c07-01-0030 Pfender.indd

Dispersal and deposition of pollen of creeping bentgrass (Agrostis stolonifera L.) was estimated by using CALPUFF, a complex model originally developed to simulate dispersal of particulates and other air pollutants. In fi eld experiments, peak pollen emission rates (8 × 106 pollen grains per min per m2 of a creeping bentgrass stand) occurred between 1000 and 1200 h. Pollen survival under outdoor conditions decreased exponentially with time, and only 1% survived for 2 h. CALPUFF simulations showed deposition of 100,000 viable pollen grains per m2 at distances of 2 to 3 km from the source fi eld, and deposition of one pollen grain per 10 m2 at distances of 4.6 to 6.7 km from the source fi eld. Pattern of simulated deposition varied with weather conditions and, to a lesser extent, source fi eld size. Simulation of dispersal by a small thermal vortex produced deposition of one grain per 10 m2 at 15.3 km from the source fi eld. Overall, the deposition modeling results suggest that pollen-mediated gene fl ow is likely at distances of 2 to 3 km from a source fi eld, and possible at distances up to 15 km. W. Pfender and M. Carney, USDA-ARS National Forage Seed Production Research Center, 3450 SW Campus Way, Corvallis, OR 97331; R. Graw, USDA Forest Service Pacifi c Northwest Region, P.O. Box 3623, Portland, OR 97208; W. Bradley and L. Maxwell, Dep. Botany and Plant Pathology, Oregon State Univ., Corvallis, OR 97331. Received 19 Jan. 2007. *Corresponding author ([email protected]). Abbreviations: GMO, genetically modifi ed. Published in Crop Sci. 47:2529–2539 (2007). doi: 10.2135/cropsci2007.01.0030 © Crop Science Society of America 677 S. Segoe Rd., Madison, WI 53711 USA All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher. R e p ro d u c e d fr o m C ro p S c ie n c e . P u b lis h e d b y C ro p S c ie n c e S o c ie ty o f A m e ri c a . A ll c o p y ri g h ts re s e rv e d . 2530 WWW.CROPS.ORG CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007 largest distances tested in each case). In contrast, another study using a relatively small number of source plants detected pollination at a distance of 150 m, but not 200 m (Wang et al., 2004). An experiment with bentgrass (Agrostis spp.) produced observations of pollination at distances of approximately 300 m (0.02% of the progeny positive for the marker gene) (Wipff and Fricker, 2001). Fitting a simple exponential decline with distance to these latter data suggests recipient plants at 1.3 km from the 286 source plants could have been fertilized. In another experiment, nontransformed Agrostis spp. plants, naturally occurring or intentionally placed as sentinels, produced transgenic seed after being exposed up to 14 to 21 km from a 162-ha source of GMO creeping bentgrass (A. stolonifera L.) in Oregon (Watrud et al., 2004). Attempts have been made to generalize or model eff ective dispersal distances for wind-blown pollen. Some pollination distance models have been concerned only with short-distance data, a maximum of 15 m (Belanger et al., 2003; Meagher et al., 2003), but others have attempted to describe longer-distance phenomena. Rognli et al. (2000) noted that pollen-mediated gene fl ow depends on dispersal distance of pollen grains, their viability, and the environment, and on the presence of competing pollen at the recipient plants. All published reports have shown that pollination decreases with distance from the source plants, generally as some form of a negative exponential function displaying leptokurtosis (higher probability distributions in the tails than predicted by a normal distribution) (Gleaves, 1973; Rognli et al., 2000). Giddings et al. (1997a) attempted to model pollen dispersal from a perennial ryegrass source to trap plants located at distances up to 80 m. He found that previouslyreported equations for dispersal (i.e., negative exponential with a general factor to refl ect wind turbulence) were not very useful to describe his observations. Adding factors for wind direction (Giddings et al., 1997b) did not improve results greatly, and the author noted that dispersal did not always decrease smoothly with distance and that multiple factors are likely needed to explain this complex phenomenon. Nurminiemi et al. (1998) fi t several models, built on exponential decrease with distance, to data derived from a pollination experiment with marker genes in tall fescue. They selected a model that could predict the main patterns in the data (i.e., dispersal inversely proportional to distance and with distinct leptokurtosis, an eff ect of wind direction, and a strong eff ect of competing pollen at recipient location), but there were discrepancies at some of the greater distances (i.e., 160 m) tested. Rognli et al. (2000) likewise found it diffi cult to predict pollen distribution >155 m from the source, but could generally model pollination to be more than exponentially leptokurtic with distance, and to depend on source characteristics (i.e., size, distribution, and density) and wind direction. A review of the physical factors involved in pollen dispersal ( Jackson and Lyford, 1999) notes that atmospheric instability has a major eff ect on dispersal distances. Adding to the complexity is the fact that pollen is likely emitted from a source fi eld as a series of puff s, rather than as a steady Gaussian plume which would be simpler to characterize ( Jackson and Lyford, 1999). Dispersal modeling has been addressed in areas of biometeorology other than pollination. For example, dispersal of plant pathogenic fungal spores from infested fi elds has received much attention (Aylor, 1986, 1999). We recently published the use of a complex air pollution model (CALPUFF) to estimate dispersal of fungus spores from grass-seed fi elds infested with the stem rust fungus (Pfender et al., 2006). CALPUFF is an air pollution modeling system originally developed for estimating movement and deposition of air pollution contaminants, including particulates, for both short and long distances (Scire et al., 1990). It is in widespread use by air quality regulation agencies, and has been validated in several studies with controlled releases of tracer gases. In one study, where the tracer was released in a rural setting and detectors were arranged in 12 arcs ranging from 0.5 to 50 km from the source, mean concentrations of the tracer modeled by CALPUFF/CALMET were 98% of the actual, and the modeled maximum concentrations were 79% of actual (Hurley and Luhar, 2005). In a diff erent study conducted in a region with desert basins and mountains and detectors up to 20 km from the source, 50 to 60% of the CALPUFF predictions were within a factor of two of observed tracer levels when the release was from a point source, and 25 to 30% were within a factor of two when the release was from line sources (Chang et al., 2003). CALPUFF has been used to model dispersion of particulates, also. It was able to reproduce the observed time series of 10-μm particulates recorded at surface monitors in a New Zealand study (Barna and Gimson, 2002). The CALPUFF modeling system allows the inclusion of such realistic elements as variations in wind speed, direction, and turbulence. It is a non-steady-state Lagrangian Gaussian puff model with modules for gridded, time-varying, three-dimensional meteorological conditions, complex terrain eff ects, and wet and dry deposition. Technical details are available in Scire et al. (1990), and a narrative summary of its salient features was presented in Pfender et al. (2006). The meteorological pre-processor, CALMET, uses prognostic output from the Penn State Mesoscale Meteorological Model (MM5) as an initial estimate for the windfi eld, which is then modifi ed to account for eff ects of complex terrain (Scire et al., 2000). The results are interpolated to 2.5 km resolution, adjusted based on surface and upper-air observations, and used as input for CALPUFF. CALPUFF also allows the use of local weather observations, including wind turbulence measurements, obtained at the release site. In addition to weather information CALPUFF uses inputs for particle (pollen) size and settling velocity, the source fi eld size and the pollen R e p ro d u c e d fr o m C ro p S c ie n c e . P u b lis h e d b y C ro p S c ie n c e S o c ie ty o f A m e ri c a . A ll c o p y ri g h ts re s e rv e d . CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007 WWW.CROPS.ORG 2531 and variance of the settling rate for pollen grains were calculated based on the time required to fall the length of the settling tower. The experiment was conducted four times. Pollen Survival Dynamics Plants of cultivar Seaside were grown in pots outdoors. When anthers fi rst emerged in the morning, the pots were taken into a greenhouse and observed continuously for anther dehiscence. Within 15 min of dehiscence, pollen was collected for testing. To test pollen survival under conditions that would mimic those during aerial dispersal (exposure to ambient radiation and relative humidity in free air), yet permit us to recover the pollen for viability testing, we exposed pollen on traps made of bird feathers. Each trap consisted of a plastic pot label (9.5 by 2.2 by 0.1 cm, Hummert International, Earth City, MO) with a rectangular window (5.0 by 1.5 cm) cut out. Three individual downy barbs from down feathers of barn owl (Tyto alba) were arranged side-by-side 2 mm apart and glued to the pot label such that each feather spanned the 1.5-cm opening width and the hookless barbules of the adjacent feathers overlapped. Before use, the down feathers were washed with a 1:1 mixture of methanol and methylene dichloride, rinsed three times in water (the fi rst rinse for 8 h) to remove any fatty acids and alcohols that might interfere with biological processes (Shawkey et al., 2003). Freshly-shed pollen was applied to each trap by holding it beneath creeping bentgrass fl owers with newly-dehiscent anthers and tapping the fl owers to release pollen. Immediately after loading, each trap was taken outdoors and exposed in an unshaded location by clipping the plastic pot label frame to a rack. At 0, 10, 30, 60, 90, 120, and 180 min after beginning exposure, viability of pollen was tested on Petri plates of a germination medium described by Fei and Nelson (2003). Pollen was transferred by briefl y pressing the feathers against the surface of the medium. Plates were incubated at 20 ± 2°C for 90 min, then pollen grains were examined microscopically to determine percent germination. A grain was considered germinated if the germ tube was at least two pollen diameters in length. The test was conducted on eight diff erent days between 14 June and 5 July 2005. On each day of testing, there were three exposure traps per exposure time. Across the diff erent days of testing, outdoor temperatures during exposure ranged from 17 to 24°C and greenhouse temperatures ranged from 22 to 27°C. Field Experiments for Pollen Emission Pollen emission rates (pollen grains emitted per m2 fi eld per min) and associated weather conditions were measured in fi eld experiments conducted at Hyslop-Schmidt Experiment Farm (44° 38′ N, 123° 12′ W) near Corvallis, OR. Creeping bentgrass cultivar Seaside was established from transplants as a circular plot, 6 m in diameter, in a 1.2 ha fi eld in January 2005. The remainder of the fi eld surrounding the creeping bentgrass plot was planted to oats (Avena sativa L.). The oats were mowed during June and July as needed to maintain a canopy height similar to that of the creeping bentgrass. There was no other creeping bentgrass within 1 km of the study area, or within 2 km upwind on sampling days. Pollen emitted from the creeping bentgrass plot by the action of naturally-occurring wind was measured with an array of samplers set in an arc downwind from the plot (Fig. 1), similar emission rate for each time unit of the modeled period. The model tracks the mass of particles emitted from the source, the amount deposited at any selected receptor sites in the modeling domain, and the amount remaining in the atmosphere (surface mixed layer and the air above the mixed layer) for each model time interval. The inclusion of both deposition (Pleim et al., 1984) and dispersion algorithms in CALPUFF, combined with the three-dimensional meteorological and land-use fi eld, should result in more accurate model-predicted results compared with simpler models based on a steady-state Gaussian plume description. In addition to information about dispersal distance of pollen, survival dynamics must be known to evaluate probability of pollination at various distances from the source (Luna et al., 2001). Grass pollen is relatively short-lived, typically <3 h (Huang et al., 2004; Teare et al., 1969). Fei and Nelson (2003) collected pollen of creeping bentgrass cultivar Crenshaw, stored it in a desiccator, and tested germination at 20-min intervals. Germination was approximately 80% during the fi rst hour after shedding, then dropped to 20% at 80 min and to 0% by 140 min after shedding. In this paper, we use CALPUFF to estimate dispersal distances and deposition rates of viable pollen from fi elds of creeping bentgrass. Additional supporting objectives were to estimate survival dynamics and the settling rate of bentgrass pollen grains, and to determine the rate of emission of pollen from fl owering stands of bentgrass. MATERIALS AND METHODS Pollen Characteristics Pollen was collected from plants of creeping bentgrass cultivar Seaside obtained from a fi eld near Corvallis, OR. Flower heads were cut in the morning before fl owers opened, the stems were placed in water and transported to the laboratory. Flower heads were placed on glassine paper and the pollen was collected as the anthers opened. Average weight per pollen grain was determined by weighing a several-milligram sample, then determining number of grains in the sample and dividing weight by number of grains. Number of grains was determined by suspending the weighed sample in a known volume of water, and counting the number of grains per 0.1 mm3 with the use of a haemocytometer. The size of hydrated pollen was determined by measuring diameters of 50 pollen grains after mounting in a solution of glycerol/water (30:70) (Stanley and Linskens, 1974). Settling velocity was measured with the use of a settling tower, as previously described (Pfender et al., 2006). A very small amount of pollen was released through a pinhole at the top of a glass tube (45 cm tall, 5 cm diam.), and collected on a series of greased microscope slides moved at 1-s intervals through a 3-mm high gap below the tube. Light directed upward through the tube from a cool fi ber-optic source allowed us to observe that the pollen fell uniformly, without turbulence. Each slide was examined microscopically to count pollen grains that reached the bottom of the tube in that time interval. Average R e p ro d u c e d fr o m C ro p S c ie n c e . P u b lis h e d b y C ro p S c ie n c e S o c ie ty o f A m e ri c a . A ll c o p y ri g h ts re s e rv e d . 2532 WWW.CROPS.ORG CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007 to the method previously reported for fungus spore emission measurements (Pfender et al., 2006). The arc was located at a radius of 10 m from the plot center, with samplers mounted on poles and set at intervals of 22.5° (approximately 3.8 m) along the arc. There were seven sampler poles (135° of arc), and the array was adjusted for each run to be centered on the downwind direction. The fi ve middle poles of the array had samplers at heights of 0.5, 1.5, and 2.5 m. The center pole had an additional sampler at 4.0 m height. The two outer poles each had only a single sampler, at 0.5 m height (Fig. 1). Four additional samplers, to detect spores entering the study site from upwind, were deployed on the 10-m radius arc, centered 180° from the downwind center pole. The samplers were rotary impaction devices (Aerobiology Research Laboratories, Ottawa, ON K2E 7Y5) that collect airborne particles on square polystyrene rods (1.6 mm by 1.6 mm by 28 mm) coated on the leading edge with silicone grease. There are two rods per sampler, located at the ends of a 9-cm long arm that spins at 2400 rpm. The greased rods are retracted (protected from contamination) until the unit is switched on and again after it is switched off . For each sampling, the 22 samplers were turned on simultaneously and allowed to run for 15 or 30 min before being turned off . The number of pollen grains per sampling rod was counted by examining with a microscope at 320× magnifi cation. Because the samplers were located near the pollen source, and there was no other grass pollen source in the immediate vicinity, most pollen grains matched the size and morphology of creeping bentgrass pollen, but any pollen of inappropriate morphology or size was not counted. The background pollen counts for a given run, estimated by the average of all upwind samplers, was subtracted from each downwind sampler reading before further analysis. The total number of pollen grains in the plume emanating from the plot during the sampling period (15 or 30 min) could be estimated for those runs in which the sampling array intercepted all or most of the plume, as evidenced by low or zero pollen numbers on samplers at the lateral and upper margins of the array. The estimate of pollen grains in the plume was made by fi rst calculating the aerial concentration of pollen measured by each sampler. Concentration was calculated as the number of pollen grains on both rods divided by the volume of air sampled (circumference of the path of rod movement × collection rods’ surface area × number of revolutions during the sampling). Pollen counts were also corrected for the collection effi ciency of the sampler, 60%. The theoretical effi ciency was calculated (Aylor, 1993), from the pollen settling velocity and the collector size and speed, to be 86%. This value was adjusted downward to 60% based on a report (Ogden and Raynor, 1967) that ragweed pollen is collected by rotary impaction samplers at about 65 to 70% of the theoretical effi ciency. Although there is no comparable information on actual vs. theoretical collection effi ciency for grass pollen, the ragweed pollen is of similar size (20 μm diam.) to that of bentgrass and should therefore have similar impaction properties. The pollen fl ux through the sampling array was calculated by using a simple numerical integration of the observed concentrations and wind speeds, as follows. The area represented by each sampler in the array was assumed to extend half the distance to the next sampler, both horizontally and vertically (Fig. 1). The pollen fl ux for each sampler was calculated as pollen concentration (pollen grains/m3) × volume of air (m3) moving past the sampler during the sample period. This air volume was calculated as the cross-sectional area of the conceptual rectangle surrounding the sampler (Fig. 1) multiplied by the wind run during the sampling period (wind speed in m/s × total seconds). Average wind speed during the sampling period for each sampler height was obtained by logarithmic interpolation (Thom, 1975) of wind speed measurements at 0.5, 2.0, and 6.7 m height. Total fl ux of pollen through the cross-section of the pollen plume was obtained by summing the fl uxes from all of the samplers. Weather data at the pollen sampling site were collected at 5-min intervals with automated weather instrumentation (Campbell Scientifi c Instruments, Logan, UT). Wind speed was measured with rotating cup anemometers at 0.5, 2.0, and 6.7 m above ground level, and wind direction was measured at 6.7 m height. Air temperature was measured at 0.5, 1.5, and 6.5 m height. Sensors for rainfall, total solar radiation and relative humidity were placed at 4.6, 3.9, and 1.5 m height, respectively. Measurements for computing turbulence parameters (standard deviation of the vertical and horizontal wind speed) were obtained with a sonic anemometer (model CSAT3, Campbell Scientifi c Instruments, Logan, UT) mounted at 1.7 m above ground level and facing into the wind. The sonic anemometer was operated at 1 Hz, and was located in the oats approximately 40 m away from the creeping bentgrass plot, 45 to 90° from the downwind direction. Figure 1. Sampling array for measuring number of pollen grains in a plume emitted from a plot of creeping bentgrass. Rotary impaction samplers were mounted on poles at heights of 0.5, 1.5, 2.5, and 4.0 m. The poles were set at 22.5° intervals along a 135° arc with a radius of 10 m, downwind from the center of a 6by 6-m plot of creeping bentgrass. Two additional poles with samplers at 0.5 and 2.5 m height were placed upwind of the plot to measure background pollen concentrations. All samplers were run simultaneously during the sampling period of 15 or 30 min. To estimate total pollen passing through the downwind sampling array during the sampling period, the plume cross-section was divided conceptually into rectangles (indicated here in dashed lines). Pollen fl ux through each rectangle was derived by multiplying the spore concentration measured at the appropriate sampler by air volume moving through that rectangle. Air volume was calculated from measured or interpolated wind speed at sampler height, sampling duration, and cross-sectional area of the rectangle. The pollen fl uxes of all rectangles were summed to produce the total fl ux. R e p ro d u c e d fr o m C ro p S c ie n c e . P u b lis h e d b y C ro p S c ie n c e S o c ie ty o f A m e ri c a . A ll c o p y ri g h ts re s e rv e d . CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007 WWW.CROPS.ORG 2533 The diurnal pattern of pollen release was measured with a Burkhard 7-d recording suction sampler at 2-h resolution and operating at a height of 0.75 m. It was not desirable to place the sampler in the 6-m circular plot where the plume was sampled, because of the eff ect such a large object would have on turbulence and thus on the pollen release being measured. Therefore the sampler was placed within a commercial fi eld of creeping bentgrass (Seaside) approximately 2 km from the plume sampling site. Pollen counts from this sampler were used to construct a temporal profi le of pollen release on each day, which was taken to represent the diurnal pattern at the plume sampling site on the same day. Running the CALPUFF Model The procedure and model settings used in running CALPUFF were as described previously (Pfender et al., 2006). As a brief description, a 420 by 445 km modeling domain approximately conterminous with the state borders of Oregon was created. The MM5 output for each day was processed with CALMET to account for terrain eff ects and to format the data for use by CALPUFF. CALPUFF was run with a 1-h time step for a 24h duration for each model run. There were no precipitation events during the days we modeled, so only the dry deposition (not the wet deposition) module of CALPUFF was used. We modifi ed the deposition reference height in CALPUFF, from 10 to 0.5 m, to better capture the near-source concentration of the pollen, which is released at canopy height. Inputs were used for the average and the variance of pollen settling velocity. CALPUFF requires an input of area-source emission rate, or pollen grains emitted per unit area of the fi eld per unit time. CALPUFF can incorporate a diff erent emission rate for each 1-h modeled time step. We used data from the plume-sampling plot to estimate emission rate during the sampling period, then assigned emission rates to each hour of the day according to results from the Burkhard sampler. For the sampling-period emission rate, we fi rst estimated the total fl ux of pollen through the sampling array as described in the previous section. Using an iterative procedure, we then found an emission rate in CALPUFF that would produce the observed mass fl ux at the sampling array located 10 m downwind of the plot center. This approach matched the simulated emission rate to the observed pollen fl ux, without the need to consider quantifying pollen production, escape fraction or near-source ( <10 m) deposition. In this way, dispersal and deposition could be modeled beyond the 10-m sampling boundary. The input for hourly emission rate was varied over the 24-h simulation run by reference to the diurnal pattern of pollen release obtained from the Burkhard sampler for each respective day’s simulation. The daily total pollen fl ux was determined as the quotient of observed 15or 30-min pollen fl ux divided by the Burkard-derived proportion of the daily pollen total that occurred during that time interval. The emission rate for each hour was adjusted as a proportion of the total according to the Burkhard sampling results. For simulations from an area source, CALPUFF uses a 2-dimensional integration algorithm (cross-wind and along-wind) (Scire et al., 2000) that avoids the errors within and near the area source that would be obtained if the total emission were assigned to a single virtual point in the center of the fi eld. The time-varying emission rates and the observed and modeled weather conditions were used to conduct model runs for several scenarios. We simulated dispersal for two specifi c dates, selected to span a range of weather conditions in favorability for dispersal. To select dates for the simulations, we fi rst ran CALPUFF for the 15 consecutive days between our fi rst and last plume sampling dates (25 June and 8 July) with a standardized daily emission rate profi le, so that deposition results for diff erent days would refl ect diff erences only in weather conditions, not emission rates. The sum of simulated pollen deposition at a ring 3 km from the plot center was evaluated for each day, and the day with the lowest and highest deposition sums were selected. For each of these two dates, two source fi eld sizes were selected for the simulations (2.4 and 25 ha) to match the 10th and 90th percentile sizes of creeping bentgrass seed production fi elds in Oregon. We also modeled dispersal and deposition of pollen occurring from a “dust devil”, a type of vortex thermal updraft (Hess and Spillane, 1990; Sinclair, 1969) that occurs commonly in the grass seed-producing region in summer. For each simulation scenario, CALPUFF model outputs were obtained for mass balance, with compartments for pollen emitted, deposited on the surface, and airborne in or above the mixed boundary layer. Outputs were produced also for spatially explicit (gridded) deposition to the surface, allowing us to map deposition isopleths in units of pollen grains per m2. Pollen survival dynamics were incorporated into the model by estimating travel time to each deposition location. The surface distance from the source center to each deposition grid point was calculated, and divided by mean wind speed (6.7 m height) recorded during each 1-h time step to produce the estimate of time required for the pollen grains to reach that location. A negative exponential equation for pollen survival as a function of time, derived from our experiments, was then used to calculate the fraction of the pollen still viable when it reached that location. Survival fractions thus obtained were applied to estimates of pollen deposition at each grid point to produce the gridded data of the viable pollen for fi nal mapping.