MONTHS_TO_SEASONS

Name

MONTHS_TO_SEASONS

Purpose

This function extracts seasonal data from monthly data.

Calling Sequence

Result = MONTHS_TO_SEASONS( Data, Season [, Seasonlen] )

Inputs

Data: A vector containing the input time series of monthly data.
It can also be an array where the last dimension is time. The
time dimension must start with a January and end with a
December. Also see OUTPUT below.
Season: The index (January=0,...) value of the middle month of the
season.
Seasonlen: Optional. The number of months in a season. The default
value is "3". If Seasonlen is even the extra month is added
at the end.

Keyword Parameters

ANOMALY: If set, then the anomaly to the mean over BASEPERIOD is
calculated. If DESEASONALISE is set, then the anomaly from
the mean annual cycle over BASEPERIOD is calculated. The
default is to calculate total values.
BASEPERIOD: A vector of [YEARSTART,YEAREND] defining the period over
which to estimate the seasonal cycle when the ANOMALY or
DESEASONALISE keywords are set. YEARSTART is the first year
index value (month/12) while YEAREND is the last year index
value. e.g. using [10,19] would use the month indices 120
through 239.
CYCLE: If DESEASONALISE is set, then cycle contains a vector or array
(with calendar month as the last dimension) containing the
calculated seasonal cycle. Note this is the total seasonal
cycle if ANOMALY is set, or the seasonal cycle anomaly from
the annual mean if ANOMALY is not set.
DESEASONALISE: If set then the seasonal cycle is removed from the
data before the seasonal values are calculated. This has an
effect if NGOOD is set, because it then removes the bias
arising from missing data in certain months. It can also
effect the end values if the season overlaps years.
NBASEGOOD: The number of years in the base period required to have
good (non-NaN) values in order for the base value to be
calculated. The default is 1.
NGOOD: The number of months in a season required to have good
(non-NaN) values in order to calculate a seasonal value. The
default is NGOOD=Seasonlen.

Output

Result: Returns the seasonal time series as a vector or an array,
depending on the dimensions of Data. The time dimension will
be shorter by a factor of 12.
CYCLE: See above.
Data: If the DEASONALISE and/or ANOMALY options are set then Data is
returned in a deseasonalised and/or anomaly form.

Uses

var_type.pro

Procedure

This function calculates seasonal values based on monthly
values.

Example

Create a random monthly time series ten years long.
x = randomn( seed, 10*12 )
Calculate values for the summer (June-August).
result = months_to_seasons( x, 6, 3 )

Modification History

Written by: Daithi A. Stone (stoned@atm.ox.ac.uk), 2000-08-22.
Modified: DAS, 2000-10-02 (debugged).
Modified: DAS, 2004-12-30 (added BASEPERIOD, CYCLE,
DESEASONALISE, NBASEPERIOD, NGOOD keywords; this
required some altering of main algorithm)
Modified: DAS, 2005-04-25 (can now deal with large number of
spatial points)
Modified: DAS, 2005-08-25 (removed use of constants.pro)
Modified: DAS, 2006-01-06 (altered DEASONALISE keyword to not
include taking anomaly; added ANOMALY keyword)
Modified: DAS, 2006-02-14 (streamlined some of the code)
Modified: DAS, 2008-03-18 (allowed larger time dimensions through
long integer indices)