Introduction

The District divides southern SLO County into zones for the purposes of forecasting and reporting air quality. While ozone is typically assumed to be uniform across the zones, levels of particulate matter (PM10 and PM2.5) are more heterogeneous, especially during wind event days when onshores winds entrain dust from the Oceano Dunes SVRA. Each forecast zone is associated with air quality measurements made at a specific monitoring station, and the current zone boundaries were informed by the results of the District’s South County Community Monitoring Study, in which PM10 data was collected by a dense but temporary network of EBAMs on the Nipomo Mesa in 2012. The current map is depicted below and is available online here.

Since 2012 study, PM10 monitoring at the District’s CDF, Mesa2, NRP, and SLO permanent sites has continued, and in 2015 a new permanent site called Oso Flaco was established downwind of the southern extent of the ODSVRA. Short-term monitoring with both BAMs and EBAMs has also been conducted near Dorothea Lang Elementary School (DLES) in Nipomo, behind Lopez Continuation High School in Nipomo, off of North View Ave in Nipomo, and along Highway 1 in Oceano. See the zoomable map, below, for the locations of these monitors. At that same time, dust mitigation projects within the ODSVRA have likely impacted dust generation and changed the PM10 concentration gradient downwind. A recent analysis of the data from the temporary monitors near DLES concluded that the forecast zone map should be redrawn to put the school in NRP zone. In light of this, it was decided to examine all of the temporary site data and consider whether additional changes to the map are warranted. That is the goal of this analysis.

South County AQI Forecast Zones and Monitoring Sites: Forecast Zones: Red - CDF; Orange - Mesa2; Tan - NRP; Green - SLO.

Dataset

This report analyzes data collected from 2017 through early 2021. Other than short interruptions for maintenance, repairs, and QA/QC, the District’s permanent sites (CDF, Mesa2, NRP, SLO, and Oso Flaco) have run continuously during this period, with the exception of Oso Flaco, which offline in early 2017. The temporary sites in South County have operated more sporadically. The dataset also includes the permanent monitor in Santa Maria, which is operated by the Santa Barbara APCD. Most of the PM10 data was collected using FEM BAM monitors, but some data was generated with an EBAM and a non-FEM BAM. These data were adjusted using correction factors derived from collocation at CDF. The correction factors and their derivation are discussed in a separate report. The table below summarizes the sampling periods and instruments for the sites included the analysis.

Location Short Name First Period Instrument Second Period Instrument
CDF CDF 1/1/2017 - 2/5/2021 BAM
Mesa2 Mesa 1/1/2017 - 2/5/2021 BAM
NRP NRP 1/1/2017 - 2/5/2021 BAM
Oso Flaco Oso 3/25/2017 - 2/5/2021 BAM
San Luis Obispo SLO 1/1/2017 - 2/5/2021 BAM
Santa Maria SMaria 1/1/2017 - 2/5/2021 BAM
Calle Cielo Cielo 11/16/2018 - 4/11/2019 BAM 4/22/2020 - 12/1/2020 non-FEM BAM
Via Alta Mesa Alta 4/21/2020 - 7/6/2020 EBAM
Lopez Continuation HS Lopez 7/19/2017 - 7/22/2019 BAM 11/21/2019 - 2/5/2021 BAM
North View NorthV 3/7/2018 - 9/24/2018 BAM 4/12/2019 - 11/30/2020 BAM
Oceano CSD Oceano 4/4/2019 - 10/10/2019 BAM

The forecast map was designed primarily to capture the PM gradient caused by wind-blown dust from the ODSVRA. Other potential sources—like wildfire smoke and dust transport from the San Joaquin Valley—have much more homogeneous impacts on the area’s air quality. The following days were influenced by these sources and were therefore removed from the dataset: 9/2/2017, 12/4/2017, 12/16/2017, 12/17/2017, 11/17/2018, 11/18/2018, 12/4/2018, 12/16/2018, 12/17/2018, 10/27/2019, 10/28/2019, 10/28/2019, 10/29/2019, 10/29/2019, 10/30/2019, 10/30/2019, 10/31/2019, 11/1/2019, 11/2/2019, 7/6/2020, 8/18/2020, 8/19/2020, 8/20/2020, 8/21/2020, 8/22/2020, 8/23/2020, 9/11/2020, 9/12/2020, 9/13/2020, 9/14/2020, 9/15/2020, 9/16/2020, 10/2/2020, 10/3/2020, 10/4/2020.

Analytical Methods

The fundamental question is: Which permanent site is most similar to each of the temporary sites? The previously mentioned analysis of sites near DLES compared the Calle Cielo and Via Alta Mesa temporary sites to NRP and Mesa2 and determined that the area was better represented by NRP than by Mesa2. That analysis looked at ratios of the PM10 values between the temporary and permanent sites. For example, it was found that for Calle Cielo 24-hour averages were 76.0% of those at NRP and 57.7% of those at Mesa2, indicating that both permanent sites tend to overestimate concentrations at Calle Cielo, but NRP is a better surrogate than Mesa2. That approach is expanded on here.

Concentration Ratios

When comparing two sites, if the “average” or “typical” concentration ratio is close to one, then the sites are very similar; progressively larger or smaller ratios indicate progressively greater differences. There are a few considerations when calculating concentration ratios:

  • Whether to work with hourly values or 24-hour averages.
  • Whether to only include wind-event days or use all days in the dataset. (Here, a wind-event day is defined as any day with an exceedance of the state PM10 standard at either CDF or Mesa2.)
  • How to calculate the average ratio.

As discussed below, these issues were explored by calculating the ratios multiple ways—using hourly versus daily averages, including versus excluding non-wind-event days, and using two different ratio calculation methods—to test their impact on the calculated concentration ratios.

The ratio calculation method bears some explanation. Intuitively, calculating the “average ratio” of concentrations between two sites seems as easy as calculating the ratios for each hour (or each 24-hour average) and then averaging the ratios, i.e., calculating the average ratio of \(x\) and \(y\) as \(\frac{1}{n} \sum_{i=1}^{n}\frac{x_i}{y_i}\). This procedure can be unstable, as a small number in the denominator can yield a very large ratio which can skew the average. Instead, it is better to calculate the “average ratio” as the average concentration at one site divided by the average concentration at the other, i.e., as \(\frac{\sum_{i=1}^{n} x_i} {\sum_{i=1}^{n} y_i}\). Another approach to calculating “average” or “typical” ratio is use linear regression, but with the regression line forced through the origin (i.e. with the y-intercept set to zero.) The ratio is then the slope of the resulting fit. This method was also explored here, but using Deming regression rather than ordinary least squares, since both variables are subject to measurement error.

Concentration Bias

In addition to assessing concentration ratios between sites, the bias between sites was also examined. Bias was calculated as the average difference between concentrations, i.e., \(\frac{1}{n} \sum_{i=1}^{n} {(x_i-y_i)}\). A concentration bias of zero would indicate that on average the sites measure the same concentration.

Visualization of Results

Colored matrices were used to visualize the results. For each comparison method (e.g., daily averages compared using ratios calculated via Deming regression and including all data), each site was compared to every other site and the resulting concentration ratio or bias values are displayed as a color-code matrix. For concentration ratios, values close to 1 are colored purple, and values far from 1 are colored yellow, with intermediate values shaded accordingly. The values on the diagonal are all 1, since these represent the site compared to itself. For example, in Figure 1, below, the ratio for CDF/Mesa2 is 1.146, while the value for Mesa/CDF is 0.872; note that these values are reciprocal, as expected, and are colored the same shade for clarity. For concentration biases (e.g. Figure 5), values close to zero are colored purple, and values with large absolute values are shaded yellow. The monitoring periods for Oceano CSD and Via Alta Mesa did not overlap, so these site cannot be compared to each other and the matrix entries for this pair are blank.

Results

Deming Regression Versus Ratio of Means

As noted above, there are different ways to calculate the “average” or “typical” ratio between sites—the two explored are here are the ratio of means (or, equivalently, sums) and using Deming regression through the origin. In practice, these two methods give nearly identical results, as shown in the examples below. These two figures both show concentration ratios calculated for daily averages using all data in the dataset (i.e., both wind-event days and non-wind-event days, but excluding days influenced by wildfires or San Joaquin Valley dust.) In Figure 1, ratios were calculated via Deming Regression; Figure 2 used the ratio of means method. While the numbers change slightly, the overall pattern is nearly identical, as indicated by the coloring of the cells. This is also true when comparing the methods using hourly data or looking only at wind-event days (not shown). Therefore, only results using the ratio of means method will be considered.

Daily Averages Versus Hourly Days

If no data were missing, then whether using hourly data or daily averages would give identical results when calculating concentration ratios by the ratio of means method. In practice, however, at least a few hours of data are lost at each site each month due to QA/QC activities, maintenance and repairs, and power failures. As long as the day is at least 75% complete, the daily average calculated from the remaining valid hours is considered valid. Thus, a multi-day average of daily averages is not necessarily the same as a multi-day average of the underlying hourly values. In effect, hours from incomplete days receive slightly greater weight in the grand average than hours from complete days.

The effect of this on concentration ratios was examined by calculating them using both hourly values and daily averages. Figure 2, above, shows ratio for all data calculated on the 24-hour averages. Figure 3, below, is analogous, but shows the results calculated using hourly values. As can be seen, the results are nearly identical, therefore only daily averages are considered.

All Data Versus Only Wind-Event Days

Next, the effect of including only wind-event days was investigated. Figure 2, above, shows ratios calculated via the ratio of means method on 24-hour averages, using all data in the dataset (except days influenced by wildfires and San Joaquin Valley dust). Figure 4, below, shows the same except only wind-event days were included. Here, the differences between sites are more pronounced. For example, when assessed with all data, the CDF/Oso Flaco ratio is 1.472 (see Figure 2), but when restricted to wind-event days, it increases to 1.993 (Figure 4).

Concentration Bias Versus Ratio

Figures 1 through 4, above, displayed the average ratio of concentrations at one site versus another. In this section, concentration biases are explored. As was done above for concentration ratios, the effect of calculating bias on daily averages versus hourly values was evaluated. Results for hourly and daily averages were nearly the same (not shown) so only daily averages are displayed and discussed below. The figures below show concentration bias calculated with daily averages using all data (Figure 5) and only wind-event days (Figure 6). Note that these matrices are antisymmetric, since the mean of \(x - y\) equals minus the mean of \(y -x\). For example, Figure 5 shows that on average, the average of daily PM10 at CDF minus daily PM10 at Mesa2 is 3.012 ug/m3, while the average of Mesa2 minus CDF is -3.012 ug/m3.

Discussion

Figures 2 and 5 tend to agree about which permanent site best represents each temporary site in general, and Figures 4 and 6 tend to agree for wind-event days. In all cases, the best surrogate for wind-event days is different from the best surrogate for all days. This is not unexpected, since non-ODSVRA/background sources might influence sites differently than the ODSVRA. This is summarized in the following table:

Temporary Site Best Permanent Site (All Days) Best Permanent Site (Wind Events) Notes
Calle Cielo SLO NRP Overestimated during wind events
Via Alta Mesa NRP Oso Flaco Underestimated during wind events
Lopez Cont. HS CDF Mesa2 Overestimated during wind events
North View Oso Flaco NRP During wind events, inbetween Oso Flaco and NRP
Oceano Mesa2 NRP Underestimated during wind events

It is recommended that the forecast zone be redrawn based on which permanent site best represents each temporary site on wind-event days, since those days tend have the highest concentrations and largest concentration gradients. These recommendations are visualized on the map below, where the color of marker for each temporary site is the same as the permanent site that represents it—i.e., all markers of the same color should be in the same zone.

Proposed New Site Mapping for South County Forecast Zone

Conclusion

The current forecast zones were informed by the saturation monitoring campaign of the South County Community Monitoring Study. It is recommended to start with that map and then refine the zone boundaries as follows:

  • Shift the northern boundary of the Mesa2 zone southward to put the DLES area into the NRP zone; currently this area is in the Mesa2 zone. (Oso Flaco is a slightly better surrogate for Via Alta Mesa than NRP, but for simplicity it is recommended to use NRP.)
  • Shift the eastern boundary of the NRP zone further east, to put the area near the Oceano CSD in the NRP zone. Currently it is in the SLO zone. The Community Monitoring in Oceano report tentatively concluded that there was a “west-to-east gradient of PM10 values during windblown dust events” in the area, and this observation could be used to further refine the zone boundary.
  • Consider defining an Oso Flaco forecast zone for the southern portion of the region.

Is worth noting that this analysis confirms that North View Ave is correctly in the NRP zone and that Lopez Continuation HS is correctly in the Mesa2 zone.