Assessment of Trends in Hg-Related Data Sets

1
Assessment of Trends in Hg-Related Data Sets
Critical Assessment of Cause Effect for Trends
in Hg Concentrations in Florida Biota

• C.D. Pollman, Tetra Tech
• D.B. Porcella, Environmental Science Management
• R. Husar, Lantern Corporation
• J. Husar Lantern Corporation
• R. Roberson, RMB Associates
• P. Frederick, University of Florida
• M. Spalding, University of Florida
• 13 December 2001

2
HOW IT ALL STARTED Biota Trend 1 - Recent
Declines in Hg in Largemouth Bass collected from
the Florida Everglades. Data from T. Lange
(2000).
3
HOW IT ALL STARTED Biota Trend 2 - Concomitant
Declines in Hg in Feathers of Everglades Great
Egret Nestlings. Data from P. Frederick (2000).
4
HOW IT ALL STARTED Biota Trend 3 -
Comparatively Rapid Response By Aquatic Biota in
Everglades Predicted to Occur in Response to
Reductions in Atmospheric Loading Rates of Hg.
5
HOW IT ALL STARTED Comparision of measured MSW
Hg emissions (concentrations) in south Florida
with the national inventory on MSW emissions
compiled by Franklin and Kearney for USEPA
6
Primary Objective

• BASIC PREMISE If changes in Hg emissions and
  deposition have occurred, and if this in turn
  produces a comparatively rapid change in biota,
  contemporaneous trends in a number of media also
  should be evident, including
• trends in emissions
• trends in atmospheric chemistry and deposition
  and
• trends in Hg accumulation rates in recent
  sediments.
• OBJJECTIVE Examine whether the trends in Hg
  concentrations in south Florida biota are
  significant, whether the trends (or lack thereof)
  are consistent with the available data on trends
  in Hg emissions, atmospheric deposition, and
  accumulation, and whether other factors need to
  be invoked to explain the data.

7
Secondary Objectives

• Examine data pertaining to long-range transport
  and deposition of Hg and establish whether there
  have been changes in the signal to south Florida
  from these sources.
• Using E-MCM, verify through model hindcasting
  whether the trends in largemouth bass are
  consistent with our understanding of changes in
  atmospheric loadings of Hg in south Florida.
• Explore alternative hypotheses using E-MCM For
  example, do other perturbations to the Everglades
  such as hydrologic changes and changes in sulfate
  loadings offer equally plausible or better
  explanations of the observed biotic trends.

8
Overall Approach

• Test the hypothesis that recent trends in biota
  Hg concentrations in south Florida are a direct
  result of emissions reductions in the region as
  well.

RECONSTRUCT TRENDS IN LOCAL EMISSIONS
CRITICALLY ANALYZE TRENDS IN BIOTA
ESTABLISH LINK BETWEEN THE TIMING AND MAGNITUDE
OF EMISSIONS AND THE TIMING AND MAGNITUDE OF
BIOTIC RESPONSE.
9
Material Flows Hg Trend Reconstruction - Update

• Phase I work supported by FDEP and conducted
  March May 2001.
• Developed trends in Hg emissions drivers for
  period 1980 2000.
• At May 10 meeting in WPB, scope expanded from
  analysis of emission drivers to a broader scale
  Hg budgeting.

10
Material Flows Hg Trend Reconstruction - Update

• Independent estimate of atmospheric Hg emissions
  in S. Florida.
• Accounting for the total mercury flow in Florida
  (air, land and water).
• Re-examining, expanding and updating the national
  Hg budget with new data.

11
Approach Scaling of Development of Hg Budgets
National
Florida
South Florida
12
National Hg Trends in Hg Flow

• Much of analysis derived from Sznopek and Goonan,
  2000 which defines recent trend of the national
  mercury flow, including the primary production,
  consumption, recycling as well as mercury flow
  from stocks.

13
Production Consumption Cycle Showing Secondary
Consumption
14
National Hg Flow Trends

• Apparent Hg supply includes 1? and 2? production,
  net imports, and government stockpile releases.
• 1970 1986 Main contributors to Hg flow were
  1? mine production and imports.
• 1986 1992 Rapid decrease of apparent Hg
  demand caused by reductions in mercury demand for
  batteries, paint and fungicide industries.
• 1993 to present 1? mine production negligible
  2? production (recycling) increased and stock
  releases were terminated.

15
National Hg Flow Trends
16
National Hg Releases to Atmosphere, 1996
From Sznopek and Goonan, 2000 USEPA, 1997.
17
National Hg Releases to Atmosphere, 1996
Materials flow schematics for 1996. Blue lines
are atmospheric emissions from the EPA (1997)
report, adding to 144 Mg/yr atmospheric
emissions. The right hand portion of the
schematics depicts mercury flow in goods, From
Sznopek and Goonan, 2000.
18
Hg in US Goods and Fuel, 1940 1995.
19
Hg in US Goods and Fuel, 1940 1995.

• 1940-1970, consumption of Hg in consumer goods
  was not well characterized. In 1940 74 of Hg
  consumption was categorized as "Other".
• 1970-1990, electrical and electronic instruments
  category (including batteries) dominant Hg
  industrial use.
• Hg demand for consumer goods drastically reduced
  beginning ca. 1989. Hg consumption in consumer
  and industrial goods declined from ?1500 Mg/yr in
  1989 to ?500 Mg/yr in 1995 (U.S. Bureau of Mines,
  1940-1995).
• Hg mobilization in coal has increased since 1940,
  from ca. lt6 of the total to ca. 20 in 1995.
• Mercury mobilization by petroleum products (using
  5 to 50 ppb Hg concentration for crude oil)
  increased since 1940. Wilhelm (in press)
  indicates range is more likely 5 15 ppb.

20
Trend of Hg Consumption in Coal in Florida
21
Trend of Hg Consumption in Petroleum in Florida
22
Trend of Hg Consumption in Electrical Sector

• On a national scale, electrical uses of Hg
  increased steadily between 1947 (lt 200 Mg/yr and
  1986 (1,000 Mg/yr).
• 1996 use of Hg was 78 Mg
• Hg use in electrical sector increased from 14 in
  early 1960s to its peak in 1985 of 63 of the
  total United States consumption.

23
Trend of Hg Consumption in Electrical Sector
Approach (Lighting)

• lt 1978 U.S. Bureau of Mines (BOM) reported a
  single mercury consumption number for electrical
  uses.
• 1978 Present BOM gives inside electrical
  consumption, battery, switches/wiring, and
  lighting categories with their cumulative value
  representing the electrical consumption.
• Reconstruct Hg in lighting using U.S. Bureau of
  Census fluorescent and HID lamps domestic
  shipments. Multiply by Hg content of fluorescent
  lamps (0.75, 0.55, 0.30) for different time
  periods, and HID mercury content (0.33 and 0.25)
  (EPA, 1992, Benazon Environmental, 1998).

24
Trend of Hg Consumption in Electrical Sector
Approach (Batteries)

• Approach I (Bureau of Mines Hg consumption data)
• Electrical switches with Hg not manufactured lt
  1960 (USEPA, 1997)
• Switching/wiring use gt 1960 relatively stable (?
  100 Mg/yr)
• Prior to 1978 Calculate Battery Hg use
• Hgbattery HgTotal Electrical Hgswitches,
  lighting
• Prior to 1960
• Hgbattery HgTotal Electrical Hglighting

25
Trend of Hg Consumption in Electrical Sector
Approach (Batteries)

• Approach II (retail battery sales)
• Based on work published by USEPA (1992)
• Use battery retail sales data from National
  Electrical Manufacturers Association (NEMA) for
  1983-1988, and estimates for 1989, and 1992
• Based on, trend was established and used to
  estimate sales 1967-1982 and 1993-2000,
  respectively. The amount of mercury battery
  import was assumed at 15 .

26
Comparison of Hg Use in Batteries National Scale
Note BOM data do not reflect the exports and
imports of batteries.
BOM data
NEMA data
27
Trend of Hg Consumption in Electrical Sector -
National
28
Trend of Hg Consumption in Electrical Sector -
Florida
Note Scaled from National level data by annual
population.
Batteries
Lighting
Switches/Wiring
29
Trend of Hg Consumption in Agricultural/Fungicide
Use Approach

• Hg consumption in agriculture compiled and
  reported by Jasinski (1994) and Murphy and Aucott
  (1999) based on BOM Yearbook data.
• Hg consumption in fungicide compiled by
  Sharvelle (1961) and Murphy and Aucott (1999).
• USDA annual Agricultural Statistics provides
  drivers for small seed (wheat, barley, oats, rye)
  consumption in USA.
• Hg use in fungicides applied to golf courses was
  estimated by number of golf courses in USA
  (Scharff, 1970, Ross, 1979, NGF web site).
  Assume 80 acres/golf course treated with 43 g
  Hg/acre.

30
Trend of Hg Consumption in Agricultural/Fungicide
Use Approach

• Golf courses routinely treated with fungicides
  late 1950s and 1960s.
• Before late 1950s, literature all suggests only
  affected areas were treated with much higher
  concentrations of fungicides. In 1970s mercury
  fungicides were gradually substituted with
  non-mercury fungicides.
• Hg use for foliage treatment started in 1942.
  Apples and pears were mainly treated with
  organomercury compounds. Annual acreage of apples
  was obtained from USDA Agricultural Statistics
  and 4.5 g/acre (Murphy and Aucott, 1999) was
  applied since middle 1940s.

31
Trend of Hg Consumption in Agricultural/Fungicide
Use Approach

• Hg use for potato and vegetable seeds (tomatoes,
  watermelon, beets), soil treatment for cabbage
  and cauliflower, and flower bulbs and corms is
  not well documented to apportion the mercury use.
  Therefore, the remaining mercury (after
  subtracting small grain, apple, and turf use) in
  the 1950 and 1960 was apportioned to vegetables
  and others.
• Production of wheat, barley, oats, rye and apples
  is not significant in Florida. Thus, only the
  Vegetable and Other portion of USA Hg consumption
  was prorated to Florida, using vegetable acreage
  ratio of Florida compared to USA vegetable
  acreage.
• The golf course mercury fungicide was estimated
  using Florida golf course statistics (Bureau of
  Census, annual 1959-2000).

32
Trend of Hg Consumption in Agricultural/Fungicide
Use Data Sources
33
Trend of Hg Consumption in Agricultural/Fungicide
Use - National
34
Trend of Hg Consumption in Agricultural/Fungicide
Use - Florida
35
Trend of Hg Consumption in Paints Approach

• Late 1950s organomercury compounds (principally
  phenylmercuric acetate) were added to the water
  type paints to prolong the paint's shelf life.
• Use the paint shipment driver and apply the
  estimated Hg concentration factor to paints.
• Information on US paint shipments obtained from
  U.S. Bureau of Census annual Current Industrial
  Reports. The historical total paint shipments
  were obtained from U.S. Bureau of Census, 1975.

36
Trend of Hg Consumption in Paints National
Trend in Driver
37
Trend of Hg Consumption in Paints National
38
Trend of Hg Consumption in Paints Florida
Note Paint Hg estimate for Florida based on U.S.
BOM Hg use and Florida population data. Data are
subject to change as better information is
obtained.
39
Synthesis of Florida Trend in Hg Use

• Use US data and apportion Hg consumption to
  Florida.
• Cl2 and caustic soda manufacturing use of Hg
  omitted from Florida trends, since manufacturing
  is external to Florida and the products do not
  contain Hg.
• Other category also omitted in Florida trend
  reconstruction. Category described in 1960
  Bureau of Mines Yearbook as containing the
  quantity of Hg to initiate new capacity of
  chlorine-caustic soda plants.
• Hg in coal for Florida included upper and lower
  estimates. Used average of the two estimates for
  trend reconstruction.

40
Synthesis of Florida Trend in Hg Use

• Hg in petroleum for Florida was reconstructed
  using Florida consumption data for petroleum
  products and mercury content of the particular
  product (Wilhelm, 2001).
• The mercury in paints, pharmaceutical,
  electrical, control, and dental uses were
  prorated to Florida population.
• The agricultural mercury was deconstructed, and
  only Florida specific uses were estimated.

41
Trend of Hg Consumption in Florida
42
Synthesis of Florida Trend in Hg Emissions

• Use coal and petroleum consumption as described
  previously
• Paint Assume 75 is emitted to the atmosphere
  (Benazon Environmental, 1998).
• Laboratory, Electrical, Dental, and Control
  products Assume disposal as municipal solid
  waste. In 1980 there were 500 open dumps in
  Florida. Open burning of waste was a common
  method to save landfill space (Solid Waste
  Management in Florida, FL DEP, 1996 p63). In
  1982 one WTE was operating in Florida. In 1995
  there were 13 WTE.
• For 1990-1998 percentage of municipal waste
  burned in WTE is available (Solid Waste
  Management in Florida, FL DEP). However, prior
  to 1980 percentage of municipal solid waste
  burned is not available. A 16 burning rate was
  assumed.

43
Trend in Hg Emissions in Florida 1930 1999
44
Trend in Hg Emissions in Florida 1990 1999
45
Comparison of material flow-based estimates of Hg
emissions with FDEP direct emissions estimates
Note FL DEP estimates of Hg emissions from
municipal waste combustion facilities courtesy of
Yi Zhu, FDEP.
46
Major Florida Hg in Biota Data Sets

• Lange et al. (FGFWFC) largemouth bass data (n
  sites 12, total n 2001)
• Frederick et al. Great Egret nestling feather
  data (n sites 7 sites total n 558 period
  1994 - 2001)
• Zillioux et al. raccoon data

47
Summary of Fish Data Used in Trend Analysis

• Data File Task Through 2000.xls (Lange June 2001
  Subsetted.xls)
• Data Locations
• Big Lostmans Creek, East Lake Tohopekaliga,
  Fowlers Bluff, Indian Camp Creek-Rogers, L-35B
  Canal, Lake Tohopekaliga, Marsh-15, Marsh-GH,
  Marsh-OM, Marsh-U3, Miami Canal and L-67A, North
  Prong
• Number of Data Points
• 1464 measurements of Hg in Largemouth Bass
• Number of Age Groups
• 0, 1, 2, 3, 4, 5, 6, 7, 8, 9
• Period Of Time
• 10/31/1988 03/14/2001

48
Statistical Analysis Methods

• Simple Linear Regression versus Time
• Mann-Kendall Slope Test-of-Sign
• Median Slope (computed Sen Slope estimate)

49
Mann-Kendall Slope Test-of-Sign
Mann-Kendall routine is a non-parametric test for
zero slope
For all j and k such that j gt k
where Cj is the mercury concentration at time Tj.
If a total of n samples are collected over time,
then there are n(n-1)/2 slope calculations. Sens
slope estimator is equal the median slope from
that of the n(n-1)/2 slope calculations.
50
Why Use the Mann-Kendall Routine?

• Standard t-test of simple linear regression is
  misleading when
• Seasonal cycles are present
• Data are not normally distributed
• Data are serially correlated
• The above conditions over predict the
  significance of the slope when the true slope is
  actually zero

51
Hg Concentration Data Availability for Largemouth
Bass
52
Trend-Adjusted Results of Mann-Kendall
Test-of-Sign
53
Typical Examples of Sens Non-parametric
Estimator of Slope (µg Hg/g/yr)
54
East Lake Tohopekaliga Data
55
Marsh-U3 Data
56
Location of Frederick et al. (2001) Great Egret
and White Ibis Sampling Sites
Orange Lake
Seahorse Key
Lake Griffin
St. Martins
Lake Mary Jane
Teneroc
Cocoa Beach
Lake Kissimmee
Terra Ceia Bay
Clewiston
Alley North
JW1
L-67
Hidden
TTW
?Long-term Great Egret monitoring stations, 1994
- 2000
57
Hg Concentrations in Great Egret Nestling
Feathers. Data courtesy of P Frederick (2001).
N sites per year 7.
Relative Hg Concentration
Average Relative Hg Concentration
Year
58
Zillioux et al. South Florida Raccoon Hair MMHg
Study - Hypotheses

• MMHg levels in raccoon hair have changed between
  museum and modern collections.
• Raccoon hair MMHg levels are the same when
  collected from different sites having the same
  deposition.

59
South Florida Raccoon Hair MMHg
60
South Florida Raccoon Hair MMHg Conclusions

• Cannot conclude that there has been a temporal
  change in MMHg raccoon hair levels.
• Strong differences between sites with relatively
  similar deposition.
• Location is more important than time
• Methylation
• Feeding behavior
• Other factors do not seem important size,
  life-stage, sex, other elements, analytical
  artifacts

61
Study Tentative Conclusions

• Estimates of statewide emissions of Hg (based on
  materials flow) indicates a large, rapid decline
  in Hg releases to atmosphere, driven
  predominantly by reductions in electrical use and
  cessation of paint product use.
• Kendall tau test demonstrates significant
  declining trends LMB Hg concentrations for a
  number of sites across the state, but patterns
  are not ubiquitously observed. Canals appear to
  be the most consistent showing declining trends,
  while trends for marsh sites are not uniform.
• Only one site, U3, in the entire data set shows
  increasing trends, and only for year classes 0 to
  2

62
Study Tentative Conclusions

• The widespread distribution of declining trends
  suggests a regional factor, such as atmospheric
  deposition, or perhaps long-term climatic
  variables (e.g., changes in hydrology and
  concomitant effects on geochemistry.
• Likewise, the within region variability in trends
  evidenced both in the LMB and raccoon data
  indicate local biogeochemistry and hydrology
  appears to play a strong role, perhaps more than
  deposition.
• Declines may have stabilized, and limited data
  from one bird colony suggests perhaps an uptick
  in the trends.

63
Project Timeline
Jan 1, 2002 Submit 1st Draft To FCG
Mar 15, 2002 Submit Final To FCG
Aug 1 to Oct 31, 2001. Main Hg emission trend
analysis. Submit trends to Tt.
Feb 23, 2002 Comments Back From FCG
Nov 1 to Dec 31, 2001. Refine analysis. Draft
Report
Dec. 1 to Jan 1, 2002. Integrate biota and
emission trend results. Prepare draft white
paper
Jul 1-31, 2001. Preliminary analysis of LMB data
Oct 1 to Nov 15, 2001. Seasonal Kendall analysis
of fish trends. Analyze bird data. Integrate
raccoon results.
64
Revised Project Timeline
Materials Flow Analysis of Hg Trends. National
and Florida scale
Complete Regional (south Florida) analysis.
Update and revise report
?
?
Meet with FDEP and submit data request
Work with FDEP to try and expedite data transfer
Preliminary analysis of LMB data
Seasonal Kendall analysis of fish trends.
Obtain and analyze raw bird data. Integrate
raccoon results.
65
Assessment of Changes in Large-Scale Hg Signal -
Phase II Analysis
NADP/MDN Site
66
Comparison of Glass Sorensen (1999) results for
Ely, MN with MDN results from same site. Period
of overlap is for 1/16/96 through 10/8/96 (n 25)
Precipitation Depth
67
Comparison of Glass Sorensen (1999) results for
Ely, MN with MDN results from same site. Period
of overlap is for 1/16/96 through 10/8/96 (n 25)
Hg in Wet Deposition
MDN Hg (ng/L)
Glass Hg (ng/L)
68
Annual Hg Deposition Rates, Ely, MN 1990 - 2000
Glass Sorensen
MDN
Hg Deposition (ng/m2-yr)
Year
69
(No Transcript)
70
Results for ANOVA Model
71
Plot of ANOVA Deposition Model Residuals vs. Time
for Ely, MN. Data from Glass and Sorensen (1999)
and MDN.
m 0.025 p 0.101
Residual Ln Hg Dep
Time
72
NADP Trend Analysis - Model Residuals vs. Time
m -0.018 p 0.065
Residual Ln SO4 Dep
Time