Biomonitoring of magnetic properties of tree leaves has been postulated to be a good approach to measure particulate matter (PM) pollution levels. We studied the variation of magnetic hysteresis parameters on leaves of Quercus ilex, an evergreen oak previously used for magnetic biomonitoring of air pollution in Rome (Italy). The hysteresis parameters (MRS, MS, BCR and BC) measured on specimens collected at a close spacing on the surface of two single leaves show variances that are smaller than those observed on a collection of Q. ilex leaves sampled from several trees distributed along high-traffic roads. The variability is higher for magnetizations than for coercivities. This suggests a uniform source for the magnetic particles, such that variations are due mainly to changes in concentration. The normalized hysteresis cycles are remarkably similar for all the specimens. Normalization of magnetic moments by mass appears however more efficient than normalization by volume.
 Particulate matter (PM) is not a new problem, but an increase in emissions, particularly along major roadways, has increased the awareness of the general population, especially with respect to potential health hazards. Many European cities have promoted advances in monitoring PM levels following the Council Directive No 99/30/EC relating to value limits for PM in ambient air. In order to establish guidelines for monitoring and controlling PM output, a better understanding of its origin, composition and dangers to health is required. Using the magnetic properties of tree leaves as a biomonitor to delineate PM content is a powerful tool that has been used in previous studies in different European cities [Matzka and Maher, 1999; Hanesch et al., 2003; Moreno et al., 2003; Urbat et al., 2004]. However, results between cities cannot be easily compared to one another due to different trees and methods used. Moreover, at present there is a poor understanding of the detailed distribution of magnetic particulate matter in tree leaves. Magnetic homogeneity is a necessary requirement in order to assess the significance of magnetic results from small specimens taken from leaves. An improved understanding of content, composition and grain size distribution of magnetic particles in a tree leaf would help in the planning of magnetic measurements to be carried out in magnetic biomonitoring studies. In this paper we report on the analysis of variations in magnetic hysteresis parameters on the surface of two individual leaves of Quercus ilex, a tree formerly used for magnetic biomonitoring of PM in Rome (Italy) [Moreno et al., 2003]. We show that hysteresis parameters are characterized by a relatively narrow range of variation, with standard deviations significantly smaller than those observed on a larger data set of Q. ilex leaves collected from several trees distributed along high-traffic roads throughout Rome. The data provide the first experimental evidence to assess the homogeneity of magnetic parameters throughout leaf samples and indicate uncertainties associated with detailed magnetic measurements on small size tree leaf specimens.
2. Location and Methods
Quercus ilex is an evergreen oak tree very common in the central and southern parts of Italy and around the Mediterranean region. It grows in mild climates and is a typical perennial constituent of the Mediterranean shrub and woodlands commonly known as macchia. The domestic Italian name of Q. ilex is Leccio and it is a resistant plant that can grow on poor soils or in adverse conditions, such as polluted city areas. It was previously used for forest cultures or furniture manufacturing. Since the 16th century, Q. ilex was planted as ornamental tree in Rome and its surroundings.
 Magnetic measurements were carried out at the Istituto Nazionale di Geofisica e Vulcanologia (INGV) in Rome. For a comprehensive study (manuscript in preparation), magnetic mass susceptibility was measured on an AGICO Kappabridge KLY-3 instrument on batches of six leaves each. For this study, we selected two Q. ilex leaves taken from a single tree (M50) along a road with high traffic volume in the center of Rome (Via Druso, 41°52.9′N 012°29.8′E). The average magnetic mass susceptibility from the M50 tree is χ = 4.5 · 10−7 m3/kg. Hysteresis cycles were measured on square specimens cut from these two leaf surfaces with a dimension of about 4 × 4 mm, using a Princeton Measurements Corporation (PMC) MicroMag 2900 instrument with the Alternating Gradient Magnetometer (AGM) coil. We also selected further 33 leaf specimens from various Q. ilex trees (one specimen per tree) widely distributed in the city of Rome. Due to the sensitivity limitations of the instrument, we took specimens selected from leaves with magnetic susceptibility values χ ≥ 2 · 10−7 m3/kg only. These leaves were collected from trees located along high-traffic roads.
 The AGM measurements provide data on the coercive force (BC), remanent coercive force (BCR), saturation remanent magnetization (MRS) and saturation magnetization (MS). Magnetization values were normalized by surface area of each specimen. For the specimens taken from one of the two leaves studied in detail (called “leaf 2” hereafter) we also computed mass-specific magnetizations. In order to improve the analysis of coercivity distribution for the magnetic PM accumulated on the leaf, we carried out analyses of first order reversal curves (FORC) on one of the samples of the entire data set. FORCs are a series of minor loops made after the sample magnetization is saturated [Pike et al., 1999; Roberts et al., 2000]. To better visualize the produced set of loops, they are transformed into contour plots by calculating the second derivative, usually referred to as FORC diagrams.
 The hysteresis loops of all the different specimens are very similar. The loops are characterized by a low coercivity phase, whose magnetization is saturated between 300 mT and 500 mT (Figure 1a). Values of BC and BCR for all specimens are in the range of 5–9 mT and 32–43 mT, respectively (Figures 2 and 3) . These values are in agreement with hysteresis parameters reported in earlier studies (Muxworthy et al.  referring to dust samples). The hysteresis ratios (BCR/BC and MRS/MS) show that all leaf specimens lie in the same region of the Day plot [Day et al., 1977]. This region is found between the theoretical curves for mixture of single domain (SD) and multidomain (MD) grains and of SD and superparamagnetic (SP) grains of magnetite [Dunlop, 2002] (Figure 1b).
 The individual hysteresis parameters show a larger variability in the collection of specimens from various trees than in the specimens cut from a single leaf (Table 1). With regards to specimens from a single leaf, the standard deviation of hysteresis parameters is on the order of 4–10% of the mean value for coercivities (BC and BCR) and of the order of 29–38% of the mean value for magnetizations normalized by the surface of the specimens (MS and MRS). In “leaf 2” normalization of remanence intensity by mass results in a decrease of the variability, i.e. of about 21% of the mean value for both MS and MRS (Figure 4). On the other hand, data from the collection of specimens from different trees show that the standard deviation of hysteresis parameters is on the order of 6–14% of the mean value for coercivities (BC and BCR) and on the order of 66–67% of the mean value for magnetizations (MS and MRS).
Table 1. Mean and Standard Deviation of Hysteresis Properties
Leaf_1 From Tree M50, N = 31
Leaf_2 From Tree M50, N = 32
Collection of Leaves From Various Trees, N = 33
 The data from the detailed study of single leaves indicate that the variability in the concentration of magnetic particles within a single leaf, though smaller than that found within the entire data set (Table 1), may result in differences up to a factor 7 for the MS and MRS values (Figures 2–4). No particular pattern is observed for hysteresis parameters measured for leaf 1 (Figure 2). Instead, a progressive increase in coercivity (BC) is evident for specimens taken from leaf 2, from “bottom” to “top” of the leaf surface (Figure 3). Moreover, sample g3 of leaf 2, which includes the midrib of the leaf, and i2, which includes a major vein, have surface MS and MRS values that are around double the mean of other samples in the leaf. The remaining specimens show less variation between their MS and MRS values. Note, however, MS and MRS values (in Am2/kg) are close to the mean of the leaf data set when considering magnetization values normalized by mass for leaf 2 (Figure 4).
 The FORC data from the selected specimen show that there is a broad density distribution of the interaction field close to the origin of the FORC diagram (Figure 5). Open contours that diverge toward the Bb axis are characteristic for MD material, while the main peak close to the origin of the diagram can indicate a significant presence of SP particles [Roberts et al., 2000]. In addition there is a tail that extends to higher coercivities (BC) which suggests the occurrence of SD-pseudo single domain (PSD) grains. The FORC data suggest that there is a broad distribution in grain sizes due to the large spread in the coercivity range. The narrowness of the FORC distribution suggests that these grains are not interacting with one another. It should be noted that the coercivity (BC) distribution extends to values that are higher than those expected for magnetite.
4. Discussion and Conclusion
 The results from this detailed analysis of the distribution of hysteresis parameters on two individual leaf surfaces of Q. ilex show that magnetic properties are very similar for the entire data set. Earlier studies examining variations in susceptibility and remanence of tree leaves assumed magnetite as the main carrier of the magnetic signal [Matzka and Maher, 1999; Moreno et al., 2003; Urbat et al., 2004]. The range of coercivity values are indicative of low-coercivity phases, which are similar to results reported by Muxworthy et al. , who suggested that maghemite and metallic iron (the latter linked to tram lines) are responsible for the observed magnetic properties. Muxworthy et al.  then proposed that the magnetic particles in particulate matter collected in Munich (Germany) consist of magnetite-like grains covered with an oxidized rim. Gómez-Paccard et al.  measured AC susceptibility and initial magnetization curves between 4.6 K and 300 K. They did not identify a Verwey transition in the AC susceptibility measurements, and attributed their signal to highly oxidized magnetite and/or maghemite. Urbat et al.  suggested that the iron sulphide pyrrhotite may be present together with magnetite on a study of pine needles in Cologne, Germany.
 The coercivities observed in this study extend beyond the range typical for magnetite. This may be due to oxidation of the particles, or incorporation of heavy metals into the crystal structure. Further work is needed to identify the magnetic phases on the leaves. Although the exact phase responsible for the magnetic parameters cannot be uniquely identified, the mean hysteresis properties on the collection of leaves are the same.
 We conclude that magnetic particles passively collected by leaves show comparatively smaller variation throughout the leaf surface and most likely originated from the same pollution sources. The comparison of surface and mass-specific magnetizations indicates that the magnetic particles are not only spread on the leaf surface but have to be more deeply incorporated in the leaf tissue. Magnetic biomonitoring provides useful proxies to delineate the degree of air pollution, as indicated by the amount of magnetic PM. The magnetic parameters from small specimens taken from large leaves were assumed to be representative of their overall magnetic PM content. However, this study shows that the natural variability of hysteresis parameters for small specimens cut from the same leaf is on the order of 5–10% for coercivities (as percentage of standard deviation with respect to the mean value) and of about 30–40% for surface magnetizations and 20% for mass-specific magnetizations. This should be taken into account when detailed rock magnetic experiments are carried out on small size leaf specimens. Sampling and measurement of large size samples (i.e., a collection of 6–8 leaves for each sampled tree) should average out the small scale variability observed in the individual leaves and appears as the best approach to delineate spatial PM air pollution in urban environments.
 MS thanks all the people who were involved with his original diploma thesis for their help and support; in particular Jaume Dinarès-Turell and the whole geomagnetics group of the INGV in Rome, Friedrich Heller for his initial engagement in the biomonitoring study, and Marco Bär for helpful suggestions and an instructive discussion. This is ETH contribution 1484.