3.1. Summertime Warmth Linked to Improved Air Quality
The twentieth century warming hole was especially notable during the summer (
Figure 2a) when anthropogenic aerosol concentrations in the SEUS are highest [
49]. Summertime temperatures (May–September) in the SEUS decreased by a negligible −0.01 ± 0.35
C century
between 1900 and 2008, compared with a substantial +0.99 ± 0.35
C century
increase in Western US (35–48
N; 125–110
W) temperatures during the same time period (
Figure 2c). Time series analysis using LOESS (locally weighted scatterplot smoothing) curve smoothing shows that much of the 20th century decrease in SEUS summertime surface temperature occurred before 1975, plateauing between 1975 and 1990, and with a slight rebound beginning afterward; indeed, the 1900–1975 trend in SEUS surface temperature was a robust −0.27 ± 0.25
C century
(
Figure 3). This period roughly coincided with the peak in SEUS tropospheric aerosol concentration [
14].
In recent years, the summertime warming hole has not only disappeared but reversed (
Figure 2b). The 2001–2015 temperature trend in the SEUS was +0.54 ± 0.52
C decade
, compared with +0.18 ± 0.62
C decade
in the Western US over the same period. Using the alternative Theil–Sen linear trend-fitting algorithm [
50], the reversal is still notable and significant, with a positive +0.50
C decade
trend in SEUS versus a negative −0.18
C decade
trend in the Western US. When only the most recent decade (2001–2010) is considered, ensuring that the average contains input from all three primary surface temperature records, the contrast is more striking: +0.69 ± 0.68
C decade
in SEUS and −0.48 ± 0.74
C decade
in the Western US.
Contemporary (2001–2015) satellite AOD measurements from the Multi-angle Imaging SpectroRadiometer (MISR) instrument [
41] (Version 22, Level 3) show high summertime aerosol loading in the SEUS, especially when compared with the Western US where population densities are lower (
Figure 4a). Summertime AOD (from MISR) in the SEUS was, on average, 0.05 to 0.10 greater than the long-term annual mean. In contrast, summertime AOD in the Western US was only 0.00 to 0.04 greater during the summer months.
The trend in regionally averaged summertime AOD in the SEUS was −0.05 decade
between 2001 and 2015 (
Figure 2d,
Figure 4b). In contrast, summertime reductions in Western US AOD were minimal, where they existed at all (
Figure 4b). Using a bootstrapping technique for error calculation (
n = 1,000), we estimate that the 14-year reduction in SEUS AOD was statistically significant (−0.05 ± 0.03). However, a decomposition of the linear trend reveals a sharp drop in AOD between 2007 and 2008. The linear trend from 2001–2007 was near-zero (+0.01 decade
), and the linear trend from 2008 to 2015 was insignificantly negative (−0.03 decade
), suggesting that the majority of the decrease occurred between 2006 and 2009; indeed, the trend from 2005 to 2010 was −0.16 decade
. The smoothed curve, fit using LOESS smoothing techniques (
n = 5 years), further illustrates this sharp transition. We note that an EPA rules change to the Clean Air Act (CAA) National Ambient Air Quality Standards (NAAQS) for Fine Particulate Matter (PM
), implemented beginning in 2006 (Environmental Protection Agency, 40 C.F.R. § 50, 71 FR 2620) corresponds well with the aerosol changes resolved in the data. The rule lowered the 24-h PM
standard from 65
g m
to 35
g m
. That the timing of this rule and its subsequent implementation matches well with the sudden trend reversal represents a circumstantial link that may reconcile the observed relationship [
51], though we acknowledge that we have only limited evidence indicating a steep drop in surface-level particulate matter concentration from 2007 to 2009 for several southeast US stations.
The 2001–2015 AOD decrease in the eastern United States (
Figure 4b) coincided with a marked increase in average surface temperature (
Figure 2d). The spatial correlation of AOD decrease and temperature increase implies that warmer regional temperatures were driven in some manner by improved air quality. To reconcile this relationship, we used vertical profiles of aerosol extinction combined with a one-dimensional radiative transfer model (
Section 2.2 and
Section 2.3) to assess whether the link could be corroborated quantitatively or was mere coincidental correlation.
Summertime monthly mean profiles of SEUS aerosol light extinction coefficient were derived for 2006–2014 using data from the CALIOP instrument [
38] (see
Section 2). Similar to our results using MISR, CALIOP data show significant reductions in aerosol burden even over the shorter 2006–2014 time series (
Figure 4c). More importantly, CALIOP data show that reductions in aerosol burden were not restricted to the surface layer but persisted as high as 5 km above the surface (though not much higher). The mean summertime (May–September) aerosol extinction coefficient decreased by an average of 35% from 2007 (the first full year of data) to 2014 (
Figure 4c). In fact, the mean integrated AOD above the ABL was 41% lower (from 0.13 ± 0.00 to 0.07 ± 0.01) during 2008–2014 when compared with 2006–2007. This contrasts with a 26% reduction for the same time period for the entire tropospheric column (from 0.30 ± 0.02 to 0.23 ± 0.02) (
Figure 4c). Furthermore, this decrease was not the result of a positive trend in ABL height, which can enhance the near-surface contribution to total aerosol loading. The ERA interim data show that, while the mean summertime regional ABL experiences interannual fluctuations, the overall trend was neither positive or negative.
Seasonal increases in the amount of secondary aerosols present above the boundary layer in the summer have been proposed to explain the apparent discrepancies noted between the total column aerosol burden from satellites and the measured near-surface pollution reported in EPA pollution measurements [
49]. Our data support these findings, but, more notably, the CALIOP data show no apparent trend in aerosol composition from 2007 to 2014, from the surface through the mid-troposphere, indicating a continuity in relative airmass physical properties approaching the surface despite an overall reduction in their relative magnitudes.
3.2. Modeling Results Corroborate Observations
The FLG radiative transfer model was used to estimate the direct effect of these reductions in aerosol burden on the surface energy budget and column radiative heating profile, and to examine how that effect compares with the observed temperature trends previously discussed. We forced the model (see
Section 2.3) with observed changes in vertically-resolved, CALIOP-derived aerosol extinction, broadband surface reflectance from MISR, and standard meteorological conditions, and estimated the direct radiative forcing of the reduced aerosol burden from 2007 to 2014 at solar noon. We performed two experiments, which are described in
Section 2.3: FULL (direct forcing effect of aerosols) and AERO (direct forcing effect of aerosols minus a control simulation with no aerosols). Here, we present results from both, but focus on the FULL simulation.
The observed 35% decrease in CALIOP-measured aerosol extinction between 2007 and 2014 resulted in a 29 W m
(30 W m
for AERO) increase in solar noon net surface energy flux (R
) (
Figure 5, inset). When we consider the respective CALIOP aerosol profiles for every year (2007 to 2014), the linear trend, which factors in all the intermediate years (2008–2013; see
Table 1) in R
was 2.3 W m
year
(2.5 W m
year
for AERO), which, when multiplied by seven (the number of years from 2007 to 2014), was less than the 29 W m
absolute difference between 2014 and 2007, but was nevertheless similar. These values are similar to what we derived using an empirical model generated for a Rayleigh-only atmosphere with a direct and diffuse component and gaseous absorption and the observed total column change in AOD (−0.09) [
52]. Despite being the average for the entire SEUS, the positive surface energy fluxes also correspond remarkably well with observations from the Goodwin Creek, MS, USA SURFRAD (Surface Radiation Budget Network) 34.25
N, 89.87
W;
https://www.esrl.noaa.gov/gmd/grad/surfrad/goodwin.html) surface radiation monitoring station (
Figure 5, inset,
Figure 6), which is located in a region of particularly large aerosol reductions (e.g.,
Figure 4b), is located outside of urbanized area in an area of representative vegetation, and is the only SURFRAD site within our region of interest.
Curiously, while surface energy fluxes increased in the radiative transfer model over the temporal study period, total top of atmosphere (TOA) fluxes decreased between 2007 and 2014: −53.3 W m in the FULL simulation and −50.5 W m in the AERO simulation. As described below, this was probably a result of decreased aerosol absorption in the middle troposphere. In fact, while the trend in TOA radiative forcing was negative, the magnitude was still positive (e.g., 114.8 W m in 2007 to 61.6 W m in 2014 in the FULL simulations).
The FLG-estimated positive change in surface energy flux is a result of less solar absorption and scattering by aerosols in the atmospheric column. In the FULL simulation, lowered aerosol burdens from 2007 to 2014 resulted in a −3.4 K day
average decrease in solar noon heating rates between the surface and 4 km (
Figure 5). The maximum decrease in heating rates occurred between 0.5 and 2.5 km, corresponding with large observed temperature reductions (2014–2007) at 0.7 km in the temperature sounding data from Birmingham, AL and Jackson, MS (
Figure 5). The slight misalignment of the modeled and observed peaks is likely due to differences in the vertical resolution of the sounding data and the model as well as temporal differences; the soundings were taken at 7:00 a.m. (1200 UTC) and 7:00 p.m. (0000 UTC next day) local time, while the model simulations were performed for solar noon (1:00 p.m. local time). Atmospheric heating rates (and, subsequently, temperatures) throughout the lower troposphere were lower in 2014 because the atmospheric aerosol burden was lower. The model results indicate that a decrease in aerosol burden allowed more radiation to reach the surface and atmospheric absorption and semi-direct heating were suppressed. The results are consistent with the observed surface temperature trends, indicating that aerosols played a substantial role in the increasing temperatures observed in this region.