Anthropogenic aerosols mask increases in US rainfall by greenhouse gases

A comprehensive understanding of human-induced changes to rainfall is essential for water resource management and infrastructure design. However, at regional scales, existing detection and attribution studies are rarely able to conclusively identify human influence on precipitation. Here we show that anthropogenic aerosol and greenhouse gas (GHG) emissions are the primary drivers of precipitation change over the United States. GHG emissions increase mean and extreme precipitation from rain gauge measurements across all seasons, while the decadal-scale effect of global aerosol emissions decreases precipitation. Local aerosol emissions further offset GHG increases in the winter and spring but enhance rainfall during the summer and fall. Our results show that the conflicting literature on historical precipitation trends can be explained by offsetting aerosol and greenhouse gas signals. At the scale of the United States, individual climate models reproduce observed changes but cannot confidently determine whether a given anthropogenic agent has increased or decreased rainfall.


Model
hist pi 1pct Model hist pi 1pct 1 Sufficiency means that the output from each ensemble member available from the Earth System Grid Federation (ESGF; [5]) includes the daily and monthly precipitation rates and the monthly variables needed to calculate the low-frequency drivers in main text Eq. 2.
2 Models with at least one ensemble member from historical, pi-Control (with at least 500 years), and 1pctCO2 (25 total). 3Models whose pi-Control runs are less than 500 years. 4Models with errors in emiso2 (vertically-integrated total emission of SO 2 ) files.
Table 3 Number of years for each Global Climate Model with daily precipitation data for each PDRMIP experiment [6].Here, "base" corresponds to present-day conditions and "sulx5" corresponds to present-day SO 4 (sulfate aerosol) concentrations multiplied by a factor of 5; each of these configurations has both a prognostic SST ("coupled") and fixed SST ("fsst") experiment.Fig. 1 Signal-to-noise ratio (SNR) estimates for the effect of anthropogenic aerosols on seasonal precipitation as simulated by individual ensemble members of the single-forcing hist-aer experiment [7].Here we show the CONUS-average precipitation difference between 1851-1880 (representing pre-industrial conditions) and 1952-1981 (the 30 years in which SO 2 emissions were at their largest in CONUS).The plotted points show the ensemble average for each model (black) and the multimodel ensemble average (red), while the lines show the minimum and maximum SNR from each ensemble.The dashed vertical lines are at ±1.645, the SNR threshold corresponding to a significance level of α = 0.1.Note that the multimodel ensemble average has very low SNR, failing to exceed the ±1.645 threshold in all seasons and for both mean and extreme precipitation, while individual ensemble members have statistically significant SNR that indicates both drying (< −1.645) and wetting (> 1.645) for seasonal precipitation at the CONUS scale.Fig. 3 Nested attribution regions, as defined in [10], that subsequently divided the CONUS into two, four, 13, and 75 subregions.The attribution regions correspond to spatial scales of ≈ 8Mm 2 (all of CONUS), ≈ 4Mm 2 (two subregions), ≈ 2Mm 2 (four subregions), ≈ 0.5Mm 2 (13 subregions), and ≈ 0.1Mm 2 (75 subregions), where 1Mm 2 = 1 million km 2 ; the grid boxes are ≈ 600km 2 .Hatching indicates that the 90% bootstrap confidence interval does not include zero.Calculations are described in "Fast versus slow precipitation response to aerosols" in Methods; the GCMs used are listed in Supplemental Table 3. Fig. 7 Regionally-averaged time series of seasonal precipitation anomalies from the preindustrial (1900) climate of mean and extreme precipitation for the two ≈ 4Mm 2 subregions shown in the rightmost column.Each panel shows the isolated effect of anthropogenic forcing agents on seasonal precipitation (GHG, solid red line; AER-glob, the slow precipitation response to aerosols, solid blue line; AER-local, the fast precipitation response to aerosols, solid green line) as well as the combined anthropogenic response (ANT; solid black line) with a 90% bootstrap confidence band.Dashed vertical lines denote the year of emergence for isolated GHG signal (red) and combined ANT response (black), where emergence is defined as the first year in which the 90% confidence band departs from zero.Fig. 8 Regionally-averaged time series of seasonal precipitation anomalies from the preindustrial (1900) climate of mean and extreme precipitation for the four ≈ 2Mm 2 subregions shown in the rightmost column.Each panel shows the isolated effect of anthropogenic forcing agents on seasonal precipitation (GHG, solid red line; AER-glob, the slow precipitation response to aerosols, solid blue line; AER-local, the fast precipitation response to aerosols, solid green line) as well as the combined anthropogenic response (ANT; solid black line) with a 90% bootstrap confidence band.Dashed vertical lines denote the year of emergence for isolated GHG signal (red) and combined ANT response (black), where emergence is defined as the first year in which the 90% confidence band departs from zero.
Fig. 11 The spatial distribution of the n = 2480 GHCN stations with a minimum of 66.7% of existent, quality-controlled daily precipitation measurements during the period spanning December 1899 to November 2020.

Fig. 2
Fig.2Reconstructed best-estimate time series of external anthropogenic forcings used for the various analyses in this paper over 1900 to present day for well-mixed greenhouse gases (sum-total GHG forcings in W m −2 ; panel a.) and CONUS-average SO 2 emissions, obtained from[8,9], for each season (panel b.; March/April/May (MAM) and September/October/November (SON) are nearly overlapping).

Fig. 4 Fig. 5
Fig.4Comparison of the estimated effect of SO 2 emissions on in-situ rainfall measurements when using a global CONUS-wide emissions time series (top row of panel a. and b.) versus the stochastically-regionalized emissions (bottom row of panel a. and b.).Note that the bottom row of each panel are as in Figure2in the main text.Hatching indicates where the changes are statistically significant, where a − (+) indicates moderate (strong) significance.

Fig. 6
Fig.6Comparison of the summertime fast precipitation response to aerosols for our GHCN station-based analysis versus corresponding patterns from PDRMIP GCM simulations (showing the multi-model mean from eight GCMs).Thumbnail maps taken from Figure2in the main text and Supplemental Figure5; note that the color bar limits in panels (a) vs.(b) and (c) vs.(d) are different due to the differing magnitude of SO 4 loadings in the simulations versus observations.(Note: hatching has different meanings in the GHCN vs. PDRMIP plots.) Change in precipitation rate: 2014 forcing minus 1900 forcing (b) Change in 20−year return value: 2014 forcing minus 1900 forcing

Fig. 12
Fig.12Sum-total forced changes to seasonal mean (panel a.) and extreme (panel b.) precipitation for 2014 versus 1900 forcing levels (GHGs and SO 2 emissions) for the GHCN analysis (top row in each panel) and the weighted CMIP6-historical multimodel mean (bottom row in each panel).Each panel includes the area-weighted CONUS-mean change with uncertainty (best estimate and lower/upper 90% confidence bounds).Stippling indicates where the grid-box 90% confidence intervals include zero, i.e., the changes are indistinguishable from zero.Note that the color bar limits are different for the GHCN vs. CMIP6 maps.