Andrea Toreti 1,2, Franco Desiato 1, Guido Fioravanti 1, Walter Perconti 1 1 APAT – Climate and Applied Meteorology Unit 2 University of Bern [email protected]

Andrea Toreti1,2, Franco Desiato1, Guido Fioravanti1, Walter Perconti1

1 APAT – Climate and Applied Meteorology Unit 2 University of [email protected]

Sixth seminar for homogenization and quality control in climatological DB– Budapest 2008

59 stations – Air Force Weather Service and some Regional Environmental Agencies

Weak climatological control

Consistency control

www.scia.sinanet.apat.it

Outlier detection (using filter constructed in function of season, latitude and altitude + spatial comparison)

Correction or elimination of wrong values Monthly Series


Monthly Series

KZA – Kolmogorov Zurbenko Adaptive Filter1

Moving Window SNHT – Standard Normal Homogeneity Test2

MRPP – Multi Response Permutation Procedure3

1. Zurbenko et al. 1996. Detecting discontinuities in time series of upper air data: development and demonstration of an adaptive filter technique. J. of Climate 9, 3548 – 3560.

2. Alexandersson and Moberg. 1997. Homogenization of Swedish temperature data. Part I: homogeneity test for linear trends. Int. J. of Climatology 17, 25 – 34.

3. Mielke et al. 1981. Application of multi response permutation procedure for examining seasonal changes in monthly mean sea level pressure patterns. Monthly Weather Rev. 109, 120 - 126


KZA filter does not involve a reference series

Let Xt be the ‘candidate’ …

)(

)(

1)()( )(

tH

tT

q

qiittTtHt XqqY

0)())((

0)()(

'

'

tDqtDf

tDqtqH

0)())((

0)()(

'

'

tDqtDf

tDqtqT

)()1()(

|)()(|)(' tDtDtD

qtZqtZtD

))(max(

)(1))((

tD

tDtDf

and Z is the Kolmogorov Zurbenko filter of X


An example of KZA filter


Moving Window SNHT

Reference Series: at least three stations chosen using best correlation criterion applied to first difference series.

Window length: 12 years.

Inhomogeneities identified in the first/last four years were discarded.

MRPP

It is non parametric

Window length: 96 values; Centred at the inhomogeneity point

Two groups: before and after inhomogeneity

Comparison of Euclidean distances


a partition into g+1 disjoint subgroups: S1,…, Sg+1

MRPP

|| ii Sn

g

iinK

1 K

nC i

i

JIIJ

ii

n1

2

g

iiiC

1

where ΔIJ is the Euclidean distance between two members (I and J) of Si

TThe Statistic


An example: Enna

November 1964


Homogenized series

Standardized anomaly series1. Reference period: 1971 - 2000

Annual Series since December until November: Northern IT, Central IT and Southern IT

Seasonal Series: DJF / MAM / JJA / SON

1 Jones and Hulme. 1996. Calculating regional climatic time series for temperature and precipitation: methods and illustrations. Int. J. of Climatology 16, 361 – 377.


Northern Italy. Comparison

North - Winter

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

North - Spring

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

North - Summer

-1.5

-1

-0.5

0

0.5

1

1.5

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

North - Autumn

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig


Northern Italy. Comparison


Central Italy. ComparisonCentre - Spring

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

Centre - Winter

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

Centre - Autumn

-2

-1.5

-1

-0.5

0

0.5

1

1.5

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

Centre - Summer

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig


Central Italy. Comparison


Southern Italy. ComparisonSouth - Winter

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

South - Spring

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

South - Summer

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

South - Autumn

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig


Southern Italy. Comparison


Annual series. Comparison

North

-2

-1.5

-1

-0.5

0

0.5

1

1.5

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

Centre

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig

South

-1.5

-1

-0.5

0

0.5

1

1.5

2

1961

1964

1967

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Hom

Orig


Annual series. Comparison


Ongoing Research

Application of other methods

Pairwise comparisons / correction method

Daily homogenization and analysis


A spatial analysis of the homogenized DB

S mode PCA applied to monthly gridded series

Varimax Rotation

Hierarchical Cluster Analysis with the average linkage method

2/1

1

2

1

2

1

n

ii

n

ii

n

iii

AB

ba

baG A, B are two PCs and a, b are the

associated EOFs


7 significant clusters were identified.


Thank you!

Documents

Andrea Toreti 1,2, Franco Desiato 1, Guido Fioravanti 1, Walter Perconti 1 1 APAT – Climate and Applied Meteorology Unit 2 University of Bern [email protected]