index 
integer, intent(in)                           ::  state_index 
integer, intent(in), optional                 ::  obs_state_ind 
integer, intent(in), optional                 ::  obs_state_max 
real(r8)                                      ::  var 

Description

Returns a weighting factor given regression factors from each group of a group filter or retrieves a factor generated by previous group filter runs from a file.

num_groups     Number of groups. Set to 1 when using information from previously run group filter from file.
regress     Regression factor from each group for a given state variable and observation variable pair.
time_index     Integer index of which time in observation file is being processed by filter. Not used in currently released version.
obs_index     Integer index of the observation being processed. Not used in current implementation .
state_index     Integer index of state variable being processed. Not used in current implementation.
obs_state_ind     Index into file generated for Bgrid model which could be duplicated in other large models.
obs_state_max     Maximum number of observation state variable pairs with non-zero impacts for a given model and observation sequence. Used for generating Bgrid statistic files.




NAMELIST

We adhere to the F90 standard of starting a namelist with an ampersand '&' and terminating with a slash '/'.

namelist / reg_factor_nml /  select_regression, input_reg_file
Contents Type Description
select_regression integer Selects the method for computing regression factor. 1 = compute using sampling theory for any ensemble size. 2 = low order model format. Works from archived time mean or time median regression files generated by low-order models like Lorenz-96. 3 = selects bgrid archived file. This is not currently supported in released versions. Default is 1.
input_reg_file character(len = 129) File name from which statistics are to be read for select_regression = 3. Default is "time_mean_reg".






REFERENCES


ERROR CODES and CONDITIONS




KNOWN BUGS




FUTURE PLANS

Need to implement large model archive capability in efficient fashion.