MODELING TSI VARIATIONS USING Autoclass SOFTWARE ON …

Download Report

Transcript MODELING TSI VARIATIONS USING Autoclass SOFTWARE ON …

A piece of humble
pie
This map shows the state
of the southern California
desert about 10,000
years ago.
The presence of these
lakes is confirmed by oral
histories, packrat nests,
water-line marks on
mountain-sides, fish
traps, lakeside campsites
and other evidence.
Is this much climate
change really due to
variations in the Earth’s
orbit?
Do we really think we
understand climate
change?
R.K. Ulrich1 · D. Parker1 · L. Bertello1 · J. Boyden1
1 Department of Physics and Astronomy, University of
California, Los Angeles 90095 email: [email protected]
email: [email protected] email: [email protected]
email: [email protected]
MODELING TSI
VARIATIONS USING AUTOCLASS
SOFTWARE ON MWO DATA
How AutoClass Works
 AutoClass works on a set of observations.
 Each observation has attributes which are
values of observed parameters.
 Each observation is referred to as an instance.
 In our case the instance is a single pixel.
 The attributes are the absolute value of the
magnetic field and a line intensity ratio Ir.
 The line intensity ratio is:
Sample Images
Intensity ratio
Absolute magnetic field
How AutoClass Works (cont.)
 AutoClass takes all observed instances and
uses posterior Bayesian statistics to
determine a set of classes to which all
observations can be assigned.
 The output of an AutoClass search is a
number of classes.
 Each class is described by probability
distribution functions for the values of the
attributes.
 For our application we get central values and
gaussian widths for Ir and |B|.
The PDF’s for an 18-class
classification.
AutoClass finds that 18
classes describe a set of
observations consisting
of 12 image pairs.
Each image pair
consists of an absolute
field magnetogram and
in intensity ratiogram.
The image pairs were
selected as one per year
for the period 1996 to
2008.
Application of AutoClass to
MWO Data
 There are J classes denoted by index j.
 Each pixel i is assigned a probability that it
belongs to class j.
 We remember which image the pixels come
from and denote that image by index n.
 The sum over all pixels on image n of the
probabilities each belongs to class j gives us
an index Ajn which is effectively the fractional
area of solar image n covered by class j.
Properties of the Indicies
 The indicies obey:
 The TSI is reproduced by:
 The sj represent the TSI the sun would have if
entirely covered by class j.
 A deviation in sj is usefully defined as:
with
Properties of the classes
Modeling the TSI, part 1
Modeling the TSI part 2
Modeling the TSI part 3
Simulation of a TSI image
Comparison between an SBI
image and a TSI simulation
Quiet Groups Q0 and Q1
Groups Network, Plage 0, Plage 1
Time trends of the groups
Log of the fractional areas
Final Scatter Diagram
The final cross
correlation reaches
0.97. The Virgo data
has been detrended for
the drift from the
previous time of solar
minimum to the
present condition.
The comparison has
also been smoothed
with a three-point wide
Gaussian.
Points at the beginning
of the series are brown
while those at the end
are green.