Site description, air pollution levels, and tree species
Kyoto City, Japan, is located in the Kyoto basin surrounded by mountains to the east, north, and west. The mean temperature, relative humidity, and cumulative precipitation during the growing season, from April to November, in an average year (1981–2010) is 20.7°C, 66.0%, and 1210 mm, respectively. Concentrations of air pollutants have been monitored continuously by the Kyoto government at 14 monitoring stations in Kyoto City since 1970, in which the data are obtained every hour (http://www.nies.go.jp/igreen/td_down.html). We selected 21 study sites for collecting branches from roadside trees (Fig. 1a). The traffic data of 12 h of daytime were obtained for the years 2005, 2010, and 2015 from the Ministry of Land, Infrastructure, Transport and Tourism. The data were measured once a year at about 2000 sites in Kyoto Prefecture (https://www.mlit.go.jp/road/census/h17, h22-1, h27). We classified the 14 monitoring stations into three levels of pollution, low, medium, and high, using cluster analysis. The levels of air pollution of the study sites were determined by applying the pollution levels of the nearest monitoring stations.
We used four major tree species planted as roadside trees in Kyoto Prefecture in this study. The tree species used were two evergreen shrub trees, Rhododendron × pulchrum Sweet 'Oomurasaki' (R. pulchrum, azalea) and Rhaphiolepis indica (L.) var. umbellata (R. indica, yeddo hawthorn), and the two broadleaved deciduous tall trees, G. biloba L. (G. biloba, ginkgo) and Prunus × yedoensis Matsumura (P. yedoensis, cherry). R. pulchrum was the most common shrub species (62.7 × 106, 45% of the shrub trees) planted in Japan in 2017, followed by R. indica (9.7 × 106, 6.9%). G. biloba was the most common tall tree species (5.5 × 105, 8%), followed by P. yedoensis (5.2 × 105, 7.8%, Iizuka and Funakubo 2018). The leaves at the top of the crowns of R. pulchrum and R. indica trees are considered annual leaves after July because they are pruned annually between May and June, and after that, new annual leaves then subsequently expand. G. biloba and P. yedoensis trees are neither watered nor pruned during the growing season by the Kyoto government. The phenology of G. biloba and P. yedoensis is such that the leaves expand in early April, fully expand in May, begin senescence in late November, and fall in December.
Leaf gas exchange measurements and photosynthetic limitations
The study sites were selected such that they had as wide a range of traffic volumes as possible within one species in one sampling year. The longitude, latitude, and altitude of the study sites were 135° 42' − 49', 34° 56' − 35° 03', and 20–80 m, respectively. We selected 4–5 trees at each study site, at which the mean tree heights were about 0.5–1 m for R. pulchrum and R. indica, and 5–7 m for G. biloba and P. yedoensis. Then, the branches of the roadside trees were collected from late May to early November. Sunlit healthy branches were harvested in the morning, from the tops of the canopies for R. pulchrum and R. indica and from the sunlit crowns for G. biloba and P. yedoensis, avoiding rainy days. These branches were excised from each individual tree at each site (n = 4–5). The ends of the branches were immediately immersed in water, brought back to the laboratory, and then photosynthesis measurements were performed for the healthy fully expanded mature leaves within 10 h. Leaf photosynthesis was measured using a photosynthesis system constructed in our laboratory (in 2007, Kagotani et al. 2013) or using Li-6400XT (Li-Cor, Lincoln, NE, USA, from the years 2014 to 2019), which had the same performance for leaf photosynthesis measurement. We previously confirmed that the leaf photosynthetic parameters of these four species were unchanged for 10 h after excision of the branches of the four tree species (p > 0.1, n = 4).
We obtained light-saturated photosynthesis (Amax) and stomatal conductance (gs) at ambient CO2 of 400 µmol mol− 1 and at 1000 µmol mol− 1 of PPFD (in 2007) or at 1500 µmol mol− 1 of PPFD (from 2014–2019). We confirmed that leaf photosynthesis was almost light-saturated at 1000 µmol mol− 1 for the four species, where the ratio to the maximum value was 94–95%. The photosynthetic biochemical parameters, such as the maximum carboxylation rate (Vcmax), were obtained from A/Ci curves (Ethier and Livingston 2004) using Li-6400XT (Kiyomizu et al. 2019; Kinoshita et al. 2021) for the years 2014 to 2019, where Ci means intercellular CO2. The measurements were conducted under 1500 µmol mol− 1 of PPFD. In these photosynthesis measurements, the leaf temperature and vapor pressure deficit (VPD) were set at 26–28°C and 1.2–1.7 kPa, respectively. The stomatal and non-stomatal limitations were calculated using A/Cc curves obtained from the A/Ci curves, following a previous study (Grassi and Magnani 2005), where Cc means chloroplast CO2.
We measured the SPAD values as an indicator of leaf chlorophyll content (Parry et al. 2014), because high levels of air pollutants cause a decline in the leaf chlorophyll content (Joshi and Swami 2009), which is related to the photosynthetic biochemistry. SPAD values were evaluated for the leaves on which leaf gas exchange measurements were conducted, as mean values of the three replicated measurements. The SPAD value was measured using a SPAD meter (SPAD-502 Plus, Konica-Minolta, Tokyo, Japan).
The difference in annual climate should affect absolute values of leaf photosynthesis. To evaluate the effect of air pollution on the limitations of photosynthesis across the different years, photosynthesis values need to be ‘normalized’ using a reference of each year. We used the maximum value of Amax of each year for each species as a reference value (Aref), following a previous study (Grassi and Magnani 2005). The restriction in Amax in each year can then be estimated as follows:
Restriction in Amax = (Aref – Aobs) / Aref (1),
where Aobs is the observed Amax.
Carbon isotope discrimination
After the photosynthesis measurements were finished, the leaves were dried at 60–70°C for 24 h using an oven (MOV-112, SANYO Electric, Osaka, Japan), and were stored in a desiccator at room temperature to analyze the leaf stable carbon isotope ratio (δ13C). Leaf carbon isotope discrimination, Δ13C, is a good indicator of long-term leaf water use efficiency (Broeckx et al. 2014), which can be used for the evaluation of air pollution effects (Kagotani et al. 2013). Δ13C is determined as:
Δ13C (‰) = (δa – δp)/(1 + δp/1000) (2),
where δa is the δ13C of atmospheric CO2 and δp is leaf δ13C. We collected air samples of the study sites from 2007 to 2019 to obtain δa, which depends on the air pollution level, and was used to calculate Δ13C. Air samples of 2 liters near the tree canopies were collected in aluminum air bags using an air pump, and then δa was measured using an isotope mass spectrometer (Finnigan MAT 252, Thermo Fisher Scientific, MA, USA) or an isotope analyzer (G2101-i, Picarro, Inc., Santa Clara, CA, USA). The averaged values of δa were − 11.0 (0.3) ‰, − 11.7 (0.4) ‰, and − 12.0 (0.4) ‰ for the low-, medium-, and high-pollution sites, respectively (mean (se), n = 14–23). In the years 2007 and 2014, δp was measured using powdered leaf samples using the combined system of an elemental analyzer (EA1108, Carlo-Erba, Italy), interface (Conflo II, Thermo Fisher Scientific, MA, USA), and isotope mass spectrometer (Delta S, Thermo Fisher Scientific, MA, USA) at the Center of Ecological Research of Kyoto University corrected with standards (Tayasu et al. 2011). From 2015 to 2019, δ13C was measured using an elemental analyzer (Flash 1112, Thermo Fisher Scientific, MA, USA), interface (Conflo IV, Thermo Fisher Scientific, MA, USA), and isotope mass spectrometer (delta V advantage, Thermo Fisher Scientific, MA, USA) at the Research Institute for Humanity and Nature.
To compare Δ13C values across the different years, we calculated deviations in the Δ13C value of each year using the maximum value as a reference (Δ13Cref):
Deviations in Δ13C = Δ13C ref – Δ13Cobs (3),
where Δ13Cobs is the observed Δ13C. The larger deviations in Δ13C mean smaller Δ13Cobs, and in turn, higher long-term water use efficiency.
The two data sets used in this study
We used two sets of data for evaluation of the photosynthetic performance of urban trees. In the first data set, leaf gas exchange data for the four tree species G. biloba, P. yedoensis, R. pulchrum, and R. indica were used to obtain the species-specific response to air pollution. The photosynthesis measurements were performed in 2016 or 2017 for the leaves of four branches collected from the three - six study sites, in which each species had one - two study sites of low-, medium-, and high-level air pollution (n = 4–8). Using the first data set, we compared the restrictions of photosynthesis, stomatal and non-stomatal limitations, and SPAD values that reflect the leaf chlorophyll content among these three pollution levels.
In the second data set, we focused on the species R. pulchrum for the evaluation of photosynthesis and water use efficiency in response to atmospheric NO and NO2 concentrations. The variation ranges in annual atmospheric NO and NO2 concentrations from 2007 to 2019 were from 0.4 to 47 ppb and 6 to 37 ppb, which are larger than the ranges within a year, from 0.7 to 17 and 6 to 19 ppb, respectively (Japan National Institute for Environmental Studies, http://www.nies.go.jp/igreen/td_down.html). We can detect the photosynthetic response of roadside trees to a wider range of NO and NO2 by using annual averaged data. The annual averaged values for the restriction of Amax and deviations in Δ13C were calculated using the data for the 4–35 leaves that were collected in 2007 and 2014–2019. Part of the second data set, the data for the years 2014 and 2015, were redrawn from the published data of our previous study (Kiyomizu et al. 2019). We obtained the dependence of leaf traits on NO and NO2 concentration for the pooled data.
Finally, we made a map of photosynthetic restrictions and deviation in Δ13C of R. × pulchrum in Kyoto City, using the second data set, traffic volumes, and NO2 concentrations. First, the central part of Kyoto City, which is about 100 km2, including the study sites, was divided into areas of a 1 km2 mesh. We obtained 121 sections. Then, we obtained traffic volumes for each 1 km2 section using traffic volume data of the road census stations from 2015 (http://www.mlit.go.jp/road/census/). If multiple census stations were included in a 1 km2 section, the higher value was applied. Twenty three of the 121 sections (19%) contained no census station, so the value of the nearest station was applied to these sections. Then, the NO2 concentration of each section was estimated using regression between the 12-h daytime traffic volumes and NO2 as in Fig. 1e (n = 41):
NO2 (ppb) = 4.93 + 0.521 × 12-h daytime traffic volume (103 cars) (4).
Finally, restrictions of Amax and the deviation in Δ13C of each section were estimated using the correlations between NO2 and restriction of Amax or deviation in Δ13C in Fig. 3 (n = 17–19):
Restrictions of Amax (%) = 16.7 + 1.48 × NO2 (ppb) (5),
Deviation in Δ13C (‰) = 0.700 + 0.115 × NO2 (ppb) (6).
Statistical analysis
For the cluster analysis of the air monitoring stations, the data of nitrogen oxides (NO and NO2) and traffic data for the years 2005, 2010, and 2015 were used, in which the traffic data of the nearest sites were applied to the air pollution monitoring stations. The NO and NO2 data used for the analysis were monthly means from April to November (n = 8). For the cluster analysis, the data of each year were normalized so the means and standard deviations were 1 and 0, respectively. Differences in the photosynthetic traits among the three levels of pollution, low-, medium-, and high-level, were analyzed using analysis of variance (ANOVA) tested by Dunett’s test, in which the values of low-level were treated as a control. The absolute values for the photosynthetic traits of each year were tested by two-way ANOVA, to evaluate the effect of pollution level and species. The dose-dependent effects of NO and NO2 on some leaf traits were analyzed using regression analysis. These statistical analyses were performed using EZR (Kanda 2013) and Rcmdr (ver. 2.7–1) for R software (R ver. 4.0.3).