Mapping the effects of ozone pollution and mixing on floral odour plumes and their impact on plant-pollinator interactions

27 The critical ecological process of animal-mediated pollination is commonly 28 facilitated by odour cues. These odours consist of volatile organic compounds (VOCs), 29 often with short chemical lifetimes, which form the strong concentration gradients 30 necessary for pollinating insects to locate a flower. Atmospheric oxidants, including ozone 31 pollution, may react with and chemically alter these VOCs, impairing the ability of 32 pollinators to locate a flower, and therefore the pollen and nectar on which they feed. 33 However, there is limited mechanistic empirical evidence to explain these processes within 34 an odour plume at temporal and spatial scales relevant to insect navigation and olfaction. 35 We investigated the impact of ozone pollution and turbulent mixing on the fate of four 36 model floral VOCs within odour plumes using a series of controlled experiments in a large 37 wind tunnel. Average rates of chemical degradation of α-terpinene, β-caryophyllene and 38 6-methyl-5-hepten-2-one were slightly faster than predicted by literature rate constants, 39 but mostly within uncertainty bounds. Mixing reduced reaction rates by 8-10% in the first 40 2 m following release. Reaction rates also varied across the plumes, being fastest at plume 41 edges where VOCs and ozone mixed most efficiently and slowest at plume centres. 42 Honeybees were trained to learn a four VOC blend equivalent to the plume released at 43 the wind tunnel source. When subsequently presented with an odour blend representative 44 of that observed 6 m from the source at the centre of the plume, 52% of honeybees 45 recognised the odour, decreasing to 38% at 12 m. When presented with the more 46 degraded blend from the plume edge, recognition decreased to 32% and 10% at 6 and 12 47 m respectively. Our findings highlight a mechanism by which anthropogenic pollutants can 48 disrupt the VOC cues used in plant-pollinator interactions, which likely impacts on other 49 critical odour-mediated behaviours such as mate attraction.


55
Floral odours are used by many pollinating insects to locate floral resources. Upon 56 landing on a flower, many species of pollinating insect can learn to associate the unique 57 blend of chemical compounds that make up the flower's odour profile with the nectar 58 reward that it provides, facilitating them to locate rewarding flowers of the same species 59 in the future (Jones and Agrawal, 2017). When an insect uses floral odours to locate a 60 flower, that odour must fulfil certain criteria. For example, it must be relatively short lived, 61 to ensure it does not accumulate, making the cue from individual flowers indistinguishable 62 from the ambient background. At the same time, it must persist for sufficiently long to (i) 63 reach insects and (ii) remain recognisable. Chemical communication is therefore a trade-64 off between short-lived cues with strong concentration gradients, and long-lived cues that 65 travel further with weak concentration gradients (Williams and Ringsdorf, 2020). 66 Monoterpenes and sesquiterpenes are groups of volatile organic compounds 67 (VOCs) which typically fulfil these criteria and are common components of floral odours 68 (Knudsen et al., 2006).They have relatively short atmospheric lifetimes in the order of tens 69 of minutes to several hours with respect to the hydroxyl radical (OH), the principal oxidant 70 in the troposphere during the daytime, and with respect to NO 3 at night (Atkinson, 2000). 71 Yet, these compounds can also be removed via reactions with ozone (O 3 ), a powerful 72 oxidant formed at ground level when VOCs and oxides of nitrogen react in the presence 73 of sunlight (Zhang et al., 2019). In the Northern hemisphere background concentrations 74 of O 3 are in the range of 25 to 50 ppb and are further increasing at a rate of between 0.2 75 and 2% per year due to an increase in the emissions of its chemical precursors (Vingarzan,76 2004), and local pollution episodes can see concentrations in excess of 200 ppb, 77 particularly within or downwind of large conurbations (Group, 2021). For many 78 monoterpenes and sesquiterpenes, reaction with O 3 is more efficient than reaction with 79 OH, resulting in significantly shorter lifetimes. For example, α-terpinene has a lifetime of 80 45 minutes with respect to OH, but less than 30 s when O 3 concentrations exceed 70 ppb 81 (Atkinson et al., 1986). 82 Reaction rates found in the literature assume complete mixing between VOC and 83 oxidant, yet this may not be the case outside of the laboratory. This is particularly true for 84 highly reactive compounds because their chemical lifetime is of the same order as the 85 mixing time scale. This effect of spatial segregation between oxidant and VOC is often 86 overlooked and our understanding of the applicability of reaction rates derived in a 87 laboratory for use in the atmosphere is, therefore, uncertain. 88 An odour plume consists of a series of filaments, which can be considered as 89 strands of higher concentrations of odour, that are formed by turbulent mixing. Recording 90 VOC concentrations at any one stationary point in a plume as it moves and shifts with air 91 movements would reveal bursts of high concentration as a filament is encountered, in-92 between absences of VOC, defined as the intermittency of the plume (J Murlis et al.,  This ensured the ratio between electric field and number density of molecules (E/N) within 207 the drift tube was maintained at 122 Td and that fragmentation of ions remained 208 consistent. An assessment of ion fragmentation and instrument calibration is shown in 209 Section S1.2 of the SI. 210 7 Air was sampled from the wind tunnel at a rate of ~ 6 L min -1 (at 500 mbar) from a 211 ~25 m long ¼" O.D (1/8" I.D) PFA sampling line. To avoid sampling at reduced pressures, 212 air for analysis by PTR-QiTOF was subsampled using a Teflon headed pump (KNF) to 213 pass the air past the instrument inlet. The sample line was mounted directly to the three-214 dimensional traverse inside the wind tunnel and the tube heated to 80 o C to limit absorption 215 of VOCs to the tube walls. A PTFE filter (0.45 μm pore size) was placed in line after the 216 pump to prevent particulate matter entering the instrument. The temperature of the air 217 within the sample line was monitored continuously and was to within 1 ˚C of the recorded 218 wind tunnel temperature. permitted. Following each wind tunnel measurement, the background concentration was 239 monitored by sampling at the wind tunnel edge, outside of the plume for 1 minute. The 240 final step was to measure the concentration of VOCs coming from the diffusion source for 241 1 minute, to ensure any minor deviations in release concentration could be corrected. In 242 total, each plume characterisation experiment took approximately 9 hours to complete. 243

Plume intermittency and odour filament width 244
The intermittency of an odour plume is defined here as the fraction of time (0-1) 245 the VOC concentration falls below the threshold of perception (ToP). This threshold is 246 insect dependent, so here an arbitrary value defined as 1% of the 98 th percentile of 247 observed concentrations was used. The width of individual odour filaments was calculated 248 by counting the number of consecutive measurements falling below the ToP at each 249 location within the wind tunnel and averaging the results. The sampling rate of the PTR-250 QiTOF was 10 Hz and, therefore, each measurement was equivalent to a filament width on a pollinating insect's ability to successfully recognise that odour, proboscis extension 255 response (PER) assays were performed using the western honeybee (Apis mellifera). 256 Liquid standards of the four VOCs used in the wind tunnel study were mixed into 257 a blend, with volumes initially based on the vapour pressure of each compound (Table  258 S1). Concentrations were assessed using a solid phase microextraction fibre (SPME) in 259 combination with GC-MS and the methods used are outlined in Section 1.3 of the SI. 260 Having established the initial ratio of the blend, two sets of three further odour blends were 261 prepared aiming to replicate the ratios observed in the wind tunnel under the 150 ppb 262 treatment at distances of 2, 6 and 12 m from the source. The first set represented the 263 ratios seen in the plume centre and the second represent those seen at the plume edge. 264 SPME collection and GC/MS analysis were repeated on two random blends to ensure that 265 individual compounds were within 2% of the calculated fractions. 3. Results and discussion 300

α-terpinene plume 302
A single component release of α-terpinene was used to map out the spatial 303 distribution of an odour plume and to determine how this changed under differing levels of 304 O 3 pollution. Figure 2 shows the measured odour plumes under O 3 fields of 0 (Fig. 2a), 50 305 (Fig. 2b), and 150 ppb (Fig. 2c). Background O 3 levels within the wind tunnel were between 306 8 and 12 ppb and therefore a true 0 ppb field could not be achieved. Instead, Fig. 2a  307 shows a plume of propane, which does not react with O 3 and can therefore be considered 308 representative of the α-terpinene plume injected into a zero O 3 concentration field.  Figure 1b) 318 which have subsequently been interpolated using the "natural neighbour" approach and 319 Voroni tessellation (See "image_interpolate" function, Igor Pro, Version 8.0.3.3, 320 Wavemetrics). Similar plots showing the average O 3 concentration are shown in the 321 Supplementary Information (figure S1). 322  The addition of O 3 to the wind tunnel changed the plume shape causing a 329 narrowing as the α-terpinene mixed with the O 3 rich air at the plume edges. Figure 2b  330 shows that ozonlysis removed almost all of the α-terpinene in the 16 s that it took for plume 331 to travel the length of the wind tunnel at 50 ppb of O 3 , and at 150 ppb (Fig. 2c), all of the 332 α-terpinene had reacted away by 9.5 m (13.5 s). These two-dimensional cross sections 333 demonstrate how a plumes spatial extent, both length and width, can be significantly 334 reduced under increasing levels of O 3 pollution at concentrations regularly observed in the 335 lower troposphere at background levels (50 ppb) and also during elevated pollution 336 episodes (150 ppb) (Group, 2021). 337 The α-terpinene plumes were also modelled by subtracting the expected chemical 338 loss from the conserved propane plume. Theoretical first-order loss rates were calculated 339 as 340 ).
(3) 344 Here, k AT,O3 , is the ozonolysis rate constant (see Table S1) and [O 3 ] was the 345 concentration of O 3 measured at each location in the wind tunnel in molecules cm -3 . The 346 time since release was calculated based on the straight-line distance between the 347 effective release and measurement point and the fixed wind speed. 348 theoretical plumes for the ~50 and ~150 ppb O 3 fields, respectively. In both cases, the 350 measured plume is degraded more quickly than would be expected based on the literature 351 rate constant: an average of 21% for 50 ppb and 24% for 150 ppb, based on the measured 352 data points, only. This was partially explained by the additional time available for reaction 353 as sampled air was drawn into the PTR-QiTOF. A cross-covariance function applied to the 354 propane signal measured in situ by the FFID (and used for the theoretical plume 355 calculations) and PTR-QiTOF showed this delay to be ~1.6 s (see Figure S2 of the SI). 356 Accounting for this additional delay reduced the average difference to 18% and 12%, for 357 50 ppb and 150 ppb scenarios, respectively. 358 The overall increased reaction rates may be influenced by additional chemical 359 sinks in the wind tunnel including reactions with either OH or NO 3 , which were not 360 measured during the study. The spatial behaviour of the plume supports the view that 361 imperfect mixing affects the rate constant: the difference plots highlight the increased 362 reaction rate at the plume edges, where mixing between O 3 and the VOC is particularly 363 efficient, but less change is seen at the plume centre. The reaction rate appears slowest 364 during the first 0.5 to 2 m after release, which is consistent with the limited opportunity for 365 mixing shortly after release. Incomplete mixing reflects the intensity of segregation 366 between the O 3 and VOCs, 367 The (5) 383 This intensity of segregation can be used to derive an effective rate constant as: 384 , , 3 = , 3 (1 + , 3 ).
(6) 385 The literature rate constant for α-terpinene is 2.1 x 10 -14 cm 3 molecule -1 s -1 , and the 388 effective rate constant is roughly equivalent to this at the plume centre between 4 and 8 389 m. However, at the effective release point (y=0, x=0.5), k eff,AT,O3 is decreased by ~8%, 390 thereby increasing the lifetime of α-terpinene at that point. Towards the plume edges, from 391 3 m down the tunnel and beyond, k eff,AT,O3 is closer to k AT , than at the plume centre where 392 there the α-terpinene has less opportunity to mix with the O 3 . 393 Repeating the same experiment at the lower height of 0.5 m (Figure 4b), where 394 turbulence is slightly increased, showed a somewhat steeper gradient in the effective rate 395 constant, with up to a 10% reduction at the effective release point. This is counter to 396 expectation because the increased turbulence and longer travel time should enhance 397 mixing and reduce the effects of segregation. Figure 3c shows the effective rate constant 398 for both measurement heights at each location within the tunnel. There is a small 399 difference between the effective rate constants measured at 0.5 and 1 m but this was 400 within the uncertainty bounds which was based on the standard deviation of the 401 measurements. This potentially reflects the fact that differences in the turbulence 402 intensities between the two heights was relatively minor (4.5 % and 3.2 % for the vertical 403 component and 5.2% and 3.2 % for the lateral). Further measurements at lower heights 404 were not possible due to the potential for the plume to interact with the wind tunnel's 405 surface. 406 A maximum 8-10% decrease of the rate constant may appear modest, but it should 407 be viewed as a lower limit. This is because our instruments were limited to a frequency 408 response of 10 Hz. Comparison of the PTR-QiTOF measurements with those of propane 409 made by the FFID at a frequency of 200 Hz, revealed fine scale variation in concentration 410 that is lost when using a 10 Hz measurement (see Fig. S3). Therefore, the effects of 411 segregation could be greater; particularly for an insect encountering the plume, whose 412 antennae can detect changes at a rate far greater than 10 Hz (Szyszka et al., 2014). 413 This result indicates that mixing plays an important, but often overlooked, role in 414 the lifetime of VOCs. For those VOCs used as chemical cues, this is particularly important 415 because the short lifetimes necessary to generate strong concentration gradients mean 416 the lifetime of the signal compound is invariably similar to, or shorter than, the mixing time 417 scale. 418 which is more representative of a floral odour plume, showed variations in the reaction 428 rates of the four VOCs: α-terpinene, β-caryophyllene, MHO and linalool (see Fig. S4 a-d).

429
The difference between the theoretical and measured plumes are shown in Figure 5 and 430 a more detailed comparison that incorporates the uncertainty of the literature rate 431 constants is shown in Figures S5-S8. For α-terpinene, β-caryophyllene and MHO, the 432 measured plumes are again degraded faster than would be expected based on their 433 respective O 3 rate constants, with the largest differences occurring at the plume edge. 434 Applying a correction for the additional sampling delay accounted for 7% and 4% of the 435 difference for α-terpinene, β-caryophyllene, respectively, and less than 1% for MHO. As  436 was the case for a-terpinene, the reason for the reaction rates exceeding those expected 437 based on literature rate constants, even despite the effects of segregation, likely relates 438 to the presence of OH and or NO 3 within the wind tunnel. 439 Linalool is degraded more slowly than implied by its rate constant and the 440 difference becomes larger towards the plume edge. It is likely that an oxidation product, 441 from either α-terpinene or β-caryophyllene is formed which contributes to the linalool 442 signal, most likely as a fragment ion. This is supported by the fact that the difference is 443 largest at the plume edge where the mixing and hence reaction rates are at their fastest. Plume intermittency is considered to be a function of source strength and turbulent mixing. 455 Yet. Figure 6 shows that higher levels of O 3 caused an increase in intermittency and 456 reduction in odour filament width of an α-terpinene plume when released under identical 457 turbulent profiles. These changes are likely to have behavioural consequences for a flying 458 insect using the odour plume to navigate to a source. Both factors are critical in eliciting 459 insect flight behaviours and small changes to plume structure can have dramatic effects 460 on flight behaviour and the ability of a foraging insect to locate an odour source (Beyaert 461 and Hilker, 2014; Mafra-Neto and Cardé, 1994) However, we did not directly investigate 462 the effects of the changes we observed to intermittency and filament number on insect 463 flight behaviour and therefore the degree of impact of those changes remains unclear. 464 Again, the largest changes in intermittency and odour filament width were observed 465 towards the outer edges of the plume where mixing with O 3 is most efficient. 466 467 468 the blend to which they had been trained. However, for the less degraded central plume 486 (figure 7 c), when honeybees were presented with a VOC blend representative of that 487 recorded at 6 m from the tunnel odour source only 52% of honeybees tested exhibited a 488 positive PER response, decreasing to 38% for bees presented with the VOC blend 489 representative of 12 m from the odour source. For the honeybees presented with VOC 490 blends representing the more degraded plume edge (Figure 7 d) It should be noted that this assay purely assessed the honeybee's abilities to 505 recognise the odour and therefore, the results do not capture any additional effects 506 associated with direct exposure to O 3 . Furthermore, here the honeybees are responding 507 to a mean concentration, and therefore, the additional impact of O 3 on the spatial extent 508 and intermittency of the plume as well as the odour filament width are not captured within 509 the honeybees response.  compound (α-terpinene, β-caryophyllene, MHO and linalool) were found to be broadly 531 similar to that expected based on literature rate constants, but in most cases slightly faster, 532 likely due to additional chemical sinks (e.g. OH and NO 3 ) for the VOCs within the wind 533 tunnel. The notable exception was linalool, where reaction rates were slower than 534 expected. In this case, fragmentation of an oxidation product to m/z 155 was thought to 535 be the most likely cause. 536 An analysis of the intensity of segregation between the concentrations of α-537 terpinene and O 3 revealed that the effective rate constant (k eff , AT,O3 ) was reduced by up to 538 10% in the first 2 m following release. The effect of segregation was expected to decrease 539 at lower measurement heights, but no statistically significant difference was observed 540 between the 0.75 and 0.5 m measurements. This reflected the small differences in 541 turbulence intensity between these two measurement heights. Reaction rates were fastest 542 at the plume edges attributed to the increased opportunity to mix with ozone. Ozone was 543 also found to increase the intermittency of the plume and decrease odour filament width, 544 two properties used by insects for navigation. 545 Replication of the average plume composition in a proboscis response assay 546 clearly showed a rapid decline in honeybees' ability to recognise the floral odour following 547 simulated degradation by O 3 . These results, although based upon a synthetic odour, 548 provide an important insight into the mechanism by which anthropogenic pollutants can 549 disrupt the chemical cues used by insects. Importantly, our experimental approach, where 550 the effects of O 3 degradation were replicated and then presented to the insect, allows its 551 effects to be de-coupled from potential olfactory coding impairment caused by direct 552 oxidative stress. In this instance, the effects are pronounced within just a short distance 553 from the point of release. 554 However, outside of the laboratory, the extent to which ozonolysis impacts on 555 pollinators will ultimately depend on the reactivity of the components within a given floral 556 odour, the ambient O 3 concentration (and NO 3 if at night-time) and wind speed.  The authors declare that they have no known competing financial interests or personal relationships 577 that could have appeared to influence the work reported in this paper.

581
Data is available upon reasonable request.