Content uploaded by Jonas Sommar
Author content
All content in this area was uploaded by Jonas Sommar
Content may be subject to copyright.
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104, NO. D17, PAGES 21,879-21,888, SEPTEMBER 20, 1999
Increases in mercury emissions from desert soils
in response to rainfall and irrigation
S. E. Lindberg, H. Zhang, M. Gustin, A. Vette, F. Marsik, J. Owens, A. Casimir,
R. Ebinghaus, G. Edwards, C. Fitzgerald, J. Kemp, H. H. Kock, J. London,
M. Majewski, L. Poissant, M. Pilote, P. Rasmussen, F. Schaedlich,
D. Schneeberger, J. Sommar, R. Turner, D. Wallschl•iger, and Z. Xiao
Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee
Abstract. As part of an international Hg flux intercomparison at the Steamboat Springs,
Nevada, geothermal area, several dynamic soil flux chambers and micrometeorological
gradient systems were operated over desert soils in early September 1997. A series of
unanticipated convective rain cells impacted the site with the first rainfall in -90 days, and
the initial 4-cm rainfall increased soil moisture from -0.01 to 0.06% (vol/vol). Several
chambers were operating prior to the events, and two were deployed over wet soils
following rainfall. Rainfall resulted in an immediate and steep rise in ambient air Hg
concentrations and soil Hg emissions which persisted for 12-24 hours. Fluxes increased
most quickly and to a greater degree over the wettest soils, and the rate of increase was
related to chamber design and flushing rate. The flux response was also apparent in the
micrometeorological data. In general, soil emissions increased by an order of magnitude
following the rain, and reached levels -6 times above those at the same time the previous
day. These fluxes were significantly correlated with temperature, radiation, humidity, wind
speed, and soil moisture. After drying for -40 hours, selected soil plots were manually
irrigated with low-Hg-distilled water. Mercury emissions responded similarly across the
2 1
three treated sites, uniformly increasing from -60 ng m- h- pretreatment to -650 ng
m -2 h -1 posttreatment, which was a factor of-6 higher than adjacent control soils.
Possible causes of the increases in flux include soil gas displacement, desorption of Hg ø by
water molecules, and desorption of Hg(II) and subsequent reduction in solution. The
kinetics of the flux response, combined with local soil and climatic conditions, suggest that
Hg emissions were responding primarily to soil moisture and solar radiation. These data
have interesting implications for the role of changing regional climates on biogeochemical
cycling of Hg.
1. Introduction
The ability of Hg to volatilize from soils has long been
known [e.g., McCarthy et al., 1969], and the natural emission of
Hg from soils is an important contributor to the global Hg
cycle [Carpi and Lindberg, 1998]. Mercury exists in soils in
several forms, many of which (particularly Hg ø) exhibit rela-
tively high elevated vapor pressures at background tempera-
ture. The primary sources of Hg in soils which have not been
directly contaminated (e.g., by mining activities) include parent
and weathered minerals, geothermal precipitates, and atmo-
spheric deposition. Soils enriched in mercury by natural geo-
logic processes may contain concentrations of the order of
10ø-102/•g/g [e.g., Gustin et al., 1994], while background soils
are generally considered to contain Hg at levels <0.5/•g/g.
The factors influencing Hg emissions from soils range from
the apparent (soil temperature and mercury speciation) to the
unexpected (humidity and solar radiation [e.g., Poissant and
Casimir, 1998; Carpi and Lindberg, 1997]. The effect of tem-
perature has been most widely reported and quantitatively
studied. The vapor pressure of Hg ø increases exponentially
Copyright 1999 by the American Geophysical Union.
Paper number 1999JD900202.
0148-0227/99/1999 JD900202509.00
with temperature, and many studies have reported exponential
relationships between temperature and emission rates for
background, contaminated, and geologically enriched soils as
well as buried wastes [e.g., Lindberg and Turner, 1977; Lindberg
et al., 1979; Xiao et al., 1991; Lindberg et al., 1995; Gustin et al.,
1996]. Because of the strong temperature effect, few other
relationships have been as well documented, and the recent
suggestions of a direct effect of solar radiation on fluxes are
complicated by indirect temperature effects frdm soil heating
[Gustin et al., 1996; Carpi and Lindberg, 1998].
Among the "other" factors, soil moisture and rainfall effects
have been mentioned in recent studies, often anecdotally via
their influence on air concentrations. Wallschlager [1996] and
Schmolke et al. [1999] have measured increased Hg concentra-
tions in air immediately above the ground following rain
events, while Carpi and Lindberg [1998] reported a strong in-
crease in Hg emissions from background field soils following
the first rain after a 2-month drought. In the only known soil
moisture manipulation study, Hg fluxes were found to change
significantly over soils which were deprived of natural rainfall
for several weeks. Although adjacent control soils showed no
trends, the treated soils went from being a net Hg source prior
to drying (>30% moisture), to a Hg sink while dry (i.e., dry
deposition was measured to soils of <10% moisture), and back
21,879
21,880 LINDBERG ET AL.: MERCURY EMISSIONS FROM DESERT SOILS
to a net source after rewatering with distilled water to 25%
moisture [Advokaat and Lindberg, 1996].
The development of field-portable teflon flux chambers [Kim
and Lindberg, 1995] operated with automated analyzers for
Hg ø [e.g., Poissant and Casimir, 1998; Lindberg and Price, 1999]
has increased research on soil fluxes, and an intercomparison
of field methods was an important research need [Iverfeldt et
al., 1996]. From September 1-4, 1997, an international team of
researchers descended on the Steamboat Springs desert geo-
thermal area for a 4-day Hg flux intercomparison over soils
geologically enriched with Hg (the Nevada Study and Tests of
the Release of Mercury From Soils (STORMS) campaign
[Gustin et al., this issue (a)]. One objective of the study was to
determine the climatic factors influencing Hg fluxes over en-
riched soils, including soil temperature and solar radiation. An
unexpected factor for this desert setting was rain-induced ef-
fects on soil moisture. This paper reports results of flux cham-
ber measurements under prerainfall and postrainfall moisture
regimes, and of the response to manual irrigation of several
treatment plots for comparison with controls. Limited mi-
crometeorological data are included for comparison, but these
data and their relationship with the flux chamber data are
described in detail elsewhere in this volume [e.g., Gustin et al.,
this issue (a)].
2. Experimental Design, Methods, and Sites
Details on the Steamboat Springs, Nevada, site, measure-
ment locations, and methods used by each group, including the
flux chamber designs, are presented elsewhere in this volume
[e.g., Gustin et al., this issue (a)]. Each group operated either or
both flux chambers (FC) and micrometeorological systems to
quantify air/surface exchange rates for Hg; however, this paper
primarily describes the flux chamber data because the artificial
moisture manipulations were only possible over the limited
surface areas afforded by the FC. Although several FC designs
were used in the intercomparison, all groups used automated
mercury analyzers to measure Hg ø levels continuously at the
inlet and outlet of each FC, with a 5-min resolution [e.g.,
Lindberg and Price, 1999]. All but the Goteborg University
(GU) group used a Tekran 2537A analyzer (GU used a Gardis
analyzer [see Gustin et al., this issue (a)]. The FC data were
derived from four clusters of sites (FC within clusters were
located <5 m apart): (1) Frontier Geosciences (FG), Goteborg
University (GU), and GKSS Institute (GKSS); (2) Oak Ridge
National Laboratory (ORNL) and University of Michigan
(UM); (3) Environment Canada, Atmospheric Environment
Service (EC); and (4) University of Guelph (UG) (detailed
locations of all measurement sites are shown by Gustin et al.
[this issue (a)]. Briefly, there were four basic FC designs: the
ORNL all-teflon rectangular "box" (deployed by GKSS, GU,
UM, and ORNL [e.g., Carpi and Lindberg, 1998], a low rect-
angular polycarbonate pastry cover (FG), a square plexiglass
box lined with teflon (UG), and an opaque teflon-lined alumi-
num hemisphere (EC [Poissant and Casimir, 1998]). The latter
two designs employed small internal circulation fans. The FC
differed in surface area, volume, and air exchange rate, which
influenced their response times, as discussed below. An exten-
sive discussion of FC blank methods and results for the ORNL
and EC FC are published (blanks are generally <1% of the
mean daytime fluxes reported in this paper [Poissant and Cao
simir, 1998; Carpi and Lindberg, 1998]. Total Hg levels in the
soils at •ach plot were as follows (in /•g/g): FG/UG/GKSS
equal to 1.4-2.9; ORNL/UM equal to 3.6-4.1; EC equal to
4.6-4.7; UG equal to 2.8-4.0 [see Gustin et al., this issue (a)].
2.1. Uncontrolled Irrigation (Rain Event)
Although "it never rains in the desert" (M. S. Gustin, per-
sonal communication, 1996), it became apparent by 1100 hours
(local time) on September 2 that storms were approaching, and
a rain event of -2 cm occurred from -1220-1240. All FC were
in place and operating over soils by -0930-1000, prior to rain.
However, after some initial soil measurements, three of these
systems (ORNL, UG, UM) were removed from soils for sev-
eral hours of blank measurements. Because of the approaching
rain, the ORNL and UM FC were not again deployed until
some time after the rain event, over wet soils (1330 for ORNL,
1530 for UM), while the UG FC was again operating over still
dry soils by 1130. Hence, when the rain began, five FC were
already in place over dry soils (FG, UG, EC, GU, GKSS),
while two FC were deployed over wet soils after the event
(ORNL, UM). Once underway, all FC ran undisturbed for
-20 hours, which included a second rain period after sunset of
-1 cm, from 1900 to 1945. Visual inspection of the surround-
ing soils indicated that the wetting front from the first event
was isolated to the upper 2-3 cm, and all soils were still clearly
wet the following morning but quickly dried under full Sun by
midafternoon (by visual inspection).
2.2. Controlled Irrigation
Because of the response seen in the FC data after the first
rain event, a controlled soil wetting experiment was performed
on the final day, -40 hours after both rain events and when
soils were again thoroughly dry. This experiment was designed
in part to test the hypothesis that the Hg flux was due to
reduction of Hg 2+ delivered in the rain event itself. After -20
hours of undisturbed operation, and beginning at 1000 on
September 4, four of the FC plots were rapidly watered by
hand with low-Hg (<0.5 ng/L) distilled water to simulate the
moisture delivered by a 1-cm rain event (using the same wa-
tering device, plots were watered between 1000 and 1015 in
this order: UM, UG, EC, GKSS). Each FC was carefully lifted
from the soil, and water was applied uniformly for -1 min
followed by sealing of the FC over the wet soil surface. The
ORNL plot immediately adjacent to the UM plot was left
unwatered as a control. All FC were then operated without
further disturbance for 2-3 hours in full sunlight. Adjacent
open soil plots were watered similarly and monitored visually
for drying rate. Visible moisture was still apparent after 2
hours at a depth of 1 cm. Inspection of the original plots below
the FC prior to watering indicated that all visible traces of soil
moisture from the events on September 2, 1997, were gone.
3. Results
3.1. Response of Hg Fluxes to Precipitation Events
From the onset of the first rain event it was apparent from
the mercury analyzer readouts that precipitation was having an
effect on airborne Hg at the Steamboat Springs field site. The
effect was initially manifested in rising air concentrations as
illustrated in Figure l a which shows the 5-min Tekran data for
two sites •20 m apart. Trends in local climatology during the
rainfall events are shown in Figure lb. Each research group at
five different locations observed similar responses during and
immediately following the initial rain event, and Hg levels in
air remained above those measured during the prerain period
LINDBERG ET AL.' MERCURY EMISSIONS FROM DESERT SOILS 21,881
70
60
•E 50
& 40
m 30
o
o
• 20
•0
FC Inle_.•/2/97)
FC Inle_.•./1/97)
Gradient @ 100 cm
(9,•97•)
0 0 0 0 0 0 0 0 0 0 0 0 0 •0 0 0 0
Time
Figure la. Trends in ambient air concentrations of Hg be-
fore and after the rain event at ---1220 on September 2, 1997.
Flux chamber (FC) data provided by the UG were measured
15 cm above the soil at the FC inlet on September 1 prior to
the rain event and on September 2. Gradient data from EC
were measured at 100 cm above the soil on September 2 (also
at 245 cm, not shown).
for at least 30 hours following the second rain event. The mean
chambers, showing a remarkably similar pattern in the timing
of the fluxes following rain; a sharp increase then two or three
generally smaller peaks, followed by a gradual decline at night
to fluxes comparable to prerain values, and a slower increase
after sunrise the following day. Beginning at comparable prer-
ain fluxes throughout the site (overall mean 47 +_ 19 ng m -2
h -•, n = 27, for the period 0930-1200, see Table 1), soil Hg
emissions measured by all FC increased dramatically over the
next 2-3 hours beginning at ---1230. The mean postrain flux
measured across the area was ---290 ng m -2 h -• (___250, n =
86 for the period 1300-2000), but some sites responded more
dramatically. The responses fell into two groups: three FC at
three different sites (FG, EC, ORNL) exhibited peak fluxes of
---500-600 ng m -2 h -h, while the other three FC (GU, GK,
UG) reached peak fluxes of ---100-200 ng m -2 h -• over the
same period. The UM FC (not shown) operated only from
1530 to 1800 but measured a peak flux of ---1000 ng m -2 h -•
at 1530. These initial peaks were all followed by variable but
generally decreasing fluxes throughout the day. Following sun-
set at 1815, the fluxes all decreased during the night, but the
EC and FG FC exhibited an additional peak following the
second rain event at ---1930 (Figure 2).
As suggested by the peak fluxes, the overall means for the
air concentration measured throughout the Steamboat Springs ' 7-hour postrain period of elevated fluxes differed significantly
area by eight different systems increased from 7.6 _+ 9.3 ng/m 3
at 1200 prior to the event, to 31 _+ 25 ng/m 3 at 1630. There was
not a similar consistent increase in air concentrations following
the second rain event which fell on already wet soils after
sunset (mean Hg ø equals to 32 ___ 20 ng/m3). By 1830 the
following day the area mean air concentration had decreased
to 13 ___ 14 ng/m 3, but concentrations did not return to prerain
levels until ---1200 of the final day (September 4). Although
other climatic variables also varied over this period (Figure
lb), it is clear that the rain event was at least partially respon-
sible for the measured increase in ambient Hg concentrations.
As reflected in the concentration data, Hg emissions from
these desert soils responded rapidly to the initial rain event.
Figure 2 illustrates the trends measured with six different flux
between two groups of sites, with one group around 300-500
ng m -2 h -• and the second around 50-120 ng m -2 h -• (p •
0.01). The afternoon fluxes on September 2, 1997, were about
sixfold higher on average compared to the same period on the
previous day for the same sites (only GU, GKSS, and FG
operated on September 1; mean equal to 29 ___ 19, n = 34 for
1300-2000). The moisture effect appeared to extend into the
following day as well, with the mean flux across the area still
well above that before the rain event (mean equal to 190 +_
180, n = 64 for 1300-2000 on September 3, 1997). The soils
retained a wet appearance until midday on September 3, 1997.
Given the differences in sampling locations, FC design, soil
Hg levels, and flux measurement approaches, the similarity in
trends after the rain event are a strong indication of a common
150
100
5O
-5O
0.07
--- RH
• Global radiation (Rg)
-•- Soil heat flux (Os)
_ -&'- Soil water
//"-
//
-
I I I [ I I I [ I I I I I I t I
0.065
0.06
0.055
0.05
Figure lb. Trends in climatologic variables measured at the Steamboat Springs site by several groups
(postrain soil moisture data measured at EC site and prerain data estimated from Gustin et al. [1996],
atmospheric variables recorded at UM/ORNL site). Unfortunately, no group thought to operate a recording
rain gage in the desert. Soil moisture from L. Poissant (personal communication, 1997).
o o oo o o •o o o oo o o o 8 o o o o o• o o o o o • o • o o o •
o e• e• o o e,• e• o e• e• o e• o o e0 o • o o o e• o
Time (9/2/97)
-100 0.045
21,882 LINDBERG ET AL.: MERCURY EMISSIONS FROM DESERT SOILS
700
600 L
5OO
"' 400
cn300
x 200
m 100
-100
-200
-.-m- GU [] ORNL
0 w- w- w- w- w- w- w- w- w- w- (• (• (• (• 0 0 0 0 0 0 0 0 0 0
Time (9/2 to 9/3/97)
Figure 2. The response of Hg emission from soils to rainfall recorded by six flux chambers on different soil
plots at the Steamboat Springs geothermal area on September 2, 1997 (see text for site codes). For compar-
ison, mean fluxes during the previous day were --•30 ng m -2 h -• (during --•1300-2000) at the GU, GKSS, and
FG plots.
Table 1. Statistical Summary of Fluxes Measured in Response to Increased Soil Moisture Resulting From Natural Rainfall
Events (--•3 cm on September 2, 1997, at --•1215, See Text) and Manual Irrigation (--•2 cm Equivalent Precipitation on
September 4, 1997, at 1000)
Hg Flux, ng m-2 h-t
Preevent Mean Postevent Mean Peak Time of
Data Event ( + s.d.) ( + s.d.) N Flux Peak
Data from flux chambers
EC
GKSS
ORNL
UM b
FG
UG
GU
Micrometeorological data
EC c
UNR/USGS
UG
Data from flux chambers
EC
GKSS
UM
UG d
ORNL
GU
Micrometeorological data
ORNL
UG
Rain" (Sept. 2, 1997) 62 (28) 516 (60) 13 570 1500
Rain" (Sept. 2, 1997) '" 122 (42) 13 131 1530
Rain" (Sept. 2, 1997) 35 (.-.) 496 (66) 10 580 1500
Rain" (Sept. 2, 1997) '" 1058 (151) 4 988 1530
Rain" (Sept. 2, 1997) 42 (10) 283 (114) 13 461 1430
Rain" (Sept. 2, 1997) 61 (16) 54 (35) 13 119 1400
Rain" (Sept. 2. 1997• 42 (1o) lax rxzt• 13 189 1400
Rain" (Sept. 2, 1997)
Rain" (Sept. 2, 1997)
Rain" (Sept. 2, 1997)
Irrigation" (Sept. 4, 1997)
Irrigation" (Sept. 4, 1997)
Irrigation" (Sept. 4, 1997)
Irrigation" (Sept. 4, 1997)
Irrigation" (Sept. 4, 1997) (control) •'
Irrigation" (Sept. 4, 1997) (control)
Irrigation" (Sept. 4, 1997) (control)
Irrigation" (Sept. 4, 1997) (control)
ß " 1033 (685) 6 2364 1530
340 (810) 1400 (1782) 13 2028 1430
340 (360) 831 (402) 9 637 1500
160 (13) 768 (122) 5 842 1130
19 (15) 670 (196) 4 812 1130
50 (20) 602 (72) 5 590 1100
15 (.-.) 582 (305) 2 797 1100
21 (7) 114 (17) 5 132 1230
15 (13) 85 (44) 5 147 1100
250 (130) 439 (184) 4 670 1200
570 (220) 692 (450) 4 1240 1200
The preevent data represent the period 0930-1200 on September 2, 1997. Values should be considered accurate to two significant figures only
and are derived from 30-min mean fluxes.
"Rain response (postevent) period is 1300-2000 on September 2, 1997; irrigation postevent period is 1030-1230 on September 4, 1997.
bUM sampled from 1530 to 1700 only.
cEC sampled from 1530 to 1800 only.
dUG sampled from 1030 to 1100 only.
eControls received no irrigation (ORNL adjacent to MU; GU adjacent to GKSS).
LINDBERG ET AL.: MERCURY EMISSIONS FROM DESERT SOILS 21,883
400 1500
350 - • -m- FC Mean
ß
• 300 - •
• 1000 •)
• 250 - •
x 200 - •-
• 150 - •
ß -r 500 •
u.. 100 •
50
0 • • • • • 0
0930-1200 1200-1230 1230-1500 1530-1730 1830-2000 2000-0700 0700-0930
Time Period (9/2 to 9/3/97)
Figure 3. Trends in mean fluxes for seven flux chambers (FC) and three micrometeorological systems
operated during a period of rain. Shown are data for several sequential time intervals representing different
behavior (in order: prerain, rain, initial peak, second peak, second rain event, dark, postsunrise). Sample N
values are ---10-100 for FC data and ---5-20 for micrometeorological data. Relative standard errors of the
mean fluxes are of the order of 20% (FC) to 40% (micrometeorological). The large difference in fluxes
reported by the FC and micrometeorological approaches appears to be a function of the FC design, in
particular the flushing rates used which caused FC to underestimate soil fluxes due to extended FC turnover
times (H. Zhang and S. E. Lindberg, manuscript in preparation, 1999).
response of Hg flux to rainfall and the resulting increase in soil
moisture in these desert soils. The responses of the different
FC/soil plot combinations were significantly intercorrelated (p
<0.01), with fluxes from the most complete data period
(1430-0900, n = 38) showing correlations among FC ranging
from 0.44 (UG-EC, the two most different FC designs) to 0.92
(ORNL-GKSS, same FC design with two different flow rates).
Overall, the mean r values for each FC with all five other FC
ranged from 0.63 (UG) to 0.81 (ORNL) indicating that the FC
were all responding in a qualitatively similar fashion to the
effects of the rain event, despite the fact that the ORNL FC
was placed on already wet soil, while the others were in place
during the rain event.
Because of the similar behavior of the FC data, the possi-
bility of an artifact must be considered. Flux chambers by their
nature of being enclosed systems influence the exchange of
gases over the surface, particularly for static designs. From the
start it was readily apparent to the participants that the field
FC and micrometeorological data were providing quantita-
tively different fluxes over these soils (FC fluxes were consis-
tently lower, Figure 3). This is the subject of the accompanying
paper by Gustin et al. [this issue (a)]. Despite the differences in
magnitude, however, it is clear that both FC and micrometeo-
rological measurement systems responded similarly over time
to the rain event. Three of the FC designs showed significant
but low correlations with the most complete micrometeoro-
logical data from the UNRAJSGS site: r values ranged from
0.34 for ORNL (p < 0.05) to -0.47 for FG and UG (p
< 0.01). The temporal patterns are basically the same, but the
micrometeorological fluxes peaked before those measured by
the FC, probably as a result of the delayed response times of
the FC (turnover times ranged from 0.02 to 0.33 hours). How-
ever, the magnitudes of the fluxes are clearly different, with the
FC yielding mean fluxes generally 15-20% of the micrometeo-
rological fluxes during daylight periods and 40-100% after
dark. Recent chamber experiments in our laboratory have
demonstrated a strong dependence of measured fluxes on FC
flushing rates. Fluxes over these same soils in our laboratory
increased in proportion to flushing rate suggesting that FC
underestimate actual fluxes at low turnover frequencies (H.
Zhang and S. E. Lindberg, manuscript in preparation, 1999).
These differences are discussed in detail elsewhere [Gustin et
al., this issue (a)]. Despite these differences, the drastic effect
of the rain event on Hg fluxes is real and apparently occurred
across the entire Steamboat Springs sampling area, increasing
mean Hg fluxes by an average of 4-8 times, and persisting for
several hours. This is somewhat surprising considering that the
soil depth visibly affected by the initial rainfall was limited to
the upper few centimeters.
Although the trends in FC responses to the rain event were
similar, there were important differences in the magnitudes of
fluxes measured by the chambers, as well as in the rate of flux
increase following rainfall, dF/dt. These differences could be
the result of soil chemistry and Hg content, but our data sug-
gest they are not. We have previously noted that the response
of a FC to an outside stimulus (e.g., solar radiation) is limited
by the FC flushing rate and that systems with a slow turnover
time exhibit a delayed response [Carpi and Lindberg, 1997].
The computed values of dF/dt for the initial peak seen by each
chamber are strongly related to turnover time (t = FC volume/
flushing rate). Excluding the UG FC which suffered blank
problems, the variance in turnover time among FC explains
over 80% of the variance in dF/dt (Figure 4). In this regression
the ORNL, GKSS, and GU FC illustrate this trend particularly
well, as they are each of identical design [Kim and Lindberg,
1995], but operated with different flushing rates (5, 16, and
-20 Lpm, respectively).
Another important factor is the degree to which soil mois-
21,884 LINDBERG ET AL.: MERCURY EMISSIONS FROM DESERT SOILS
250
200
150
lOO
50
•ORNL/UM ß data
-- regression line (r2=0.9)
•EC
aUG
I I I
0 5 10 15 20 25
FC Turnover Time (min)
Figure 4. The rate of increase in soil flux following rain as
measured by several flux chambers plotted as a function of FC
turnover time (see text for site codes).
ture was increased. The OR FC falls above the regression line
as a result of its deployment over already wet soils following
the rainfall. The ORNL and UM chambers, which were placed
on adjacent soil plots after rainfall had ceased, exhibited
among the highest rates of increase and flux maxima (Figure 4
and Table 1). This suggests that the Hg response was related to
the degree of increased soil moisture. Following rainfall, soil
moisture in the upper 2 cm of open plots increased from
extreme aridity (<1%, vol/vol) to ---5-6% (Figure lb), but no
data are available for the plots covered by FC. Since soil water
appears to drive the increased fluxes, it is surprising that the
other FC plots exhibited temporally similar responses to the
ORNL and UM FC. However, the design of the various FC
that were in place prior to the rain probably influenced the
degree to which moisture penetrated the underlying soils. The
EC FC, which exhibited the highest mean flux, and also a peak
after the second rain event (unlike the other FC), was a 0.4 m
diameter hemisphere, a design conducive to transporting fall-
ing rain into the immediate vicinity of the underlying soil. This
site was also situated on a former stream outwash which ex-
hibited a coarse and highly porous texture which was amenable
to rapid diffusion of water into the soils beneath the chamber.
The FG FC was situated near the EC site, on finer soils, but
was of a design with a minimal footprint beyond the FC walls,
while the GKSS and GU FC both had a teflon skirt for sealing
the FC to the soil. These skirts reduce the transport of rain-
water to soils below the FC surfaces (the ORNL and UM
designs have the same skirts but were on already wet soils).
The fact that all chambers did respond to moisture increases in
the surrounding soils to some extent suggests that the effective
"footprint" of these chambers may extend beyond their actual
dimensions (this idea is also supported by reported correla-
tions between flux and solar radiation for the EC FC which is
opaque [Poissant and Casimir 1998]).
Once soil moisture was increased, the resulting fluxes may
have been influenced by a number of variables. Soil Hg content
was apparently not one of them. We found no significant re-
lationships between total soil Hg or soil Hg ø content and Hg
flux (we regressed both the mean and maximum fluxes for the
postrain period against soil concentrations for each plot; both
r < 0.45, p > 0.10, N = 7). However, for the 31-hour
period following rainfall (1230-0930), the fluxes measured by
all but the EC FC showed significant correlations (p < 0.01,
N > 40) with these variables: air temperature (mean r =
0.76, N - 5 FC), global radiation (r = 0.54), wind speed
(r = 0.61), soil moisture (r = 0.75), and relative humidity
(r = 0.73). The UNR/USGS micrometeorological data
showed lower but still significant correlations with these vari-
ables, while the opaque EC FC exhibited a comparable corre-
lation only with temperature (r = 0.49). Of the parameters
we measured, temperature is the one variable most often
found to show a strong relationship with soil Hg emissions
[e.g., Lindberg et al., 1979; Poissant and Casimir, 1998], and air
temperature consistently showed the best postrain correlations
(soil temperatures were not recorded continuously, but the
ORNL/UM soil plots were measured intermittently with a
probe at 2 cm depth both inside and outside of the FC, and
these temperatures exhibited a correlation coefficient of 0.99
with air temperature). Figure 5 shows the plot of soil flux
versus temperature for selected plots and suggests that Hg may
exhibit a hysteresis effect during this period (this effect is most
apparent for the FC with the fastest turnover times: FG,
ORNL, EC, and UG). Mercury flux over wet soils increases
with rising soil temperatures along the upper curve (---1300 to
1530) but shows a different response to falling temperature
along the lower curve (---1530 to 2000). This behavior suggests
that there may be two different processes influencing the re-
sponse to temperature, one during sunlight (rising limb) and
one in the dark, or that the kinetics of one process are light
sensitive.
It is also possible that these various correlations simply re-
flect similar diel cycles among the variables. When the daylight
period (1230-1930, N = 14) was examined alone, the corre-
lations generally decreased, and were significant for fewer FC
(primarily GKSS and ORNL); but those with solar radiation
were most consistent (mean r = 0.66, p < 0.01, N-- 5).
Flux increased in parallel with radiation after the first event
(compare Figures lb and 2), but the role of solar radiation in
controlling flux in these soils is uncertain [see Gustin et al., this
issue (b)].
700
600
500
• 400
E
•c 300
x
u. 200
lOO
-lOO
[] ORNL (r=.94)
e FG (r=.69)
ß GKSS (r=.82)
5 10 15 20 25 30
T (air)(øC)
Figure 5. The relationship between soil flux and air temper-
ature for three flux chambers (the correlations shown in the
text were computed from all data for each FC; only three sites
are illustrated here). The FG and ORNL data suggest a hys-
teresis effect. The low flux values along the x axis near 25øC are
from the prerain period (see text for site codes).
I I I I
LINDBERG ET AL.: MERCURY EMISSIONS FROM DESERT SOILS 21,885
900
8OO
700
600
500
400
300
200
lOO
-lOO
ORNL (control) GKSS UM
EC Tsoil
/
./
./
//
/ ß ß
I 45
40
35
- 30G
o
- 25 I_
- 2O
15
0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 ,- ,- ,- ,- w-
Time (9•4•97)
Figure 6. The response of Hg emission from soils to manual irrigation with low-Hg distilled water (<0.5
ng/L) recorded by three flux chambers on different soil plots, plus data from an adjacent control soil. The plots
were irrigated at 1000 hours. Soil temperature (T soil) within the flux chamber at 2 cm depth is also shown
(see text for site codes; the UG chamber reported a similar trend but did not operate continuously and is not
shown).
3.2. Response of Fluxes to Controlled Soil Irrigation
It has been hypothesized that rain can deliver a fresh pool of
reactive (e.g., easily reducible) Hg 2+ to soils or surface waters
which can be converted to Hg ø by organic matter (R. R.
Turner, Frontier Geosciences, Seattle, personal communica-
tion, 1998), and increased Hg evasion from water surfaces has
been noted after rain events in subtropical wetlands [Lindberg
et al., 1998]. To test the possibility that the composition of the
rain itself influenced the measured Hg fluxes, a simple test was
performed at several soil plots. On the final day of the inter-
comparison (---46 hours postrain), four plots were irrigated
with a volume of low-Hg distilled water (<0.5 ng/L) compara-
ble to the first rain event (UM, UG, GKSS, EC), while one plot
was maintained as a control (ORNL, adjacent to UM), and
fluxes were monitored for 2-3 hours. The response to irriga-
tion was remarkably similar among the four plot/FC combina-
tions (Figure 6) and paralleled that seen following the rain 2
days earlier. Within an hour of irrigation, fluxes in each plot
increased by an order of magnitude and by a nearly identical
factor over the control soil for the same period (treatment/
control equal to 6.1 _+ 0.6). This observation is not consistent
with a strong effect of rain chemistry and supports the concept
that moisture addition alone significantly enhances Hg fluxes.
Figure 7 illustrates the response of Hg flux to both natural
rainfall and artificial irrigation for three "matched" FC plots,
two of which had FC over dry soils prior to the rain (GKSS,
EC), and one which had the FC placed on wet soil after rain
(ORNL/UM). For the GKSS and EC plots the data show that
Hg flux over irrigated soils increases more rapidly and reaches
a higher plateau compared to the response after rainfall. This
difference is most dramatic for the GKSS plot and supports the
idea that much of the variance influx response after rainfall was
related to the amount of moisture actually reaching the soils
beneath the FC. Although the EC and GKSS FC remained
over the same plots throughout the study, the design of the EC
FC and greater soil porosity at that site combined to promote
diffusion of the rainwater into the underlying soils, while the
GKSS FC design and soil texture were less conducive to this
process, resulting in a much reduced response. However, upon
direct irrigation, the Hg flux over both soils responded simi-
larly. The ORNL and UM FC represent an interesting "pair"
because the positions of these FC were exchanged on Septem-
ber 3 between the rain and irrigation events, meaning that the
flux versus time response curve for each event (rain on Sep-
tember 2 and irrigation on September 4) represents the same
soil plot. These curves are plotted together in Figure 7 where
the response of Hg flux after irrigation (UM FC, 0 to 2.5 hours)
matches surprisingly well with that after rain on the same plot
(ORNL FC, 2 to 4.5 hours). This suggests that the delayed start
of the ORNL FC after rainfall on September 2 probably
missed the peak soil flux which occurred at most sites ---1.5-2.5
hours after the rain ended (Figure 2).
4. Discussion
There are several processes which could be responsible for
the increase in Hg flux following the addition of water to these
desert soils. The irrigation experiment with low-Hg-distilled
water does not support the hypothesis that a form of easily
reduced, reactive Hg is added to soils by rainfall in sufficient
amounts to cause the response. Rather, the increase in flux
appears to be related to soil physical or chemical interactions.
The shapes of the response curves of flux versus time (Figures
2, 6, and 7) suggest that the initial response to moisture may
exhibit first-order behavior. The 30-min mean fluxes suggest
that the response plateaus within 30-60 min after water is
added, but that most of the increase occurs in the first 30 min.
However, these data were averaged from 5- or 10-min fluxes,
and the 5-min raw data available to us for the irrigation ex-
periments (EC, UM, GKSS) show that the fluxes plateau at
---20 min and appear to follow a first-order function to that
point (data not shown).
Table 2 summarizes rate constants from the initial flux re-
sponse curves for the most complete data sets (EC (rain/
21,886 LINDBERG ET AL.: MERCURY EMISSIONS FROM DESERT SOILS
8OO
700
"' 600
E
• 500
x
= 400
z 300
• EC-rain GKSS-rain UM-irr
ORNL-rain EC-irr. GKSS-irr
2OO
lOO
0 I I I [ I
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time since moisture added to soil (h)
Figure 7. The relationship between Hg flux over soils and time since moisture was added to soils for both
rain and irrigation events (all sites showed these trends; a subset of sites are shown for comparison; see text
for site codes). Note that the fluxes for the 2 and 2.5 hour time periods from the UM-irr and ORNL-rain data
sets represent the same soil plot (the UM and ORNL FC swapped positions between the rain and irrigation
events; see text).
irrigation), GKSS (rain/irrigation), ORNL/UM (control/
irrigation), and FG (rain); we used the highest time resolution
data available to compute these values, generally 10-min flux-
es). These coefficients fall into two general ranges (Table 2):
k 2 < 0.04 (mean 0.033 + 0.007) for plots where moisture was
added indirectly by rain to soils surrounding the FC, and k 2 >
0.07 (mean 0.086 _+ 0.023) for plots where water was added
directly by irrigation of soils beneath the FC. The nonirrigated
control soil exhibited a much lower rate constant of (0.005) for
the response to increasing solar radiation and soil temperature.
Given the initial fluxes (k•) measured at each site prior to
increased soil moisture, we can compute flux doubling times
for each response curve:
t2• = [Ln2Fo) - Lnkl)]/k2.
Following direct irrigation, the soil fluxes doubled within 2 to
12 min (mean 7 _+ 4 min), compared to a much slower response
for the dry. control soils (t2x -'-90 min). When rain wetted the
soils indirectly outside of the FC footprint, fluxes doubled in
---20-30 min (Table 2).
Considering the rapid response time (---10 min) and its du-
ration (> 12 hours), both physical and chemical soil processes
probably contributed to the observed fluxes. We will briefly
consider the possible contributions of the following processes:
(1) physical displacement of Hg-enriched soil gas by the per-
colating rain water front, (2) exchange of Hg(0) adsorbed on
dry soil particle surfaces with (rain) water molecules, and (3)
desorption of Hg(II) adsorbed on soil solid particle surfaces
and its subsequent reduction. The hypothesis that Hg 2+ in
rainwater was reduced to create the observed fluxes is not
supported by the irrigation data.
Our data for Hg concentrations in ambient air showed two
separate increases during the rain event (Figure la), a short
(15 min) intense spike, followed by a second increase of longer
duration (80 min). We hypothesize that the first is the signal of
a physical displacement of Hg-enriched soil gas by the perco-
lating water front, while the second is the primary response to
increased soil moisture and represents a different process. The
transient nature of this initial spike make it difficult to measure
the resulting flux, and the FC with the highest time resolution
(FG) actually recorded a decreased flux during this spike (Fig-
ure 2). This decrease was probably an artifact of the rapid spike
in the ambient air Hg concentration at the FC inlet compared
to the slower FC turnover time.
We can, however, roughly approximate the flux that could
have resulted from the physical displacement of soil gas during
the initial spike. Concentrations of Hg ø in soil gas were not
measured in this study as there are no accepted field measure-
ment methods, and few comparable data exist for comparison.
Hence we estimated the contribution of the prerain soil gas
Hg ø using Fick's diffusion law [Hillel, 1982] as applied by John-
son and Lindberg [1995]. We assumed that meteorological ef-
fects are slight inside an enclosed flux chamber and estimated
the contribution of the soil gas Hg ø which was displaced by the
rain water as follows:
Table 2. Rate Constants for the Increase of Hg Flux Over
Wetted Soils
Soil Plot and Flux Doubling
Flux Chamber Event k• k2 Time, min
EC rain 4.7 0.036 20
FG rain 4.3 0.038 18
GKSS rain 2.2 0.025 27
EC irrigation 4.9 0.073 12
GKSS irrigation 4.2 0.119 2
UM irrigation 4.2 0.111 6
ORNL control 4.4 0.0052 90
Constants were computed from the initial slope of Ln (flux) versus
time response curves for both rain and irrigation events. The equation
is of the form F = k• x ek2t (see text).
LINDBERG ET AL.: MERCURY EMISSIONS FROM DESERT SOILS 21,887
F = -0.66(p - s)Do(C.• .... -- CHg .... lgas)/d
where F is the soil Hg flux (ng m -2 h-1), p is the soil porosity
(Vpore/VsoiO, S is the soil moisture saturation, D O is the Hg(0)
diffusion coefficient in the ambient air (0.13 cm 2 s -1 or 0.047
m 2 h -1 at 25øC) [Thibodeaux, 1996], CI-Ig-air is the Hg(0) con-
centration in the air immediately above the soil surface (ng
m-3), CHg-soil gas is the Hg(0) concentration in soil gas (ng
m-3), and d is the soil depth considered for the diffusion
gradient (cm) (taken as the depth of the rain wetting front).
The constant, 0.66, is a tortuosity coefficient, suggesting that
the apparent path is about two thirds the length of the real
average path of diffusion in the soil [Hillel, 1982]. Using prer-
ain measured mean values of CI•g-air (prerain) = 5 ng m -3 and
F (prerain) = 40 ng m -2 h -1, and assuming p = 50%, s =
1%, and d = 2 cm, we estimate the Ci-ig-soil gas (prerain) was
---60 ng m -3. The rain wetting front was seen to extend to ---2
cm in depth, and we estimate that the Hg ø concentration inside
the flux chamber attributed to displaced soil gas (CHg_chamber)
was '--3 ng m -3, a small fraction of the initial measured spike
(---35 ng m -3, Figure la). These data lead to an estimated flux
of the order of 10 ng m -2 h -1 for soil gas displacement which
would be only a few percent of the mean Hg emission mea-
sured over the area for the 2.5-hour period following the rain
event (260 _+ 190 ng m -2 h -•, n = 27). Although only a crude
estimate, this calculation suggests that gas displacement alone
could not account for a significant fraction of the emitted Hg.
It is more likely that the second process listed above has a
major contribution. The soil analyses [Gustin, et al., 1996, this
issue (a)] show that most of the soil plots contained a large
proportion of total Hg as Hg ø, ---30-50%. Since the soil mer-
--2
cury flux over prerain dry soils was relatively low (---40 ng m
h-•), we propose that before the rain, most of the Hg ø in the
soils was adsorbed to dry soil particle surfaces and not directly
available for emission. Various kinds of oxygen surface func-
tional groups on soil mineral particle surfaces have a higher
affinity for water molecules than for Hg ø atoms (as a class B
soft acid, Hg ø favors S and N groups [Schuster, 1991]), and the
adsorption of Hg ø on dry soil particles has been reported
[Fang, 1978; Klusman and Matoske, 1983]. It follows that per-
colation of rainwater into the soil pores led to exchange of the
water molecules with the Hg ø adsorbed on the previously dry
soil particles. As a result, the Hg ø was desorbed into soil gas
(and/or overlying air) creating a pool of "available" Hgø,"
which was emitted during the broad postrain peak of elevated
Hg flux (Figure 2). A similar phenomenon is known to influ-
ence volatile pesticides, which are also more strongly bound to
dry than wet soils [e.g., Spencer and Cliath, 1974]. In a study of
competitive sorption between VOCs and water in Nevada soils,
small additions of water drastically decreased VOC sorption,
leading to increased volatilization [Steinberg and Kreamer,
1993].
In laboratory studies, adsorption of Hg ø on soil particle
surfaces was found to depend on soil mineralogical composi-
tion and surface area [Fang, 1978; Klusman and Matoske,
1983]. The observation that measured Hg concentrations in the
clay and silt fractions of the soil samples from most sites were
higher than those of all other fractions (P. Rasmussen, Geo-
logical Survey, Canada, personal communication, 1998) pro-
vides further evidence that desorption of Hg ø could have been
a major contributing process to account for the sustained in-
crease in Hg flux. Upon drying, more soil particle surfaces
became available for Hg ø adsorption. Consequently, the Hg ø
readily available for diffusion and emission would be read-
sorbed, resulting in decreased flux upon soil drying, which
agrees well with our observations (Figure 3). Manual irrigation
2 days after the rain again liberated the adsorbed Hg ø, leading
to another emission pulse. Recent laboratory studies have con-
firmed our hypothesis: manual addition of water to dry Steam-
boat Springs soils (as well as background soils from Tennessee)
liberated significant Hg ø, while a similar addition of less polar
methanol to the Steamboat Springs soils elicited no emission
response (M. S. Gustin and H. Zhang, unpublished data,
1998).
Aqueous reactions involving Hg(II) in soil solution are also
a possible explanation. If there were sufficient Hg(II) present
in a form readily reduced by either biotic or abiotic processes,
then the addition of water to soils could enhance the produc-
tion of Hg ø. Such reactions would have been suppressed in the
prerain dry soils. The role of possible biotic activities stimu-
lated by rainwater still remains debatable, but the extreme
aridity of the prerain soil combined with the rapid flux re-
sponse seems to rule out a significant role of microbial pro-
cesses. However, there are several possible abiotic pathways,
and desorption of soil-bound Hg(II) and subsequent reduction
to Hg ø in soil solution must also be considered as contributing
processes. Once rainwater was added to the dry soils, desorp-
tion of Hg(II) could readily occur into the initially low-Hg
solution. In the presence of soil organic acids, dissolved dioxy-
gen, and photoactive Fe(III), Hg(II) could be reduced via
various pathways by reactions enhanced to different degrees in
sunlight (discussed by Zhang and Lindberg [this issue]). How-
ever, we suspect that these contributions should be smaller
than the desorption of Hg ø by the percolating water molecules
because the water-soluble Hg(II) fractions for all the sites were
shown to be quite small [Gustin et al., this issue (a, b)]. How-
ever, another candidate process could involve photosolubiliza-
tion of the cinnabar (HgS) present in these soils [Gustin et al.,
this issue (b)]. While this reaction directly generates Hg(II), it
has also been reported to result in reduction of Hg(II) and
subsequent volatilization of Hg ø in the laboratory during UV
photoirradiation [Okouchi and Sasaki, 1983]. Because of the
role of irradiation, this process would seem to be limited to the
surface most soil layer, but without further information this
process could not be ruled out as having contributed some Hg ø.
Clearly, further research is required to reveal the mecha-
nisms by which soil moisture affects mercury emission. These
observations illustrate the gaps in our current understanding of
Hg soil emission mechanisms, Hg ø desorption rates, and
Hg(II) reduction processes in soils. It will be difficult to rea-
sonably estimate the contribution of these various processes to
the rainfall response measured at Steamboat Springs without
much better data on soil Hg speciation and its role in soil
emissions. Even with such data, quantitative attribution of the
elevated flux to any single process or species would be tenuous
since the overall postrain Hg flux (although clearly elevated)
represented <<0.1% (flux/total Hg equal 1.4 x 10 -2 to 4 x
10 -6) of the total Hg pool available in the upper 2 cm of the
soils. This study and the other recent observations cited above
have clarified the importance of soil moisture in influencing Hg
emissions from both background and geologically enriched
soils. The dramatic increase in emissions of Hg following irri-
gation of these desert soils has interesting implications for the
role of changing climate regimes on regional biogeochemical
cycles of Hg. This is especially true if the response is generally
as reproducible as we demonstrated with our irrigation studies
21,888 LINDBERG ET AL.: MERCURY EMISSIONS FROM DESERT SOILS
after 2 days of soil drying, and considering the large pool of
apparently available mercury in these soils.
Acknowledgments. Research is sponsored by the Electric Power
Research Institute under contract with Oak Ridge National Labora-
tory (ORNL). ORNL is managed by Lockheed-Martin Energy Re-
search, Inc. for the U.S. Department of Energy. This paper was pre-
pared in part while the author was a visiting scientist at the Institute for
Physical and Chemical Analysis at the GKSS Research Center, Gees-
tacht, Germany. Authors affiliations during this study: S. L., H. Z.,
J. O. (ORNL); M. G. (University of Nevada, Reno); F. M., A. V.
(University of Michigan); A. C., L. P., M.P. (Atmospheric Environ-
ment Service, Toronto); R. E., H. K. (GKSS, Germany); C. F., J. K.,
G. E. (University of Guelph, Canada); P. R. (Geological Survey Can-
ada); M. M. (USGS, California); F. S., D. S. (Tekran Inc., Toronto);
J. S., Z. X. (Goteborg University, Sweden); J. L., R. T., D. W. (Frontier
Geosciences, Seattle). The authors would like to thank the Electric
Power Research Institute for the support for development and testing
of Hg flux methodologies which made this study possible. We would
also like to thank the UNR students and staff for significant field
assistance during Nevada STORMS, and G. Southworth for helpful
comments on the manuscript. This is publication 4870 Environmental
Sciences Division, ORNL.
References
Advokaat, E., and S. E. Lindberg, Effect of rainfall exclusion on air/
surface exchange rates of mercury over forest soils, in Abstract Book,
Fourth International Conference on Mercury as a Global Pollutant,
Hamburg, Germany, p. 458, GKSS Inst., Geestacht, Germany, 1996.
Carpi, A., and S. E. Lindberg, The sunlight mediated emission of
elemental mercury from soil amended with municipal sewage
sludge, Environ. Sci. Technol., 31, 2185-2091, 1997.
Carpi, A., and S. E. Lindberg, Application of a teflon dynamic flux
chamber for quantifying soil mercury fluxes: Tests and results over
background soils, Atmos. Environ., 32, 873-882, 1998.
Fang, S.C., Sorption and transformation of mercury vapor by dry soil,
Environ. Sci. Technol., 12, 285-288, 1978.
Gustin, M. S., G. E. Taylor Jr., and T. L. Leonard, High levels of
mercury contamination in multiple media of the Carson River drain-
age basin of Nevada: Implications for risk assessment, Environ.
Health Perspect., 102, 772-778, 1994.
Gustin, M. S., G. E. Taylor, T. L. Leonard, and R. E. Keislar, Atmo-
spheric mercury concentrations associated with geologically and an-
thropogenically enriched sites in central western Nevada, Environ.
Sci. Technol., 30, 2572-2579, 1996.
Gustin, M. S., et al., Nevada STORMS project: Measurement of mer-
cury emissions from naturally enriched surfaces, J. Geophys. Res.,
this issue (a).
Gustin, M. S., P. Rasmussen, G. Edwards, W. Schroeder, and J. Kemp,
Application of a laboratory gas exchange chamber for assessment of
in situ mercury emissions, J. Geophys. Res., this issue (b).
Hillel, D., Introduction to Soil Physics, pp. 142-144, Academic, San
Diego, Calif., 1982.
Iverfeldt, A., S. E. Lindberg, S. Karamata, G. Anoshin, M. Horvat,
T. Laperdina, A. Obolenskiy, K. Osmonbetov, C. Ramel, N. Roslya-
kov, and V. Tausen, Working group report on terrestrial mercury
cycling, in Global and Regional Mercury Cycles: Sources, Fluxes attd
Mass Balances, edited by W. Baeyens, R. Ebinghaus, and O. Vasil-
iev, pp. 543-546, Kluwer Acad., Norwell, Mass., 1996.
Johnson, D. W., and S. E. Lindberg, The biogeochemical cycling of Hg
in forests: Alternative methods for quantifying total deposition and
soil emission, Water Air Soil Pollut., 80, 1069-1077, 1995.
Kim, K.-H., and S. E. Lindberg, Design and initial tests of a dynamic
enclosure chamber for measurements of vapor-phase mercury fluxes
over soils, Water Air Soil Pollut., 80, 1059-1068, 1995.
Klusman, R. W., and C. P. Matoske, Adsorption of mercury by soils
from oil shale development areas in the Pieceance Creek Basin of
Northwestern Colorado, Environ. Sci. Technol., 17, 251-261, 1983.
Lindberg, S. E., and J. Price, Measurements of the airborne emissions
of mercury from municipal landfill operations: A short-term study in
Florida, J. Air Waste Manage. Assoc., 49, 520-532, 1999.
Lindberg, S. E., and R. R. Turner, Mercury emissions from chlorine
production solid waste deposits, Nature, 268, 133-136, 1977.
Lindberg, S. E., D. R. Jackson, J. W. Huckabee, S. A. Janzen, M. J.
Levin, and J. R. Lund, Atmospheric emission and plant uptake of
mercury from agricultural soils near the Almaden mercury mine, J.
Environ. Qual., 8, 572-578, 1979.
Lindberg, S. E., K.-H. Kim, T. P. Meyers, and J. G. Owens, A mi-
crometeorological gradient approach for quantifying air/surface ex-
change of mercury vapor: Tests over contaminated soils, Environ.
Sci. Technol., 29, 126-135, 1995.
Lindberg, S. E., P. J. Hanson, T. P. Meyers, and K.-Y. Kim, Microme-
teorological studies of air/surface exchange of mercury over forest
vegetation and a reassessment of continental biogenic mercury emis-
sions, Atmos. Environ., 32, 895-908, 1998.
McCarthy, J. H., Jr., W. W. Vaughn, R. E. Learned, and J. L. Meus-
chke, Mercury in soil gas and air--A potential tool in mineral ex-
ploration, Geol. Surv. Circ., 609, 16, 1969.
Okouchi, S., and S. Sasaki, Photochemical behavior of mercury ore in
water, Environ. Int., 9, 103-106, 1983.
Poissant, L., and A. Casimir, Water-air and soil-air exchange rate of
total gaseous mercury measured at background sites, Atmos. Envi-
ron., 32, 883-893, 1998.
Schmolke, S., W. H. Schroeder, J. Munthe, H. H. Kock, D. Schnee-
berger, and R. Ebinghaus, Simultaneous measurements of total gas-
eous mercury at four sites on a 800 km transect: Spatial distribution
and short time variability of total gaseous mercury over central
Europe, Atmos. Environ., in press, 1999.
Schuster, E., The behavior of mercury in the soil with special emphasis
on complexation and adsorption processes--A review of the litera-
ture, Water Air Soil Pollut., 56, 667-680, 1991.
Spencer, W. F., and M. M. Cliath, Factors affecting vapor loss of
trifluralin from soil, J. Agric. Food Chem., 22, 987, 1974.
Steinberg, S. M., and D. K. Kreamer, Evaluation of the sorption of
volatile organic compounds by unsaturated calcareous soil from
southern Nevada using inverse gas chromatography, Environ. Sci.
Technol., 27, 883-888, 1993.
Thibodeaux, L. J., Environmental Chemodynamics, Movement of Chem-
icals in Air, Water, and Soil, John Wiley, New York, 1996.
Wallschliiger, D., Speziesanalytische untersuchungen zur abschiitzung
des remobilisierungspotentials von Quecksilber aus kontaminierten
elbauen, Ph.D. thesis, Univ. of Bremen, Bremen, Germany, 1996.
Xiao, Z. F., J. Munthe, W. H. Schroeder, and O. Lindqvist, Vertical
fluxes of volatile mercury over forest soil and lake surfaces in Swe-
den, Tellus, Set. B, 43, 267-279, 1991.
Zhang, H., and S. E. Lindberg, Processes influencing the emission of
mercury from soils: A conceptual model, J. Geophys. Res., this issue.
A. Casimir, R. Ebinghaus, G. Edwards, C. Fitzgerald, M. Gustin,
J. Kemp, H. H. Kock, S. E. Lindberg (corresponding author), J. Lon-
don, M. Majewski, F. Marsik, J. Owens, M. Pilote, L. 'Poissant,
P. Rasmussen, F. Schaedlich, D. Schneeberger, J. Sommar, R. Turner,
A. Vette, D. Wallschliiger, Z. Xiao, and H. Zhang, Environmental
Sciences Division, Oak Ridge National Laboratory, P.O. Box 2008,
Oak Ridge, TN 37831. (sllQ•ornl.gov)
(Received September 29, 1998; revised March 17, 1999;
accepted March 22, 1999.)
