1 Introduction

In temperate zones, the reproductive cycle of plants is greatly controlled by temperature and day length while at lower latitudes rainfall and evapotranspiration must be taken into account. The timing of spring events in mid- to high-latitude plants (budding, leafing, and flowering) is mainly regulated by temperature after dormancy. Moreover, while the climate signal controlling spring phenology is quite well understood, autumn phenology is less clearly explained in terms of climate. Data from phenological observations are considered to be the most sensitive for identifying how plant species respond to regional climatic conditions and to climatic changes (Chuine et al. 1999; Schwartz 1999; Rötzer and Chmielewski 2001). Long-term phenological monitoring over a wide range of latitudes and altitudes is an essential component of earth observation programs and global change monitoring. Phenology value as an indicator assumes even greater importance when these processes greatly affect biodiversity, agriculture, forestry, and human health (Schaber 2001; Van Vliet and De Groot 2001).

Phenological observations applied to agriculture are used to determine significant relationships between meteorological variables and plant behavior. Plant-growing models can be created during the vegetative and reproductive annual cycles in relationship to meteorological variables (Réaumur de 1730, 1738). While the relationships among temperature, precipitation, solar radiation, and phenological phases are considered for modeling (Fornaciari et al. 2000a; Galan et al. 2004), knowledge of the physiological mechanisms that regulate the biological trends is clearly lacking. Plants, particularly arboreal species, are very sensitive to climatic changes and can be used to show even slight temperature variations; therefore, plant species can be considered as climatic indicators (Kramer et al. 1996; Orlandi et al. 2005a).

Several studies have convincingly demonstrated that plants are already responding to global climate change by earlier leafing out, flowering, and fruiting, and later leaf drop; earlier flowering is one of the most sensitive biological indications of climate change (Dixon 2003; Chen et al. 2005; Menzel et al. 2006).

The particular advantage of phenological observations of plants regarding climatic change is that their reaction to even slight temperature variations can enable us to predict future time variations of the main reproductive phenomena and geographical shifts of productive areas.

In recent years, different hypotheses have been advanced concerning future trends in climatic change; most regional studies of this use data output from global circulation models (GCMs) of the atmosphere and oceans (Rahmstorf and Ganopolski 1999; UN-IPCC 2001; Seager et al. 2002; Arnell 2003). Most of the scientific community hypothesizes a more or less rapid temperature increase (with very hot summers and mild winters) and rainfall decrease in the Mediterranean area. This will lead to an increased risk of forest fires, water shortages, loss of agricultural land, and a northward shift of tree species. Another aspect of this scenario could be a geographical shift of some plant species; specifically the olive.

The aim of this study was to verify the relationships between meteorological trends and the biological response of the olive (Olea europaea L.) in a large central Mediterranean area by considering the flowering phenomenon of a typical species of the Mediterranean scrub (Orlandi et al. 2005a). Phenological observations of the anthesic phases of olive flowering were recorded over a 9-year period and then compared to the results of a longer 26-year study period.

2 Materials and methods

Since the 1980s, experimental techniques for pollen monitoring have been used in Europe for anemophilous plant species of agricultural and environmental importance (grape, olive, citrus, etc.) in order to determine the periods of anthesis by directly recording pollen emission by means of particular pollen monitoring instruments (pollen traps). In this study, the anthesic phenophase in the olive was recorded by using a pollen-monitoring methodology that involves the capture of pollen emitted and dispersed into the atmosphere. The aerobiological results, expressed as daily pollen concentrations (pollen/m3), were used in previous studies to determine the flowering phenophase (Chuine et al. 1998, 1999; Fornaciari et al. 2000a, 2000b; Osborne et al. 2000). Complete aerobiological stations with a volumetric pollen trap (VPPS 2000, Lanzoni model), a solar panel, four electrical batteries, and an automatic control system to assure a 24-h autonomy to the monitoring system were installed in each of the studied olive grove areas (Fig. 1).

Fig. 1
figure 1

Pollen sampler, olive monitoring, and meteorological areas

Full size image

The aerobiological sampling instrument consists of two parts: a lower fixed part and an upper mobile part that has a small winged portion that positions the trap with the slit opening facing into the wind. The “mouth", through which the airborne material enters, allows a constant air flow volume (10 l/min) and when particles enter they stick to a 7-day rotating drum that moves at 2 mm/h. By knowing the volume of air that enters, the concentration of particles/m3 can be expressed hourly, daily, or weekly (Gagnon and Comtois 1992; Tomas et al. 1997). A 7-day transparent adhesive strip is fixed on the drum (Melinex® with silicone oil); once removed, this strip is cut into seven pieces, corresponding to the days of the week, and mounted on microscope slides for identification and counting (Ogden et al. 1974).

From 1999 to 2007, pollen monitoring was conducted in 15 optimal olive growing zones located in four southern Italian regions (Sicily, Puglia, Calabria and Campania) well known for their olive oil production and chosen by a two-stage sampling method considering the most representative areas of each local olive cultivation zone.

A longer data series (1982–2007) of olive pollen monitoring in a large area of central Italy (Umbria Region) was also considered. In this way, the overall study area covered 16 stations in a large geographical zone between 35° 17’ N. and 34° 15’ N. within the greater Mediterranean macro-climate area. The Mediterranean climate is generally moderate with a relatively small range of temperatures between the winter lows and summer highs. The daily range of temperatures during the summer is wide, except along the coasts. The winter temperatures rarely reach freezing (except in high altitude areas). In summer, the temperatures are warm (with maximum values of 30–32°C) but do not reach the high temperatures of inland desert areas. The meteorological data obtained from Italian meteorological network stations of the Central Office of Ecology (UCEA) were the most representative of the meteorological conditions of the geographical areas where the pollen samplers had been installed. The following daily meteorological variables were considered: the maximum (TMax), minimum (Tmin), average (TAve) temperatures (°C) and precipitation (Rain) in mm.

Temperature (with daily GDD derived from a Single Sine Method and different threshold temperatures) were calculated from 1 January to full flowering dates for each study area to establish the variability of heat requirements during the years (Zalom et al. 1983) as high variability might demonstrate the low reliability of biological phenomena in relationship to climatic trends.

The start, the full flowering (maximum) and the end of flowering, in terms of pollen emission, were calculated in order to interpret the essential phenomena of pollen emission during flowering. The start was considered as the day where 5% of the total pollen amount during the entire flowering season was reached, while the end of flowering was considered as the day where 95% of total pollen amount was reached. Full flowering was the day where the maximum value of daily pollen concentration was recorded. This value was chosen because it corresponds to the moment when the majority of olive trees are involved in the full flowering phenomenon, representing in an indirect way the full flowering periods in the different study areas (Orlandi et al. 2005b). All flowering dates were calculated on the basis of the number of days since 1 January; this provided a simple chronology of the phenomena that occurred in the study areas. Graphs were made to show the presence of hypothetical trends during the study periods and to compare the trend of the Perugia station (the longer series) with that of the other monitoring stations during the same period (1999–2007).

Furthermore, the quarterly and monthly mean values of minimum, maximum, average temperatures and precipitations for all 16 study areas were calculated for statistical purposes to show the relationships between climatic trends and flowering data. Quarterly values were arrived at considering the four mean temperatures of December-January-February (DJF), March-April-May (MAM), June-July-August (JJA), and September-October-November (SON).

Correlation and regression analyses were made using the full flowering dates as the dependent variable and the monthly and quarterly mean meteorological values as the independent variable in each olive grove. The full flowering dates were the only phenological data used for statistical analyses considering that pollen emissions can be affected by the oasis phenomena (anticipated or delayed olive flowering in micrometeorological areas) during the first and last emission phases. For a suitable cause-effect evaluation, correlation analyses were made between flowering dates and meteorological variables recorded from the late spring (May) of the previous year (t-1) to the late-spring of the flowering year (t) to consider all the possible phenomena (flower induction-chilling-forcing) related to the fertilization process in plants with a 2-year reproductive cycle such as the olive.

Moreover, data from the NASA Goddard Institute for Space Studies (GISS), which is involved in predicting atmospheric and climate changes, were used in order to have a basic general database to test local climatic trends. In this way, European temperature maps based on the GISS databank were created using the monthly updated Land GISS 2001 Analysis (Hansen et al. 2001). Temperature change over a specified mean period, considered as a trend, was used with a smoothing radius of 1,200 km.

3 Results and discussion

Figure 1 shows the overall study area of the 16 Italian stations and their coordinates. Variability among temperature requirements calculated from 1 January to full flowering dates during the study years for each area were evaluated with RMSE calculations.

Figure 2 shows the threshold temperatures that minimize the RMSE calculated with yearly GDD amounts in each study area. The lowest threshold values (7°C) were found in two hilly areas (Benevento and Perugia). In three other areas (Foggia, Avellino, Bari) the threshold was 9°C, while in the other stations threshold temperatures ranged from 10–12°C. The highest value of 14°C was in Messina, Sicily. Thirteen monitoring areas had low variability values (an average value of 32.4) while in three stations of the Puglia Region (Brindisi, Taranto, Bari) the values were almost twice as high as the others (average 70.4). This difference very probably is related to non-meteorological factors, such as agronomic techniques, olive cultivars, orography, etc.

Fig. 2
figure 2

Threshold temperature utilized to calculate the yearly heat requirements and RMSE values obtained considering all the monitoring years in each area of investigation

Full size image

In Fig. 3, the dates of the start, full, and end of flowering at each of the 16 monitoring stations are shown along with the mean value of all dates. Flowering was first observed in the Sicilian region with the first pollen emissions registered at the end of April at the station near Palermo. Shortly afterwards, the olives at Trapani flowered while in the two other Sicilian areas of Agrigento and Messina pollen emission started at the beginning of May. In Calabria, pollen emission in the Cosenza area was 2 weeks earlier than in Reggio Calabria and Catanzaro. Regarding flowering trend, the Catanzaro area (especially for the beginning and maximum flowering) is somewhat different from the delay registered in the last study years. Indeed, the flowering trend at the Puglia stations confirmed this tendency toward delay of pollen emissions during the last study years. In three olive areas of the Campania region, the presence of pollen in the atmosphere was registered later than in the other sites. In particular, the values registered in the interior areas of Avellino and Benevento are close to those of the Perugia station, which is representative of central Italian olive areas. However, even in these areas (Avellino, Benevento, and Perugia) the flowering shift showed its major anticipation in 2001.

Fig. 3
figure 3

All flowering dates (start, max, end) in the 16 study areas during the 9 years of monitoring (1999–2007)

Full size image

In general, a common flowering trend was observed. In the first study years (1999–2001), pollen emission occurred progressively earlier, with the earliest date in 2001, while in the following five study years, until 2006, the trend seems to be inverted, showing a progressive delay of flowering. Only in 2007 did flowering dates return as they previously were.

Figure 4 compares the dates of full flowering for the Perugia olive area (with its monitoring activity from 1982) to the mean dates for the other Italian areas. The upper chart of the figure shows the complete trend of the Perugia station described by the line that reaches a good R2 value, considering the long period of analysis and the presence of some outliers. The overall trend of Perugia shows that the period from 1982–1995 was characterized by variable flowering dates (varying by 15–20 days); after that, the dates became more homogeneous, with differences of only a few days among the years. The period starting from 1995 corresponds to the beginning of a series of successive date anticipations, such as the full flowering date at the end of May in 2001. After 2001 there are more delayed dates until returning to the mean values registered at the beginning of the 1990s (flowering in the first 10 days of June). The particularly warm winter and spring temperatures of 2007 caused the advance of flowering dates in the whole series.

Fig. 4
figure 4

Full flowering dates for the Perugia olive grove area (1982–2007) and for the mean of the 15 southern Italian olive grove areas (1999–2007). Comparison between the Perugia series and the mean of the 15 stations during the overlapping period

Full size image

The lower chart of Fig. 4 compares the pollen emission dates in the Perugia olive area and the mean values of the other 15 southern Italian stations during the overlapping period from 1999–2007. The polynomial trend lines of the second order for both series of values are given. These lines illustrate quite adequately the flowering trends and show their similarity, allowing us to compare flowering in different geographical territories.

Statistical correlation among the yearly full flowering dates, the monthly and the quarterly mean values of the minimum, maximum and average temperatures and rainfall demonstrates the strong relationship to all temperature and rain values recorded in spring (T MAM, T March, T Apr …), and also to winter temperatures (T Jan, T Feb …) of the year. Moreover, temperature trends and rain recorded in the previous autumn (t-1) exercised a great influence on flowering dates. From the biological point of view, this result emphasizes the importance of the forcing phenomenon during the winter and spring of the year of flowering. However, the previous autumn phase should not be underestimated since it is here that the delicate equilibrium at the base of the flowering bud induction is determined.

Correlation analysis results were considered as functional to the successive linear regression analysis, the results of which are presented in Table 1. The results of the regressions between the effective full flowering dates (dependent variable) and the more correlated variables (independent variables) are reported for each of the 16 study areas. The regressions were carried out considering the best one- and two-variable models, utilizing independent variables affected by low inter-collinearity.

Table 1 Regression analysis results considering the 16 olive grove areas
Full size table

In seven olive grove areas (Foggia, Cosenza, Reggio C., Avellino, Benevento, Salerno, and Perugia) the best variable useful to make one-variable models was the temperature trend recorded in the spring period (March-April-May), which taken alone can interpret about 80% of the flowering date variability. The best two-variable models were used in nine olive grove areas: Agrigento, Messina, Palermo, Trapani, Bari, Brindisi, Lecce, Taranto and Catanzaro. In these areas the most correlated and affected by minor inter-correlation problem variables were those of temperature (above all, those of the winter of the year of flowering) and precipitation. In four cases (Bari, Brindisi, Lecce and Catanzaro) the variables useful for the regression analysis were those related to meteorological elements registered in October-November of the year before flowering (t-1). Generally, the highest correlation values between temperatures and flowering dates were negative, showing that the temperature increase corresponds to the anticipation of pollen emission. In particular, the mean anticipation was calculated at 6.7 days per degree, with a minimum value for the stations of Bari, Palermo and Lecce (< 3 days) and a maximum value for the stations of Cosenza, Messina and Salerno (> 13 days). Another phenomenon to highlight is that of the Sicilian area, where the best interpretative variables were those of winter temperature and spring rain, in accordance with the typical climate features of the area. At that latitude, in fact, the only cold period was recorded in January, while the limiting effect of precipitation should be considered until March.

The correlation and regression results indicate that in almost all study areas the spring temperature is one of the most influencing variables in determining flowering time. Figure 5 shows the trends of TAve_MAM and flowering dates, using the mean values of the 15 southern Italian stations from 1999 to 2007 and the value of the longer Perugia series. In the upper chart of the figure, the linear trend lines for the Perugia series show a good level of interpretation (the significant R2), even if obtained using rather alternating values of the 26-year survey. The main phenomenon of this “historical series” is the ascertained interconnected variation of temperature means and flowering times from 1996 that confirms the trend partially evidenced by the TAve mean values of the first 6 months of the year. Generally, in this long period (1982–2007), the trend lines appear as opposite, defining their complementary behavior. In the lower chart of the figure, the shorter time series related to the mean values of the 15 olive areas do not evidence a clear trend either for flowering or meteorological phenomena.

Fig. 5
figure 5

Trends of spring quarterly mean of average temperature (TAve_MAM) and flowering dates for the 15 southern Italian stations (1999 to 2007) and for the longer Perugia series (1982–2007) with related trend lines

Full size image

In Fig. 6, the spring (March-May) quarterly mean temperature trends considered as a temperature change over a specified mean period extrapolated from the GISS European maps are shown for 1982–2001 (A) and 2002–2006 (B). The comparison between these two maps shows a cooling phenomenon limited in time in comparison to the larger previous period in the area considered. The second period (B) had a cooler trend than the first period (A), with cold air currents from northern Russia breaking into the Mediterranean area; however, the last year of investigation (2007) appeared as an interruption of that mini-trend, with flowering dates returning to a decreasing (warming) trend.

Fig. 6
figure 6

Quarterly (Mar-May) mean temperature trends based on local linear trends in Europe extrapolated from the GISS maps for two periods: 1982–2001 (A), 2002–2006 (B)

Full size image

4 Conclusions

The present study evidenced during the years a high homogeneity of temperature requirements in almost all the monitoring areas, as related to flowering dates. Full flowering mean values (in days), calculated at all the southern Italian monitoring stations, were obtained considering dates with very little temperature variations year by year. Thus, the mean trend reflects the climatic behavior of a large geographical area. However, in bio-meteorological studies, plant responses may be more or less proportional to environmental changes, exalting or hiding the natural phenomena, and inducing us to take extreme caution in reaching conclusions on climatic trends, above all in short investigation periods.

The results of this investigation also showed the necessity to have a long series of data in order to interpret correctly the climatic trend and to avoid partial and/or incorrect knowledge of such complex behavior. The investigation in the southern Italian areas was for only 9 years and consequently too brief from a climatic point of view. It only revealed a meteorological mini-trend from 2001 to 2006 that once inserted in a larger database, such as that recorded in Perugia for 26 years, may be interpreted in a different way and may suggest more important biological and meteorological trends. Considering future meteorological trends, the suitable area for olive cultivation could be enlarged due to temperature and precipitation pattern changes. In a Mediterranean climate warming scenario, for example, a northward shift of the potential olive cultivation area could be expected (Bindi and Howden 2004; Maracchi et al. 2005).

Statistical analyses evidenced that the spring quarter temperature trends were strictly related to the trends of olive flowering phenomena, demonstrating this dependence by forcing temperature of the olive growing buds which must prepare their reproductive structures (Galan et al. 2004; Orlandi et al. 2005c).

Development of reproductive phases in the olive is concentrated in a 5–6-week pre-anthesis period, from the end of April to the end of May. The climate trend during this period may strongly influence the biological development and consequent flowering, which is the last event within the complex biological and physiological reproductive cycle. Moreover, in arboreal species such as the olive, the reproductive cycle takes place over a 2-year period. In the first part (September-October t-1) the buds are induced to flower, and this depends greatly on the plant’s inner balance between developing fruit and photosynthetic products, but also on meteorological conditions. In the second part, the meteorological variables play a direct and decisive role in the micro-gametophytic development. The olive, being a typical Mediterranean species, is sensitive to low temperatures and resistant to water shortage. In this manner, the northern and the southern limits of olive cultivation in the Mediterranean basin are conditioned by low temperature and low rainfall (Bindi et al. 1992).

Due to the short observation period, above all in the southern Italian areas, identification of clear trends is difficult and therefore the scale of natural variability is uncertain. Continued observations of the olive will allow the current trend to be monitored. This type of study with its interpretative potential could be useful in showing the vegetative-reproductive response of a particular plant species to local climatic changes.