Homogenisation of temperature time series in Croatia Dubravka Rasol, Tanja Likso, Janja Milković Meteorological and Hydrological Service of Croatia

Outline Homogenisation in Croatia Croatian meteorological stations network MASH and SNHT methods applied to Croatian data and the comparison of the results Closer look - Karlovac temperature time series

Homogenisation in Croatia Carried out only sporadically (B. Volaric 1982, G. Galekovic 1995, T. Likso 2004) K. Pandzic: homogenisation of 22 temperature series for the Climate Atlas needs – modified SNHT method – to be published Joining the HOME COST Action ES0601 (2007-2011) Bilateral cooperation with the Hungarian Meteorological Service – the homogenisation issue (2007-2009)

Croatian meteorological stations network Temporal distribution of all meteorological stations

Croatian meteorological stations network Spatial distribution of main and climatological stations 41 main meteorological stations 116 climatological stations 336 precipitation stations some with more than 100 years of observations 2 upper-air stations 8 radar stations 34 automatic stations

Temperature series chosen for analysis Monthly mean air temperature series 9 stations from the NW part of Croatia Bjelovar Karlovac Koprivnica Krizevci Sisak Slavonski Brod Zagreb Gric Zagreb Maksimir Varazdin Period: 1961 - 2006

MASH and SNHT method Both methods are for relative homogeneity testing MASH – Multiple Analysis of Series for Homogenisation (Szentimerey, 1994) Reference series do not need to be homogeneous Mutual comparisons of series within the same climatic area Weight factors of reference series based on distance from the test series Software: MASH 3.02 (Many thanks to Tamas!) SNHT – Standard Normal Homogeneity Test (Alexandersson, 1986) Ideally, reference series should be homogeneous Likelihood ratio test Weight factors of reference series are correlation coefficients Software: AnClim (Stepanek, 2005)

Breakpoints detected by MASH and SNHT Bjelovar 1969 1981 1983 1989 1998 2001 2003 2004 1969 1983 1989 1998 2001 2004 Karlovac 1992 1993 1994 1999 2000 2001 1965 1984 1992 2000 2001 Koprivnica 1964 1967 1999 1965 1967 1980 1998 Krizevci 1980 1982 1984 1994 1995 2000 1981 1986 1993 1997 2000 Sisak 1964 1969 1971 1984 1989 1996 1964 1969 1979 1984 1996 2001 Slavonski Brod 1969 1974 1986 1991 1994 1996 1969 1986 1991 1996 1998 Zagreb Gric 1982 1989 1992 1998 1967 1972 1984 Zagreb Maksimir 1981 1987 1989 1991 2002 1981 1991 1995 2002 Varazdin 1971 2000 1970 1987 1989 2000

MASH and SNHT homogenised temperature series

MASH and SNHT homogenised temperature series 1981 1991

MASH and SNHT homogenised temperature series 1992 2001

MASH and SNHT differences

Karlovac Huge inhomogeneity! Two breakpoints detected by both methods - both in accordance with metadata: 1992 - relocated from the urban to the suburban area 2001 - moved around 50 m

Karlovac Differences between yearly mean temperatures at Karlovac station and at all the reference stations 1992 differences (°C) 2001 year

Karlovac y = 0.0108x + 10.645 y = 0.0316x + 9.3661

Conclusions Both MASH and SNHT methods detected all breakpoints known from metadata, as well as additional ones. The differences between homogenised series obtained by each method were negligible. The question is what to do with the Karlovac series: Is it better to divide it in two series or to homogenise it?