An official website of the United States government
Abstract The increase in Greenland Ice Sheet (GrIS) surface runoff since the turn of the century has been linked to a rise in Greenland blocking frequency. However, a range of synoptic patterns can be considered blocked flow and efforts that summarize all blocking types indiscriminately likely fail to capture consequential differences in GrIS response. To account for these differences, we employ ERA5 reanalysis to identify summer blocking using two independent blocking metrics: the Greenland Blocking Index (GBI) and the blocking index of Pelly and Hoskins (2003,https://doi.org/10.1175/1520-0469(2003)060<0743:ANPOB>2.0.CO;2). We then conduct a self‐organizing map analysis to objectively classify synoptic conditions during Greenland blocking episodes and identify three primary blocking types: (a) a high‐amplitude Omega block, (b) a lower‐amplitude, stationary summer ridge, and (c) a cyclonic wave breaking pattern. Using Modèle Atmosphérique Régional output, we document the spatiotemporal progression of the surface energy and mass balance for each blocking type. Relative to all blocking episodes, summer ridge patterns produce more melt over the southern ice sheet, Omega blocks produce more melt across the northern ice sheet, and cyclonic wave breaking patterns produce more melt in northeast Greenland. Our results indicate that the recent trend in summer Greenland blocking was largely driven by an increase in Omega patterns and suggest that Omega blocks have played a central role in the recent acceleration of GrIS mass loss. Furthermore, the GBI exhibited a relative bias toward Omega patterns, which may help explain why it has measured stronger trends in summer Greenland blocking than other blocking metrics.
more »
« less
Using Explainable AI and Transfer Learning to Understand and Predict the Maintenance of Atlantic Blocking With Limited Observational Data
Abstract Blocking events are an important cause of extreme weather, especially long‐lasting blocking events that trap weather systems in place. The duration of blocking events is, however, underestimated in climate models. Explainable Artificial Intelligence are a class of data analysis methods that can help identify physical causes of prolonged blocking events and diagnose model deficiencies. We demonstrate this approach on an idealized quasigeostrophic (QG) model developed by Marshall and Molteni (1993),https://doi.org/10.1175/1520‐0469(1993)050<1792:taduop>2.0.co;2. We train a convolutional neural network (CNN), and subsequently, build a sparse predictive model for the persistence of Atlantic blocking, conditioned on an initial high‐pressure anomaly. Shapley Additive ExPlanation (SHAP) analysis reveals that high‐pressure anomalies in the American Southeast and North Atlantic, separated by a trough over Atlantic Canada, contribute significantly to prediction of sustained blocking events in the Atlantic region. This agrees with previous work that identified precursors in the same regions via wave train analysis. When we apply the same CNN to blockings in the ERA5 atmospheric reanalysis, there is insufficient data to accurately predict persistent blocks. We partially overcome this limitation by pre‐training the CNN on the plentiful data of the Marshall‐Molteni model, and then using Transfer learning (TL) to achieve better predictions than direct training. SHAP analysis before and after TL allows a comparison between the predictive features in the reanalysis and the QG model, quantifying dynamical biases in the idealized model. This work demonstrates the potential for machine learning methods to extract meaningful precursors of extreme weather events and achieve better prediction using limited observational data.
more »
« less
- Award ID(s):
- 2004572
- PAR ID:
- 10579855
- Publisher / Repository:
- DOI PREFIX: 10.1029
- Date Published:
- Journal Name:
- Journal of Geophysical Research: Machine Learning and Computation
- Volume:
- 1
- Issue:
- 4
- ISSN:
- 2993-5210
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract The joint influence of the stratospheric quasi‐biennial oscillation (QBO) and the El Niño Southern Oscillation (ENSO) on the polar vortex, subtropical westerly jets (STJs), and wave patterns during boreal winter is investigated in 40 years (1979–2018) of monthly mean ERA‐Interim reanalyses. The method of Wallace et al. (1993),https://doi.org/10.1175/15200469(1993)050<1751:ROTESQ>2.0.CO;2is used to conduct a QBO phase angle sweep. QBO westerly (W) and easterly (E) composites are then segregated by the phase of ENSO. Two pathways are described by which the QBO mean meridional circulation (MMC) influences the northern winter hemisphere. The “stratospheric pathway” modulates stratospheric planetary wave absorption via the Holton‐Tan mechanism. The “tropospheric pathway” modulates the tropical and subtropical upper troposphere and lower stratosphere. QBO MMC anomalies exhibit a checkerboard pattern in temperature and arched structures in zonal wind which extend into midlatitudes, and are stronger on the winter side. During QBO W, the polar vortex and STJs are enhanced. QBO signals in the polar vortex are amplified during La Niña. During El Niño and QBO W, the strongest STJs occur, and a warm pole/wave two pattern is found. During El Niño and QBO E, a trough is found over Eurasia and a ridge over the North Atlantic, in a wave one pattern. El Niño diminishes QBO anomalies in the tropical stratosphere and reduces the poleward extent and amplitude of the QBO MMC, thereby influencing the stratospheric pathway. Effects on the boreal winter hemisphere are attributed to the combined influence of the QBO and ENSO via both pathways.more » « less
-
Abstract The polarFregion ionosphere frequently exhibits sporadic variability (e.g., Meek, 1949,https://doi.org/10.1029/JZ054i004p00339; Hill, 1963,https://doi.org/10.1175/1520‐0469(1963)020<0492:SEOLII>2.0.CO;2). Recent satellite data analysis (Noja et al., 2013,https://doi.org/10.1002/rds.20033; Chartier et al., 2018,https://doi.org/10.1002/2017JA024811) showed that the high‐latitudeFregion ionosphere exhibits sporadic enhancements more frequently in January than in July in both the northern and southern hemispheres. The same pattern has been seen in statistics of the degradation and total loss of GPS service onboard low‐Earth orbit satellites (Xiong et al. 2018,https://doi.org/10.5194/angeo‐36‐679‐2018). Here, we confirm the existence of this annual pattern using ground GPS‐based images of TEC from the MIDAS algorithm. Images covering January and July 2014 confirm that the high‐latitude (>70 MLAT)Fregion exhibits a substantially larger range of values in January than in July in both the northern and southern hemispheres. The range of TEC values observed in the polar caps is 38–57 TECU (north‐south) in January versus 25–37 TECU in July. First‐principle modeling using SAMI3 reproduces this pattern, and indicates that it is caused by an asymmetry in plasma levels (30% higher in January than in July across both polar caps), as well as 17% longer O+plasma lifetimes in northern hemisphere winter, compared to southern hemisphere winter.more » « less
-
Abstract Along‐shelf wind stresses drive substantial along‐coast variations in sea level that result in significant along‐coast pressure gradients in the Middle Atlantic Bight (MAB) at time scales from days to years. Forty years of sea‐level data and reanalysis wind stresses are examined to determine the characteristics and dynamics of pressure gradients along the New England and Central MAB coasts. Along‐coast dynamic sea level (pressure) gradients often exceed 5 cm/100 km at daily time scales, 2 cm/100 km at monthly time scales and 0.2 cm/100 km at yearly time scales. Along‐shelf wind stresses account for more than 50% of the along‐coast pressure gradient variance at daily and monthly time scales and more than 25% at yearly time scales. Pressure gradients along the New England coast are primarily driven by local wind stresses along the New England shelf, while pressure gradients along the Central MAB shelf are driven by both local wind stresses along the Central MAB shelf and remote wind stresses along the New England shelf. A steady depth‐average model (Csanady, 1978,https://doi.org/10.1175/1520‐0485(1978)008<0047:tatw>2.0.co;2) accurately reproduces the wind‐driven along‐coast pressure gradients in both regions. The along‐coast pressure gradients typically oppose the local wind stress and, in the along‐shelf momentum balance, are 50%–80% of the along‐shelf wind stress over the inner shelf (water depth 15 m).more » « less
-
Abstract Analysis of 40 years of tide gauge data and reanalysis wind stresses from the Middle Atlantic Bight (MAB) indicate that along‐shelf wind stresses are a dominant driver of coastal dynamic sea level (sea level plus atmospheric pressure) variability at daily to yearly time scales. The sea‐level response to along‐shelf wind stress varies substantially along the coast and is accurately reproduced by a steady, barotropic, depth‐averaged model (Csanady, 1978,https://doi.org/10.1175/1520‐0485(1978)008<0047:tatw>2.0.co;2, Arrested Topographic Wave). The model indicates that the sea‐level response in the MAB depends primarily on the along‐shelf distribution of the along‐shelf wind stress, the Coriolis frequency, the bottom drag coefficient, and the cross‐shelf bottom slope. The along‐shelf wind stress varies along the MAB shelf due primarily to changes in the shelf orientation. The sea‐level response depends on both the local and upstream (in the sense of Kelvin wave propagation) along‐shelf wind stresses. Consequently, sea‐level variability at daily, monthly and yearly time scales along much of the central MAB coast is more strongly driven by upstream winds along the southern New England shelf than by local winds along the central MAB shelf. The residual coastal sea‐level variability, after removing the wind‐driven response and the trend, is roughly uniform along the MAB coast. The along‐coast average of the residual sea level at monthly and yearly time scales is caused by variations in shelf water densities primarily associated with the large annual cycle in water temperature and interannual variations in salinity.more » « less
