Sampling

In order to satisfy the objectives of characterizing deposited particles and determining their spatial deposition patterns, four sets of leaf and filter samples were collected (Table 3-1): 1) mesquite leaves from twenty trees within the Phoenix metropolitan area on two days; 2) eight leaves from one tree at one time; 3) six sets of filter and leaf samples from one tree over a period of three weeks; and 4) filters that were placed in ten trees for six days. For each sample, leaves or filters were collected or placed near the edges of the trees at approximately 2.5 m above the ground with few or no leaves above them. All samples were analyzed with an electron microprobe and subjected to multivariate analysis.

Mesquite leaves were collected on June 19 and 29, 2001, from twenty locations within the Phoenix metropolitan area (Figure 3-1). The leaves were dried at room temperature for at least two weeks. Square portions of each leaf, approximately 2 mm per side, were affixed to an aluminum stub with conductive carbon tape. One leaf per tree was analyzed.

In order to determine the spatial variability of deposited particles on leaves within one tree, eight leaves were collected from a tree on June 8, 2002. The leaves were collected from the north, northeast, east, southeast, south, southwest, west, and northwest sides of the tree. A Wilcoxon Signed Ranks test was performed on the concentrations of 66

the analyzed particle groups. We chose the Wilcoxon Signed Ranks test because it is a nonparametric test in which the distribution of the sample does not need to be known.

The third data set involved the simultaneous collection of leaves and filters from one tree during April 2001. The purpose of this data collection was to compare the particle types between the leaves and filters in order to identify the particles that were deposited. Because the filter was clean when it was placed in the tree, the particles on the filter were assumed to be deposited from the area surrounding the exposed filter. Particles that were observed on both the filter and leaves were considered deposited. Particle types that differed between the two surfaces were examined to determine their origins.

For this collection, a square piece of Nuclepore polycarbonate filter, approximately 5 mm on a side, was affixed to an aluminum stub and placed on a mesquite tree branch. Polycarbonate filters were chosen because they are commonly used in particle samplers for individual-particle analysis (Post and Buseck, 1984; Katrinak et al., 1995; Anderson et al., 1996; Gao and Anderson, 2001). The filter remained in the tree from two to seven days. When the filter was collected, leaves near the filter were also collected, and a clean filter was placed into the tree. Six filter and six leaf samples were collected over a three-week period. The resulting compositions and sizes of the particle groups on the leaves and filters were compared.

The purpose of the fourth set of data was to obtain spatial deposition patterns of particle groups collected on filters. These results were then compared with those 67

collected on leaves. Polycarbonate filters were affixed to aluminum stubs and placed in trees at ten sampling locations from October 29 to November 4, 2002 (Figure 3-1).

Automated Electron Microprobe Analysis

Before analysis in the microprobe, all leaf and filter samples were carbon-coated to improve conductivity. The individual-particle analysis technique was similar to that used in previous studies (Anderson et al., 1992, 1996; Katrinak et al., 1995). We used an automated JEOL Model JXA-8600 electron microprobe with NORAN Voyager-IV image analysis and automation and equipped with a NORAN Be-window energy-dispersive x-ray spectrometer (EDS).

We used the following operating conditions: accelerating voltage - 15 kV, beam current - 400 pA, working distance - 11 mm, and magnification - 4,000x. X-ray spectra were acquired for 60 seconds at energies between 0 - 20 keV; relative dead times were less than 30%. We analyzed 50 to 400 particles per sample.

Back-scattered electron (BSE) images were obtained with an annular, split-ring, semi-conductor detector, mounted 11 mm from the sample. BSE images were preferred over secondary-electron images because the atomic number contrast is greater, little topographic information is revealed, and the effects of electron-beam charging are reduced, all of which are important in the analysis of individual particles (Germani and Buseck, 1991). The greater atomic number contrast was especially important on the leaves, where the greater contrast allowed for the analysis of particles with similar 68

compositions to the leaves. The resolution for each image frame was 1,024 by 1,024 pixels, and each pixel width was 0.022 μm. The final BSE image was produced from the average of three images, each with dwell times of 3,200 microseconds per pixel.

The analyses were performed on the smooth portions of the leaves to maintain the focus in the images, a necessary condition for the analysis. Although in previous studies with oak leaves, more particles were observed on surface irregularities, such as veins, than on smooth portions of leaves (Lindberg et al., 1982; Coe and Lindberg, 1987), we did not notice a significant difference on mesquite leaves. The upward-facing surfaces of the leaves were analyzed, as more particles were observed there, as opposed to the downward-facing surfaces (Lindberg and Lovett, 1985; Coe and Lindberg, 1987; Nicholson, 1988).

The particles were distinguished from the leaf and filter surfaces by segmentation of the BSE images into binary images. In this process, the gray-level for each pixel was compared with a user-defined threshold gray-level. The pixels with values greater than the threshold value were assigned one; particles appeared in these pixels. The pixels with gray-levels less than the threshold value were assigned zero; these pixels contained the leaf and filter surfaces. At times, the segmentation procedure produced artifacts from the leaf, because, in some cases, leaf material was slightly above the threshold; in these instances, the composition data was useful for distinguishing artifacts from true particles.

Size and chemical analyses were performed for each particle. In the size analysis, we determined the area, length, width, perimeter, aspect ratio, circularity, and location for each particle. For the chemical analysis, a rectangular area surrounding the particle was 69

analyzed; the longest side of the rectangle equaled the length of the particle. Because a small fraction of the rectangular area consisted of the sampling substrate, either leaf or filter, x-rays were emitted from the sampling surface. For most analyses, these x-ray counts were low compared to those for the particles; although, for some of the smaller particles on the leaf samples, elements from the leaf did appear in the analysis results.

The x-ray spectra were quantified by comparing each particle spectrum with standard spectra for 24 elements: Na, Mg, Al, Si, P, S, Cl, K, Ca, Ti, Cr, Mn, Fe, Br, I, Ba, Pb, Ni, A, Sn, Sb, Se, Zn, and Cu. The quantified spectra were reported in unnormalized weight percents. With the Be window on the EDS, energies from elements lighter than Na could not be detected.

Detection limits were applied to the weight percents of each particle before input to a cluster routine. The weight percents below the specified detection limits were given a value of zero, whereas those above the limits did not change. The use of detection limits is essential for cluster analysis of the data.