Field investigations for evaluating green infrastructure effects on air quality in open-road conditions

Many people live, work and spend time during their commute in near-road environments ( 2 m) road conditions. These configurations gave us a total of six different real-world scenarios for evaluation. The changes in concentrations of PM10, PM2.5, PM1, BC and PNC at all six sites were estimated by comparing simultaneous measurements behind and in front of GI (or adjacent clear area). A portable battery-operated experimental set-up was designed for measuring the pollutant concentrations for 30 full days over a field campaign period of three months. On each day, around 10 h of continuous data were recorded simultaneously behind and in front of GI/adjacent clear area, capturing both morning and evening traffic peaks. Our objectives were to: (i) assess the effectiveness of different types of GI in reducing various pollutants; (ii) evaluate the impact of wind directions and density of vegetation on reducing different pollutant concentrations behind GI; (iii) investigate the changes in fractional composition of sub-micron (PM1), fine (PM2.5) and coarse (PM2.5-10) particles; and (iv) quantify the elemental composition of collected particles before and after the GI. In away-road conditions, all three configurations showed reductions behind the GI for all pollutants. The ‘hedges only’ configuration showed higher pollutant reductions than the other two configurations, with maximum reductions of up to 63% shown for BC. In close-road conditions, the results were mixed. The ‘trees only’ configuration reported increases in most of the pollutant concentrations, whereas the combination of trees and hedges resulted in reduced pollutant concentrations behind the GI. Among all pollutants, the highest relative changes in concentration were observed for BC (up to 63%) and lowest for PM2.5 (14%). Categorising the data based on wind directions showed the highest reduction during along-road wind conditions (i.e., parallel to the road). This was expected due to the sweeping of emissions by the wind and the wake of road vehicles whilst the barrier effect of GI enhanced this cleansing, limiting lateral diffusion of the pollutants. However, cross-road winds that took vehicular emissions to pass through the GI allowed us to assess their influence, showing up to 52, 15, 17, 31 and 30% reduction for BC, PM10, PM2.5, PM1 and PNC, respectively. The largest reductions were consistently noted for the mixed ‘trees and hedges’ configuration in close-road conditions and the ‘hedge only’ configuration in away-road conditions. The assessment of various fractions of PM showed that ‘hedges only’ and a combination of trees and hedges lowered fine particles behind GI. The SEM-EDS analysis indicated the dominance of natural particles (50%) and a reduction in vehicle-related particles (i.e., iron and its oxides, Ba, Cr, Mn) behind GI when compared with the in-front/adjacent clear area. The evidence contributed by this work enhances our understanding of air quality modifications under the influence of different GI configurations, for multiple pollutants. In turn, this will support the formulation of appropriate guidelines for GI design, to reduce the air pollution exposure of those living, working or travelling near busy roads.

Among all pollutants, the highest relative changes in concentration were observed for BC (up 34 to 63%) and lowest for PM2.5 (14%). Categorising the data based on wind directions showed 35 the highest reduction during along-road wind conditions (i.e., parallel to the road). This was 36 expected due to the sweeping of emissions by the wind and the wake of road vehicles whilst 37 the barrier effect of GI enhanced this cleansing, limiting lateral diffusion of the pollutants. 38 However, cross-road winds that took vehicular emissions to pass through the GI allowed us to 39 assess their influence, showing up to 52, 15, 17, 31 and 30% reduction for BC, PM10, PM2.5,     (Table 2).

160
All six measurement locations were near to residential areas containing two-storey buildings   Table 2.

208
Sampling location had two sets of instruments (includes GRIMM, P-TRAK, and MicroAeth) 209 mounted on a tripod stand at a 1.5m height to sample air from a typical breathing height. One  the data was divided based on the wind flow direction with respect to street and GI alignment.

222
The dataset was divided into three wind direction sectors: 'along-road' (parallel to road), instruments were calibrated prior to fieldwork. One in each pair of the instruments was 255 calibrated later than the other, and was considered as a base instrument to harmonise the data. 256 For quality assurance of the data collected by instruments, we implemented the following  We co-located both sets of instruments side-by-side for at least 30 minutes prior to start and 259 after the GI monitoring campaigns each day. On some days, we carried out this co-location 260 exercise in the middle of the monitoring period, when instruments were restarted after a battery 261 change. The total period of co-location data accounted for ~10% of total field campaign data, 262 enabling us to inter-compare results from two identical instruments and assess the relative 263 difference. All our instruments performed well against their counterpart and obtained a good 264 agreement (Fig 4). We obtained (i) a minimum R 2 value of 0.85 for BC measurements by   (Table 3).

289
Similar to ΔBC, ΔPM10 behind the GI also exhibited a similar trend in both close-road and 290 away-road sites, but the magnitude of ΔPM10 was lower compared to ΔBC. The highest 291 improvement in ΔPM10 was observed for trees with hedge in away-road (THIB; -24%) and  In summary, the HIB site presented better improvement in air quality behind GI across measured were seen for BC and PNC (rapid decay) and the least for PM2.5 (gradual decay). Finally, we 336 observed that hedges, and the combination of trees with hedges, provided the better reduction 337 potential.

Effects on wind direction 339
In order to understand the influence of wind direction on concentrations behind the GI, 340 we separated the wind conditions into three main categories: along-road, cross-road and cross-341 vegetation (Fig 2), as explained in Section 2.3. For some sites, we did not have enough data 342 points available; for example, during cross-road winds at THIB and cross-vegetation winds at 343 both the TCB and HIB sites (Table S2). ΔPNC in three investigated wind directions were lower 344 than that of ΔBC and were similar to ΔPM1. Along-road wind conditions resulted in a 345 maximum reduction between wind categories. HIB and HCB in both close-road and away-road 346 sites showed the highest reduction in ΔPNC of -30% and -50%, respectively (Fig 6). In cross-347 road conditions, HIB displayed a maximum reduction (-30%) in PNC, followed by TCB (-13%) 348 and HCB (-12%). The highest deterioration in PNC among all wind conditions was reported 349 during cross-road winds, although less than 5% at sites TCB and THCB (Table S2) conditions. ΔPNC in cross-road wind conditions were comparable and along-road wind 356 direction displayed higher ΔPNC than cross-road winds in both studies.

357
Highest relative changes between measurements taken behind GI and in front of GI/clear areas 358 were observed with BC compared other investigated pollutants. Furthermore, the maximum 359 percentage differences in BC were comparable across different wind directions (Fig 6).  ΔPM2.5 concentrations were lower than all other measured pollutants in this study (Fig 6). HCB 384 and TCB sites showed deterioration in PM2.5 concentration behind the GI for all wind directions.

385
In along-road wind direction, the highest improvements were revealed by THCB (-17%) at ranging from 2% to 7% in the cross-vegetation wind category (Table S2). Past studies

396
In most of the wind categories, influences on ΔPM1 were positive (Fig 6). The magnitude of 397 differences was similar to PNC and higher than PM10 and PM2.5 (Table S2) HCB site (Fig 6).

404
In summary, the magnitude of percentage differences followed the following trend:

422
In order to assess the effect of vegetation density on percentage differences in pollutant 423 concentration behind the GI, the correlation coefficient (R 2 ) between LAD and relative 424 pollutant concentration were drawn (SI Fig S3). As mentioned in Section 3.2, a full dataset was 425 not available for cross-road and cross-vegetation wind directions and such scenarios were 426 therefore excluded in this analysis. While analysing the overall data, we observed R 2 well 427 below 0.8 at close-road and away-road sites for more than half of the cases and were considered 428 as insignificant (Fig 7).    Table 3).

509
Other particles in the vehicle category were dominated by Ba, followed by Mn, Cr, V, and Ti.

510
Although the percentage difference of vehicle group between behind and in front of or clear 511 area adjacent to vegetation were smaller, these elements are toxic even in lower concentrations.

512
When comparing identified particles from behind GI with those from the other monitoring 513 locations, natural (+7%) and NaCl (+5%) particles were higher behind GI than in front of or in 514 a clear area adjacent to GI (Fig 11). Conversely, a significantly lower percentage (-7%) of 515 vehicle particles were found behind GI than in the other monitoring locations (Fig 11). In terms        The inner circle shows PM fractions behind the GI; the outer circle shows PM fractions in-801 front/clear areas. Blue, orange and grey colours denote PM1, PM1-2.5 and PM2.5-10, respectively.

802
Line shading represents a lack of data available in particular situations.  Table 2. Details of six monitoring locations. Note the clear area and behind (CB) and in-front and 820 behind (IB) monitoring points refer to measurements taken at a clear location adjacent to and in front