Atmospheric Research 122 (2013) 16–31

Contents lists available at SciVerse ScienceDirect

Atmospheric Research
journal homepage: www.elsevier.com/locate/atmos

Changes in temperature and precipitation extremes observed in

Modena, Italy
M. Boccolari ⁎, S. Malmusi
Department of Materials and Environmental Engineering, University of Modena and Reggio Emilia, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Climate changes has become one of the most analysed subjects from researchers community,
Received 13 February 2012 mainly because of the numerous extreme events that hit the globe. To have a better view of
Received in revised form 12 October 2012 climate changes and trends, long observations time series are needed.
Accepted 14 October 2012 During last decade a lot of Italian time series, concerning several surface meteorological
variables, have been analysed and published. No one of them includes one of the longest record
Keywords: in Italy, the time series of the Geophysical Observatory of the University of Modena and Reggio
Temperature Emilia. Measurements, collected since early 19th century, always in the same position, except
Precipitation for some months during the second world war, embrace daily temperature, precipitation
Trends
amount, relative humidity, pressure, cloudiness and other variables.
Climate indices
In this work we concentrated on the analysis of yearly and seasonal trends and climate extremes
Time series
of temperature, both minimum and maximum, and precipitation time series, for the periods
1861–2010 and 1831–2010 respectively, in which continuous measurements are available.
In general, our results confirm quite well those reported by IPCC and in many other studies over
Mediterranean area. In particular, we found that minimum temperature has a non significant
positive trend of +0.1 °C per decade considering all the period, the value increases to 0.9 °C per
decade for 1981–2010. For maximum temperature we observed a non significant +0.1 °C trend
for all the period, while +0.8 °C for the last thirty years.
On the other hand precipitation is decreasing, -6.3 mm per decade, considering all the analysed
period, while the last thirty years are characterised by a great increment of 74.8 mm per decade.
For both variables several climate indices have been analysed and they confirm what has been
found for minimum and maximum temperatures and precipitation. In particular, during last
30 years frost days and ice days are decreasing, whereas summer days are increasing. During
the last 30-year tropical nights and warm spell duration indices are characterised by a
particular strong increment, if compared to the ones of the entire period.
Finally, a cursory comparison between winter precipitation and NAO index was done, showing
a high anti-correlation, especially since the second half of 20th century.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction as hurricanes, droughts, floods and so on, that increasingly hit

the globe. The study of these events and their frequency is
The subject of climate changes has become more and very important in order to understand why they happen and
more important, due to the numerous extreme events, such if there are some signs or particular periodicities. Moreover
this kind of researches are fundamental in order, first of all, to
save human beings, but also to avert or to limit damages
⁎ Corresponding author at: Department of Chemical and Geological Sciences,
University of Modena and Reggio Emilia, Modena, I-41100, Italy. Tel.: 39 059
done by these extreme events.
2056215; fax: 39 059 2056243. In 2007 IPCC published its Fourth Assessment Report
E-mail address: mauro.boccolari@unimore.it (M. Boccolari). (IPCC, 2007), where numerous results, obtained by studies

0169-8095/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.atmosres.2012.10.022
 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31 17

done all over the globe from different instruments observa- Brunetti et al., 2000a; Piervitali and Colacino, 2003; Pavan
tions (both from singular stations, satellites and models), are et al., 2008), but also on other meteorological variables, as for
collected, to give a global view of climate changes and of the example pressure (Maugeri et al., 2004). Brunetti et al. (2006b)
anthropogenic effects on the Earth system; it reported also recently improved their previous studies and they analysed
some projection models results and the potential impacts on several mean, minimum and maximum temperatures and
the life of human beings. precipitation time series that almost cover all Italian territory.
IPCC described that during the last one hundred years Another important work has been done by Auer et al. (2007),
(1906–2005) the global mean surface temperature increased in which they took into account the HISTALP database and
by almost + 0.74 °C, particularly during two periods: be- they homogenised and analysed temperature, precipitation,
tween 1910 and 1945 and between 1970 and 2005. In the last pressure, sunshine and cloudiness records. They reported an
analysed 50 years the warming rate is equal to + 0.13 °C per increasing trend of about +1 °C per century all over Italy for
decade, almost double the one over the last 100 years mean temperature, higher for minimum temperature than for
(+ 0.07 °C per decade). Moreover it said that, for the northern maximum one. On the other hand precipitation is decreasing
Hemisphere, the second half of the 20th century was warmer of about −5% per century, but trends are low and rarely
than any other 50-year period in the last 500 years, with significant.
the greatest warming during winter and spring. It reported a So far, none of these cited works took into account the
reduction of daily cold extremes and cold nights, while an relatively long temperature and precipitation time series of
increase of warm extremes and warm nights. Finally, diurnal Geophysical Observatory of Modena (GOMO) located in the
temperature range (DTR) has decreased of about − 0.07 °C Italian Po Plain, and that have been studied in the present
per decade in the second half of the 20th century, even if in paper. GOMO dataset includes different meteorological vari-
the last decades the change has been smaller. ables (temperature, precipitation, pressure, humidity, wind,
On the other hand, IPCC underlined the difficulty to give a snow, cloudiness, fog and solar radiation), but in this work
general global trend for precipitation, because its changes only daily maximum and minimum temperatures, during
are more spatially and seasonally variable than temperature the period between 1861 and 2010, and daily accumulated
ones. It reported that, in the considered period, it has become precipitation, for the period between 1831 and 2010, have
wetter in northern Europe, while drier in the Mediterranean. been analysed. Data before 1861 for temperature and before
Moreover an increase of heavy precipitation for many land 1831 for precipitation also exist but they represent sparse
areas has been found, also for regions where the total amount observations, so it was decided to not considered them in this
of precipitation is decreasing. These interannual and inter- analysis.
seasonal precipitation variations could be caused by the These time series have already been studied and published
influence of different atmospheric oscillations, as for example since the early years of the establishment of Geophysical
the North Atlantic Oscillation, which influences precipitation Observatory. However, do not exist recent international cli-
over Europe. matological studies that analyse the GOMO time series. Among
To better analyse long-term climate changes, in addition the less old, it could be cited for example Marseguerra et al.
to variations of minimum, maximum and mean temperatures (1979), in which daily temperature for the period 1869–1977
and precipitation, several studies have defined some climate has been analysed, estimating minimum, mean and maximum
indices to underline the variation of intensity of extreme temperature trends equal to +0.14, +0.08 and −0.05 °C per
events (e.g. Klein Tank and Können, 2003; Frich et al., 2002; decade respectively, or Aprilesi et al. (1977), in which the
Moberg et al., 2006; Ramos et al., 2011). Different groups of stochastic behaviour of daily temperatures during 1892–1975
indices have been created, using thresholds, percentiles was studied. Recently, a collection of several information about
and in general values tied with the low or high probability GOMO history, dataset from all its time series and meteoro-
that certain events occur. One of the most used is the group logical events, can be found in Lombroso and Quattrocchi
of indices of the Expert Team on Climate Change Detection (2008).
and Indices (ETCCDI), sponsored by the Commission for The main aim of this work is to analyse temperature and
Climatology (CCI) of the World Meteorological Organization precipitation trends and the time variability of their climate
(WMO), the JCOMM (Joint Commission for Oceanography extremes since the beginning of Modena time series, to
and Marine Meteorology) and by the CLIVAR (Climate understand if the climate of this area has changed.
Variability and Predictability) project. The paper starts with a brief history of Geophysical
For the investigation of climate changes and trends long Observatory of Modena and the description of dataset, its recon-
time series are fundamental. Italy has a significant historical struction and its homogenisation. Thereafter, both minimum
role in meteorological observations and their improvement, (TN) and maximum (TX) temperatures extreme indices are
underlined by the creation of several important meteorological given. Following the analogous analysis for precipitation (PRCP)
instruments and by the development of the “rete of Cimento”, time series is reported. At the end the conclusions, in which
the first observations network in the world (Brunetti et al., obtained results are summarised.
2006b). Several long-term time series are available, thanks
to the numerous observatories and institutes in the Italian 2. History, description of data and methodology
area; few started to collect data since the 18th century, other
since the 19th one. Several climatological works have been 2.1. History of Geophysical Observatory
done taking into account these stations: studies on minimum,
maximum or mean temperatures (Maugeri and Nanni, 1998; On 14th January 1826 in the east tower of the Ducal
Brunetti et al., 2000b), on precipitation (Buffoni et al., 1999; Palace of Modena (Italy), an astronomic observatory was set
18 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31

up by the astronomer Giuseppe Bianchi, following the Duke In 1897 the observatory takes the current official name of
Francesco IV and his brother Archduke Massimiliano d'Este “Geophysical Observatory”. In 1907 new instruments were
will. The astronomic observatory was allowed in perpetual bought and Cesare Bonacini, that in 1906 took Chistoni place
use at the University of Modena (now University of Modena as director, modified in June 1908 the window, adding
and Reggio Emilia) and the management was assigned to grids in the lower part of the parapet to improve aeration,
Bianchi. Since its creation, in addition to astronomic obser- modifying the roof (Bonacini, 1909). In 1944, because of the
vations, some meteorological measurements (air pressure, second world war, researchers had to change place to the
temperature and precipitation) were also collected, even if instruments and from 21 December 1944 until 21 April 1945
not systematically, in fact only in 1830 measures began to be they took observations from the terrace in the old Institute of
recorded regularly. In that period temperature was measured Physics of Modena University, always in the old town centre
putting thermometers outside at different fixed hours; in this (Barbanti-Silva, 1952).
way there was no kind of shield to protect instruments from Another milestone was the change of instruments in 1984;
direct sun radiation and, moreover, no minimum and maxi- in fact from that year until 1988 recorder SIAP instruments
mum temperature measurements were taken. From 1860 also were placed side by side with minimum/maximum tempera-
minimum and maximum temperature were recorded and ture thermometers and from July 1988 the meteorological
since 1863 a Rutherford thermometrograph has been used. station was completely automated, with PT100 sensor. In 1999
After a brief period led by Pietro Tacchini (1859–1863), the observatory was hit by fire, but both temperature and
Domenico Ragona arrived in 1863 as new director; with precipitation instruments had no damages.
Ragona the observatory was improved and meteorological There were only two interruptions in the temperature
observations became the main aim of the institute (Ragona, time series, the first one was between 13th August 2001 and
1865). He understood that measurements were taken in a 7th February 2002 due to problems tied with the structure
wrong way, so he projected and then built a sort of meteo- of the observatory, after which also a new SIAP MICROS
rological window, in the northern-eastern side of the tower, meteorological station was installed. The second interruption
that started working in 1865. He adapted one of the tower was between 24th June and 29th October 2002, due to a crack
windows, putting a sort of wood box to protect instruments of the instrument. Considering precipitation, only measure-
from the direct sun radiation. Ragona bought some Salleron ments for nine days, between 08 and 16 September 1943,
instruments, among which there were the first recorder were missed and, due to the length and the meteorological
devices; interesting was the fact that some of the instruments situation of that period, those days have been considered dry.
were created by himself, as for example the hygrothermograph
built in 1869 by Salleron, following Ragona's project, that 2.2. Location and dataset description
became operative since May of the same year. In 1876 the
astronomical observatory changed its name in “Meteorological Modena is located in a central position in the Po Plain, about
Observatory”. In 1886 a Richard recorder was introduced. 100 km away from both the Adriatic and the Ligurian Sea
Ragona, before his death in March 1892, understood that the (the presence of the Apennine Mountains limits the influence
window wasn't suitable, due to the lack of the right ventilation. of the second one), in a continental position (Fig. 1).
Prof. Ciro Chistoni, that became the new director of the According to the recent Köppen–Geiger climate classifi-
observatory after Ragona's death, decided to improve it, cation (Kottek et al., 2006), based on temperature and
projecting a new one that became operative in June 1898. A precipitation data for the period 1951–2000, the climate of
milestone for the observatory was, in 1895, the publication of Modena is classified as warm temperate, fully humid, hot
the first yearbook (Chistoni, 1895) containing all the meteoro- summer (Cfa). Rubel and Kottek (2010) updated the climate
logical observations of years 1892–1894. Since that time, the classification analysing temperature and precipitation since
publication of the yearbook has continued almost every year 1901 and using also models projection, in order to cover the
until nowadays. period between 1901 and 2100. Modena climate classifica-
Analysing metadata a discrepancy was found for the first tion has not changed from 1901 till nowadays, while in the
three months of 1892, where published data were corrected future, considering the A1FI (Mitchell et al., 2004) emission
respect to values reported in meteorological observations scenario (a world with quick economic growth, that is
volume; probably Chistoni tried to correct some mistakes maximum climate shift), the climate could become Csa
done before his coming in March. The new long window (from fully humid to summer dry) in the last period of the
lined balcony, used for measurements till today, is near the 21th century (2076–2100).
old Ragona's window, at the same height, and it has wood As said before, Modena observatory is located at the last
walls with bigger openings and so more ventilated. A com- floors of the right tower of Ducal Palace. Instruments have
parison between the two measurements locations was done always been positioned at 64.2 m (above the sea level), that
later in 1923 by Michele Baruzzi (1928). He built a copy of is almost 31.07 m from the surface (lat. 44°38′53′′ N and
the old window and for two years he took measurements long. 10°55′47′′ E).
from both positions. He found that temperature from the old Modena observatory is in the old town centre and it has
window was always greater than that measured in the new been always in an urban context since its creation. Until the
balcony, with a mean difference around 0.25 °C. This discrep- 1950s the northern area was mainly agriculture terrain, after
ancy is due mainly to the different ventilation between the two it has been urbanised.
places. He compared this difference with wind data and he Modena old town centre had no great changes during the
found that when wind was stronger difference was minimum analysed period, however the population has incremented
and vice-versa. since 1861 from 52629 to 175502 (in 2001) and, consequently,
 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31 19

Fig. 1. Geographical location of Modena station; map has been built up with GTOPO30 DEM of U.S. Geological Survey Department of the Interior (USGS).

suburb areas have grown as appear in the statistical data of the with lacking data and the homogenisation tests have been
Modena municipality (ANCM, 2010); in fact referring to 2001 performed.
about 63% of houses have been built between 1946 and 1981. Although all the time series have been controlled from the
Even if the analysed station is positioned in an urbanised foundation of the observatory, an additional quality control
area, it is possible to study the climatological evolution of this was performed; in particular for temperature, the physical
zone. Conditions around the station have not been affected consistency between maximum and minimum values and
by great changes, primarily because it has been positioned in then the identification of any outliers have been applied.
an historical building that has always been in the centre of Outliers have been searched by checking if any value falls
the city. In order to consider an urban station representative within a range pre-defined, equal to four standard deviations
for the climatological study of the area, instruments must be (decided analysing if found outliers were or not confirmed by
positioned over the upper limit of the roughness sublayer, metadata) from the mean climatological value for that day;
that is characterised by the strong roughness of the urban values outside this range have been empirically modified
surface, due to the presence of buildings and trees. In- case-by-case based on data and meteorological situation of
struments must be located at least twice the mean height of the closest days. Only a small number of these outliers have
the roughness elements, in order to not be subjected to their been found and then modified; they regarded only maximum
influence and, in that way, datasets could represent a general temperature.
response at local-scale (Rotach, 1999, 2001; Kastner-Klein After the quality check, the reconstruction of periods with
and Rotach, 2003). The initial choice of Modena observatory lacking data has been done. As already said, such periods are
instruments location follows these requirements. Moreover 13 August 2001–07 February 2002 and 24 June 2002–29
IPCC (Trenberth et al., 2007) affirmed that the effects of October 2002, due to problems at the observatory structure
urbanisation on global temperature are negligible, only and to an instrumental crash respectively. The reconstruction
+ 0.006 °C per decade. of temperature time series has been obtained by means a
The first meteorological time series analysed in this work multiple linear regression procedure based on predictors of
is temperature (°C), more precisely both minimum and observations collected in an area inside Modena's scientific
maximum ones. The analysed time series starts in 1861 and university campus, that is far about 3 km from geophysical
data have been collected from GOMO; in the entire archive observatory. The period in which data has been contempo-
there are also previous observations but taken irregularly. rary measured in both stations was 1997–1999.
The second meteorological time series analysed in this work About precipitation, it has been decided to reconstruct the
is the daily accumulated precipitation in mm; it has been lacking data using exactly the precipitation measurements
collected from GOMO since the 1st April 1830. collected in the same university campus during the period of
missing data. As already noted in the previous Section 2.1,
missing total precipitation data, for the week in 1943, have
2.3. Quality check, data reconstruction and homogenisation test been set to zero.
Talking about time series homogeneity, different causes
For all datasets, both daily temperatures and precipita- can be sources of inhomogeneities in the data as changes
tion, the quality check, the reconstruction of short periods in measurement instruments, in station relocation and its
20 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31

closest environment and so they should be taken into (with opposite sign, for example adding +0.1 to TX and −0.1
account. Modena time series has the advantages that it has to TN or viceversa), unless the available metadata indicate
been positioned in the same location (apart during the that the correction should be applied only to TX or TN, in this
second world war between 1944 and 1945) and it has been case the temperature is corrected with +/− 0.2 °C. If the
always controlled by at least one operator. Metadata, that are correction makes worse test statistics, the test run is done
information about station's history, are very useful in order again but with correction values with opposite sign. Procedure
to homogenise time series, and fortunately some works, in terminates when at least two of four tests pass.
particular of the early period, are still available and changes The general algorithm is composed by the following
in instruments and locations are recorded. steps:
Despite of several methods that could be applied to
1) Calculation of mDTR and vDTR from daily TX and TN time
homogenise a time series (Peterson et al., 1998), in order to
series
assess the possible inhomogeneities it was decided to follow
2) Perform 4 tests and compute test statistics
the procedure indicated in Wijngaard et al. (2003). In this
3) Is it the first loop? If YES go to 6.
method the number of wet days in a year has been chosen as
4) Test statistics are worst than those of the previous loop? If
the testing variable for precipitation (a wet day is defined as a
NO go to 6)
day with precipitation above 1 mm), and two variables
5) Delete last correction to TX and TN and recompute mDTR
derived from diurnal temperature range (DTR), i.e. its annual
and vDTR, and go to 7)
mean (mDTR) and its annual mean absolute day-to-day
6) Almost two test statistics are within the respective critical
differences (vDTR), as testing variables for minimum and
levels? If Yes then STOP
maximum temperature.
7) Found the year of break; add ΔT to TX and subtract ΔT to
For a specific year with n days, vDTR for that year is defined
TN from the beginning of the series to the break year;
as vDTR= (∑i|DTR(i) − DTR(i − 1)|)/(n − 1) , i = 2,…, n.
then goto 1)
Again, following Wijngaard et al. (2003), SNHT for a single
break (Alexandersson, 1986), Buishand range test (Buishand, At the end of testing procedure, 17 breaks have been
1982), Pettitt test (Pettitt, 1979) and Von Neumann ratio test detected (see Table 1). Among the found breaks, seven (four
(Von Neumann, 1941) have been selected as homogenisation between 1861 and 1882) are undocumented; the first two
tests, without reference series, to examine the departure of (1862 and 1863) are comprised in a period in which changes in
homogeneity in the mDTR and vDTR time series and indirectly measuring techniques and observing practices have frequently
to correct daily TN and TX time series using as much as possible happened. For these breaks, the DTR correction was rather
the available metadata. Wijngaard et al. (2003) classify time large as it can be noted in column 4 of Table 1. The last four
series in ‘useful’, ‘doubtful’ and ‘suspect’ depending if one or undocumented detected breaks after 1900 have been corrected
zero, two or three or four tests reject the null hypothesis (that with low values. The documented breaks have been found in
the annual values of the testing variable are independent and 1869 when a new hygrotermograph began to be operative; in
identically distributed) at the 1% level respectively. 1886 with the introduction of a new Richard recorder; in 1898
Original precipitation time series past directly the previ- when the new observation window was built; in 1908 when
ous set of four tests, so no correction procedures were further a new ventilated grid was set; in 1944 with the temporary
conducted, while the original mDTR and vDTR didn't pass all relocation of the instruments; in 1967 and 1968 with TN
test and the homogenisation procedure was needed. Several and TX instrument calibration (values reported exactly on a
breaks have been found and then corrections have been per- document by an operator), and in 1985 and 1988 with the
formed to obtain two new useful time series for the climate introduction of a new automatic instruments together with
investigation, in particular temperature series result ‘useful’ the old instruments (Lombroso and Quattrocchi, 2008). The
for the period 1980–2010, while for the entire period maximum correction has been made before 1883, as it could
1861–2010 result ‘doubtful’. In Section 3.1 for temperatures, also be noted in Fig. 2, in which differences between the
and Section 4.1 for precipitation further descriptions of corrected TX and TN (TXcorr and TNcorr, respectively) and the
homogenisation procedure will be given. original TX and TN (TXorig and TNorig, respectively) are shown.
Apart for 1967 and 1968, when it seems clear the attribution to
3. Temperature TX/TN, for the other breaks, where it is not sufficiently clear if
the shift would affect TX or TN, it was decided to assign half
3.1. Homogenisation part of the contribution to TX and the other to TN.
Referring three periods (1861–2010, 1861–1980 and
After the transformation of original daily TX and TN time 1981–2010), mDTR has passed all 4 tests (Buishand, SNHT,
series in mDTR and vDTR series, the four tests (Buishand, Pettitt and Von Neumann) respect their critical values at 0.01
SNHT, Pettitt and Von Neumann) have been launched significant values. Considering vDTR all tests passed for
through an iterative procedure, in which at each loop test 1981–2010, while SNHT statistic exceeded the critical value
statistics are checked; if at least two of four tests are outside for 1861–1980 and for 1861–2010 and Buishand one only for
the corresponding critical levels, the major break is found 1861–2010. Therefore, the two sub series (1861–1980 and
and a correction on daily TX and TN is applied; again mDTR 1981–2010) should be considered ‘useful’ while the entire
and vDTR are calculated and four tests are executed. The period 1861–2010 should be considered ‘doubtful’. It has
procedure is stopped when time series could be considered at been decided also to divide the time series in two sub series
least ‘doubtful’ (at 0.01 significant level). The correction, because around 1980 vDTR has a significant decrease (about
equal to 0.1 °C, at each loop is assigned equally to TX and TN 0.1 °C) respect the previous years. The origin of such shift has
 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31 21

Table 1 trend followed by an abrupt increment after late 1980s.

List of years’ breaks, the corresponding mDTR, TX and TN corrections, and Before 1980s annual smoothing curves indicate prominent
the suspected origins (Lombroso and Quattrocchi, 2008).
positive anomalies, that took place between mid-1940s and
Break mDTR TX TN Suspected origin mid-1950s and between late 1960s and late 1970s for both
Year correction correction correction TX and TN. Such periods are quite confirmed for (mean)
(°C) (°C) (°C) temperature time series for Mediterranean area in Giorgi
1862 +0.8 +0.4 −0.4 Undocumented (2002). The warm period during 1960s and 1970s is clearly
1863 +0.8 +0.4 −0.4 Undocumented observed also for winter season and corresponds with a
1869 +0.6 +0.3 −0.3 New hygrotermograph
decrement phase in precipitation. Winter is also the season
1872 +1.2 +0.6 −0.6 Undocumented
1882 −1.2 −0.6 +0.6 Undocumented showing the most pronounced relative minimum, in partic-
1886 −0.2 −0.1 +0.1 Undocumented ular during 1890 s and 1940s; in correspondence with these
1898 −0.2 −0.1 +0.1 Undocumented periods the other seasons seem to have an opposite
1908 −0.6 −0.3 +0.3 New ventilated grid behaviour. The whole behaviour of all temperature is in
1919 +0.2 +0.1 −0.1 Undocumented
good agreement with that shows in Brunetti et al. (2000b)
1944 +0.2 +0.1 −0.1 Temporary relocation
1953 −0.4 −0.2 +0.2 Undocumented over north Italy.
1967 0 −0.3 −0.3 TN and TX instrument Seasonal TX and TN anomalies before last 30-year show
calibration both common and different features respect the annual ones,
1968 −0.5 0 +0.5 TN instrument calibration
and respect themselves. Spring TX series resemble quite well
1971 −0.2 −0.1 +0.1 Undocumented
1979 −0.2 −0.1 +0.1 Undocumented
the corresponding annual one, also with positive anomalies
1985 +0.2 +0.1 −0.1 New automatic station during 1860 s and in early year of 1920s; winter TX has
1988 +0.2 +0.1 −0.1 New automatic station positive anomaly that took place around 1900, mid 1910s
to mid 1920s, early 1970s to early 1980s; autumn TX from
mid 1920 to mid 1930, in late years of 1860s, and in1980s;
not yet been understood and it represents an objective of summer TX in early 1870 to early 1890, last 1920 to last 1930
investigation. and early 1950 to early 1970. Sometimes summer and
Critical values are obtained for SNHT from Khaliq and autumn TX series seem anti-correlated with winter and
Ouarda (2007), for Buishand test from simulations using spring ones, for example from late 1920s to late 1930 and
bootstrap procedure, for Von Neumann test from the normal from early 1980 to early 1990. About seasonal TN series their
approximation equation in Buishand (1981) and for Pettitt behaviour is often similar to the annuals ones, apart winter in
test from the expression for significant extreme (Rutishauser which a strong negative anomaly is found in early 1890 s,
et al., 2009). 1940s, and late 1960, and a strong positive anomaly is
observed during 1970s.
3.2. Annual and seasonal temperature trends Table 2, reporting the averages of yearly and seasonal data
for all the period 1861–2010, including their standard
The anomalies (relative to 1961–1990) of annual and deviations, it could be noted that, both for TX and TN, spring
seasonal TN and TX are shown in Fig. 3, together with the and autumn have values nearly equal to the corresponding
11-year smoothing curves. For all series, and for both TN and yearly average. During the last 30-year, respect to 1861–
TX, it can be noted a more or less slight oscillating positive 2010, there was primarily an increase in minimum and

a
12
TN (°C)

11
10
9
8 Orig.
Correct
7
1870 1890 1910 1930 1950 1970 1990 2010
Year

b
19.5

18.5
TX (°C)

17.5

16.5
Orig.
15.5 Correct

1870 1890 1910 1930 1950 1970 1990 2010

Year

Fig. 2. Annual differences of TN (a) and TX (b) before and after the homogenisation.
22 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31

Minimum temperature anomaly Maximum temperature anomaly

3 3
YEAR YEAR
1.5 1.5

0 0

−1.5 −1.5

−3 −3
1860 1890 1920 1950 1980 2010 1860 1890 1920 1950 1980 2010

3 3
DJF DJF
1.5 1.5

0 0

−1.5 −1.5

−3 −3
1860 1890 1920 1950 1980 2010 1860 1890 1920 1950 1980 2010

3 3
MAM MAM
1.5 1.5

0 0

−1.5 −1.5

−3 −3
1860 1890 1920 1950 1980 2010 1860 1890 1920 1950 1980 2010

3 3
JJA JJA
1.5 1.5

0 0

−1.5 −1.5

−3 −3
1860 1890 1920 1950 1980 2010 1860 1890 1920 1950 1980 2010

3 3
SON SON
1.5 1.5
0
0
−1.5
−1.5
−3
1860 1890 1920 1950 1980 2010 1860 1890 1920 1950 1980 2010

Fig. 3. Anomalies, relative to 1961–1990, of yearly and seasonal anomalies (°C), for (a) TX and (b) TN, with 11-year smoothing curves.

maximum temperatures (about 1 °C), in particular in sum- In Table 3 for the entire time series, but also for the last
mer and in winter respectively. Both for TX and TN the 30-year period and 1861–1980, annual and seasonal tem-
difference between all the period and the last 30-year is peratures trends are reported.
about 1 °C. From Table 3 it could be noted that, for all the period
Trends of averaged annual and seasonal time series have 1861–2010, both annual and seasonal minimum and maxi-
been analysed, after the calculation of averaged monthly mum temperatures have all significant positive trends
values. To assess the temperature and precipitation trends, (except yearly TX) even if very small (0.1 °C per decade),
the linear regression has been chosen. The existence of trends that become practically zero during the period until 1980.
has been detected with both parametric t-test and by the Looking at the last 30-year period, both annual and seasonal
non-parametric Mann–Kendall test at 5% level but, because TX and TN trends have significant increment, lesser extent
results are comparable (Önöz and Bayazit, 2003), in the rest in autumn.
of this paper only Mann-Kendall outcomes (passed or not Winter values are confirmed in Luterbacher et al. (2004),
passed) are reported. where temperature trend over Europe for the 20th century
 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31 23

Table 2
Average of yearly and seasonal mean values with their standard deviation for minimum and maximum temperature (°C), for the entire period, the last 30 years
period and 1861–1980.

YEARLY DJF MAM JJA SON

1861-2010 TX 17.0 ± 0.8 5.6 ± 1.6 17.1 ± 1.1 27.8 ± 1.0 17.2 ± 1.1
TN 9.9 ± 0.8 0.8 ± 1.5 9.1 ± 1.1 18.6 ± 1.0 10.9 ± 1.1
1981-2010 TX 17.9 ± 1.0 6.7 ± 1.5 17.9 ± 1.4 28.8 ± 1.2 17.8 ± 1.1
TN 10.9 ± 0.9 1.8 ± 1.5 9.9 ± 1.2 19.8 ± 1.1 11.7 ± 1.0
1861-1980 TX 16.7 ± 0.6 5.3 ± 1.5 16.9 ± 1.0 27.6 ± 0.9 17.0 ± 1.0
TN 9.7 ± 0.5 0.5 ± 1.4 8.9 ± 0.9 18.3 ± 0.7 10.7 ± 1.0

Table 3
Annual and seasonal trends (°C per decade) of annual TN and TX time series, for the entire period, the last 30-year period and for 1861–1980, with their 95%
confidence intervals.

YEARLY DJF MAM JJA SON

1861-2010 TX +0.1 ± 0.0⁎ +0.1 ± 0.1 +0.1 ±0.0 +0.1 ± 0.0 +0.1 ± 0.0
TN +0.1 ± 0.0 +0.1 ± 0.1 +0.1 ± 0.0 +0.1 ± 0.0 +0.1 ± 0.0
1981-2010 TX +0.8 ± 0.3 +1.1 ± 0.5 +1.0 ± 0.5 +0.8 ± 0.4 +0.5 ± 0.4
TN +0.9 ± 0.2 +1.2 ± 0.5 +1.0 ± 0.4 +0.8 ± 0.4 +0.6 ± 0.4
1861-1980 TX +0.0 ± 0.0⁎ +0.0 ± 0.1⁎ −0.0 ± 0.1⁎ −0.0 ± 0.0⁎ +0.0 ± 0.1⁎
TN +0.0 ± 0.0 +0.1 ± 0.1⁎ +0.0 ± 0.0⁎ +0.0 ± 0.0⁎ +0.0 ± 0.1⁎
⁎ not significant with Mann-Kendall test at 5% level.

(1901 to 2000) is + 0.08 °C per decade, while the winters indices; cold spell duration index (CSDI), warm spell
between 1973 and 2002 were the warmest period of the duration index (WSDI) and growing season length (GSL) as
last half-millennium. In Klein Tank and Können (2003) mean duration indices. None of the absolute indices have been
temperature trends over Europe during 1946–1999, calculated because their results are quite redundant com-
1946–1975 and 1976–1999 increase 0.11, -0.03 (n.s.) and pared to the others. The indices calculation has been
0.43 percentage per decade respectively vs. 0.2 , 0.1 (n.s.) and implemented in MATLAB, based on Zhang and Yang (2004).
0.6 percentage per decade in the same periods. For threshold indices, to avoid possible inhomogeneities
across the in-base and out-base periods, for the calculation of
3.3. Climate indices the base period (1961–1990) a bootstrap procedure (Zhang
et al., 2005) has been implemented.
Climate extremes indices allow the assessment of changes Even for all the indices, trends have been evaluated by
in extreme weather and climate events. They can be divided means a linear regression analysis, and the Mann-Kendall
mainly into four different classes (Alexander et al., 2006): test has been used to estimate the statistical significance of
percentile-based indices, absolute indices, threshold-based these trends. In Table 4 trends for the entire period, last
indices and duration indices. Percentile-based indices repre- 30-year, and 1861–1980 are shown. Confidence intervals at
sent the amount in percent of temperature/precipitation falling 95% are included.
above a fixed percentile, generally during each year. Absolute From Table 4 appears to be clear that for the period
indices refer to counts of days crossing a specified absolute 1861–1980, trends are very small and n. s. (apart TN90p),
value that can be for example a minimum or a maximum value while for the period after 1980 values, although the same
within a month or a season or a year. Threshold-based indices
consist in the number of days on which a temperature/
Table 4
precipitation value falls above or below a fixed threshold. Climate indices trends, including 95% confidence intervals, calculated from TX
Finally, duration indices define periods of extreme warmth, and TN time series for the entire period, for the last 30-year, and for 1861–1980.
cold, wetness or dryness. In addition to these there are other All indices trends are expressed in number of annual days per decade, except
indices as for example the diurnal temperature range (DTR), percentile-based indices in annual percentage of days per decade.

that has been used indirectly to homogenise the time series, or 1861-2010 1981-2010 1861-1980
the simple daily intensity index (SDII), which are not classified
FD −1.3 ± 0.5 −14.1 ± 4.9 −0.8 ± 0.7
in the four previous classes. TR +1.5 ± 0.5 +11.2 ± 5.6 +0.2 ± 0.6⁎
Different set of climate indices have been proposed, but in TN10p −0.5 ± 0.2 −3.8 ± 1.5 −0.2 ± 0.2⁎
this work most of the climate indices recommended by the TN90p +0.9 ± 0.2 +8.1 ± 2.6 +0.2 ± 0.2
ETCCDI (http://cccma.seos.uvic.ca/ETCCDMI/list_27_indices. SU +1.0 ± 0.4 +6.4 ± 5.9⁎ +0.1 ± 0.6⁎
ID −0.3 ± 0.2 −1.4 ± 1.0 −0.1 ± 0.4⁎
shtml) (Peterson et al., 2001) has been chosen. For minimum
TX10p −0.4 ± 0.1 −3.0 ± 1.4 −0.1 ± 0.2⁎
and maximum temperature time series the following climate TX90p +0.7 ± 0.2 +7.0 ± 2.6 +0.2 ± 0.2⁎
indices have been considered: frost days (FD), tropical nights CSDI −0.7 ± 0.4 −4.8 ± 3.4 −0.2 ± 0.6⁎
(TR), summer days (SU) and ice days (ID) as threshold-based WSDI +1.2 ± 0.4 +14.1 ± 7.0 +0.3 ± 0.4⁎
indices; cold nights (TN10p), warm nights (TN90p), cold GSL +1.5 ± 0.9 +15.9 ± 12.7 +0.8 ± 1.2⁎

days (TX10p) and warm days (TX90p) as percentile-based ⁎ not significant with Mann-Kendall test at 5% level.
24 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31

sign, are in absolute significantly higher than the previous FD and TN, with a strong anti-correlation (from −0.54 during
ones. In particular it could be noted an increase of tropical 1861–1980 to −0.72 during 1981–2010).
nights (almost 40%) and of warm spells duration (46%). All trends indices seem to be consistent with what has
Percentage is still higher for summer days, but it is n.s. with been found in other studies, as in the global analysis of Frich
Mann-Kendall test at 5% level. et al. (2002) for 1946–1999, Kostopoulou and Jones (2005)
In Fig. 4(a) smoothing curves (11-year span) of frost days on eastern Mediterranean between 1958 and 2000, Tebaldi
and tropical nights anomalies (relative to 1961–1990) are et al. (2006) in an intercomparison model-simulated of
shown. Apart the just mentioned significant increase in extreme events covering all the globe, for a period between
absolute value, during the last period it could be noted the 1880 and 2000.
particularly evident anti-correlation between the two indices The TN10p negative trend in the last century is confirmed
(with the linear correlation coefficient raising −0.62 for in several works: for example, by Alexander et al. (2006) in
1981–2010 period). Similarly, in Fig. 4(b) the smoothing curves a global analysis during 1951–1993 where they found a
of TN10p and TN90p anomalies show the same behaviour of general decrease all over the world; in Kostopoulou and Jones

a b

Anomaly (Percentage of days per year)

FD TN10p
40
TR 20 TN90p
Anomaly (number of days)

20

10

0
−20

−40 −10
1860 1890 1920 1950 1980 2010 1860 1890 1920 1950 1980 2010
Year Year

c d
20 20
Anomaly (Percentage of days per year)

TX10p

15 TX90p

10
Anomaly (number of days)

10

0 5

−10
SU
−5
ID

−20 −10
1860 1890 1920 1950 1980 2010 1860 1890 1920 1950 1980 2010
Year Year

Fig. 4. Smoothing curves (span 11 year) of anomalies, relative to 1961–1990, of (a) frost days and tropical nights (number of annual days), (b) cold nights and
warm nights (percentage of days), (c) summer days and ice days (number of annual days), (d) cold days and warm days (percentage of days).
 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31 25

40 northern Italy for 1952–2008 and 1980–2008 have equal sign,

CSDI
but values are greater respect Modena in the same periods.
WSDI
In Fig. 5 smoothing curves of CSDI and WSDI anomalies
30
are shown. Again, a certain anti-correlation between cold
spell and warm spell is found (ranging between − 0.26 for
Anomaly (number of days)

the whole period, − 0.06 for 1861–1980 and − 0.38 for the
last thirty years), but it is evident a higher increase of warm
20 spell respect the cold ones
Growing Season Length that defines how long is the
“warm” period of every year is very similar to the WSDI. In
Ramos et al. (2011) for 1941–2006, 1945–1975 and 1976–2006
10
GSL, CSDI and WSDI, prevalent trend signs over Portugal are in
accordance with Modena in the same period.

0 4. Precipitation

4.1. Homogenisation

−10
The number of wet days in a year has been chosen as the
1860 1890 1920 1950 1980 2010
Year testing variable for precipitation (Wijngaard et al., 2003),
where a wet day is defined as a day with precipitation above
Fig. 5. Smoothing curves (span 11-year) of anomalies (number of annual 1 mm. Again, the four tests, SNHT, Buishand range test,
days), relative to 1961–1990, of cold and warm spell duration indices. Pettitt test and Von Neumann ratio test, have been selected to
test the homogeneity of the time series. All test statistics values
(2005) in the eastern Mediterranean for 1958–2000. For fall inside their respective critical values both for the whole
TN90p, both above mentioned studies found a large significant period (1831–2010) and for 1831–1980 and 1981–2010.
positive trend mainly in summer pattern (JJA), underlining a Therefore, the total precipitation series has been analysed
general tendency for warmer summer nights throughout the without any further corrections.
eastern Mediterranean.
Smoothing curve of SU index and ID index anomalies 4.2. Annual and seasonal trends
(relative to 1961–1990) are shown in Fig. 4(c). Again,
especially for SU index significant increases appear during In Table 5, total averages of yearly and seasonal total
last thirty years. In Fig. 4(d) the smoothing curves of TX10p precipitation for the whole period (1831–2010) and for the
and TX90p anomalies are shown. They resemble the sub-periods are reported. For all three periods, annual, vernal
smoothing curves of TN10p and TN90p: in fact TX10p and and autumnal values are fairly close to each other; for
TX90p show a strong anti-correlation between them (− 0.59 summer and winter 1981–2010 values, compared with those
for the entire period, − 0.35 for 1861–1980 and − 0.76 for of the entire period, a considerable decrease during winter
the last 30 years). (− 20%) and an increase in summer (+12%) could be noted.
In Ramos et al. (2011) for 1941–2006, 1945–1975 and As expected, annual values are very similar to those found
1976–2006 percentile-based and threshold-based indices for nearby stations (Bologna, Ferrara, Mantova) analysed in
have prevalent trend signs over the most analysed stations Brunetti et al. (2000c). Seasonal average precipitation for the
in Portugal in accordance with Modena in the same periods, same period indicates that summer is characterised by the
but not for TX10p and TN10p, during 1945–1975, for which it smallest amount, followed by winter, spring and autumn,
is opposite. Similar results are given by the comparisons with the highest precipitation amount. So it is not surprising
of the same trend indices with Klein Tank and Können that the highest correlation between the annual series and
(2003) ones, over Europe during 1946–1999, 1946–1975 and seasonal ones is found for autumn followed by spring,
1976–1999. In Wulfmeyer and Henning-Müller (2006) during summer and winter, with linear correlation coefficients
1878–2002, for Hohenheim, Germany, ice days decline by given by 0.72, 0.55, 0.52 and 0.18 respectively.
1.2 days per decade and frost days decline by 2.2 days The annual and seasonal precipitation amount anomalies
per decade while for Modena in the same period, values are (relative to 1961–1990) smoothed over 11 years, in Fig. 6, at
−0.5 days per decade and −1.5 days per decade respectively. a first glance show a sequence of wet periods, alternated with
In Simolo et al. (2010) percentile-based indices trends over dry periods. Looking at the annual filtered curve, positive

Table 5
Mean annual and seasonal accumulated precipitation (mm), with their standard deviation, for the whole period and for 1831–1980 and 1981–2010 sub-periods.

YEARLY DJF MAM JJA SON

1831-2010 664.8 ± 158.7 142.6 ± 66.8 172.4 ± 64.1 138.7 ± 63.7 211.8 ± 91.0
1981-2010 660.7 ± 159.0 114.1 ± 47.0 172.1 ± 56.6 155.4 ± 73.0 218.9 ± 93.1
1831-1980 665.6 ± 159.1 148.3 ± 68.9 172.5 ± 65.7 135.4 ± 61.3 210.4 ± 90.8
26 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31

600
YEARLY
300
0
−300

1850 1875 1900 1925 1950 1975 2000

200 DJF

−200
1850 1875 1900 1925 1950 1975 2000
PRCP anomaly (mm)

200 MAM

−200
1850 1875 1900 1925 1950 1975 2000

200 JJA

−200
1850 1875 1900 1925 1950 1975 2000

200 SON
0
−200
1850 1875 1900 1925 1950 1975 2000
Year

Fig. 6. Anomalies (mm), relative to 1961–1990, of annual and seasonal precipitation amount (PRCP) for 1831–2010 period. Solid lines represent the 11-year
smoothing curves.

anomalies take place until the last 1880 s, between 1890–1920, Brunetti et al. (2000a) in their annual precipitation analysis
1930–1940, from late 1950s to late 1970, and after 1990, while over northern Italy; they have also found a significant
negative anomalies take place between 1880 and 1890, 14–26 years long period that it has not been found in this
1920–1930, 1940–1950, and 1970–1990. analysis. The short period could be related with Quasi-
A preliminary spectral analysis with a Lomb periodogram, Biennal Oscillation, while the quasi 60-year period is found in
together with a wavelet analysis (Torrence and Compo, numerous multi-secular climatic records (Scafetta, 2010).
1998), over annual PRCP indicates that the significant (at Considering all the period, yearly and seasonal trends for
95%) contributions come from the harmonics at 2.3 and total precipitation result negative, and in particular for the
2.8 years, and about 55–60 years. The previous first two are sub-period until 1980 only the winter one is not significant
in agreement with short periods (2–2.8 years) found by (Table 6). Instead, annual and seasonal trends for the last

Table 6
Linear trends (with 95% confidence intervals) of annual and seasonal accumulated precipitation time series (mm per decade) for all the period and for 1831–1980
and 1981-2010.

YEARLY DJF MAM JJA SON

1831-2010 −6.3 ± 4.4⁎ −1.8 ± 1.9 −1.8 ± 1.8 −0.8 ± 1.8 −2.1 ± 2.6
1831-1980 −11.1 ± 5.7⁎ −0.5 ± 2.6 −3.3 ± 2.4⁎ −3.2 ± 2.2⁎ −4.5 ± 3.3⁎
1981-2010 +74.8 ± 63.6⁎ +13.1 ± 20.0 +18.8 ± 23.8 +13.1 ± 31.7 +30.5 ± 39.2
⁎ Significant with Mann-Kendall test at 5% level.
 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31 27

Table 7
Climate indices trends for all the period and for the last 150-year, 120-year, 90-year, 60 year and 30 yr. All indices are expressed in number of annual days per
decade, except R75, R95 and R99 that are in annual percentage per decade and SDII that is in (mm/day) per decade.

1831-2010 1861-2010 1891-2010 1921-2010 1951-2010 1981-2010

R75p −6.2 ± 4.1⁎ −1.0 ± 5.1 1.6 ± 7.0 +15.8 ± 9.7⁎ +23.0 ± 18.9⁎ +61.2 ± 59.8
R95p −3.4 ± 2.9⁎ −0.2 ± 3.4 1.9 ± 4.9 +12.8 ± 6.8⁎ +16.4 ± 14.1⁎ +21.4 ± 48.1
R99p −1.1 ± 1.9 +1.1 ± 2.3 +3.4 ± 3.3 +10.1 ± 4.7⁎ +14.6 ± 9.7⁎ +31.4 ± 34.5
R10mm −0.3 ± 0.2⁎ −0.1 ± 0.2 −0.1 ± 0.3 +0.5 ± 0.4 +0.7 ± 0.9 +2.7 ± 2.5⁎
R20mm −0.1 ± 0.1⁎ −0.1 ± 0.1 −0.0 ± 0.2 +0.2 ± 0.2 +0.2 ± 0.4 +0.6 ± 1.1
CDD +0.1 ± 0.3 +0.1 ± 0.4 +0.2 ± 0.6 +0.2 ± 0.9 +0.3 ± 1.6 −2.8 ± 5.4
CWD +0.0 ± 0.0 +0.0 ± 0.0 +0.0 ± 0.1 −0.0 ± 0.1 +0.0 ± 0.2 +0.7 ± 0.6
SDII −0.1 ± 0.0⁎ −0.0 ± 0.1 +0.1 ± 0.1 +0.2 ± 0.1⁎ −0.3 ± 0.2⁎ +0.6 ± 0.7
⁎ significant with Mann–Kendall test at 5% level.

30-year period, 1981–2010, are all positive; also in this case R20mm gives very similar results to that of R95p percentile-
only yearly trend is significant, but with a higher value based index (that corresponds to 27.6 mm). The behaviour of
(+ 74.8 mm per decade). these indices anomalies reflect quite closely the one of total
A great phenomenon, that could be seen in Fig. 6, is the precipitation, seen before (Fig. 6). Similar results have been
negative winter rainfall trend since the 1960s (Cullen and reported by Ramos and Martìnez-Casasnovas (2006) in the
deMenocal, 2000; Goodess and Jones, 2002; Xoplaki et al., north-east Spain region; they found that extreme events are
2004) and finishing in 1990s (the biggest pattern that could increasing in the analysed periods (one time series between
be seen in the graph), when it is switching from one wet 1923 and 2002, the other two of 40-year periods). In IPCC
phase to a dry phase. This negative trend can at least partly report the number of events with heavy precipitation (95th
be explained by the positive trend of NAO oscillation percentile) has increased over the period 1900–2005, also
(e.g. Dünkeloh and Jacobeit, 2003; Xoplaki et al., 2004). in continental regions where the total amount of rainfall
In Brunetti et al. (2006a) during 1865–2003, in the geo- has actually dropped (Trenberth et al., 2007), and this is
graphical area to which Modena belongs, there is accordance in confirmed for Modena (about + 5% per decade but n.s.). R75p
the sign of trends except for winter. In Klein Tank and Können and R95p calculated by Klein Tank and Können (2003) for the
(2003), for the period 1946–1999, the annual precipitation period 1946–1999, for the Po Valley, seem to underestimate
trend over the Po valley is positive but n.s. at 5%, that is in the value found in this work.
accordance with the value for Modena. Even Ramos and Consecutive dry days (CDD) and consecutive wet days
Martìnez-Casasnovas (2006) found n.s. positive trend for (CWD) represent the greatest number of consecutive days,
annual precipitation in their work on three time series in for each year, in which daily precipitation amount is b 1 mm
Penedès-Anoia region, in the north-east of Spain. and ≥ 1 mm respectively. From Table 7 it can be noted that
IPCC analysed precipitations between 1900–2005 and it CDD and CWD trends for all the period are always n.s.,
found both increases and decreases in precipitation over practically indicating no variation since 1831. In the last
Eurasia (Trenberth et al., 2007), concluding that it is difficult thirty years CDD trend is negative but n.s., while CWD trend
to give an unique trend. They reported that it has become is slight positive but n.s. Also Moberg and Jones (2005) found
wetter in northern Europe while drier in the Mediterranean. that CDD index trend is mostly n.s. in summer and winter
IPCC annual precipitation trend graph for the Mediterranean seasons over central and western Europe, during 1901–1999.
Basin (Trenberth et al., 2007) is similar to the one found Simple daily intensity index (SDII) is quite constant
for Modena, with precipitation decrement during almost all during all period, even if an alternation of trend sign. The
the 20th century, with a minimum in the 1940s, and then n.s. trend of precipitation around 1900 is balanced by the
an increment after the 1990s, in correspondence of a great positive trend in number of rainy days, so SDII trends for this
positive NAO index value. period remain significantly negative. SDII trends for the last
100 years are positive, due mainly to the negative trend in
4.3. Climate indices the number of days. This is confirmed by Frich et al. (2002)
where they reported that SDII has increased during the
Also for the precipitation time series climate indices have second half of 20th century in many parts of the world; in
been calculated. Generally speaking their trend are not particular there are significant positive trends over Italy.
significant (Table 7). Statistically significant values can be Similar results, for both annual and winter periods, have been
found for R75 and R95 (and also R10mm, R20mm), for the revealed by Kostopoulou and Jones (2005) in their study
entire period 1831–2010, even if quite low, and higher for all focused on eastern Mediterranean.
three percentile-based indices and SDII during the 90 years
since 1921 (and 60 years since 1951). 4.4. Correlation between NAO and precipitation
In Fig. 7 the smoothing curves of percentile-based indices
(R75p, R95p and R99p) anomalies (with respect 1961–1990) Hurrell and Folland (2002) discussed a possible link
are shown. Heavy precipitation days (R10mm) and very between sea-level pressure (SLP) averaged over the southern
heavy precipitation days (R20mm) threshold-based indices action centre for NAO in summer and JJA precipitation in
are not shown because the first one is very similar to R75p Europe. Xoplaki et al. (2004), from a CCA, showed relation-
percentile-based index (that corresponds to 11.1 mm), while ship between variations in the NAO index and time scale
28 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31

200

150

Anomaly (number of days)

100

50

−50
R75
R95
R99
−100
1840 1860 1880 1900 1920 1940 1960 1980 2000
Year

Fig. 7. Smoothing curves of anomalies (mm/day), relative to 1961–1990, of R75p , R95p and R99p indices for 1831–2010 period, with 11-year smoothing curve.

variations of the Mediterranean precipitation throughout the for Mediterranean area, negative NAO index corresponds to
20th century. It is known that temperature increment is rainiest season and vice-versa. For northern Europe, on the
connected to the change of water storage of the atmosphere, contrary, precipitation is directly correlated to NAO index
about 7% per °C. So over northern continents, as for example (Hurrel, 1995). This anti-correlation could clearly be seen
Europe, warm winters lead to more precipitation. Different from the 1990s, where a great positive NAO phase caused
situation is for the Mediterranean area, where a drier trend drier conditions over the area.
has been found (Trenberth et al., 2007). These differences are In Fig. 8 NAO index (provided by the Climate Analysis
mainly tied to NAO oscillation: for example in Po plain, as Section, NCAR, Boulder, USA, Hurrel (1995)), for the period

2
Standardized anomaly

−1

−2

−3
1860 1880 1900 1920 1940 1960 1980 2000
Year

Fig. 8. NAOI vs. DJFM precipitation comparison, for the period 1864–2010. Squares are NAO index, stars precipitation data, the thin solid and the thick solid lines
are the smoothing curves with an 11-year span of NAO index and precipitation respectively.
 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31 29

1864–2010, and DJFM accumulated precipitation over are increasing (+1.5 days per decade). For all the analysed
Modena are compared. Squares represent NAO index and period and for the last 30-year period all indices (apart SU
stars precipitation data; the two lines (thin solid line for NAO for 1981–2010) calculated from TN and TX are statistically
index and thick solid line for precipitation) are the smoothing significant. Greater trend indices values have been found
curves with 11-year span. For a better comparison all data for the period 1981–2010; FD is diminishing (− 14.1 days per
were transformed in their centred scaled version (standardised decade) while TR and SU are increasing (+ 11.2 and
anomalies). + 6.4 days per decade but n.s.). Spell duration indices follow
Fig. 8 underlines a clear anti-correlation between the two previous results: cold spells duration index is diminishing
datasets, especially after around 1920 (already noted in (− 4.8 days per decade for 1981–2010), while warm spell
Boccolari et al., 1998). Such anti-correlation reaches high duration index has a strong increase (+14.1 days per decade
values up to last of 1800s (− 0.62 during 1871–1900), then for 1981–2010).
drops and then increases again, reaching − 0.61 in the last Precipitation trends are significant for yearly data
30-year 1981–2010. One explanation could be derived from (− 6.3 mm per decade for 1831–2010 and + 74.8 mm per
the NAO centres of action shifting, but the NAO index used decade over last 30 years) but n.s. for seasonal ones. Unlike
in this context is station-based that is fixed in space, so previous temperature indices, those related to precipitation
shouldn't take into account the motion (Hurrel and Deser, show almost all significant trends over the period 1831–
2010). However, the correlation calculation using NAO index 2010, while for the last 30 years only R10mm is significant,
based on principal components (https://climatedataguide.ucar. with + 2.7 days per decade. Duration indices are practically
edu/guidance/hurrell-north-atlantic-oscillation-nao-index-pc- constant for all periods.
based) for the period 1899–2010, anti-correlation variations The comparison of DJFM precipitation with NAO index
remain the same even if limited to a narrower range. The confirm an high anti-correlation, that is at negative NAO
strong positive NAO between the late 1980's and 1990 could index corresponds a rainiest season and vice-versa.
explain the higher value of anti-correlation (Marshall et al.,
2001). The fact that the wettest seasons in this location are
Acknowledgments
autumn and spring, suggests that probably there isn't a direct
connection with the recent positive trend of climatic indices,
The authors would like to thank: the European Climate
in particular R75p, R95p and R99P, with the augmented anti-
Assessment & Dataset project for all useful documentations
correlation. In other works the correlation between precipita-
present in its site; James W. Hurrel and NCAR Climate Analysis
tion and NAO index over the Po plain has been studied
Section for NAO index database; U.S. Geological Survey
(Brunetti et al., 2000a and 2006c). Also Rodríguez-Fonseca and
Department of the Interior (USGS) for GTOPO30 DEM, the
Castro (2002) found a connection between winter precipita-
Institute of Veterinary Public Health of the University of Wien
tion and NAO index over Iberian peninsula.
for its Köppen and Geiger climate classification maps and
Luca Lombroso for GOMO daily temperature and precipitation
5. Conclusions
time series.
The maintenance of the historical stations is very
important, because these long time series are fundamental References
for the analysis of climate and its changes. Modena contin-
uous time series of maximum and minimum temperature Alexander, L.V., Zhang, X., Peterson, T.C., Caesar, J., Gleason, B., Klein Tank,
A.M.G., Haylock, M., Collins, D., Trewin, B., Rahimzadeh, F., Tagipour, A.,
and precipitation, available since 1861 and 1831 respectively,
Rupa Kumar, K., Revadekar, J., Griffiths, G., Vincent, L., Stephenson, D.B.,
have been analysed, for trends and extreme climate indices, Burn, J., Aguilar, E., Brunet, M., Taylor, M., New, M., Zhai, P., Rusticucci,
after having tested the homogeneity of their respective M., Vazquez-Aguirre, J.L., 2006. Global observed changes in daily climate
mDTR and vDTR series for temperature, and the number of extremes of temperature and precipitation. J. Geophys. Res. 111
(D05109), 1–22.
rainy days for precipitation. Only the temperature time series Alexandersson, H., 1986. A homogeneity test applied to precipitation data.
have required to be corrected for homogenisation, carried out J. Climatol. 6, 661–675.
without reference time series, through an iterative procedure ANCM, 2010. Annuario Statistico. Comune di Modena. (available online
at http://www.comune.modena.it/serviziostatistica/pubblicazioni/annuari/
'test-and- fix', giving satisfactory results. For the first two annuario2010/annuario2010.shtml).
temperature time series decades, other documented meta- Aprilesi, G., Marseguerra, M., Morelli, S., Rivasi, M.R., Saltini, G., Santangelo,
data would be necessary to improve the first part of the time R., 1977. Stochastic behavior of daily minimum, mean and maximum
temperatures in Modena, Italy. Mon. Weather Rev. 105, 1434–1441.
series. Auer, I., Böhm, R., Jurkovic, A., Lipa, W., Orlik, A., Potzmann, R., Schöner, W.,
Minimum and maximum temperature time series show a Ungersböck, M., Matulla, C., Briffa, K., Jones, P.D., Efthymiadis, D.,
significant abrupt increment during last three decades. Their Brunetti, M., Nanni, T., Maugeri, M., Mercalli, L., Mestre, O., Moisselin,
J.M., Begert, M., Müller-Westermeier, G., Kveton, V., Bochnicek, O.,
annual trends increase from + 0.1 °C per decade during all Stastny, P., Lapin, M., Szalai, S., Szentimrey, T., Cegnar, T., Dolinar, M.,
analysed period 1861–2010, to + 0.8 and + 0.9 °C per decade Gajic-Capka, M., Zaninovic, K., Majstorovic, Z., Nieplova, E., 2007.
respectively during the last thirty years, with major values in HISTALP – Historical Instrumental Climatological Surface Time Series of
the Greater Alpine Region. Int. J. Climatol. 27, 17–46.
winter and minor ones in autumn. If the last 30-year are not
Barbanti-Silva, L., 1952. Osservazioni meteorologiche fatte all'Osservatorio
considered, both annual and seasonal trends are practically Geofisico dell'Università di Modena nelle annate 1938–1951. Pubblicazioni
zero. dell'Osservatorio Geofisico dell'Università di Modena., 55. Antica Tipografia
The analysis of several temperature climate indices, for Soliani, Modena.
Baruzzi, M., 1928. Andamenti periodici della temperatura media diurna a
all the period 1861–2010, shows for example that frost days Modena. Nel primo centenario dalla fondazione dell'osservatorio
are decreasing (− 1.3 days per decade) while tropical nights (1827–1927). Antica Tipografia Soliani, Modena.
30 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31

Boccolari, M., Frontero, P., Lombroso, L., Pugnaghi, S., Santangelo, R., 1998. Europe and the globe: the observed records (1901–2000) and 16
Climate of Modena (Italy): temperature and rainfall time series. Atti Soc. scenarios (2001–2100). Tyndall Centre of Climate Change Research,
Nat. Mat. Modena 129, 5–15. Working Paper 55.
Bonacini, C., 1909. Osservazioni meteorologiche degli anni 1906–1907. Moberg, A., Jones, P.D., 2005. Trends in indices for extremes in daily
L'Osservatorio di Modena nel biennio 1905–1907. Rapporto del Prof. C. temperature and precipitation in central and western Europe, 1901–99.
Bonacini. Pubblicazioni del R. Osservatorio Geofisico di Modena n., 23. Int. J. Climatol. 25, 1149–1171.
Antica Tipografia Soliani, Modena. Moberg, A., Jones, P.D., Lister, D., Walther, A., Brunet, M., Jacobeit, J., Alexander,
Brunetti, M., Maugeri, M., Nanni, T., 2000a. Variations of Temperature and L.V., Della-Marta, P.M., Luterbacher, J., Yiou, P., Chen, D., Klein-Tank, A.M.G.,
Precipitation in Italy from 1866–1995. Theor. Appl. Climatol. 65, 165–174. Saladié, O., Sigró, J., Aguilar, E., Alexandersson, H., Almarza, C., Auer, I.,
Brunetti, M., Buffoni, L., Maugeri, M., Nanni, T., 2000b. Trends of minimum Barriendos, M., Begert, M., Bergström, H., Böhm, R., Butler, C.J., Caesar, J.,
and maximum daily temperatures in Italy from 1865 to 1996. Theor. Drebs, A., Founda, D., Gerstengarbe, F.W., Micela, G., Maugeri, M., Österle,
Appl. Climatol. 66, 49–60. H., Pandzic, K., Petrakis, M., Srnec, L., Tolasz, R., Tuomenvirta, H., Werner,
Brunetti, M., Buffoni, L., Maugeri, M., Nanni, T., 2000c. Precipitation intensity P.C., Linderholm, H., Philipp, A., Wanner, H., Xoplaki, E., 2006. Indices for
trends in northern Italy. Int. J. Climatol. 20, 1017–1031. daily temperature and precipitation extremes in Europe analyzed for the
Brunetti, M., Maugeri, M., Monti, F., Nanni, T., 2006a. Temperature and a period 1901–2000. J. Geophys. Res. 111, D22106.
precipitation variability in Italy in the last two centuries from Önöz, B., Bayazit, M.C., 2003. The power of statistical tests for trend
homogenised instrumental time series. Int. J. Climatol. 26, 345–381. detection. Turkish J. Eng. Env. Sci. 27, 247–251.
Brunetti, M., Maugeri, M., Nanni, T., 2006b. Trends of the daily intensity of Pavan, V., Tomozeiu, R., Cacciamani, C., Di Lorenzo, M., 2008. Daily
precipitation in Italy and teleconnections. Nuovo Cimento 29 (1), 105–116. precipitation observations over Emilia-Romagna: mean values and
Buffoni, L., Maugeri, M., Nanni, T., 1999. Precipitation in Italy from extremes. Int. J. Climatol. 28, 2065–2079.
1833–1996. Theor. Appl. Climatol. 63, 33–40. Peterson, T.C., Easterling, D.R., Karl, T.R., Groisman, P., Nicholls, N., Plummer,
Buishand, T.A., 1981. The analysis of homogeneity of long-term rainfall N., Torok, S., Auer, I., Boehm, R., Gullett, D., Vincent, L., Heino, R.,
records in the Netherlands. KNMI Scientific Report WR 81–7, De Bilt, The Tuomenvirta, H., Mestre, O., Szentimrey, T., Salinger, J., Førland, E.,
Netherlands. Hanssen-Bauer, I., Alexandersson, H., Jones, P., Parker, D., 1998.
Buishand, T.A., 1982. Some methods for testing the homogeneity of rainfall Homogeneity adjustments of in situ atmospheric climate data: a review.
records. J. Hydrol. 58, 11–27. Int. J. Climatol. 18, 1493–1517.
Chistoni, C., 1895. Osservazioni Meteorologiche fatte negli anni 1892, 1893 e Peterson, T.C., Folland, C., Gruza, G., Hogg, W., Mokssit, A., Plummer, N., 2001.
1894 nell'Osservatorio di Modena. Tipografia dell'Erede Soliani, Modena. Report on the activities of the working group on climate change
Cullen, H.M., deMenocal, P.B., 2000. North Atlantic influence on Tigris- detection and related rapporteurs 1998–2001. WMO, Rep. WCDMP-47,
Euphrates streamflow. Int. J. Climatol. 20, 853–863. WMO-TD 1071, Geneve, Switzerland.
Dünkeloh, A., Jacobeit, J., 2003. Circulation dynamics of Mediterranean Pettitt, A.N., 1979. A non-parametric approach to the change-point
precipitation variability 1948–98. Int. J. Climatol. 23, 1843–1866. detection. Appl. Stat. 28, 126–135.
Frich, P., Alexander, L.V., Della-Marta, P., Gleason, B., Haylock, M., Klein Tank, Piervitali, E., Colacino, M., 2003. Precipitation Scenarios in the Central-
A.M.G., Peterson, T., 2002. Observed coherent changes in climatic Western Mediterranean Basin. In: Bolle, H.J. (Ed.), Mediterranean
extremes during the second half of the twentieth century. Clim. Res. Climate Variability and Trends. Springer, Berlin, pp. 245–258.
19, 193–212. Ragona, D., 1865. Risultati delle osservazioni eseguite nel R. Osservatorio
Giorgi, F., 2002. Variability and trends of sub-continental scale surface climate di Modena nell'anno 1864, Parte meteorologica. Tipografia dell'Erede
in the twentieth century. Part I: observations. Clim. Dyn. 18, 675–691. Soliani, Modena.
Goodess, C.M., Jones, P.D., 2002. Links between circulation and changes in Ramos, A.M., Martìnez-Casasnovas, J.A., 2006. Trends in precipitation
the characteristics of Iberian rainfall. Int. J. Climatol. 22, 1593–1615. concentration and extremes in the Mediterranean Penedès-Anoia
Hurrel, J.W., 1995. Decadal trends in the North Atlantic Oscillation: regional region, NE Spain. Clim. Chang. 74, 457–474.
temperatures and precipitation. Science 269, 676–679. Ramos, A.M., Trigo, R.M., Santo, F.E., 2011. Evolution of extreme temperatures
Hurrel, J.W., Deser, C., 2010. North Atlantic climate variability: The role of the over Portugal: recent changes and future scenarios. Clim. Res. 48, 177–192.
North Atlantic Oscillation. J. Mar. Syst. 79, 231–244. Rodríguez-Fonseca, B., Castro, M., 2002. On the connection between winter
Hurrell, J.W., Folland, C.K., 2002. The relationship between tropical Atlantic anomalous precipitation in the Iberian Peninsula and North West Africa
rainfall and the summer circulation over the North Atlantic. CLIVAR and the Summer subtropical Atlantic sea surface temperature. Geophys.
Exchanges., 25, pp. 52–54. Res. Lett. 29.
IPCC, 2007. IPCC Fourth Assessment Report: Climate Change 2007 (AR4). Rotach, M.W., 1999. On the influence of the urban roughness sublayer on
(available online at http://www.ipcc.ch/publications_and_data/publications_ turbulence and dispersion. Atmos. Environ. 33, 4001–4008.
and_data_reports.shtml). Rotach, M.W., 2001. Simulation of urban-scale dispersion using a lagrangian
Kastner-Klein, P., Rotach, M.W., 2003. Mean flow and turbulence characteristics stochastic dispersion model. Bound.-Layer Meteorol. 99, 379–410.
in an urban roughness sublayer. Bound.-Layer Meteorol. 111, 55–84. Rubel, F., Kottek, M., 2010. Observed and projected climate shifts 1901–2100
Khaliq, M.N., Ouarda, T.B.M.J., 2007. Short Communication On the critical depicted by world maps of the Köppen-Geiger climate classification.
values of the standard normal homogeneity test (SNHT). Int. J. Climatol. Meteorol. Z. 19, 135–141.
27, 681–687. Rutishauser, T., Schleip, C., Sparks, T.H., Nordli, Ø., Menzel, A., Wanner, H.,
Klein Tank, A.M.G., Können, G.P., 2003. Trend in indices of daily temperature Jeanneret, F., Luterbacher, J., 2009. Temperature sensitivity of Swiss and
and precipitation extremes in Europe, 1946–99. J. Climate 16, 3665–3680. British plant phenology from 1753 to 1958. Clim. Res. vol. 39, 179–190.
Kostopoulou, E., Jones, P.D., 2005. Assessment of climate extremes in the Scafetta, N., 2010. Empirical evidence for a celestial origin of the climate
eastern Mediterranean. Meteorol. Atmos. Phys. 89, 69–85. oscillations and its implications. J. Atmos. Sol. Terr. Phys. 72, 951–970.
Kottek, M., Grieser, J., Beck, C., Rudolf, B., Rubel, F., 2006. World map of the Simolo, C., Brunetti, M., Maugeri, M., Nanni, T., Speranza, A., 2010.
Köppen-Geiger climate classification updated. Meteorol. Z. 15, 259–263. Understanding climate change-induced variations in daily temperature
Lombroso, L., Quattrocchi, S., 2008. L'osservatorio di Modena: 180 anni di distributions over Italy. J. Geophys. Res. 115, D22110.
misure meteoclimatiche. Edizioni SMI/SMS, Modena. Tebaldi, C., Hayhoe, K., Arblaster, J.M., Meehl, G.A., 2006. Going to the
Luterbacher, J., Dietrich, D., Xoplaki, E., Grosjean, M., Wanner, H., 2004. extremes: An intercomparison of model-simulated historical and future
European seasonal and annual temperature variability, trends, and changes in extreme events. Clim. Chang. 79, 185–211.
extremes since 1500. Science 303, 1499–1503. Torrence, C., Compo, G.P., 1998. A Practical Guide to Wavelet Analysis. Bull.
Marseguerra, M., Morelli, S., Saltini, G., Santangelo, R., 1979. Evidence of an Amer. Meteor. Soc. 79, 61–78.
anthropogenic climatic change in the daily temperatures of Modena. Trenberth, K.E., Jones, P.D., Ambenje, P., Bojariu, R., Easterling, D., Klein Tank,
Nuovo Cimento 2C (4), 499–503. A., Parker, D., Rahimzadeh, F., Renwick, J.A., Rusticucci, M., Soden, B.,
Marshall, J., Kushnir, Y., Battisti, D., Chang, P., Czaja, A., C, Dickson, R., Hurrell, Zhai, P., 2007. Observations: Surface and Atmospheric Climate Change.
J., McCartney, M., Saravanan, R., Visbeck, M., 2001. North Atlantic In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B.,
Climate Variability: Phenomena, Impacts And Mechanisms. Int. J. Tignor, M., Miller, H.L. (Eds.), Climate Change 2007: The Physical Science
Climatol. 21, 1863–1898. Basis. Contribution of Working Group I to the Fourth Assessment Report
Maugeri, M., Nanni, T., 1998. Surface air temperature variations in Italy: of the Intergovernmental Panel on Climate Change. Cambridge Univer-
recent trends and an update to 1993. Theor. Appl. Climatol. 61, 191–196. sity Press, Cambridge, United Kingdom and New York, NY, USA.
Maugeri, M., Brunetti, M., Monti, F., Nanni, T., 2004. Sea-level pressure Von Neumann, J., 1941. Distribution of the ratio of the mean square
variability in the Po Plain (1765–2000) from homogenised daily secular successive difference to the variance. Ann. Math. Stat. 13, 367–395.
records. Int. J. Climatol. 24, 437–455. Wijngaard, J.B., Klein Tank, A.M.G., Können, G.P., 2003. Homogeneity of 20th
Mitchell, T.D., Carter, T.R., Jones, P.D., Hulme, M., New, M., 2004. A Century European Daily Temperature and Precipitation Series. Int. J.
comprehensive set of high-resolution grids of monthly climate for Climatol. 23, 679–692.
 M. Boccolari, S. Malmusi / Atmospheric Research 122 (2013) 16–31 31

Wulfmeyer, V., Henning-Müller, I., 2006. The climate station of the Zhang, X., Yang, F., 2004. RClimDex (1.0) user manual. (Available online at
University of Hohenheim: analyses of air temperature and precipitation http://cccma.seos.uvic.ca/ETCCDMI/software.shtml).
time series since 1878. Int. J. Climatol. 26, 113–138. Zhang, X., Hegerl, G., Zwiers, F., Kenyon, J., 2005. Avoiding Inhomogeneity in
Xoplaki, E., Gonzalez-Rouco, J.F., Luterbacher, J., Wanner, H., 2004. Wet Percentile-Based Indices of Temperature Extremes. J. Climate 18, 1641–1651.
season Mediterranean precipitation variability: influence of large-scale
dynamics and trends. Clim. Dyn. 23, 63–78.