Role of atmospheric rivers in shaping long term Arctic moisture variability

Abstract

Atmospheric rivers (ARs) reaching high-latitudes in summer contribute to the majority of climatological poleward water vapor transport into the Arctic. This transport has exhibited long term changes over the past decades, which cannot be entirely explained by anthropogenic forcing according to ensemble model responses. Here, through observational analyses and model experiments in which winds are adjusted to match observations, we demonstrate that low-frequency, large-scale circulation changes in the Arctic play a decisive role in regulating AR activity and thus inducing the recent upsurge of this activity in the region. It is estimated that the trend in summertime AR activity may contribute to 36% of the increasing trend of atmospheric summer moisture over the entire Arctic since 1979 and account for over half of the humidity trends in certain areas experiencing significant recent warming, such as western Greenland, northern Europe, and eastern Siberia. This indicates that AR activity, mostly driven by strong synoptic weather systems often regarded as stochastic, may serve as a vital mechanism in regulating long term moisture variability in the Arctic.

Fig. 1. Relationships between summer atmospheric rivers (ARs) and atmospheric variables.

a–d Linear trends of June–July–August (JJA) AR frequency (days/month/decade) (a), lower to middle tropospheric (surface to 500 hPa average) specific humidity (g/kg/decade) (b), 200 hPa geopotential height (Z200, shaded) (m/decade) and 200 hPa winds (200hPa_Wind, vectors) (m/s/decade) (c), and tropospheric (surface to 200 hPa average) air temperature (K/decade) (d) in the Arctic from 1979 to 2019. e, f Spatial pattern of the leading Maximum Covariance Analysis (MCA) mode of detrended summer Z200 (e, in m) and AR frequency (f, in days/month) over the period. g Standardized time series of the leading MCA mode of summer Z200 (orange line) and AR frequency (blue line). Black dots in (a–d) denote statistically significant trends at the 95% confidence level. The scale for the wind trend is shown at the bottom-right corner of (c). Wind anomalies (vectors, m/s) at 200 hPa in (f) are obtained by linear regression of anomalous JJA winds at 200 hPa against the standardized time series of the Z200 pattern in the leading MCA mode. The scale for wind anomalies is shown at the bottom-right corner of (f). “SCF” in (g) indicates the squared covariance fraction of the leading MCA mode. “r” in (g) indicates the correlation coefficient between the time series of Z200 and AR frequency patterns in the leading MCA mode.

Fig. 2. Connections between atmospheric rivers (ARs) and specific humidity on a day-to-day time scale.

a, b Spatial pattern of the leading Maximum Covariance Analysis (MCA) mode of ARs (a, in days) and lower to middle tropospheric (surface to 500 hPa average) specific humidity (b, in g/kg) over western Greenland in summer 2012 (detrended and with climatological seasonal cycles removed). c Composite 6-hourly ARs (days) and lower to middle tropospheric specific humidity (g/kg) anomalies (from four days before to 5 days after the outbreak of ARs) within the western Greenland region from 1979 to 2019 (detrended and with climatological seasonal cycles removed). d Standardized time series of the leading MCA mode of ARs and lower to middle tropospheric specific humidity over western Greenland in summer 2012. “SCF” in (d) indicates the squared covariance fraction of the leading MCA mode. “r” in (d) indicates the correlation coefficient between the time series of the two patterns in the leading MCA mode. Numerous long horizontal blue lines in (d) indicate the absence of AR activity during the period. Both AR activity and humidity anomalies reached the maximum on July 11, which is marked in (d).

Fig. 3. Contribution of atmospheric rivers (ARs) to observed trends of specific humidity.

a, b Linear trends (g/kg/decade) of summer-specific humidity at lower to middle troposphere (surface to 500 hPa average) unrelated (a) and related (b) to the activity of ARs from 1979 to 2019. Black dots denote statistically significant trends at the 95% confidence level.

Fig. 4. Simulated influences of large-scale circulation and anthropogenic forcing on Atmospheric River (AR) changes.

a, c Differences of summer AR frequency (days/month) between the nudging run and the control run (WIN minus CTL) (a) and between the nudging+anthropogenic forcing run and the control run (WIN + CO₂ minus CTL) (c) based on the 40-year simulations. b, d Differences of summer geopotential height at 200 hPa (Z200) (m) between the nudging run and control run (b) and between the nudging+anthropogenic forcing run and the control run (d) based on the 40-year simulations. Wind anomalies (vectors, m/s) at 200 hPa in (b) are the differences of summer winds at 200 hPa between the nudging run and control run based on the 40-year simulations. The scale for winds is shown at the bottom-right corner of (b). Black dots denote statistically significant differences at the 95% confidence level for AR frequency in (a, c), and for Z200 in (b, d). WIN refers to the simulation with wind nudging used. CTL refers to the control run. WIN + CO₂ refers to the runs with both wind nudging and anthropogenic forcing imposed.

Fig. 5. Reconstructed trends of atmospheric rivers (ARs) and geopotential height at 200 hPa (Z200) in the subgroups of CESM2-LEN.

a, c Linear trends of summer AR frequency (days/month/decade) derived from the subgroup exhibiting strong increasing AR frequency trends over western Greenland (8 members) (a) and over both western Greenland and eastern Siberia (2 members) (c). b, d Linear trends in summer Z200 (m/decade) based on the subgroup with strong increasing AR frequency trends over western Greenland (b) and over both western Greenland and eastern Siberia (d). The 40 members from CESM2-LEN are used in this calculation (see “Methods”). Black dots denote statistically significant trends at the 95% confidence level.