1. Natural Haze Levels II: Application of the New IMPROVE Algorithm to Natural Species Concentrations Estimates
Final Report by the
Natural Haze Levels II Committee to the
RPO Monitoring/Data Analysis Workgroup
The Natural Haze Levels II Committee was established in the Spring of 2006 to review and refine, as appropriate, a methodology developed by Roger Ames (CIRA) for applying the new IMPROVE algorithm for estimating light extinction from aerosol species concentrations to natural species concentration estimates. The ultimate purpose of this is to determine natural haze estimates for the 20% best and 20% worst day for each of the visibility-protected class I areas. This is the final report of the committee which has completed its review and successfully refined the Roger Ames approach, and in so doing has produced natural haze estimates for each of the visibility protected area.
After the presentation and review by the RPO Monitoring/Data Analysis Workgroup a final Workgroup approved version will be forwarded to the RPOs for their consideration.
The Committee was composed of Marc Pitchford, NOAA; Bill Malm, NPS Bruce Polkowsky, NPS; Pat Brewer, VISTAS; Tom Moore, WRAP; Ivar Tombach, consultant; John Vimont, NPS; Rich Poirot, Vermont; Roger Ames, CIRA; and Naresh Kumar, EPRI.

2. These notes were prepared
Note: This presentation contains substantial additional information in the notes section of PowerPoint that can be seen in the bottom panel of the “Normal” view mode, and can be printed by selecting “Notes Pages” from the “Print what” selection of the “Print” menu (go to “Files” on the top tool bar and select “Print”). You can also use “Preview” for the “Notes Pages” to more easily read the notes on your computer monitor.
These notes were prepared
to aid those hearing the presentation by relieving them of the burden of taking as many notes;
to allow those who haven’t heard the presentation to understand it by providing the additional information that is spoken during presentations; and
to provide a more complete documentation of the natural haze levels II approach for anyone who may want to understand it
For additional information contact Marc Pitchford at
Here is the notes section where the additional material associated with each slide of the presentation has been placed.

3. Overall Goal
Estimate 20% best and 20% worst natural haze levels for visibility-protected class I areas using the new IMPROVE algorithm for estimating light extinction from aerosol species concentrations.
Needed for Regional Haze Rule (RHR) rate of progress glide slopes where the new IMPROVE algorithm is used to characterize current haze levels
Should minimize the technical problems identified in the RHR default natural haze levels that were developed using the original IMPROVE algorithm
EPA’s RHR guidance issued in 2003 concerning tracking progress and estimating natural condition were based on the original IMPROVE algorithm. This guidance defines a consistent set of instructions for implementing the RHR. However, in late 2005 a new IMPROVE algorithm was developed that mitigated some of the technical criticism of the original IMPROVE algorithm especially as it applied to the RHR. Many of the RPOs and states have indicated their preference to use the new algorithm, but to do so they need natural haze conditions for their class I areas determined in a consistent manner (i.e. by the new IMPROVE algorithm). A report describing the new algorithm is available in the Grey Literature Section of the IMPROVE web site at
Estimates of natural haze levels using either algorithm involves applying the algorithm to estimates of natural species concentrations. The natural species concentration estimate used for this purpose come from the NAPAP State of Science Report 24 by John Trijonis (1990) and are typical values for the eastern U.S. and for the western U.S. Some methodology is needed to adjust these typical values to estimate the 20% best and 20% worst values. The statistical approach used in the EPA guidance document to develop the default best and worst natural haze levels was subject to criticism. A goal in developing the new values is to avoid the problems identified in the EPA default approach. The visibility guidance documents are available on the IMPROVE web site at

4. Default Natural Haze Levels Approach
Typical haze level estimates for East and West
Typical light extinction by applying the original IMPROVE algorithm to Trijonis natural species concentration estimates for East and West
Convert to haze index (deciview units)
20% best and 20% worst haze estimate for East and West
Best = typical – 1.28(standard deviation)
Worst = typical (standard deviation)
Standard deviation is 3dv for the East and 2dv for the West (corresponds to the 10th and 90th percentile)
A complete description of the default approach for estimating natural haze levels is available in the Guidance for Estimating Natural Visibility Conditions Under the Regional Haze Rule is available at as are the results of applying it all of the IMPROVE monitoring sites.
The Trijonis estimates of the major species concentrations under natural conditions were typical or mean values for the eastern and western U.S. The methodology used by EPA in the guidance document was based on the assumptions that the deciview distributions were near normal and that the average of the 20% best and 20% worst haze index values were well approximated by the 10th and 90th percentiles respectively. These assumptions seemed to work well in estimating the 20% best and worst values for current conditions at most monitoring sites. However, these two assumptions were the subject of considerable scrutiny and criticism.

5. Criticism of the Default Approach
Limitations of the original IMPROVE algorithm
Biased light extinction estimates at the extremes
Uses an outdated organic compound mass to carbon mass ratio
No sea salt (important at a few sites)
Rayleigh scattering of 10Mm-1 used for all site
Flawed assumptions used to estimate 20% best and worst conditions
Haze index for natural conditions are not likely to be normally distributed due to inclusion of Rayleigh scattering
10th and 90th percentiles don’t correspond to the best and worst conditions if the distribution were normal
Limitations of the Original IMPROVE Algorithm:
The original IMPROVE algorithm tended to underestimate light extinction for the highest haze conditions and overestimate it for the lowest haze conditions. It used a ratio of organic compound mass to carbon mass of 1.4, though the literature indicated that the ratio is higher especially in remote areas. It also didn’t include a term for sea salt, which is important sites near the ocean coasts. As a result of these criticisms, the IMPROVE Steering Committee commissioned the development of a new IMPROVE algorithm which they adopted for its use in December Many states and RPOs have indicated their intent to use the new IMPROVE algorithm to estimate current haze condition, because it performs better in regards to bias at the extremes, uses 1.8 as the organic compound mass to organic carbon mass, it includes sea salt and uses site elevation related Rayleigh scattering. A report describing the new algorithm is available at
Flawed Assumptions:
The basis for the assumption that the natural haze conditions are normally distributed was that most pollution concentration values (including particle light extinction values) are well represented by lognormal distributions, which when logarithmically transformed (as is the case in conversion of light extinction to the haze index in deciview units) gives a normal distribution. However, this thinking did not account for the Rayleigh scattering which is a taken to be a constant for any particular monitoring site and is a substantial component of the light extinction on less hazy days (e.g. natural conditions). By adding this constant to the lognormally distributed light extinction the result is no longer lognormal so the logarithmical transformation to deciview produces a distribution that isn’t normally distributed.
This flaw assumption also distorted the selection of the standard deviations used to represent the natural distribution for the east and west (3dv and 2dv respectively). These were derived by extrapolating to the least hazy conditions in scatter plots of each sites standard deviation versus mean haze index for sites in the east and for sites in the west. These scatter plots generally showed that the sites with lower means values had lower standard deviations. However, this characteristic is principally due to the inclusion of a constant Rayleigh scattering. If Rayleigh scattering is removed so that the plots are of logarithmically transformed particle extinction, the standard deviations were not systematically smaller for the least hazy sites in a region.
The use of the 10th and 90th percentile values as estimates of the mean of the 20% lowest and mean of the 20% highest haze index values is not consistent with the assumption that the natural haze index values are normally distributed. The 8th and 92nd percentiles are the best estimates of the mean of the 20% best and 20% worst conditions which then changes the multiplicative coefficients from 1.28 to ~1.4 in the equation for estimating the extreme values from the typical values. However since the original distribution were not normal (as discussed above), this critique is less pertinent.

6. Natural Haze Levels II Approach
Adjust each of the measured major species concentrations to the Trijonis natural concentration estimates
Multiply each species concentration at a site by the site-specific ratio of the (Trijonis natural estimate) divided by the (annual mean concentration) for the species for the 5 year baseline period
If the annual mean concentration for a species is smaller than the Trijonis natural estimate, make no adjustment
Current sea salt levels are taken to be natural levels
Apply the new IMPROVE algorithm to the Trijonis-adjusted species concentrations at each site to produce a distribution of natural light extinction values
Convert to deciview and calculate the mean of the 20% best and 20% worst haze levels
Trijonis-Adjusted Species Concentrations:
Unlike the default approach which directly uses the Trijonis natural species concentration estimates to calculate haze levels, this approach the adjusts the data set of current species concentration using a multiplier applied to each species measurement that gives the Trijonis estimate for that species. Ratio of the Trijonis estimates for each species divided by the annual mean values for the species is used to transform the entire data set to what is then assumed to be the natural species concentration levels for that site and year. This process is applied to each of the complete years of data (as defined by the tracking regional haze trends guidance) in the baseline period (2000 through 2004). Sites with 3 complete years of data are treated as having sufficient data for this assessment. Provisional estimate were made for sites with fewer than three complete years (including for Breton Island were one composite year was developed).
If any of the current annual mean for any species is less than the Trijonis estimate for that species, the unadjusted species data are used. Trijonis estimates didn’t include sea salt, which is only significant at a few coastal sites. Estimates of current sea salt concentrations determined from Cl- ion data (described as part of the new IMPROVE algorithm) are take to be natural contributors to haze.
Fundamental Assumption:
In the default approach, the Trijonis estimates for the east and west were used directly to calculate the haze index, and assumptions were made about the shape and width of the natural haze distribution to permit the 20% best and 20% worst haze conditions to be extrapolated from the estimated mean values.
The fundamental assumption inherent in this Natural Levels II approach is that the Trijonis-adjusted species data set is a good estimate of the natural species data set. Thus each site has its own a natural haze distribution which is derived from the current distribution, so no assumptions about the shape and width of the natural distribution are needed because there are sufficient values to actually calculate the 20% best and 20% worst means for site and year.
Another assumption, that the baseline period of up to five years (2000 through 2004) is sufficiently long and representative of longer time periods to be used in this process to produce reasonable estimate of natural haze levels, was tested using data from long term monitoring sites.  The concern was raised about whether the characteristics of the baseline period that included an unusually intense wildfire year in the Western U.S. (2002) might bias the estimates of natural haze levels for sites that were particularly impacted by smoke from those fires.  Comparisons were made between the natural haze estimates based on the five-year baseline and the 17-year complete period of record for 22 long-term IMPROVE sites.  The mean difference among the sites between the two estimates of natural haze was 0.2 deciview, which corresponds to an average fractional change of about 2.5%.  The site with the greatest difference in the estimates of natural haze is Mesa Verde National Park in Colorado that had a difference of 0.7 deciview (higher estimate of natural haze using the baseline period compared to the long-term period), which corresponds to about a 9% change.  These differences based on the periods of record used in the approach were judged to be within the uncertainty of the methodology.  Given that only a small fraction of the sites have long-term data, the use of the 5-year baseline period for each site for the sake of consistency was thought by the committee to be appropriate.
Application of the New IMPROVE Algorithm:
The new IMPROVE algorithm is applied to the Trijonis-adjusted concentration data set to estimate the light extinction values for what is assumed to be the distribution of natural conditions for each site. These are then converted to deciview values from which the mean of the 20% best and the mean of the 20% worst haze conditions are calculated for each year of the baseline period.
Alternate Approaches Tried, but Rejected:
Log-Transformed Particle Light Scattering Approach – We used the new IMPROVE algorithm to produce particle extinction levels (i.e. no Rayleigh scattering). These were log-transformed and shown to have very nearly normal distributions for most monitoring site since the distortion caused by having a constant Rayleigh scattering term was eliminated. Then we used a statistical approach similar to that used in the default natural haze approach to estimate the 20% best and 20% worst haze conditions. These values looked reasonable. However once we realized that we had sufficient data at nearly every monitoring site so that we could adjust each sample periods data and we wouldn’t need to test for the distributions and make assumptions about the width of the distribution to estimate the best and worst conditions, so we abandoned this type of approach.
Modified Application of the New IMPROVE Algorithm – Some on the committee were concerned that the new IMPROVE algorithm’s empirical approach for partitioning the sulfates, nitrates and organic compounds into the two particle size distributions (thus affecting the extinction efficiency and water growth properties of these species) would be inappropriate under natural emissions conditions. The empirical approach uses the concentrations of the three species to determine what fraction of each is in the less efficient light scattering small particle size distribution and how much would be in the more efficient light scattering large size distribution. When man-made emissions are substantially reduced so that natural haze sources are dominant (i.e. attainment of the RHR goal), the concentrations of sulfates and nitrates, and to a lesser extent the organic concentrations will be substantially reduced, so the IMPROVE algorithm would have a higher fraction of these components in the smaller size distribution. However, the concerned committee members argued that the same meteorological conditions that produced large size distributions (e.g. in-cloud chemistry, particle growth through aging) would still apply with about the same frequency in various regions as is currently the case. For example, fog events would be just as frequent under natural conditions as currently is the case. In other words, they were arguing that the empirical approach that determines the fraction of the mass of these species to large and small sizes will not work well if the emissions are significantly different than those for which it was developed.
An approach to preserve the current splits of sulfate, nitrates organic carbon was developed. This was done by using the new IMPROVE algorithm’s empirical approach applied to current conditions for every sample period to determine the split between large and small particle size distribution. These sample-period specific splits were applied to the Trijonis-adjusted concentrations to estimate the natural haze levels. The affect of this approach was to generate natural haze level estimates that were higher, principally because the sulfate and nitrate extinction efficiencies were much larger. However, these natural haze level values were only slightly higher then those without this adjustment, principally because the higher extinction efficiencies affected sulfate and nitrate concentrations which under natural conditions are relatively minor contributors to haze.
In the end an alternative technical discussion convinced the committee that the new IMPROVE algorithm should be used without any modification. The basic point of the discussion was that while the frequency of meteorological conditions that generate large size distribution particles might be similar for current conditions and natural levels, the concentrations of the primary pollutants (including SO2, NOx, and VOCs) would be so low that particles would not generally grow to the sizes that current high concentrations of these pollutants allows. Another reason for using the new IMPROVE algorithm without modification was to avoid having to develop justify that the approach selected for the modification would produce better results than from the unmodified algorithm. In the long term, the technical community can see if the new IMPROVE algorithm continues to perform well when emission patterns and rates are significantly different than today.
Substitution for Some of the Trijonis Species Values – We use the Trijonis estimated natural species concentrations for the anthropogenically-dominated species (i.e. sulfate, nitrate and elemental carbon), but tried various approaches to use currently monitoring values for the naturally-dominated species (i.e. organic carbon, fine soil and coarse mass). A number of different variations were tried including:
Trijonis-adjusted values for each component except for organic carbon; all currently measured organic carbon was assumed to be natural.
Trijonis-adjusted values for each component except for organic carbon; 80% of the currently measured organic carbon was assumed to be natural.
Trijonis-adjusted values for each component except for organic carbon, fine soil and coarse mass; all currently measured organic carbon, fine soil and coarse mass is assumed to be natural.
Other variations of the above with all or various fractions of the natural source-dominated species being used in place of the Trijonis estimates for those species.
Ultimately these were all rejected because we could find regional and site-specific counter-examples to the underlying assumptions in each of them, and it would have greatly expanded what we saw as our charge which was to employ the new IMPROVE algorithm to the Trijonis natural species estimates. Counter-examples to the assumption that all or most of the organic carbon in remote areas is from natural sources included sites that are heavily impacted by large urban sources where transportation is a primary source of carbonaceous particulate matter (e.g. San Gorgonio Wilderness, just east of L.A., Saguaro, very near Tucson) also sites impacted by smoke from agricultural burning. Similar cases could be made for specific sites and regions of the country for fine soil and dust from construction, mining and agriculture. The effort to produce site-specific and regional rules for how much of these species should be assumed to be from natural sources would have required more time and resources than were available. Such an effort should be further explored for future refinements of natural haze level estimates.

7. Trijonis-Adjusted Specie Frequency Distributions
Current
Hanging bars
Solid - current mean
Dashed - natural estimate mean
Sipsey Alabama
Each aerosol species mass concentration frequency distribution scaled to estimated natural mass concentrations
If current species mean is less than natural estimate, the that species is not scaled
Geometric shape of species distributions is unchanged
Natural Estimate
These two plots demonstrate the Trijonis-adjustment to the measurement data for a specific site and year. Both plots are frequency distributions (number of sample periods with species concentrations in specific narrow ranges). Different colors are used to show each of the major species (ammonium sulfate, ammonium nitrate, organic mass, elemental carbon, fine soil, coarse mass and sea salt). The top plot shows the distributions of the current species concentrations, while the bottom plot shows the distributions of the adjusted species concentrations. Note that the concentrations shown on the x-axis are logarithmically scaled units of microgram/meter cubed. As a result of the log-scale, the Trijonis-adjusted distributions have the same shape as the current distribution, only they are translated to the left (this is because multiplication and division of numbers is equivalent to addition and subtraction of their logarithms).
Reconstructed total mass (RCTM) is the sum of the distributions for the major species. Notice that the shape of these distributions is not the same in the upper and lower plots. This is because each of the species is adjusted by different ratios of Trijonis to current mean conditions.
The dangling bars at the tops of the two plots show the locations of the mean values for the current conditions (short solid bars) and the natural levels (longer dashed bars). The bars are identical in the top and bottom plots. Notice that the sulfate (dark blue) and elemental carbon (green) have large displacements between their solid and dashed bars which current conditions are much higher than the Trijonis values, while coarse mass (orange) and fine soil (yellow) have their solid and dashed bars much closer together indicating that the current conditions are only a little higher than the Trijonis estimates of natural concentrations. Current sea salt (solid red bar) mean levels are assumed to equal the natural levels (dashed red bars).
Thought not called for in the RHR or guidance for implementing the RHR, species specific rate of progress glide slopes can be calculated for each monitoring site. Such glide slopes can more explicitly show which aerosol species offer the greatest opportunities for contributing to reduce haze.

8. Current and Natural Haze Frequency Distributions
Sipsey Alabama
Natural scenario joint distribution shape is derived from scaling current aerosol species mass concentrations to natural condition estimates
Allows estimation of best and worst 20% dv or aerosol species extinction
These two plot show the frequency distributions of current (solid line) and estimated natural (dotted line) haze levels expressed in particle light extinction (plotted on the left) and haze index with deciview units (plotted on the right). The 20% best and 20% worst haze current and natural conditions can be calculated directly from the values that were plotted in the right most plot by averaging the values in the lowest 20% and those in the highest 20% respectively.

9. Natural Haze Levels II
The map shows site values and contours of estimates of the 20% worst natural haze levels (deciview units) using the new approach at IMPROVE and IMPROVE protocol sites. Values based on with less than 3 year of valid baseline data are shown in red. The value for Breton Island, LA was based on four years of data with n>= 70 values, because no baseline year met RHR completeness criteria

10. Default Natural Haze Levels
This map is shown for comparison of to the previous map. It contains the values and contours of natural haze estimates using the default approach as described in the EPA guidance document for estimating natural haze conditions for the RHR. Basically it applies the original IMPROVE algorithm to the Trijonis estimate of natural species concentrations plus 3.84dv (i.e. 1.28x3dv) for the eastern sites, or plus 2.56dv (i.e. 1.28x2dv) for the western sites.
It instructive to view these in slideshow mode and to flip back and forth while looking at specific regions or sites to see the difference. The general pattern is similar except on the West coast where sea salt is responsible for a very noticeable increase using the new methodology. Sea salt also make a detectable difference at a few eastern coastal sites, but they don’t shift the contours as noticeably. Some high elevation sites have lower values using the new methodology due primarily to the lower values used at higher elevation for Rayleigh scattering in the new IMPROVE algorithm. However many other high elevation sites have higher natural estimate values using the new methodology (principally in the Pacific Northwest and Rocky Mountains), probably due to the broader distribution of current (and therefore assumed natural) haze levels associated with a higher incidence of wildfire influence.

11. Natural Haze Levels II, 10-Year Rate of Progress Glide Path
This map shows the rate of progress (deciview reduction per decade) required to reach natural levels in 60 years for site and using contours determined using the new IMPROVE algorithm and the estimates of natural haze levels II approach.

12. Default Natural Haze Levels, 10-Year Rate of Progress Glide Path
This map shows the rate of progress (deciview reduction per decade) required to reach natural levels in 60 years for site and using contours determined using the original IMPROVE algorithm and the default estimates of natural haze.
This map is shown for comparison to the previous map, which shows the new approach glide path results using the same contour intervals. Though not identical to the previous maps, the results are very similar. In fact they are much more similar than the comparisons between the default and new method natural levels maps. For example the effects of sea salt on the west coast is no longer obvious.

13. Status and Next Steps
This presentation is the final report of the Natural Haze Levels II Committee of the RPO Monitoring/Data Analysis Workgroup
Review comments received by August 25, 2006 will be considered in preparation of the RPO Monitoring/Data Analysis Workgroup approved approach
Workgroup approved approach is forwarded to the RPOs for their consideration by August 31, 2006
This presentation, including the natural haze estimates and any modifications will be made available on VIEWS
The information in this presentation constitute the final report of the Natural Haze Levels II Committee, which is scheduled to be presented to the RPO Monitoring/Data Analysis Workgroup at their scheduled call in August 2006 (10am Pacific Time Aug 14, 2006). Comment received by August 25 will be considered prior to issuing a Workgroup-approved version which will be forwarded to the five RPO for their consideration. All of this material will be made available on the VIEWS web site.

14. Appendix
Tables of Natural Haze Level II Estimates for all IMPROVE Sites by RPO and State
The tables that follow include only the IMPROVE site estimates of natural levels, as well as the current worst haze and 10-year glide slopes. These were produced from an Excel spreadsheet (NaturalLevelsII.xls) which is an excerpt of another spreadsheet (NC_summary.xls) that contains both the default and new method natural haze estimates for all sites (including Protocol sites). Both of these Excel files is available on the VIEWS web site.

15. CENRAP MANE-VU & Midwest RPO Site Name State Complete Years
Worst Haze Natural II (dv)
Worst Haze Baseline (dv)
10-year Glide Slope (dv)
Class I Area(s)
CACR1
Caney Creek
Arkansas 3
11.7
26.4
2.4
UPBU1
Upper Buffalo Wilderness 5
26.3
Upper Buffalo
BRET1
Breton*
Louisiana 4
12.3
0.0
Breton
BOWA1
Boundary Waters Canoe Area
Minnesota 2
11.6
20.0
1.4
VOYA2
Voyageurs NP #2
12.2
19.3
1.2
Voyageurs
HEGL1
Hercules-Glades
Missouri
11.4
26.7
2.6
Hercules-Glade
MING1
Mingo 1
12.5
29.5
2.8
WIMO1
Wichita Mountains
Oklahoma
7.6
23.8
2.7
BIBE1
Big Bend NP
Texas
7.2
17.3
1.7
Big Bend
GUMO1
Guadalupe Mountains NP
6.8
17.2
Guadalupe Mountains; Carlsbad Caverns
CENRAP
Site
Name
State
Complete Years
Worst Haze Natural II (dv)
Worst Haze Baseline (dv)
10-year Glide Slope (dv)
Class I Area(s)
ACAD1
Acadia NP
Maine 5
12.5
22.9
1.7
Acadia
MOOS1
Moosehorn NWR
12.1
21.7
1.6
Moosehorn; Roosevelt Campobello
GRGU1
Great Gulf Wilderness
New Hampshire 4
22.8
1.8
Great Gulf; Presidential Range-Dry River
BRIG1
Brigantine NWR
New Jersey
12.3
29.0
2.8
Brigantine
LYBR1
Lye Brook Wilderness
Vermont
11.9
24.4
2.1
Lye Brook
ISLE1
Isle Royale NP
Michigan
20.7
1.4
Isle Royale
SENE1
Seney
12.8
24.2
1.9
MANE-VU & Midwest RPO

16. Worst Haze Natural II (dv) Worst Haze Baseline (dv)
VISTAS + VI
Site
Name
State
Complete Years
Worst Haze Natural II (dv)
Worst Haze Baseline (dv)
10-year Glide Slope (dv)
Class I Area(s)
SIPS1
Sipsy Wilderness
Alabama 4
11.1
29.0
3.0
Sipsey
CHAS1
Chassahowitzka NWR
Florida 3
11.3
26.1
2.5
Chassahowitzka
EVER1
Everglades NP
12.3
22.3
1.7
Everglades
SAMA1
St. Marks 2
11.6
26.0
2.4
Saint Marks
COHU1
Cohutta
Georgia
11.2
30.3
3.2
OKEF1
Okefenokee NWR 5
11.5
27.1
2.6
Okefenokee; Wolf Island
MACA1
Mammoth Cave NP
Kentucky
31.4
3.4
Mammoth Cave
LIGO1
Linville Gorge
North Carolina
28.8
2.9
SHRO1
Shining Rock Wilderness
27.9
2.7
Shining Rock
SWAN1
Swanquarter
12.0
25.5
2.2
ROMA1
Cape Romain NWR
South Carolina
12.2
26.5
Cape Romain
GRSM1
Great Smoky Mountains NP
Tennessee
Great Smoky Mountains; Joyce Kilmer-Slickrock
JARI1
James River Face Wilderness
Virginia
29.1
James River Face
SHEN1
Shenandoah NP
11.4
29.3
Shenandoah
DOSO1
Dolly Sods Wilderness
West Virginia
10.4
3.1
Dolly Sods; Otter Creek
VIIS1
Virgin Islands NP
Virgin Islands
10.8
17.0
1.0

17. Worst Haze Natural II (dv) Worst Haze Baseline (dv)
Site
Name
State
Complete Years
Worst Haze Natural II (dv)
Worst Haze Baseline (dv)
10-year Glide Slope (dv)
Class I Area(s)
DENA1
Denali NP
Alaska 5
7.4
9.9
0.4
Denali
SIME1
Simeonof 3
15.7
18.6
0.5
TUXE1
Tuxedni
11.5
14.1
BALD1
Mount Baldy
Arizona 2
6.6
0.8
CHIR1
Chiricahua NM
7.3
13.4
1.0
Chiricahua NM; Chiricahua W; Galiuro
GRCA2
Hance Camp at Grand Canyon NP 4
7.2
11.7
0.7
Grand Canyon
IKBA1
Ike's Backbone
6.7
13.3
1.1
Mazatzal; Pine Mountain
PEFO1
Petrified Forest NP
13.2
Petrified Forest
SAGU1
Saguaro NM
6.5
14.8
1.4
Saguaro
SIAN1
Sierra Ancha
13.7
1.2
SYCA1
Sycamore Canyon
6.8
15.3
TONT1
Tonto NM
13.9
Superstition
WRAP

18. Worst Haze Natural II (dv) Worst Haze Baseline (dv)
Site
Name
State
Complete Years
Worst Haze Natural II (dv)
Worst Haze Baseline (dv)
10-year Glide Slope (dv)
Class I Area(s)
AGTI1
Agua Tibia
California 4
7.7
23.5
2.6
BLIS1
Bliss SP (TRPA)
6.2
12.6
1.1
Desolation; Mokelumne
DOME1
Dome Lands Wilderness
7.5
19.4
2.0
Dome Land
HOOV1
Hoover 3
8.0
12.9
0.8
JOSH1
Joshua Tree NP
7.3
19.6
2.1
Joshua Tree
KAIS1
Kaiser 2
14.8
1.2
Ansel Adams; Kaiser; John Muir
LABE1
Lava Beds NM
7.9
15.1
Lava Beds; South Warner
LAVO1
Lassen Volcanic NP 5
7.4
14.1
Lassen Volcanic; Caribou; Thousand Lakes; Caribou; Thousand Lakes
PINN1
Pinnacles NM
8.1
18.5
1.7
Pinnacles; Ventana
PORE1
Point Reyes National Seashore
15.9
22.8
Point Reyes
RAFA1
San Rafael
7.8
18.9
1.8
REDW1
Redwood NP
14.0
0.7
Redwood
SAGA1
San Gabriel
7.1
19.9
San Gabriel; Cucamonga
SAGO1
San Gorgonio Wilderness
22.2
2.5
San Gorgonio; San Jacinto
SEQU1
Sequoia NP
7.6
24.6
2.8
Sequoia; Kings Canyon
TRIN1
Trinity
8.3
16.3
1.3
Marble Mountain; Yolla Bolly-Middle Eel
YOSE1
Yosemite NP
17.6
Yosemite; Emigrant
WRAP

19. Worst Haze Natural II (dv) Worst Haze Baseline (dv)
WRAP
Site
Name
State
Complete Years
Worst Haze Natural II (dv)
Worst Haze Baseline (dv)
10-year Glide Slope (dv)
Class I Area(s)
GRSA1
Great Sand Dunes NM
Colorado 5
6.7
12.8
1.0
Great Sand Dunes
MEVE1
Mesa Verde NP
6.9
13.0
Mesa Verde
MOZI1
Mount Zirkel Wilderness 4
6.6
10.5
0.7
Mount Zirkel; Rawah
ROMO1
Rocky Mountain NP
7.3
13.8
1.1
Rocky Mountain
WEMI1
Weminuche Wilderness
6.3
10.3
La Garita; Black Canyon of the Gunnison; Weminuche
WHRI1
White River NF
9.6
0.5
Flat Tops; Maroon Bells-Snowmass; West Elk; Eagles Nest
HALE1
Haleakala NP
Hawaii
7.5
13.3
Haleakala
HAVO1
Hawaii Volcanoes NP
7.2
18.9
1.9
Hawaii Volcanoes
CRMO1
Craters of the Moon NM
Idaho
7.6
14.0
Craters of the Moon
SAWT1
Sawtooth NF
6.5
1.2
Sawtooth
CABI1
Cabinet Mountains
Montana
14.1
GAMO1
Gates of the Mountains
11.3
0.8
GLAC1
Glacier NP 3
9.1
20.5
Glacier
MELA1
Medicine Lake
8.0
17.7
1.6
MONT1
Monture
7.8
14.5
Bob Marshall; Mission Mountains; Scapegoat
SULA1
Sula Peak
13.4
Selway-Bitterroot; Anaconda-Pintler
ULBE1
UL Bend
8.2
15.1

20. Worst Haze Natural II (dv) Worst Haze Baseline (dv)
Site
Name
State
Complete Years
Worst Haze Natural II (dv)
Worst Haze Baseline (dv)
10-year Glide Slope (dv)
Class I Area(s)
JARB1
Jarbidge Wilderness
Nevada 4
8.0
12.1
0.7
Jarbidge
BAND1
Bandelier NM
New Mexico 5
6.3
12.2
1.0
Bandelier
BOAP1
Bosque del Apache 3
6.8
13.8
1.2
GICL1
Gila Wilderness
13.1
1.1
Gila
SACR1
Salt Creek
6.9
18.0
1.9
SAPE1
San Pedro Parks
6.2
10.2
WHIT1
White Mountain
13.7
WHPE1
Wheeler Peak
6.6
10.4
0.6
Wheeler Peak; Pecos
LOST1
Lostwood
North Dakota
8.1
19.6
THRO1
Theodore Roosevelt
7.9
17.7
1.6
CRLA1
Crater Lake NP
Oregon
Gearhart Mountain; Crater Lake; Diamond Peak; Mountain Lakes
HECA1
Hells Canyon
8.4
18.6
1.7
KALM1
Kalmiopsis
9.5
15.5
MOHO1
Mount Hood
8.5
14.9
STAR1
Starkey
9.0
Eagle Cap; Strawberry Mountain
THSI1
Three Sisters Wilderness
8.9
15.3
Three Sisters; Mount Jefferson; Mount Washington
WRAP

21. Worst Haze Natural II (dv) Worst Haze Baseline (dv)
WRAP
Site
Name
State
Complete Years
Worst Haze Natural II (dv)
Worst Haze Baseline (dv)
10-year Glide Slope (dv)
Class I Area(s)
BADL1
Badlands NP
South Dakota 5
8.1
17.1
1.5
Badlands
WICA1
Wind Cave
7.8
15.8
1.3
BRCA1
Bryce Canyon NP
Utah
7.0
11.6
0.8
Bryce Canyon
CANY1
Canyonlands NP
6.5
11.2
Canyonlands; Arches
CAPI1
Capitol Reef NP 2
5.7
10.0
0.7
Capitol Reef
ZION1
Zion 3
7.1
13.2
1.0
MORA1
Mount Rainier NP
Washington 4
8.6
18.2
1.6
Mount Rainier
NOCA1
North Cascades
14.0
North Cascades; Glacier Peak
OLYM1
Olympic
8.5
16.7
1.4
PASA1
Pasayten
8.4
15.2
1.1
SNPA1
Snoqualmie Pass
17.8
Alpine Lakes
WHPA1
White Pass
12.8
Mount Adams; Goat Rocks
BRID1
Bridger Wilderness
Wyoming
6.7
11.1
Bridger; Fitzpatrick
NOAB1
North Absaroka
6.9
11.5
North Absaroka; Washakie
YELL2
Yellowstone NP 2
6.6
11.8
0.9
Yellowstone; Red Rock Lakes; Grand Teton; Teton