Using phenology to assess urban heat islands in tropical and temperate regions

Authors

  • Susanne Jochner,

    Corresponding author
    1. Department of Ecology and Ecosystem Management, Ecoclimatology, Technische Universität München, Freising, Germany
    • Correspondence to: S. Jochner, Department of Ecology and Ecosystem Management, Ecoclimatology, Technische Universität München, Hans-Carl-von-Carlowitz-Platz 2, 85354 Freising, Germany. E-mail: jochner@wzw.tum.de

    Search for more papers by this author
  • Milene Alves-Eigenheer,

    1. Departamento de Botânica, Laboratório de Fenologia, Plant Phenology and Seed Dispersal Group, Instituto de Biologia, Universidade Estadual Paulista (UNESP), Rio Claro (SP), Brazil
    Search for more papers by this author
  • Annette Menzel,

    1. Department of Ecology and Ecosystem Management, Ecoclimatology, Technische Universität München, Freising, Germany
    Search for more papers by this author
  • Leonor Patrícia C. Morellato

    1. Departamento de Botânica, Laboratório de Fenologia, Plant Phenology and Seed Dispersal Group, Instituto de Biologia, Universidade Estadual Paulista (UNESP), Rio Claro (SP), Brazil
    Search for more papers by this author

ABSTRACT

The study of urban phenology is relevant to assess the effects of heat islands and the potential impacts of climate change on plants. Temperature is the main factor regulating plant development in moist/temperate regions. However, plants in tropical cities may respond to additional environmental cues, such as air humidity.

We examined whether differences in air temperature and humidity along an urban–rural gradient influenced phenological onset dates of trees in a temperate mid-latitude city (Munich, Germany) and a city in the neotropics (Campinas, Brazil). (Dis)similarities were identified incorporating equivalent study design based on identical methods for recording phenology and site-specific meteorological data.

We found that phenological onset dates of silver birch (Betula pendula) were able to describe local temperature variations in Munich. Correlation coefficients between mean temperature and onset dates of Betula were significant and varied between r = −0.48 and r = −0.72. However, onset dates for yellow trumpet tree (Tabebuia chrysotricha), false Brazilwood (Caesalpinia pluviosa) and rosewood (Tipuana tipu) in Campinas were rather variable, and the magnitude and significance of correlation coefficients with temperature varied greatly among species/phenophases. Nevertheless, we detected significant correlations, especially between Tipuana phenophases and temperature and between leaf unfolding of Tabebuia and humidity.

Our findings revealed that the phenology of Tipuana was sufficiently sensitive to detect urban heat island effects in Campinas and might be a useful indicator of temperature variations and, in turn, of global warming. Because Tipuana is widely used for urban arborization in South America, we foresee many applications for monitoring heat islands in the neotropics. Furthermore, the fact that most phenophases of Tabebuia were not responsive to temperature stressed the importance of humidity variables. Additional species and phenophases should be explored to identify the further potential of phenology for monitoring heat islands in tropical cities.

1. Introduction

Since Howard (1833) published the first study about the urban heat island (UHI) in London, this phenomenon has gained considerable interest among meteorologists (Kratzer, 1937; Chandler, 1964; Oke, 1976; Landsberg, 1981). The examination of the UHI effect and associated aspects is still relevant and important (McCarthy et al., 2010; Oleson et al., 2010), especially because an increasing percentage of the world population – particularly in the tropics – live in (mega)cities (Roth, 2007; Grimm et al., 2008) or are regular visitors to urban areas for inter alia employment or social/recreational reasons.

The major cause of the development of the UHI is the substitution of the natural environment by built-up and sealed areas (Landsberg, 1981). The construction material used in cities (e.g. concrete and asphalt) does not allow water to penetrate into the soil. Additionally, a high proportion of the incident shortwave radiation is absorbed and transformed into sensible heat (Landsberg, 1981). In particular, the three-dimensional urban design with tall buildings, high building density and narrow streets promotes the absorption of a high fraction of radiation (Landsberg, 1981; Oke, 1987). In contrast, vegetated urban areas represent cooler spots through evaporation processes (Upmanis et al., 1998; Shustack et al., 2009). Another factor that contributes to the UHI is the anthropogenic emission of sensible heat as well as moisture associated with energy consumption (Sailor, 2011). Urban air pollutants do have an effect not only on air quality and therefore on human health (Davidson et al., 2005; Kampa and Castanas, 2008) but also on urban energy fluxes (Landsberg, 1981). Particulates scatter and absorb incoming solar radiation leading to a reduction of direct radiation and a higher amount of diffuse radiation (Landsberg, 1981; Kuttler, 2004). In addition, absorption and reemission by infrared active gases and aerosols increase downward longwave atmospheric radiation (Kuttler, 2004).

Urban climate studies have been conducted far less in (sub)tropical areas compared with temperate regions (Roth, 2007). In general, the UHI intensity is less pronounced in sub(tropical) compared with temperate cities, whereas highest urban–rural temperature differences are observed in the dry season. In temperate cities the UHI is more pronounced in the winter season and during night (Landsberg, 1981; Baker et al., 2002; Mimet et al., 2009; Shustack et al., 2009).

In phenological research, cities represent important study areas because their warmer conditions allow an assessment of the potential future impacts of climate change on plant development. Therefore, urban areas can be used as a surrogate or experimental treatment for future global warming (Ziska et al., 2003; Luo et al., 2007; Mimet et al., 2009). This application is related to one of the original purposes of urban phenology: detecting UHI effects. The use of phenological observations in urban areas for assessing microclimatic conditions has a long tradition in temperate regions. There are plenty of studies dealing with plant development in urban areas of Europe (e.g. Zacharias, 1972; Baumgartner et al., 1984; Bernhofer, 1991; Lakatos and Gulyás, 2003; Mimet et al., 2009; Jochner et al., 2011; Jochner et al., 2012a), North America (e.g. White et al., 2002; Zhang et al., 2004) and Asia (e.g. Omoto and Aono, 1990; Lu et al., 2006; Luo et al., 2007; Jeong et al., 2011).

These studies clearly reveal that plants growing in temperate cities flower earlier than plants in rural areas because of higher local temperatures. However, urban phenology studies in tropical cities are rare (e.g. Gazal et al., 2008) and absent in the neotropics. There is still a lack of understanding of how the phenology of tropical tree species is influenced by temperature, e.g. whether there is a temperature threshold for plant activity under warm tropical climates (Clark, 2007; Colwell et al., 2008). Moreover, phenology in tropical trees is generally considered to be water or light limited (Morellato et al., 2000; Borchert et al., 2002; Singh and Kushwaha, 2005; Staggemeier and Morellato, 2011). There are suggestions that climate-driven models are not applicable for predicting plant phenology in the tropics (Borchert et al., 2005; Gazal et al., 2008) and that phenology in tropical biomes may fail to be a useful indicator of global warming (Borchert et al., 2005). Does this in turn also apply for the estimation of temperature distributions at the local scale?

This study incorporated one temperate mid-latitude city (Munich, Germany) and one tropical city (Campinas, Brazil) of almost the same size, in each of which a broad-scale network was installed measuring air temperature and humidity at the observed sites. Using consistent phenological observation methods, an equivalent study design and meteorological data collection, we address these major research questions:

  1. Do the two cities show distinctive UHI effects?
  2. How suitable are the selected phenophases and species – especially in the tropics – for detecting differences in urban–rural temperature at the local scale?
  3. Are the urban index, an estimate for the degree of urbanization, as well as relative and absolute humidity valuable explanatory variables in urban phenology?

2. Data and methods

2.1. Study area

2.1.1. Geographical location

Campinas (22°54′S, 47°3′W) is located in the Piracicaba river basin in the state of São Paulo, Brazil (Figure 1), on a plateau at about 685 m altitude, near to the Serra do Japi mountains. Munich (48°8′N, 11°35′E) is located in southern Bavaria, Germany (Figure 1), on the Isar river north of the Bavarian Alps at an altitude of around 515 m. Population sizes are 1.38 million (Munich) and 1.08 million inhabitants (Campinas). However, the cities' structures differ considerably. Although Campinas' inner city is dominated by numerous multi-storey buildings, Munich's architecture is characterized by only a few tall buildings higher than 100 m. Conversely, a number of green open areas can be found in both cities. For details of land use, see Figure 2.

Figure 1.

Location of (a) the study sites (black dots) in Brazil and Germany, (b) Campinas (22°54′S, 47°3′W), São Paulo State, Brazil and (c) Munich (48°8′N, 11°35′E), Bavaria, Germany (country boundaries: ESRI, 2011).

Figure 2.

The cities of (a) Campinas, background: land cover (INPE, Brazilian National Institute for Space Research, Vieira et al., 2010, see also: http://urlib.net/8JMKD3MGP7W/36QPBQ5), major classes: red = urban fabric, orange = sugar cane, light green = pasture and dark green = eucalyptus forest remnants, and (b) Munich, Germany, background: land cover (CORINE Land Cover 2006, EEA, 2010, see also: http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2006-raster for a complete legend), major classes: red = urban fabric, green = forest and pastures, yellow = arable land and blue = rivers and lakes. Black and grey dots represent the studied urban and rural sites within each city, respectively.

2.1.2. Climate (1971–2000)

Campinas has a seasonal tropical climate (Köppen type Cwa); it is characterized by a warm and wet season from October to March and a cold and comparably drier season from April to September, and an annual mean precipitation of 1410 mm. During the dry season, only 25% of the annual rainfall is received and the average precipitation during the driest month (August) is 33 mm. Maximum mean monthly rainfall occurs in January (250 mm). Mean annual temperature is 21.9 °C (minimum in June: 18.4 °C and maximum in February: 24.6 °C) (data source: IAC, Instituto Agronômico).

Munich is characterized by a warm temperate climate (Köppen type Cfb). The annual mean temperature is 9.5 °C with an average of 0.3 °C in the coldest month (January) and 18.9 °C in the warmest month (July). Annual precipitation averages 954 mm, with most of the rain occurring during summer and a maximum of 125 mm in July. The winter is drier with a minimum of 46 mm in January (data source: DWD, German Meteorological Service).

2.2. Phenological observations

2.2.1. Selected plants

The Brazilian study was based on flowering and leaf unfolding of three different species. As the suitability of trees for urban climate studies in the tropics has not been investigated yet, we selected species that are abundant in urban areas: Tabebuia chrysotricha (Mart. Ex DC.) Standl., known as yellow trumpet tree or ipê, is a deciduous tropical tree species of the Bignoniaceae family and native to Brazil. It is the national flower of Brazil and often used as an ornamental tree in urban settlements, parks and along streets (Souza et al., 2005). Caesalpinia peltophoroides Benth., also called sibipiruna or false Brazilwood, is a legume tree of the Fabaceae family. It originates from Brazil and is often used as an ornamental tree in cities (Corte et al., 2008). Tipuana tipu (Benth.) Kuntze, commonly called tipu tree or racehorse tree, is a large deciduous to semi-deciduous tree belonging to the Fabaceae family and originates from Argentina and Bolivia. Tipuana is widely planted in urban areas, mainly in southern Brazil (dos Santos Pereira and de Aquino-Neto, 2003). Because rural areas outside Campinas are mostly occupied by sugar cane or pasture, these three tropical tree species are mainly restricted to settlements.

As opposed to the tropics, a greater knowledge exists about species that are suitable for detecting UHIs in temperate cities (Baumgartner, 1952; Bernhofer, 1991) and especially spring phenophases (leaf unfolding and flowering) were found to be most sensitive to temperature (Menzel et al., 2006). For the German study area, we selected leaf unfolding and flowering of only one species, Betula pendula Roth (silver birch). It is a deciduous tree of the Betulaceae family and a pioneer plant that is widespread across Europe and can be frequently found both in urban and rural areas (Aas, 2000). Because of the allergenicity of its pollen, the tree has gained strong interest in studies related to phenology (Siljamo et al., 2008; Jochner et al., 2011), aerobiology (Emberlin et al., 2002; Jochner et al., 2012b) and medicine (Traidl-Hoffmann et al., 2003; Bryce et al., 2010).

2.2.2. Observation methods

The observation sites in both cities were chosen along urban–rural gradients within a ∼100-m total elevation range to avoid altitudinal and associated temperature effects on phenology. The selection of sites was mainly based on the occurrence of trees (two to nine individuals required) and depended on, e.g. free accessibility or an authorization from the property owner. We observed 83 Tabebuia trees at 19 sites, 101 Caesalpinia trees at 28 sites and 68 Tipuana trees at 17 sites in Campinas (Figure 2(a)). In Munich, we observed 130 Betula trees at 38 sites (Figure 2(b)).

To ensure consistent phenological observation methods we used the BBCH code (Meier, 2001). Initially developed for development stages of agricultural plants by the Biologische Bundesanstalt, Bundessortenamt and CHemical industry, the extended BBCH scale for monocotyledonous and dicotyledonous plant species allows the assessment of not only principal growth stages (e.g. flowering and leaf unfolding) but also short developmental steps (sub-stages and secondary growth stages) that are passed successively within the development process (e.g. bud shows green tip, inflorescence or flower buds visible). In Section 'Results' for clarity, however, we only present the data for BBCH 61 (beginning of flowering: 10% of flowers open/emitting pollen), BBCH 65 (full flowering: >50% of flowers open/emitting pollen, first petals falling), BBCH 10 (mouse-ear stage: green leaf tips 10 mm above the bud scales) and BBCH 15 (>50% leaves unfolded).

Observations were carried out by only one person per city in order to reduce subjective variation in assessing BBCH development stages and were repeated every third day. The observation period in Munich lasted from the end of March to mid-April 2011 and in Campinas from mid-July to the beginning of October 2011.

2.2.3. Meteorological measurements

One individual tree per site (Campinas: n = 31 and Munich: n = 38) was equipped with an air temperature and humidity sensor (HOBO U23-001, Onset Computer Corporation, Southern MA, USA). The loggers were mounted in a radiation shield on tree trunks at 3 m height to minimize theft or vandalism because almost all sites were communal or municipal and therefore open to the public. In both Munich and Campinas, one logger was stolen and not replaced during the observation period. Loggers were mounted on the northern (Munich) or southern (Campinas) side of trees. However, for security and safety reasons, about half of the loggers in Campinas could not be placed on this orientation prior to the end of September – until leaf unfolding progressed and acted as visual cover. To avoid possible problems, radiation shields were provided with slots allowing for ventilation and we additionally included mean nighttime temperatures in statistical analyses. Loggers in Campinas were sequentially installed: 18 in August, 28 in September and a further three loggers were mounted in October.

2.2.4. Auxiliary data

The sites in Munich were categorized as urban or rural using CORINE Land Cover (CLC) 2006 data (EEA, 2010) with a spatial resolution of 100 m. CLC data consist of 44 land cover classes, which were grouped into five main categories (artificial surfaces, agricultural areas, forest and semi-natural areas, wetlands and water bodies). Using ArcGIS 10 (ESRI, 2009, Redlands, CA, USA) we extracted particular elements of artificial surfaces (e.g. continuous and discontinuous urban fabric, industrial and commercial units) that are characterized by a high degree of impervious surfaces to create an ‘urban layer’. Because CLC data only cover European countries we applied land use data of the Brazilian National Institute for Space Research (INPE, Vieira et al., 2010) with a spatial resolution of 30 m for Campinas and extracted areas that were already classified as ‘urban’.

Following Jochner et al. (2012a), we calculated an index describing the degree of urbanization (ui = urban index) for each site using the proportion of urban land use within a radius of 2 km. The sites were classified as urban when the ui-values were greater than 0.5, otherwise as rural. Black and grey dots in Figure 2(a) and (b) indicate the sites characterized as urban and rural, respectively.

2.2.5. Statistical methods

We derived daily mean and maximum temperatures (Tmean and Tmax) as well as mean nighttime temperatures [Tnight; calculated using data from 6:00 pm to 6:00 am (Campinas) or 7:00 pm to 7:00 am (Munich)]. The diurnal temperature range (DTR) was calculated as the difference between the maximum temperature and the minimum temperature of 1 d. In addition to relative humidity [RH (%)], we also included absolute humidity [a (g m−3)] calculated using Equations (1)–(3),

display math(1)
display math(2)
display math(3)

where e is the water vapour pressure (hPa), RH is the relative humidity (%), E is the saturation water vapour pressure (hPa) and T is the temperature (°C).

The overall range of meteorological variables (range) is the difference between the warmest and coldest/most humid and least humid station. We also calculated differences between urban and rural areas for meteorological and phenological data where positive temperature/moisture differences indicate higher temperatures/humidity in the city. The UHI effect refers to differences in urban and rural temperature. Negative phenological differences – called the urban phenology effect (UPE) – indicate an earlier onset in the city compared with the countryside and vice versa.

Air temperature, humidity and phenological data were tested for differences between urban and rural areas using two-sample t-tests. We calculated Pearson's correlations to examine relationships between phenological onset dates, the urban index and meteorological variables. We used temperature and humidity data of the months before mean onset of the respective phenophases (e.g. see missing cells in Table 4).

All statistical analyses were performed using SPSS Statistics 19 (IBM SPSS, Chicago, IL, USA).

3. Results

3.1. Climate

Table 1 shows meteorological data for urban and rural areas of Campinas in August and September 2011 and for Munich in March 2011. The UHI effect as shown by mean temperatures in Campinas was small and not significant for either month. The overall range of mean monthly temperatures showed little variation in August (0.9 °C) and September (1.1 °C). In both months differences between urban and rural areas were only significant for mean nighttime temperatures (0.7 and 0.6 °C, respectively) and DTR (−1.7 and −1.9 °C, respectively). Relative and absolute humidity showed significant differences between urban and rural areas in September.

Table 1. Mean temperature (°C), humidity variables and respective standard deviation between sites for selected months in urban and rural areas in Campinas (August = 8 and September = 9) and Munich (March = 3) in 2011
 UrbanRuralDifferenceOverallRange
  1. Tm, mean temperature; Tnight, mean nighttime temperature; Tmax, maximum temperature; DTR, diurnal temperature range; RH, relative humidity (%); a, absolute humidity (g m−3); difference, urban–rural difference; range, maximum value minus minimum value in the study area. Significance, bold values: significant urban–rural differences at the 5% level; bold and italic value: significant urban–rural differences at the 10% level.

Campinas     
  Tm8 (°C)20.7 ± 0.320.5 ± 0.20.220.6 ± 0.30.9
  Tm9 (°C)21.6 ± 0.321.6 ± 0.30.021.6 ± 0.31.1
  Tnight8 (°C)18.6 ± 0.517.9 ± 0.50.718.4 ± 0.62.3
  Tnight9 (°C)19.2 ± 0.418.6 ± 0.50.619.0 ± 0.52.1
  Tmax8 (°C)26.1 ± 0.725.6 ± 0.50.525.9 ± 0.72.2
  Tmax9 (°C)27.1 ± 0.426.7 ± 0.70.427 ± 0.62.2
  DTR8 (°C)11.3 ± 1.013.0 ± 1.0−1.711.8 ± 1.24.7
  DTR9 (°C)12.8 ± 0.914.7 ± 1.1−1.913.4 ± 1.35.1
  RH8 (%)62.3 ± 2.364.2 ± 2.4−1.962.9 ± 2.59.0
  RH9 (%)54.9 ± 1.856.8 ± 1.8−1.955.5 ± 1.97.9
  a8 (g m−3)10.7 ± 0.310.9 ± 0.3−0.210.7 ± 0.30.9
  a9 (g m−3)9.9 ± 0.210.1 ± 0.2−0.29.9 ± 0.30.9
Munich     
  Tm3 (°C)6.1 ± 0.55.1 ± 0.41.05.7 ± 0.62.5
  Tnight3 (°C)4.0 ± 0.72.6 ± 0.61.43.6 ± 0.93.8
  Tmax3 (°C)11.7 ± 0.711.3 ± 0.5 0.411.6 ± 0.62.6
  DTR3 (°C)10.6 ± 0.211.8 ± 0.3−1.211 ± 1.14.6
  RH3 (%)71.0 ± 2.576.8 ± 2.2−5.872.9 ± 3.714.7
  a3 (g m−3)5.1 ± 0.15.2 ± 0.1−0.15.2 ± 0.10.5

In contrast, we found distinctive urban–rural differences in March temperature for Munich (Table 1). The UHI effect was significant at the 5% level: 1.0 °C for mean temperatures, 1.4 °C for nighttime temperatures, −1.2 °C for DTR, −5.8% for RH and −0.1 g m−3 for absolute humidity. Compared to Campinas, the range of mean temperature as well as of nighttime temperature was higher, reaching 2.5 and 3.8 °C, respectively.

Correlations between the urban index and temperature/humidity variables are summarized in Table 2. For Campinas the highest correlations with urban index were for DTR (r ∼ −0.8) and mean nighttime temperature (r ∼ 0.75). Maximum temperature, RH and absolute humidity (September only) also had relatively strong correlations (all |r| > 0.5) with urban index. However, mean monthly temperatures were not strongly correlated with urban index: the correlation coefficients were either low (August and September) and/or not significant (September).

Table 2. Pearson's correlation coefficients (r) for urban index (ui) and temperature and humidity variables for selected months in Campinas (August = 8 and September = 9) and Munich (March = 3) in 2011
  1. Tm, mean temperature; Tnight, mean nighttime temperature; Tmax, maximum temperature; DTR, diurnal temperature range; RH, relative humidity; a, absolute humidity; n, number of sites. Significance, bold values: significant coefficients at the 5% level; bold and italic value: significant coefficients at the 10% level.

Campinas Tm8Tnight8Tmax8DTR8RH8a8
uir0.4860.7500.534−0.825−0.599−0.463
n181818181818
Campinas Tm9Tnight9Tmax9DTR9RH9a9
uir0.1910.7560.633−0.794−0.678−0.623
n272727272727
Munich Tm3Tnight3Tmax3DTR3RH3a3
uir0.7980.7670.348−0.567−0.839−0.560
n373737373737

The urban index in Munich was most strongly correlated with RH (r = −0.84), mean temperature (r = 0.8) and mean nighttime temperatures (r = 0.77) in March. All other correlations were also significant.

3.2. Phenological onset dates in 2011

The mean onset dates and standard deviations (SDs) of flowering and leaf unfolding phenophases for the four studied species in both urban and rural areas are summarized in Table 3. Campinas exhibited large SDs between site means of 8.9 (Caesalpinia: mouse-ear stage) and 17.4 d (Tipuana: beginning of flowering) (Table 3). However, even trees within one site exhibited a large asynchrony in flowering and leaf unfolding: mean SD ranged between 12.1 and 15.7 d for Tabebuia, between 9.4 and 12.9 d for Caesalpinia and between 7.5 and 14.1 d for Tipuana (data not shown). The highest SD for one site (with n = 6) was 36.1 d for full flowering of Tabebuia. In contrast, the results for Betula in Munich were characterized by relatively low SDs between 1.6 (mouse-ear stage) and 3.0 d (full flowering, Table 3). At single Munich sites, mean SD did not exceed 2.4 d for all analysed phenological phases (data not shown).

Table 3. Mean (μ) onset dates (DOY, day of the year), standard deviations (SDs) between sites for flowering and leaf unfolding phenophases of Tabebuia, Caesalpinia and Tipuana in Campinas and Betula in Munich, 2011
Species (city) nBBCH 61BBCH 65BBCH 10BBCH 15
μ DateSDμ DateSDμ DateSDμ DateSD
  1. BBCH 61, beginning of flowering; BBCH 65, full flowering; BBCH 10, mouse-ear stage; BBCH 15, >50% of leaves unfolded; n, number of sites.

Tabebuia (Campinas)Overall19230.0

18.08.

13.0242.7

31.08.

10.7225.5

13.09.

13.3271.8

29.09.

13.6
 Urban14232.5

21.08.

13.8244.0

01.09.

11.8257.1

14.09.

14.5271.4

28.09.

15.0
 Rural5223.2

11.08.

7.9239.1

27.08.

6.2251.3

08.09.

9.5272.9

30.09.

9.8
Caesalpinia (Campinas)Overall28246.2

22.09.

11.5272.2

29.09.

10.6240.1

28.08.

8.9248.2

05.09.

10.4
 Urban19265.7

23.09.

10.6274.1

01.10.

8.8240.8

29.08.

8.9249.2

06.09.

10.5
 Rural9261.2

18.09.

13.4268.4

25.09.

13.5238.6

27.08.

9.3246.1

03.09.

10.6
Tipuana (Campinas)Overall17264.9

22.09.

17.4280.0

07.10.

14.6235.0

23.08.

10.3241.7

30.08.

10.9
 Urban12259.1

16.09.

16.0276.3

03.10.

14.4241.2

29.08.

7.6238.6

27.08.

10.7
 Rural5278.8

06.10.

12.8288.8

16.10.

11.9232.4

20.08.

10.5246.1

06.09.

8.0
Betula (Munich)Overall3898.6

09.04.

2.1101.0

11.04.

3.094.7

05.04.

1.6101.0

11.04.

2.5
 Urban2597.8

09.04.

1.599.8

10.04.

2.194.2

04.04.

1.3100.5

11.04.

2.3
 Rural13100.1

10.04.

2.3103.2

13.04.

3.395.7

06.04.

1.8102.0

12.04.

2.7
Table 4. Pearson's correlation coefficients (r) for flowering and leaf unfolding phenophases of Tabebuia, Caesalpinia and Tipuana in Campinas and Betula in Munich, 2011
TabebuiauiTm8Tnight8Tmax8DTR8RH8a8Tm9Tnight9Tmax9DTR9RH9a9
BBCH 61r0.284−0.1990.1730.178−0.1920.012−0.043 
n19141414141414
BBCH 65r0.162−0.1670.1740.068−0.2000.0810.077
n19141414141414
BBCH 10r0.120−0.591 −0.533 −0.5850.2340.8120.803−0.0900.180−0.3220.1640.1680.152
n18131313131313161716161616
BBCH 15r−0.112−0.455−0.557−0.5700.3710.7100.7520.0460.187−0.3440.4000.2710.348
n19141414141414171716161616
CaesalpiniauiTm8Tnight8Tmax8DTR8RH8a8Tm9Tnight9Tmax9DTR9RH9a9
BBCH 61r0.215−0.385−0.0600.014−0.1070.1760.088−0.061−0.2950.0400.0690.010−0.049
n28181818181818262625252525
BBCH 65r0.273−0.212−0.0200.176−0.0080.049−0.042−0.017−0.2930.1310.041−0.080−0.152
n28181818181818262625252525
BBCH 10r0.109−0.157−0.0520.1640.071−0.006−0.100 
n28181818181818
BBCH 15r0.144−0.203−0.0280.145−0.0040.001−0.092
n28181818181818
TipuanauiTm8Tnight8Tmax8DTR8RH8a8Tm9Tnight9Tmax9DTR9RH9a9
BBCH 61r−0.540−0.618−0.632−0.5820.5010.5590.463−0.114−0.578−0.3650.5820.5630.607
n17888888151414141414
BBCH 65r−0.388 −0.695 −0.624 −0.5500.3850.5450.393−0.234−0.590−0.2970.5480.4370.335
n17888888151414141414
BBCH 10r−0.547−0.727−0.714 −0.638 0.564 0.674 0.553 
n17888888
BBCH 15r−0.561−0.746−0.779 −0.700 0.655 0.7270.612
n17888888
BetulauiTm3Tnight3Tmax3DTR3RH3a3      
  1. ui, urban index; Tm, mean temperature; Tnight, mean nighttime temperature; Tmax, maximum temperature; DTR, diurnal temperature range; RH, relative humidity; a, absolute humidity, in March (=3), August (=8) and September (=9); BBCH 61, beginning of flowering; BBCH 65, full flowering; BBCH 10, mouse-ear stage; BBCH 15, >50% of leaves unfolded; n, number of sites; P, significance, bold values: significant coefficients at the 5% level; bold and italic values: significant coefficients at the 10% level.

BBCH 61r−0.574−0.742−0.752−0.1630.7090.6370.124 
n38373737373737
BBCH 65r−0.585−0.739−0.749−0.1940.6790.6730.219
n38373737373737
BBCH 10r−0.404−0.571−0.609−0.0300.6680.4730.031
n38373737373737
BBCH 15r−0.288−0.482−0.450−0.2590.3240.4120.095
n38373737373737

Significant urban–rural differences (UPE) were found for all phenophases of Betula in Munich and ranged between −1.5 (>50% leaves unfolded) and −3.4 d (full flowering), indicating an earlier onset in urban areas (Figure 3). In Campinas, UPE of Tabebuia varied between −1.5 (>50% leaves unfolded) and +9.3 d (beginning of flowering). However, these differences were not statistically significant. This was also true for Caesalpinia where differences ranged between +2.2 and +5.7 d. Significant differences were only found for flowering onset of Tipuana (−19.7 d) and for >50% leaf unfolding of the same species (−10.5 d, significant at the 10% level).

Figure 3.

Urban–rural differences in phenology (UPE, urban phenology effect, in days) and corresponding standard error for flowering and leaf unfolding phenophases of Tabebuia, Caesalpinia and Tipuana in Campinas and Betula in Munich, 2011; BBCH 61, beginning of flowering; BBCH 65, full flowering; BBCH 10, mouse-ear stage; BBCH 15, >50% of leaves unfolded. Significant at the ***0.1% level, **1% level, *5% level and (*)10% level.

3.3. Explanatory variables in urban phenology

Correlation coefficients between phenological onset dates of the three selected species and urban index, temperature and humidity in Campinas are summarized in Table 4. There were no significant correlations with flowering phenophases of Tabebuia. However, leaf phenophases revealed some strong correlations with mean, nighttime and maximum temperatures in August (all r < −0.5). The highest correlations were obtained with relative and absolute humidity in August (r > 0.7). Partial correlation analyses with mean temperature as a control variable underlined the significant influence of humidity variables on leaf phenophases (all r > 0.65, data not shown). For Caesalpinia there were no significant correlations. In contrast, there were several significant correlations for flowering and leaf unfolding phenophases of Tipuana. The beginning of flowering, e.g. was correlated with DTR (r = 0.58), mean nighttime temperature and humidity-related variables in September (all |r| > 0.5). For full flowering, correlations with relative and absolute humidity were less significant. We detected strong correlations for leaf unfolding phenophases and mean and nighttime temperatures (all r < −0.7). For almost all phenophases (except full flowering) statistically significant correlations with the urban index were r < −0.5.

For Betula phenology there were strong correlations with nighttime and mean temperatures in March, especially for the beginning of flowering and for full flowering (all r < −0.7), as well as with DTR (Table 4). In contrast, we did not find any significant correlations with maximum temperature or absolute humidity. The urban index was negatively correlated with onset dates and was again stronger for flowering phenophases (r < −0.57). RH was more strongly correlated with flowering phenophases of Betula than with leaf unfolding phenophases. This, however, appears to be attributable to the high correlation between RH and mean temperature (r = −0.94) because partial correlations with mean temperature as the control variable did not show any significant correlations (data not shown).

4. Discussion

4.1. UHI effect

We did not find a significant UHI in the tropical city of Campinas, contrasting with the more distinct urban–rural temperature differences in Munich, showing a mean UHI of 1.0 °C. In addition, differences were more pronounced in Munich than in Campinas for DTR, nighttime temperatures and for the absolute temperature ranges between all sites. The fact that UHI is greater for minimum than for maximum temperatures is well documented in existing literature (Landsberg, 1981; Baker et al., 2002; Mimet et al., 2009; Shustack et al., 2009), and was also shown in our study of Campinas and Munich. Furthermore, relative and absolute humidity were significantly lower in urban than rural Munich (−5.8% and −0.1 g m−3, respectively). Campinas, however, was characterized by a smaller urban–rural difference in relative and absolute humidity.

The question arises whether these negligible differences within the study area in Brazil are linked to the site selection criteria. Campinas' most rural site exhibited an urban index (ui) of 0.156 compared with 0.003 in Munich, with five additional Munich sites having ui-values smaller than 0.010. In addition, the maximum distance to the city center was 12.7 km in Campinas compared with 28.2 km in Munich. Therefore, we suggest that the dominance of sites with a higher degree of urbanization might have influenced temperature variations/ranges in Campinas. However, existing studies also demonstrated minor UHI effects in (sub)tropical cities (e.g. reviewed by Roth, 2007).

In both cities the urban index was a good predictor of local temperature and humidity conditions. The only exception was mean temperature in Campinas, that only had a correlation of r = 0.49 in August and no significant correlation in September. Nevertheless, we propose that an easily computable urban index might be useful for a quick and inexpensive estimation of the spatial structure of UHIs.

4.2. Urban–rural differences in phenology

In contrast to small SDs of Betula onset dates in Munich of between 1.6 (mouse-ear stage) and 3.0 d (>50% leaves unfolded), we observed high SDs in onset dates for the three selected species in Campinas ranging between 8.9 (Caesalpinia: mouse-ear stage) and 17.4 d (Tipuana: beginning of flowering). A high variation in onset dates of phenological phases in the tropics was also reported by Gazal et al. (2008) for three tropical cities in Asia and Africa and for tropical wet and dry forest trees (Morellato et al., 2000; Borchert et al., 2005). The high within-species variation in tropical tree phenology might partly explain small urban–rural differences. We only found significant UHI-related differences for Tipuana in Campinas, whereas almost all phenophases of Tabebuia and all of Caesalpinia had small positive urban–rural differences (albeit not significant) that indicated an earlier onset in the countryside. Gazal et al. (2008) reported earlier bud burst dates only in the tropical city of Bangkok (−23 d), but not in Korat (+9 d) or Dakar (+9 d). In addition, bud burst was delayed with increasing land surface temperature, suggesting a low temperature sensitivity, or other influencing environmental variables (e.g. humidity-related variables).

In contrast to our findings for Campinas, we detected significant differences between onset dates for Betula phenophases in Munich ranging between −1.5 (mouse-ear stage) and −3.4 d (full flowering). An urban–rural comparison in Berlin, Germany, conducted by Henniges and Chmielewski (2006) also revealed greater differences for Betula flowering (−2.6 d) compared with Betula leaf unfolding (−0.3 d). Hence, this result is in accordance with our findings: Betula flowering phenophases are more responsive to temperature variations than leaf unfolding phenophases.

In Campinas, the vertical structure of the city is much more pronounced than in Munich. Zhang et al. (2004) concluded that UHI effects on plant phenology are stronger in North America than in Europe or Asia owing to the dense, vertical urban design in North American cities. However, this cannot be tested for our European and South American study sites because species were not identical. We only found a distinctive high urban–rural difference of −10.5 to −19.7 d in Tipuana spring phenology that was much higher than the difference of −1.5 to −3.4 d in Betula phenology.

4.3. Air temperature, humidity and the phenology of trees in urban areas

We demonstrated the suitability of Betula phenophases in urban climatology applications by high correlations between phenological onset dates and DTR (exception > 50% leaves unfolded). The DTR is a good indicator for UHIs as it is smaller in the city because of the thermal energy storage of urban constructions (Landsberg, 1981). Mimet et al. (2009) also reported that the DTR is most strongly correlated with budburst onset dates of sour cherry in the city of Rennes, France. In general, UHI is clearer in minimum temperatures (Baker et al., 2002; Mimet et al., 2009; Shustack et al., 2009). Therefore, we also found slightly stronger correlations with mean nighttime temperature than with mean temperature (exception > 50% leaves unfolded) and no statistically significant correlation with maximum temperature. This is in accordance with Wielgolaski (1999) and Mimet et al. (2009) who stated that maximum temperature does not play an important role in phenological models. Furthermore, particularly for flowering phenophases of Betula in Munich, there was a strong and evident relationship between urban index and onset dates.

In Campinas, however, the significance and magnitude of the difference in onset dates varied considerably among species and phenophases. The most promising species for urban phenology applications was Tipuana as shown by high correlations, especially with mean and nighttime temperatures, and the significant relationship with the urban index for three of the four selected phenophases. Caesalpinia failed to show any significant correlations with the selected environmental variables and was thus not suitable to demonstrate temperature variations at the local scale associated with the UHI effect. This also applied for flowering phenophases of Tabebuia, although leaf unfolding phenophases were particularly sensitive to humidity and also to air temperature.

The influence of air humidity in the study area of Munich was negligible because absolute humidity was not correlated with phenological onset dates of Betula and partial correlation analyses with temperature as a control variable showed no significant correlations with RH.

On the other hand, our results suggest that humidity influences leaf phenology of Tabebuia. The influence of humidity on the other phases and species, however, was not that marked. The only other study comparing urban and rural phenology of tropical tree species also identified the potential importance of humidity (Gazal et al., 2008). Leafing phenology of tropical trees under a seasonal climate is driven mostly by precipitation and also by non-climatic parameters such as leaf longevity, water stress and increasing day length (Morellato et al., 2000; Borchert et al., 2005; Staggemeier and Morellato, 2011). Therefore, a consideration of humidity variables is especially recommended to understand urban effects in tropical cities under climates with dry and wet seasons.

5. Conclusions

We propose that phenological observations of Tipuana, a tree species widely used for planting in South American cities, are useful for a quick and inexpensive estimation of the spatial structure of UHIs in neotropical cities. Facing the high number of different species in tropical biomes, the need for further investigations becomes particularly evident. In the temperate/moist regions of Europe, possible species have already been tested for their suitability in urban applications, and greater knowledge in this respect exists (Baumgartner, 1952; Bernhofer, 1991). Consequently, additional species and phenophases should be explored to further identify the potential of phenology to monitor heat islands in tropical cities.

Acknowledgements

The research conducted in Germany was supported by the grant ME 179/3-1 of the Deutsche Forschungsgemeinschaft (DFG) and the EUROPA MÖBEL-Umweltstiftung; the research conducted in Brazil by CAPES/PROBAL (360/11), FAPESP (Fundação de Apoio à Pesquisa do Estado de São Paulo, 2009/54208-6), CNPq (Coordenação Nacional de Pesquisa) and DAAD (German Academic Exchange Service, 50752579). MAE received a master and BEPE fellowship from FAPESP; LPCM received a Research Productivity fellowship and grant from CNPq. We are grateful to Milton Ribeiro from the Ecology Department (Landscape Ecology, UNESP) for the acquisition of digital land use data. We thank our drivers Yuri Brenn, Amanda Alfonso Batista and Fernanda Zambonini (Campinas).

Ancillary