Appendix A

Emission Fluxes of On- and Off-Lake Sources

This appendix explains the approach used to compare emission fluxes between on- and off-lake sources impacting monitors around Owens Lake adjusted for meteorology. This analysis is used to assess the relative level of emissions leading to exceedances, potential for future exceedances related to historical emissions, and the relative emission impacts from on- versus off-lake sources. More detailed modeling (e.g., using dispersion modeling) could be used to better quantify emissions from specific source areas, more fully accounting for source location and system geometry, as recommended in the report.

EMISSION FLUXES OF ON- AND OFF-LAKE SOURCES

The figures in Chapter 3 (Figures 3-5, 3-8, 3-11, 3-14, 3-17, and 3-21) show that most of the PM₁₀ monitoring stations are surrounded by on- and off-lake dust sources. This suggests the application of an idea that underlies a dispersion model proposed by Gifford and Hanna (1970) and Hanna (1971) to compute pollutant concentrations in an urban area. The model assumes that a monitor measuring concentrations is surrounded by areas with pollutant emissions that are relatively uniform across an urban area. Gifford and Hanna (1970) then show that the seasonally averaged concentration at the monitor is governed primarily by the emission flux (emission rate per unit area of the source) of the upwind area in the immediate vicinity of the monitor and the wind speed at the monitor. The concentration, at the monitor is given by the simple formula

where q is the emission flux of the upwind area contributing to the concentration, and U is the wind speed at the monitor. The variable A = CU/q, derived from annual averages of particulate concentrations (total suspended particles) and 1/U, is relatively insensitive to the size of the city (or contributing area) as seen in Figure A-1. A varies by a factor of two around the value of 225 when the inferred size of the city (or contributing area) varies by a factor of 7. This indicates that the product of the PM₁₀ concentration, C, and the wind speed at a monitor, U, can be used to compare the relative magnitudes of upwind emission fluxes at different monitors as long as the extents of upwind areas affecting the monitors do not differ significantly.

The locations of the monitors around Owens Lake relative to the sources impacting them are similar to those in the urban areas studied by Gifford and Hanna (1973) in that they are surrounded by sources of dust. This sug-

gests a method to estimate fluxes from the upwind areas of the monitors using the relationship q = CU/A. The relative insensitivity of A to the extent of the upwind area that impacts a monitor can be readily confirmed with a calculation that accounts for the heights of the PM₁₀ monitors and wind measurements and the neutral stability associated with the wind speeds that gives rise to exceedances. Furthermore, the magnitude of A does not play a role in assessing emissions trends at a site if the source-receptor geometry does not change over time (i.e., the source area and monitor location are consistent).

Using hourly wind speed, wind direction, and PM₁₀ concentration data from all monitors from 2000 to 2023 (Chris Howard, personal communication, November 2024; Ann Logan, personal communication, July 2024), the panel calculated normalized emission fluxes for monitors around the lake. The wind directions used in computing UC correspond to the wind sectors that the Greater Basin Unified Air Pollution Control District (the District) associates with on- and off-lake dust sources. The emission flux averaged over a year is q(average) = average(U_iC_i), where the subscript ‘i’ corresponds to hourly concentrations and wind speeds measured from the on- and off-lake sectors over 1 year, and the factor A is absorbed in the normalization, described below. Given the sensitivity of PM₁₀ emissions to wind speed (Gillette, Ono, and Richmond 2004; Jiang, Liu, and Doyle 2011), effective emission fluxes are taken to be averages over values corresponding to wind speeds between 10 and 20 m/s, a range likely to yield peak monitor-recorded concentrations. This range is similar to the maximum average hourly wind speeds observed at Owens Lake during exceedance events (see Figure 5-2).

The annual computed values of q at the monitors were normalized by the value of q at Keeler averaged over the entire period (thus, on the graph for Keeler, the normalized emissions cross one during the period); this provides a comparison of emission fluxes relative to the value at Keeler. The choice of Keeler is made because of the relatively large number of exceedances in Keeler that have been attributed to off-lake sources. Normalizing by results from another location would only shift the values found and not change the findings.

The magnitudes of the normalized emission fluxes can be interpreted by plotting them against the associated annual exceedances at the monitors (Figure A-2); these exceedances are related to on- and off-lake dust sources

using the wind direction sectors specified by the District. They do not include exceedances that are attributed partially to both on- and off-lake sources. The plot shows that, as expected, the number of annual exceedances at a monitor increases with upwind emission flux. Note that relative emission flux (normalized to Keeler Dunes) as low as 0.2 can lead to 5 exceedances at a monitor.

Although all available data from 2000 to 2023 from all monitors was used for Figure A-2, Keeler and Dirty Socks were the only monitoring stations for which the panel had complete wind and PM₁₀ concentration records from 2000 to 2023 (Figures A-3 and A-4). The panel did not have access to complete hourly wind and concentration data for the other monitors until after 2008, and many monitoring stations were also missing additional data between 2008 and 2023. Trends for the data at the Keeler and Dirty Socks monitors were extracted with singular spectrum analysis (SSA; Elsner and Tsonis 1996; Vautard, Yiou, and Ghil 1992), a decomposition technique widely used in climatology and meteorology.

The shaded areas around the trend lines are the 95 percent confidence intervals of the trends obtained by bootstrapping the residuals between the computed fluxes and the first estimate of the trend from the SSA. The computed fluxes (based on measured concentrations and wind speeds) are assumed to be lognormally distributed about the corresponding SSA prediction so that the residual is the logarithm of the computed flux to the corresponding SSA estimate. A set of 1,000 “pseudo-fluxes” were created by adding the residuals randomly to the logarithms of the SSA estimates. Then, SSA trends were created for each of the pseudo-fluxes. The limits of the shaded areas for each year correspond to the 2.5th and the 97.5th percentiles of the 1,000 SSA predictions for each year.

At Keeler and Dirty Socks, current off-lake emission fluxes are markedly higher than on-lake fluxes and show small or no statistically significant temporal decline (Figures A3 and A4). Given these sustained off-lake levels, exceedances of the PM₁₀ standard at downwind monitors are likely unless additional controls are implemented.