Rock varnish is a thin dark coating best known from deserts, and is believed to grow extremely slowly. Varnish samples from near Socorro, New Mexico (United States), contain as much as 3.7% PbO, derived from nearby smelters operating from A.D. 1870 to 1931. Additional varnish, measuring as much as 4 μm beyond the Pb-rich layer, indicates continued growth from 1931 to 2003. Comparison with other varnish confirms that the Pb is not an artifact. Based on Pb layer thickness, and the period of smelter operation, these very young rock varnishes yield growth rates of 28–639 μm/k.y., substantially higher than previously documented fastest rates of 40 μm/k.y. These rates imply that the average 1–2 μm/k.y. rate for older varnish is not the active growth rate. Rather, it is a long-term value including periods of nondeposition, erosion, and active growth. Therefore, models of rock varnish formation should be reevaluated with consideration of much faster maximum growth rates.


Rock, or desert, varnish is a thin veneer (<200 μm) of manganese and/or iron oxides interlaminated with clays and often silica (Dorn, 1991, 2007; Liu and Broecker, 2000). Although exact mechanisms of rock varnish formation remain controversial, most workers agree on some combination of airborne dust supplying materials and microbial concentration of manganese, modified by considerable diagenesis (Liu et al., 2000; Garvie et al., 2008; Northup et al., 2010; Dorn and Krinsley, 2011). Additional background material can be found in the GSA Data Repository1.

Archaeologists are interested in dating varnish in order to date petroglyphs carved by ancient cultures (Stasack et al., 1996; Dietzel et al., 2008). Geomorphologists have attempted to date landforms and surfaces using rock varnish (Friend et al., 2000; French and Guglielmin, 2002). However, direct dating attempts have proven unsuccessful so far (Watchman, 2000).

Using careful dating of the underlying surface, Liu and Broecker (2000) documented varnish growth rates of 1–40 μm/k.y. in the southwestern United States. The oldest samples have the slowest rates, while the youngest sample (1.5 ka) has the fastest rate, 40 μm/k.y. (Liu and Broecker, 2000). Here we document extremely young varnish that records anthropogenic Pb from historic smelters with substantially higher growth rates. These rapid rates have important implications for rock varnish formation.


Six samples of black varnish on rhyolite were collected near Socorro, New Mexico (United States; Fig. 1; Figs. DR1 and DR2 in the Data Repository); four of these were chosen for detailed analysis. This location is in the vicinity of historic smelters that operated from ca. A.D. 1870 to 1931 (Fig. 1; Evelith, 1983). The study site is ∼11 km southwest of the Billing–Rio Grande Smelter site in Socorro, and ∼25 km southeast of several smaller smelters in Kelly and Magdalena (Fig. 1).

For comparison, two additional locations were analyzed (Fig. 1). A varnished granite from the Kelso Mountains, Mojave Desert (California), was collected 28 km from the nearest major highway, but the site is centrally located between two major metropolitan areas, the Los Angeles Basin and Las Vegas (Nevada). A siltstone from the Jurassic Morrison Formation, collected west of Hanksville in central Utah (Fig. 1), is the most remote site, 50 km from any major highway, with no known significant mining or smelting activity within 100 km, and no metropolitan area within 200 km.


Samples were first examined using scanning electron microscopy (SEM) on varnish and host rock fragments coated with a 60:40 alloy of Au and Pd prior to imaging in a JEOL 5800LV SEM.

Four samples, Socorro sites 2, 5, 6, and 7, were selected for electron microprobe analysis (EMP), along with one sample each from Hanksville and the Mojave Desert. Separate fragments from each sample were embedded in epoxy, cut with a diamond saw to expose undisturbed varnish, and polished. Elemental line profiles and continuous line scans across the varnish were acquired on a fully automated JEOL 8200 electron microprobe equipped with five wavelength-dispersive X-ray spectrometers and backscattered electron imaging. For elemental line profiles, quantitative analyses were taken every 1.5–2 μm across the full varnish thickness. Operating parameters include a beam current of 30 nA, with a 20 s peak counting time for major elements (Al, Si, Mn, Fe), 30–40 s on minor elements (Na, Mg, K, Ca, Co, Ni, Cu, Zn, Ba), and 60 s on Pb. Calibration and precision and/or accuracy checks used natural mineral standards. The last point of each traverse was 1–2 μm from the edge, in order to contain the X-ray interaction volume.

To refine Pb distribution, continuous stage-stepped elemental line scans utilized a reduced accelerating voltage of 10 kV and 20 nA to improve spatial resolution. The interaction volume for the Pb M-alpha line at 10 kV, using the density of the manganese mineral birnessite at 3 g/cm3, is calculated to be 0.9 μm. The stage was stepped at 0.2–0.3 μm/step, with a 1000–2500 ms dwell/step, ending beyond the varnish edge. This method yields relative but not quantitative peak heights.


Scanning Electron Microscopy

The Socorro samples are incompletely covered with varnish in a patchwork of botryoidal protrusions, flat layered sections, and chaotic mixtures of these fabrics, all varying from <10 to ∼200 μm thick. The botryoidal towers protrude a few tens of micrometers above thicker varnish areas (Figs. 2A and 2G). Layered sections are typically thinner than botryoidal sections, ranging from <40 to 60 μm (Figs. 2D and 2J). The Hanksville sample has isolated patches of botryoidal varnish as thick as 40 μm in surface pits >500 μm deep. The Mojave sample has more continuous layered varnish that thickens from <10 μm to >100 μm in broad basins.

Electron Microprobe Analysis

Two areas from the Socorro site 2 sample were analyzed: a botryoidal varnish and a layered specimen (Figs. 2A and 2D). Both show SiO2 values that vary inversely with both MnO (Figs. 2B and 2E) and Fe2O3 (Tables DR1 and DR2). The PbO values are ≤0.6%, except near the surface. In the botryoidal varnish, the PbO values climb to 3.7% at 5–6 μm from the outside edge, and then drop to 0.8% at the surface (Fig. 2B; Table DR1), while in the layered varnish, PbO values rise to 3.4% (Fig. 2E; Table DR2) in the last few micrometers near the outer edge. With the better resolution of elemental line scans, the Pb in the botryoidal varnish separates into two individual peaks, with the center between the two peaks at a higher Pb value than the background (Fig. 2C). The last 4 μm of varnish lacks Pb (Fig. 2C). In the layered varnish, there is a single 5-μm-wide Pb peak extending to the edge of the varnish (Fig. 2F). The site 6 layered varnish and site 7 botryoidal varnish (not illustrated) are similar to the site 2 layered varnish (Fig. 2F), with a PbO peak of 1.6% (site 6) and 2.3% (site 7) over the last 4 μm.

A botryoidal varnish and a layered varnish were also analyzed from the site 5 sample (Figs. 2G and 2J). Again, the SiO2 values vary inversely with both MnO (Figs. 2H and 2K) and Fe2O3 (not shown). The botryoidal varnish PbO values are ≤0.07%, except for the last 10–15 μm where the values jump to 0.9%, and then drop to 0.6% at the edge (Fig. 2H). The layered varnish has PbO values ≤0.2% for the entire thickness of the varnish (Fig. 2K). The elemental line scans show a double Pb peak for the botryoidal varnish with a total thickness of 27 μm, whereas the layered varnish has consistently low values. Both samples exhibit an ∼10 μm layer of varnish at the outer edge in which Pb and Mn are absent and only Si is present (Figs. 2I and 2L).

The Hanksville botryoidal varnish (Fig. 3A) and Mojave layered varnish (Fig. 3B) show similar SiO2 values that vary inversely with MnO and Fe2O3. The PbO values are generally ≤0.1%. The transect across the Hanksville varnish shows an increase in PbO to 0.3% within 6 μm of the edge, then a decrease back to ≤0.1% at the edge. The Mojave varnish similarly has a larger PbO peak (0.9%) within 5 μm of the edge, extending to the very edge of the varnish. Other areas scanned, but not shown here, exhibit no PbO peak at all. In the elemental line scans, the Pb peak in the Hanksville varnish is 5 μm wide, extending beyond the Mn-rich varnish into a silica layer at the outer edge (Fig. 3B). The Mojave varnish has a >5-μm-wide Pb peak that extends to the edge of the varnish (Fig. 3D).


A number of previous workers attributed small amounts of Pb, or other contaminants, in the outermost layers of varnish to anthropogenic sources (Dorn, 1998, 2010; Fleisher et al., 1999; Broecker and Liu, 2001; Thiagarajan and Lee, 2004; Hodge et al., 2005; Wayne et al., 2006; Hoar et al., 2011). The PbO concentration (to 3.7%) found in the Socorro varnish is at least four times higher than that found in either the Mojave or Hanksville samples, and requires a more concentrated local source as an adequate explanation; i.e., the historic smelters. Because of the surrounding mountain ranges, there is no dominant wind direction in Socorro. Therefore, the Pb signal could have come from either of the smelters (Kelly or Socorro; Fig. 1), and most likely from both.

The botryoidal varnish from Socorro site 2 shows two clearly separated peaks (Figs. 2B and 2C), suggesting variable Pb production over time. Smelter activity reached its highest output from A.D. 1883 to 1894, followed by a revival in smelting of Pb-rich Zn ores from 1904 to 1931 (Evelith, 1983). However, the second peak is the larger (Figs. 2B and 2C), even though the earlier Pb smelting should have produced more atmospheric Pb because it was closer to the sampling site. Other sites in the same area show a buried Pb-layer but do not show this double peak, suggesting that there is variable preservation. Therefore, individual Pb peaks cannot be conclusively attributed to specific smelters or time periods.

The Socorro sites 2 and 5 botryoidal samples also show clear Mn-rich varnish growth following Pb peaks (Figs. 2B, 2C, 2H, and 2L). This is interpreted as new growth after the 1931 smelter closure, and indicates that the Pb peak thicknesses are real in this case, and not analytical artifacts, as suggested for other situations (Broecker and Liu, 2001). The Hanksville sample and the Socorro site 5 samples show silica growth beyond the Pb layer as a glaze over the Mn-rich varnish, similar to the black opal reported by Perry et al. (2006). The Pb in the Hanksville and Mojave samples is likely from aerosol deposition of automobile exhaust, because no other airborne Pb sources are near these sample sites. Tetraethyl Pb was phased out of gasoline after 60 yr of use, and by 1986, most lead in gas was removed (Kovarik, 2005). Thus, the Pb layer in the Hanksville and Mojave samples represents Pb deposited over about the same duration as the Socorro smelter deposition, but ending in 1986. Because the silica covers this anthropogenic Pb, we propose that the silica growth has occurred very recently. The recent Pb-free growth also suggests that the Pb accumulated during growth of the varnish, rather than moving down later during diagenesis (Dorn and Krinsley, 2011). If diagenesis had scavenged Pb from the surface, there should still be Pb remaining in the surface layer. For those samples without a surface Pb-free layer, diagenetic alteration cannot be ruled out, but seems unlikely.


Liu and Broecker (2000) calculated desert varnish growth rates on well-dated surfaces. Their youngest varnish was 1.5 ka, giving their highest growth rate of 40 μm/k.y. In contrast, all of their samples older than 50 k.y. have rates ≤2.0 μm/k.y. The anthropogenic Socorro Pb layer allows us to push the youngest varnish to early 20th century mining, the only possible source of such high Pb values. This will allow a better estimate of short-term varnish growth rates.

The microprobe line plots provide better resolution for the position of the Pb peaks (Figs. 2C, 2F, 2I, and 2L), so they are used to calculate growth rates. In the site 2 example, the width of the Pb peak varies from 5 μm in the laminated area to 39 μm in one of the botryoidal areas (Figs. 2C, 2F). Considering the entire 61 yr smelting period, the laminated area yields a growth rate of 82 μm/k.y. and the botryoidal area has a rate of 639 μm/k.y. Post-smelter varnish thicknesses of 2.5 and 4 μm formed in 72 yr (1931 to our sampling in 2003), giving rates of 35 and 56 μm/k.y., respectively. For site 5, the layered sample has no Pb, while the botryoidal area has a 27-μm-thick Pb layer, followed by 11 μm of post-Pb varnish. This gives a growth rate of 422 μm/k.y. for the botryoidal varnish during the smelter era, and 153 μm/k.y. in the post-smelter years. The absence of a Pb peak in the layered sample implies no growth or at least no Pb preservation since 1870. Thus, varnish growth is highly irregular, even on adjacent samples.

These rates are substantially greater than published estimates, and lead to questions about applying long-term integrated growth rates to recent rock varnish. Within the data set of Liu and Broecker (2000), the highest growth rates (14–40 μm/k.y.) are from the youngest varnish (<10 k.y.). The data from Socorro follow this same trend, giving rates that are several times higher than the previous maximum. Liu and Broecker (2000) suggested that slower-growing varnish has a greater preservation potential. However, if growth rates vary on decadal scales, as suggested here, then any long-term growth rate likely includes periods of non-deposition, effectively microscopic disconformities. Thus, the commonly cited rates of microns per millennium (Broecker and Liu, 2001; Dorn 2007; Krinsley et al., 2009) may be an accurate average preservation rate, but can lead to incorrect assumptions about the rate of varnish-forming processes. Based on the limited samples analyzed here, actively forming varnish appears to be able to grow at rates sufficient to create the entire thickness of a typical coating (<200 μm) in mere centuries. This does not mean, however, that all areas must accumulate at these rates, as some samples apparently did not grow at all during the historic period. Krinsley et al. (2012) also found much higher growth rates for varnish in humid environments, which could also apply to wetter areas in desert environments. Thus, varnish formation models are not limited only to processes that operate at 1–2 μm/k.y., as suggested by Dorn (2007).

The total thickness of varnish rarely exceeds 200 μm (Liu and Broecker, 2000); therefore, the average calculated rate is not the actual rate during active growth. Because older varnish has such consistently low growth rates (Liu and Broecker, 2000), this can be interpreted as indicating very slow processes (Dorn, 2007), or as a long-term average net result of both growth and hiatuses (Krinsley, 1998; Dragovich, 1994; Liu and Broecker, 2000; this study). The highest rate documented here (639 μm/k.y.) is clearly not sustainable over long periods (even in these samples), but must be considered when evaluating varnish models. Even at the higher rates documented here, varnish remains the slowest of sedimentary processes (Liu and Broecker, 2000); however, like most such processes, what is recorded by the deposit is only a small part of the time of formation.


Historic lead smelters around Socorro, New Mexico, provide a time-marker layer (A.D. 1870–1931) in rock varnish that offers an opportunity to directly date varnish formation. Two samples also show additional varnish growth after the smelters closed. The growth rates determined from these varnish samples range from 28 to 639 μm/k.y., significantly exceeding the highest previously recorded rate of 40 μm/k.y. The highest growth rates found by other workers (e.g., Liu and Broecker, 2000) are on young rocks (1.5 ka), whereas older substrates yield slower growth rates (1–2 μm/k.y.). Our data suggest that active varnish growth rates are much higher than previously thought, since dating of varnish from the age of the substrate rock can provide only an average rate. The well-documented lower growth rate of older varnish is a long-term average that may include periods of deposition, nondeposition, and/or erosion, and therefore these older surfaces will yield varnish deposition rates that appear to be slower. Models of varnish formation therefore need to be reevaluated with potentially faster growth rates in mind. This is particularly important for attempts at climate reconstruction involving varnish, assessments of biological contributions to varnish growth, interpretation of archaeological art created in varnish by ancient peoples, and potentially, interpretation of rock varnish on extraterrestrial materials like Martian rocks examined in recent missions.

This work was funded by the National Science Foundation (grants EAR-0311932 to Northup and EAR-0311930 to Boston). We thank the U.S. National Park Service for a sampling permit at Mojave National Preserve (issued to Chris McKay, NASA Ames Research Center, California).

1GSA Data Repository item 2013065, supplemental background, Figures DR1 and DR2, and Tables DR1 and DR2, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.