Urbanisation induces early flowering: evidence from Platanus acerifolia and Prunus cerasus
- First Online:
- Cite this article as:
- Mimet, A., Pellissier, V., Quénol, H. et al. Int J Biometeorol (2009) 53: 287. doi:10.1007/s00484-009-0214-7
- 368 Views
The effect of towns on plant phenology, i.e. advancement of spring development compared with a rural environment, via the urban heat island (UHI) phenomenon, has been shown for many towns in many countries. This work combines experimental and observational methodology to provide a better and deeper view of climatic habitat in an urban context with a view to understanding the relationship between plant development and urban climate on the intra-urban scale (by taking into account town structure). A dense network of 17 meteorological stations was set up in Rennes, France, enabling us to identify and quantify climatic changes associated with the UHI. Meanwhile, phenological observations were made during early spring (March and April) in 2005 on Platanus acerifolia and Prunus cerasus to study the relationship between climatic and phenological data. The results show that there is both a climatic gradient and a developmental gradient corresponding to the type of urbanisation in the town of Rennes. The town influences plant phenology by reducing the diurnal temperature range and by increasing the minimum temperature as one approaches the town centre. The influence of ground cover type (plants or buildings) on development is also shown. The developmental phases of preflowering and flowering are influenced to differing extents by climatic variables. The period during which climatic variables are effective before a given developmental phase varies considerably. The preflowering phases are best correlated with the mean of the minimum air temperature for the 15-day period before the observation, whereas flowering appears to be more dependent on the mean of the daily diurnal temperature range for the 8 days preceding the observation.
KeywordsUrban heat islandExperimental phenologyEarly spring phenologyMinimum temperatureDiurnal temperature range
Now that 50% of the world’s population live in towns or cities, the influence of urbanisation on the rural environment is increasing through, for instance, urban sprawl and pollution. As a result, remnant rural areas are being engulfed by urban zones (United Nation Population Division 2008).
In addition to environmental modifications in towns due to human activities (e.g. pollution dues to industry and fuel combustion, fragmentation due to urban sprawl, photoperiod modification dues to public lighting), urban areas have horizontal and vertical surfaces that modify the physical characteristics of the lower layers of the atmosphere (temperature, wind and rainfall). This results in a variation in the net balance of radiation (Escourrou 1981), which creates a dome of warm air called an urban heat island (UHI) above the town.
Within the UHI, climate varies in space and time because of the heterogeneity in land use between the town centre and the suburbs. The centre is usually heavily built-up, with vertical surfaces, whereas the suburbs have horizontal surfaces that are damper or vegetation-covered. The magnitude of the UHI depends on the topographical context of the town, its nature (shape and density of buildings, nature of construction materials and industrial activities), regional climatic conditions, types of traffic and the weather pattern (Oke 1987; Cantat 1989; Lee 1992). The UHI increases when atmospheric conditions are stable (anticyclonic). From the centre to the edge of the UHI, minimum daily air temperature varies more than maximum daily air temperature. The diurnal temperature range (DTR: maximal temperature − minimal temperature) is therefore smaller in the town centre than in the surrounding countryside (Oke 1987).
This mesoclimatic modification is likely to be a major driver of the dynamics of urban-dwelling animal and plant species. Phenology, which can be defined as “the study of the timing of recurring phases, the causes of their timing, with regard to biotic and abiotic forces, and the interrelation among phases of the same or different species” (Global Phenological Monitoring, http://www/dow.wau.nl/msa/gpm/), is one of the first processes to be affected. Climate change constitutes a laboratory for studying species responses to increasing temperature and CO2 concentration. As such, the field of phenology is expanding rapidly (Schwartz 1999; Aerts et al. 2006). A precocious beginning of spring phenology has been shown to be strongly correlated to increases in air temperature (Menzel 2000; Sparks et al. 2001; Chmielewski and Rotzer 2001; Menzel et al. 2001) and to the lengthening of the growing period (Kvaalen and Johnsen 2007; Luo et al. 2007), mainly among early spring species (Sparks et al. 2001).
Several variables involved in spring phenology are modified in urban areas; indeed air temperature (Nieddu et al. 1990; Chmielewski and Rotzer 2001; Menzel et al. 2001; Sparks et al. 2001; Fitter and Fitter 2002; Walther 2003; Wolfe et al. 2005; Aerts et al. 2006; Linderholm 2006) as well as photoperiodicity (Halliday et al. 2003; Setiyono et al. 2006) and precipitation (Peñuelas et al. 2003; Dreyer et al. 2005) act as drivers in the course of spring development.
The literature shows a clear relationship between the UHI and the triggering of plant phenological stages. In these broad-scale or long-term studies, air temperature appears to be the main climatic driver of plant phenology (Defila 1991; Roetzer et al. 2000; Defila and Clot 2001; Menzel et al. 2006; Defila and Jeanneret 2007; Luo et al. 2007). The ecophysiological mechanisms whereby air temperature affects plant phenology are not well understood. Various phytochromes are involved, as well as many regulatory processes (Smith and Whitelam 1997; Halliday et al. 2003; Benedict et al. 2006).
While the impact of urbanisation on plant phenology is well known, at both broad spatial and time scales, there is a lack of phenology studies at the finest scales. Indeed, studies have focussed so far neither on intra-urban phenology processes nor on precise budburst over short time range.
Our study focusses on the spatial and temporal impact of UHI on phenology at the intra-urban scale in the city of Rennes (France), with an emphasis on the short period of budburst. As an answer to Shochat et al. (2006a, b), who emphasised the need for a mechanistic approach to urban ecology, we propose an experimental phenology approach to understand more accurately the link between the UHI during spring 2005 and its consequences for the early spring phenology of trees (Platanus acerifolia and Prunus cerasus).
Materials and methods
This study took place within the framework of the multidisciplinary research program ECORURB (http://w3.rennes.inra.fr/ecorurb/index.php). This program (2003–2012) centres on the understanding and prediction of town/country biological relationships (Daniel and Lecamp 2004; Clergeau et al. 2006; Vallet et al. 2008), and is ongoing in the towns of Rennes and Angers.
The inner ring (U) is characterised by a dense cover of roads and buildings. It includes the shopping centre, the historic town centre, whose plant cover is less than 15%, and the adjacent residential area including 1–3 storey flats with 20–40% green space.
The suburban ring (SU) also has a large amount of roads and buildings, but a bigger proportion of green space. It consists of residential areas made up of detached houses with gardens, and residential complexes surrounded by green areas. The planted area in this ring can reach 70%.
The outlying (periurban, PU) ring is the farming and natural zone adjacent to the town. It extends for several kilometres beyond the limits of the suburban ring.
The UHI of Rennes
A total of 17 weather stations (Davis Weather Monitor II; Davis Instrument Corp. 2000) were set up to take hourly measurements of the minimal, maximal and mean temperature, the mean relative humidity, the mean speed and direction of the wind, and the cumulative precipitation sum (data were actually measured continuously and averaged, or summed hourly by the weather station). The weather stations were checked for several weeks to test the uniformity of the measurements made by all the sensors. They were installed between the end of 2003 and 2005. To site the weather stations, the Rennes map was used to localise theoretically suitable areas for the study purpose. More precisely, since there are numerous nearby obstacles in the town, we also tried to place the stations in sites exposed towards the south (to avoid the effects of shade when measuring maximum temperatures) and towards the west to take good account of the prevailing winds. We deliberately chose to give priority to temperature measurements, assuming that the UHI is the phenomenon most likely to affect the developmental response of species.
In March 2005, after one period colder than the climatic norm (climatic mean calculated over 30 years), temperature and precipitation was equivalent to the climatic norm. In April 2005, the monthly temperature mean was higher by 1–2°C than the climatic norm (Meteo France 2005; http://france.meteofrance.com).
The data from the weather station network showed considerable spatial and temporal variability in the weather on the town/country scale but also within the town. There were substantial weather variations among urban, suburban and periurban stations. The density of buildings along the urbanisation gradient caused both day and night temperature increases, as did the type of ground cover (which had a significant impact on the local climate). The climate observed in an open green suburb can be very similar to that of the outlying countryside.
The presence of the UHI causes horizontal temperature variations and influences the distribution of relative humidity. Hence, it was not surprising to find lower humidity values in the (urban) areas where the temperature was the highest. The urban atmosphere locally influences the circulation of air masses and cloud formation. The rising of the air layer above the town forces air masses to rise, leading to rainfall below the wind of the urban dome. In general, rainfall figures are lower in the outlying zone than in the town.
Analysis of the wind speed in the urban, suburban and periurban zones also showed big variability linked to the presence of obstacles close to the weather station. Consequently, the wind speed mean was higher in the periurban (outlying) zone where the absence of buildings favours wind flow.
All the meteorological data recorded during the experimental period showed pronounced weather variations within relatively restricted areas.
Spatial phenological differences in Rennes
The biological model chosen for this experiment is the tree form—more suitable to measure the impact of urbanisation on phenology at an urban scale than herbaceous plants, which are more sensitive to microclimatic variations and their immediately adjacent environment. Indeed, the urban scale at which we recorded our climatologic measures is a mesoscale rather than a microscale. The choice of species was limited by the existence of individuals available in the field and on budburst dates, which had to be at the beginning of spring (March–April in this region).
This part of the study aimed to establish a snapshot of phenological progress at a given date in the town of Rennes. The species considered, Platanus acerifolia, is found throughout the town and is planted mainly in lines along the streets, on the pavements. The trees used in the study are at least 30 years old and have reached sexual maturity.
A sampling program at 17 sites, covering the whole town (see Fig. 5 for site locations), was carried out on 8 April 8 2005.
Branches were sampled at about 4 m height, using a pruning hook. At least four branches were sampled at each site, giving a mean for each date of 51 ± 13.27 buds per site.
BB1 = budburst, phase 1: buds are swollen.
BB2 = budburst, phase 2: buds are actually starting to burst (i.e. to open)
F1 = flowering, phase 1: buds are open and flowers are now visible
F2 = flowering, phase 2: flowers are now fully open
The percentage of buds at each developmental phase was calculated per branch for each site. These percentages were used to compute the regressions of each developmental phase as a function of distance from the town centre, using Minitab 14 software (http://www.minitab.com/).
Highly temporally resolved phenological observations along an urbanisation gradient
The purpose of this experiment was to describe the impact of the UHI on different budburst stages using highly temporally resolved data.
The experimental phenology protocol used was innovative and allowed us to avoid errors caused by intraspecific variability. Branches of cherry trees (Prunus cerasus) were sampled on a given individual at a given height. This latter precaution limited the phenological variations that could have resulted from nutritional differences in terms of sap quality and quantity within the plant depending on height. The branches were placed in plastic bags of water containing bleach at 1:1,000 to limit proliferation of microorganisms. The bags were angled to the south and placed against a wall or hedge at a height of 1.5 m so as to be in similar sunlit conditions. These conditions were observed at all sites except Gallets (in the suburban ring), which is a large area of grass with no hedges or walls, so the bags were fixed to a fence.
The phenological surveys were made every week by determining the four stages of phenological development of the buds as defined above.
In the urban zone, two stations. U1: urban non-vegetative station, and U2v: vegetative station
In the suburban zone, three stations. SU3: non-vegetative station, SU4: non-vegetative station, SU5v: vegetative station
In the periurban zone, one vegetative station (PU6v)
Station SU4 was chosen for its location in the south-east of the centre, a zone that appears from the climatic data to be warmer than the other suburban districts.
“simple” branches: 6–9 buds, 18–25 cm, unbranched
“average” branches: 11–17 buds, 15–25 cm, one branch
“large” branches: 20–31 buds, 31–43 cm, at least one branch
Five branches, i.e. one simple, two average and two large, were placed in each station. The total number of branches set out at each site was between 76 and 93 (mean 82.66 ± 6.02 SD). If there are different phenological reactions of branches according to their size or number of buds, they will be accounted for without affecting the results.
For each date, the differences in percentages of the different phenological phases between stations were tested using a Kruskal-Wallis test on the data for Prunus cerasus.
The statistical methods used in this work are based on the postulate of linearity of the phenological response to temperature (Wielgolaski 1999).
It is necessary to determine the duration of the period of influence of temperature on developmental progress. In fact, the state of advancement of buds at a precise date depends on a number of factors, in particular the weather conditions preceding the observation date. The difficulty thus resides in the determination of this period, which varies with the species, soil and weather conditions.
We chose to test different periods of influence of weather conditions within a 15-day range from the cut of branches to the set-up of the experiment. Data were processed using RDA (redundancy analysis, Stewart and Love 1968). This multifactorial analysis measures the link between a matrix of variables to be explained (here phenological data) and a matrix of explanatory variables (here climatologic data). Results indicate the fraction of the variance of the phenological data explained by the climatologic data (in other words the correlation between phenology and climatology). Monte-Carlo permutation tests allow forward selection of climatologic variables significantly correlated with the phenological variables. As in principal component analysis (PCA), the variability is divided into several components (thereafter axis) bearing a decreasing fraction of total variability. As variability of explained variables is expressed as a multivariate linear combination of explanatory variables, the RDA allows the relationship between the explained and the explanatory variables to be plotted on a single graph (hereafter called “biplot graph”). The RDAs were carried out using Canoco 3.1 for Windows (http://www.microcomputerpower.com/catalog/canoco.html).
The phenological data matrix contained the four variables BB1, BB2, F1 and F2. The climatologic variables (explanatory) tested by the RDA were the daily means of minimum temperature (Tn), maximum temperature (Tx) cumulative precipitation sum (rainfall), gusting wind speed (gust), atmospheric relative humidity (humid) and diurnal temperature range (diff). The considered climatologic variables were chosen because they were measured by the meteorological stations, they were affected by urbanisation, and the influence on phenology on some of them is well known (Nieddu et al. 1990; Chmielewski and Rotzer 2001; Menzel et al. 2001; Sparks et al. 2001; Fitter and Fitter 2002; Halliday et al. 2003; Peñuelas et al. 2003; Walther 2003; Dreyer et al. 2005; Wolfe et al. 2005; Aerts et al. 2006; Linderholm 2006; Setiyono et al. 2006).
calculation of weather means over the 15 days preceding the observation. This period represents the time elapsed between the set-up of the device and the first observation. Therefore, this is the maximum duration.
calculation of weather means for the 8 days preceding the observation
calculation of weather means for the 3 days preceding the observation
calculation of weather means for 4 days, ending 4 days before the observation date
For each period, a first RDA was carried out using every climatologic variable in order to select the significant explanatory variables (using Monte-Carlo permutation tests). Another RDA is then carried out using the above selected variables. Only these second RDAs are shown in this paper. This procedure allowed to explain the variability of the phenological matrix in a more appropriate way than the RDA with all the climatologic variables as it removed any collinear and non-significant variables.
Spatial variability of plane phenology
The developmental stage used for the map of phenological advancement of the planes in the town is the percentage of flowering, i.e. the final stage of development.
In general the results show an earlier start to flowering in the town centre, but with variability within the urban rings.
Temporal variability of cherry tree phenology and link with the UHI
Highly temporally resolved observations
Results of Kruskal-Wallis tests (H values) carried out on phenophase data for each date. BB1 Budburst, phase 1: buds are swollen; BB2 budburst, phase 2: buds are actually starting to burst (i.e. to open); F1 flowering, phase 1: buds are open and flowers are now visible; F2 flowering, phase 2: flowers are now fully open
U1a, U2v, SU3, SU5v, PU6v
U1a, U2v, SU3, SU5v, PU6v
SU5va, PU6va, U1, U2v, SU3
U1a, U2va SU3, SU5v, PU6v
Stations SU5v and PU6v, the two furthest from the town, situated in green spaces, showed a similar trend: the number of swollen buds decreased fairly steadily during the observation period. The maximum percentage of buds in first bursting phase (BB1) appeared around 25 March whereas the branches began to flower between 29 March and 1 April. Moreover, these stations had fewer visible flowers on 29 March than the other stations. The developmental maxima of these two stations were well identified. The observation period covered only the beginning of flowering. For the three first stages and the three first dates there was a difference of about 10% between the two stations, which tended to decline, and the difference had disappeared by the end of the observation period. This trend was confirmed by the graph showing the flowering buds, on which the curves for the two sites were almost superimposed. Therefore, the beginning of budburst was more consistent in the outlying zone, but later this difference has no effect on the number of buds at the start of flowering.
Stations U1 and U2v are two urban stations. U2v was located in a green space, whereas U1 was in a very urban environment. The two stations showed fairly similar random behaviour during the three first phases, during which it was difficult to discern a trend. Two maxima were identifiable at the BB1 (25 and 29 March), BB2 (25 and 29 March for U1 and 25 March and 1 April for U2v) and F2 (27 March and 1 April) stages. Although the two curves followed the same tendency during most developmental stages, this was not the case at the time of flowering. Station U1 flowered significantly earlier, and reached a flowering maximum whereas U2v had not begun to flower (29 March). The two curves crossed over at the end of the observation period, and the buds at station U2v began to flower intensely and suddenly, catching up with the flowering percentage at U1 by the end of the observation period. The two urban stations differed significantly from the other stations in their flowering on 1 April.
The behaviour of buds at station SU3 converged during the swollen bud phase with that of the two town centre stations, with two maxima (27 March and 1 April), although the two peaks were later. Flowering at SU3 also followed the same curve shape as U1 and fell between that of the urban and periurban stations. For the other stages, the curves indicate clear tendencies, similar to those of the periurban sites, only slightly earlier. In general, this station was intermediate in behaviour between that of the town centre and the outlying zone.
Two budburst maxima are identifiable at the BB1 stage for the three stations. As the second maximum of station SU3 appeared at the end the observation period, the later stages of development were not observed.
More generally, the earliest station in terms of flowering was U1. The first two stations to reach more than 25% flowering were those in the town centre, followed by SU3 (the warm side of the town periphery).
Relationships between UHI and cherry tree budburst
The analyses to be presented were chosen after selecting the most influential variables. Since minimum temperature (Tn), maximum temperature (Tx) and the diurnal temperature range (DTR) were very closely related, only the most significant of the three was used in each analysis.
Comparison of redundancy analyses (RDAs) carried out on each data set. DTR Diurnal temperature range, Tn minimum temperature, Tx maximum temperature
Duration of influence of variables
Monte Carlo tests
DTR, wind, humidity
4 days mean calculated 4 days before
Tn, wind, humidity, DTR
Thermal difference, wind, humidity
The results of the 4−4 day RDA were better than those of the 3-day RDA analysis but less good than those for 8 days. This comparison confirmed that the 3–4 days preceding the observation were not very important, since the explained variation only increased from 22.9% for the 4−4 day analysis to 25.6% of the 8-day RDA.
The 8-day RDA (Fig. 7a) explained the most phenological variation, mostly on axis 1 (22.4%). Monte-Carlo tests indicated that the climatic variable contributing most was the daily DTR (F = 19.65**).
The first phase of flowering was the phenological phase best explained by axis 1 (by the DTR) followed by the variables BB2 and F2 (around 22–23% of their variance). A large DTR, as found outside the town, was inversely correlated with a large number of flowering buds. Conversely, a large DTR was correlated with a large number of buds beginning to burst.
The variable BB1 was less well accounted for. Only 13% of this variable was explained by the primary axis, and 17% by the two primary axes. It is interesting to study this phase, which marks the beginning of budburst, and which was rather poorly explained by all the analyses. The RDA that best explained BB1 was that using the 15-day weather data, which accounted for 19.1% of the variance.
Altogether, the phenological variable best explained by the analyses was flowering (Fig. 7). The phases BB1 and F2 behaved in opposite ways, and budburst (BB2) was better explained by a 15-day period of influence of the minimum temperature, while flowering (F2) was more affected by an 8-day period of influence and by the daily DTR.
This spatial experiment showed a phenological gradient within the town between the urban and suburban rings; urban development seemed to influence the triggering of the spring developmental phases of the trees. The analyses showed variation within the rings, suggesting that distance to the town centre alone is not sufficient to explain the phenomenon of the phenological gradient within the town. Considering temperature, the UHI had two main characteristics: an increase in the town mean temperature along with a decrease in the daily differences between day and night temperatures. Green urban areas therefore experienced higher mean temperatures than those of rural areas. Within these zones, the diurnal temperature ranges are more consistent and are somewhat dissimilar to those of the countryside (Carrega 1994).
Between the outlying (periurban) zone and the green suburbs, there was a difference in the percentage of buds at each phenological stage, even though the curves were similarly shaped: there were more buds showing phenological peaks in the periurban zone. The difference decreased over the course of time, leading to a crossing over of the responses at the end of the observation period. The town spread out the first phases of development: the peaks were less clear; they applied to fewer buds and were more spread out over time. On the other hand, budburst started later in the periurban zone, with a pronounced peak.
For stations experiencing a high urban influence (urban stations and built-up suburbs), the phenological phases turned out to be more disturbed. Thus, double peaks were visible for these three stations, and the different phases were spread out over time.
Behaviour was different between green spaces and the built-up stations (U1 and SU3): in the latter, flowering took place at the same time as in other green areas, whilst the other phases were ahead.
These observations were confirmed by the results of the RDAs carried out on the cherry tree branch data. This analysis showed the importance of the type of climatic variable used, but also of the time over which the temperatures were averaged.
The main factor explaining the differences in flowering in the different urban rings was DTR measured over the 8 days before the observation. Wide DTRs favoured an early start of budburst and later slowed down the development of flowers (the rural situation), which had the effect of bringing together the flowering dates of the green zones, whether they were in the urban, suburban or periurban rings.
Our results disagree with those of Myking (1997), who found no flowering response in individuals of Betula pubescens subjected either to fluctuating temperatures over 24 h or to the corresponding mean temperature. However, in our case the plants were subjected to both fluctuating temperatures and to differences in temperature. It is likely that the combination of both these factors accentuated the observed differences between the urban rings. Moreover, studies on the initiation of flowering by phytochromes (Smith and Whitelam 1997; Halliday et al. 2003; Benedict et al. 2006) illustrate the complexity and multiplicity of the causalities involved in the phenomenon (leading to redundant or opposite answers; Heide 2008)
Although the DTR affected every phenological phase, comparison between the different RDAs indicates that the early phases of budburst (BB1 and BB2) were best explained by the mean minimum temperature for the 15 days preceding the observation. The variation in BB1 explained by the RDA increased when the period considered increases. Low minimum temperatures (periurban, station SU3 situated south of the town, and the green stations) led to earlier onset of budburst (BB1). Roetzer et al. (2000) had already observed that preflowering phases react more to temperature than flowering. These results agree with those of Wielgolaski (1999), who showed that the date of budburst of several tree species was more closely related to the minimum than to the maximum temperature. On the other hand, Luo et al. (2007) obtained opposite results in which the spring phenology was better correlated with the mean maximum temperature for the 3 months preceding flowering.
The combined action of the two weather variables (diurnal temperature range and minimum temperature) disturbed the urban phenology and turned the flowering phenomenon in the urban environment into a more or less random process. The developmental phases became more spread out over time. The peak usually characterising the progress of a phenological phase disappeared and was replaced by several peaks, each necessarily affecting fewer buds. The different phases did not appear to have the same delay in their response to weather conditions.
Peiling et al. showed in 2006 (for four tree species in China) that the period of temperature influence on flowering was that of the 30 days preceding flowering. Our work showed that a period of exposure of 15 days of the branches of a given individual to variable temperatures was sufficient to cause a time lag in development.
The simple but effective methodology used in this work revealed the direct impact of a well-known climatic phenomenon, the urban heat island, on the triggering of spring developmental phases. We overcame the issues caused by studying plant species with long lifespans by focussing only on the direct variations induced by urban weather on buds.
The contribution of this work is to demonstrate the role of the UHI and small scale urban structure on development by way of variables related to temperature. Towns appear to be a suitable places for studying development because of the existence of a gradient of climatic variables that affect the flowering phenomenon.
the minimum temperature acts over a period of at least 15 days and influences the starting date of budburst.
the diurnal temperature range acts over shorter periods (3 days) and influences the end of the flowering cycle, and especially the flowering date.
Our results lead to the conclusion that, when studying the influence of towns on phenology, it is necessary to consider at least two characteristics on interlinking scales. The first is that the town, in the wider sense, seems to disturb and spread out plant developmental phases, and the second is that the immediate environment (in our case the presence of either a planted or built-up area) may buffer the impact of the town on phenological stages. The planted area seems, in our study, to attenuate the influence of the town at flowering time, without however having had any earlier effect on spring development. Thus, as “greenness” of some part of the town is a driver of plant life, one should consider this factor in city planning design. Nevertheless, if the experiment were repeated, one may expect an increasing shift between city and countryside, as the UHI increases over decades along the urban sprawl.
This work reveals a very short-term impact of the town on phenology, leading to a phenological spatial and temporal gradient between the town centre to the suburbs. Indeed, urbanisation induces early flowering in the city centre. Moreover, as urbanisation led to a more random and less clear phenological sequence in the centre, one may expect a severe impact on plant reproduction, likely to lead to longer exposure of humans to plant pollen. As cities appear to be surrogates of global change, results of experimental phenology in town, carried out over years, may be good approximations of local plant species responses to global change.