Fusion of dual‐frequency SAR imagery of sea ice

Abstract

A new method is presented for the classification of sea ice using multi‐parametric Synthetic Aperture Radar (SAR) imagery. The local textural information, which is in essence a weighted gradient at a point, is computed in two SAR images of similar polarization but differing radar wavelength. The local information from the two images is combined at every pixel using a suggested rule for the addition of an entropy‐like measure. The resulting summation is shown to have the same negative exponential probability distribution found for the information from each separate image, confirming that the combined measure has the properties of information also. It is shown that the resulting joint information categories support a segmentation very similar to one based on consideration of the full complex scattering matrix for three wavelengths.     

Linking Arctic sea‐ice and atmospheric circulation anomalies on interannual and decadal timescales

Abstract

The relationship between Arctic sea‐ice concentration anomalies, particularly those associated with the “Great Salinity Anomaly” of 1968–1982, and atmospheric circulation anomalies north of 45°N is investigated. Empirical orthogonal function (EOF) analyses are performed on winter Arctic ice concentration from 1954 to 1990, sea level pressure and 500‐hPa heights from 1947 to 1994, and 850‐hPa temperatures from 1963 to 1994. Variability on both interannual and decadal timescales is apparent in the time series of the leading winter EOFs of all variables. The first EOF of winter sea‐ice concentration was found to characterize the patterns of ice variability associated with the Great Salinity Anomaly in the northern North Atlantic from 1968–82. Spatial maps of temporal correlation coefficients between the time series of the first EOF of winter sea‐ice concentration and the winter atmospheric anomaly fields are calculated at lags of 0 and ±7 year. Maximum correlations were found to exist when the time‐series of this ice EOF 1 leads the atmospheric anomaly fields by one year. A particularly interesting result is the connection between the presence of ice anomalies in the Greenland and Barents Seas and subsequent pressure anomalies of the same sign over the Irminger Basin and the Canadian Arctic. The main emphasis of the paper is to identify connections between Arctic sea‐ice and atmospheric circulation anomalies at interannual time‐scales.     

Role of sea surface temperature,Arctic sea ice and Siberian snow in forcing the atmospheric circulation in winter of 2012–2013

Climate Dynamics - During the 2012–2013 winter, the negative phase of the North Atlantic Oscillation (NAO) predominated, resulting in a cold winter over Europe and northern Asia punctuated by...     

Delving into the relationship between autumn Arctic sea ice and central–eastern Eurasian winter climate

北极海冰的快速减少是否已经显著地影响了最近中纬度大陆冬季极端天气气候事件引起了气候学家的广泛争论。问题的争论是来源于观测数据的年限很短以及中高纬度复杂的内部变率。在本研究中,采用气候突变检测的方法,我们将秋季海冰覆盖面积的变化分为三个阶段:1979–1986(高海冰阶段),1987–2006(海冰缓慢减少阶段)和2007–2014(海冰快速减少阶段)。然后,我们分析了与每一个阶段秋季海冰变化相联系的中-东欧亚地区冬季气候(尤其极端天气事件)是什么。结果表明北极海冰减少对西伯利亚西部和东亚极端天气事件影响的信号是稳健可测的。伴随着海冰的快速减少,高低空急流速度的减弱和急流位置的南移;波动振幅的加强、乌拉尔山阻塞频率的增多。这些导致了寒潮事件从亚洲中部到中国东北部地区显著增多。并且,与北极海冰的快速减少相关的环流异常与观测到的环流异常基本一致。相反地,在高海冰阶段,与海冰相关的环流异常和观测的异常并不一致。这个阶段的环流异常是与北极涛动处于持续的负位相有关的。     

Comparison and fusion of co‐occurrence,Gabor and MRF texture features for classification of SAR sea‐ice imagery

Abstract

Image texture interpretation is an important aspect of the computer‐assisted discrimination of Synthetic Aperture Radar (SAR) sea‐ice imagery. Co‐occurrence probabilities are the most common approach used to solve this problem. However, other texture feature extraction methods exist that have not been fully studied for their ability to interpret SAR sea‐ice imagery. Gabor filters and Markov random fields (MRF) are two such methods considered here. Classification and significance level testing shows that co‐occurrence probabilities classify the data with the highest accuracy, with Gaborfilters a close second. MRF results significantly lag Gabor and co‐occurrence results. However, the MRF features are uncorrelated with respect to co‐occurrence and Gabor features. The fused co‐occurrence/MRF feature set achieves higher performance. In addition, it is demonstrated that uniform quantization is a preferred quantization method compared to histogram equalization.     

An exceptional winter sea‐ice retreat/advance in the Bellingshausen sea,Antarctica

Abstract

The exceptional sea‐ice retreat and advance that occurred in the Bellingshausen Sea, Antarctica during August 1993 was the largest such winter event in this sector of the Antarctic during the satellite era. The reasons for this fluctuation of ice are investigated using passive microwave satellite imagery, ice motion vectors derived from the satellite data, in‐situ meteorological reports and near‐surface winds and temperatures from the European Centre for Medium‐range Weather Forecasts (ECMWF) numerical weather prediction model. The ice edge retreat of more than 400 km took place near 80°W from approximately 1–15 August, although the southward migration of the ice edge was not continuous and short periods of advance were also recorded. Between 16 August and 2 September there was almost continuous sea‐ice recovery. The rate of change of the ice edge location during both the retreat and advance phases significantly exceeded the southward and northward velocity components of ice within the pack, pointing to the importance of ice production and melting during this event. During the month, markedly different air masses affected the area, resulting in temperature changes from +2°C to ‐21°C at the nearby Rothera station. ‘Bulk’ movement of the pack, and compaction and divergence of the sea ice, made a secondary, but still significant, contribution to the observed advance and retreat. The ice extent fluctuations were so extreme because strong meridional atmospheric flow was experienced in a sector of the Southern Ocean where relatively low ice concentrations were occurring. The very rapid ice retreat/advance was associated with pronounced low‐high surface pressure anomaly couplets on either side of the Antarctic Peninsula.     

Decadal‐to‐interdecadal fluctuations of arctic sea‐ice cover and the atmospheric circulation during 1954–1994

Abstract

The relationship between the Arctic and subarctic sea‐ice concentration (SIC) anomalies, particularly those associated with the decadal‐scale Greenland and Labrador Seas “Ice and Salinity Anomalies (ISAs) “, and the overlying atmospheric circulation fluctuations is investigated using the singular value decomposition (SVD) and composite map analysis methods. The data analyzed are monthly SIC and sea level pressure (SLP) anomalies, which cover the northern hemisphere poleward of 45°N and extend over the 41‐year period 1954–1994.

The SVD₁ (first) mode of the coupled variability, which accounts for 57% of the square covariance, is for the most part an atmosphere‐to‐ice forcing mode characterized by the decadal timescale. The aforementioned ISA anomalies are clearly captured by this mode whose SIC anomalies are dominated by a strong dipole across Greenland. However, as part of the same mode, there is also a weaker SIC dipole in the northern North Pacific which has opposite‐signed anomalies in the Sea of Okhotsk and the Bering Sea. It is also shown that there exists a significant negative correlation between the decadal SIC variability in the Greenland‐Barents Seas region associated with this mode and the North Atlantic Oscillation, whose spectrum also exhibits a quasi‐decadal signal.

The SVD₂ mode accounts for 12% of the square covariance and shows no evidence of a dominant forcing field of either SIC or SLP. This SVD mode exhibits very low frequency (interdecadal) variability, and its co‐variability is mainly concentrated in the northern North Pacific. It appears to be a high‐latitude extension of the recently investigated interdecadal North Pacific Oscillation. The spatial structure of the second mode complements the case of the first SVD mode whose co‐variability mainly occurs in the northern North Atlantic.     

Cross‐shore variations of near‐surface wind velocity and atmospheric turbulence at the land‐sea boundary during CASP

Abstract

Airborne measurements of mean wind velocity and turbulence in the atmospheric boundary layer under wintertime conditions of cold offshore advection suggest that at a height of 50 m the mean wind speed increases with offshore distance by roughly 20% over a horizontal scale of order 10 km. Similarly, the vertical gust velocity and turbulent kinetic energy decay on scales of order 3.5 km by factors of 1.5 and 3.2, respectively. The scale of cross‐shore variations in the vertical fluxes of heat and downwind momentum is also 10 km, and the momentum flux is found to be roughly constant to 300 m, whereas the heat flux decreases with height. The stability parameter, z/L (where z = 50 m and L is the local Monin‐Obukhov length), is generally small over land but may reach order one over the warm ocean. The magnitude and horizontal length scales associated with the offshore variations in wind speed and turbulence are reasonably consistent with model results for a simple roughness change, but a more sophisticated model is required to interpret the combined effects of surface roughness and heat flux contrasts between land and sea.

Comparisons between aircraft and profile‐adjusted surface measurements of wind speed indicate that Doppler biases of 1–2 m s^?1 in the aircraft data caused by surface motions must be accounted for. In addition, the wind direction measurements of the Minimet anemometer buoy deployed in CASP are found to be in error by 25 ± 5°, possibly due to a misalignment of the anemometer vane. The vertical fluxes of heat and momentum show reasonably good agreement with surface estimates based on the Minimet data.     

Ice‐band characteristics of the Antarctic seasonal ice zone observed using MOS MESSR images

Abstract

Ice‐band characteristics for the region off East Queen Maud Land in Antarctica were examined and their relationship with the wind conditions was assessed using a large number of Marine Observation Satellite (MOS) Multispectral Electronic Self Scanning Radiometer (MESSR) images received at Syowa Station during the period 1989–93. Analyses from 43 examples of bands captured from August to December suggest that ice‐band formation and band scale are affected by both wind speed and direction over approximately the preceding four days (defined as the effective wind). Ice‐band width and spacing are significantly correlated with the effective wind speed and the maximum wind speed during that period. The long axis of ice bands tends to be oriented at 70°‐90° (mean of 75°) to the right of the effective wind direction. The band scales decrease from winter (August) to summer (December) with typical band spacing of 4–6 km in winter and 1–2 km in summer. This seems to be primarily due to a decrease in ice floe size and partly due to a decrease in the effective wind speed from winter to summer. Band scale decreases from the ice interior to the ice edge under conditions of off‐ice winds.     

Arctic sea ice and the Madden–Julian Oscillation (MJO)

Arctic sea ice responds to atmospheric forcing in primarily a top-down manner, whereby near-surface air circulation and temperature govern motion, formation, melting, and accretion. As a result, concentrations of sea ice vary with phases of many of the major modes of atmospheric variability, including the North Atlantic Oscillation, the Arctic Oscillation, and the El Niño-Southern Oscillation. However, until this present study, variability of sea ice by phase of the leading mode of atmospheric intraseasonal variability, the Madden–Julian Oscillation (MJO), which has been found to modify Arctic circulation and temperature, remained largely unstudied. Anomalies in daily change in sea ice concentration were isolated for all phases of the real-time multivariate MJO index during both summer (May–July) and winter (November–January) months. The three principal findings of the current study were as follows. (1) The MJO projects onto the Arctic atmosphere, as evidenced by statistically significant wavy patterns and consistent anomaly sign changes in composites of surface and mid-tropospheric atmospheric fields. (2) The MJO modulates Arctic sea ice in both summer and winter seasons, with the region of greatest variability shifting with the migration of the ice margin poleward (equatorward) during the summer (winter) period. Active regions of coherent ice concentration variability were identified in the Atlantic sector on days when the MJO was in phases 4 and 7 and the Pacific sector on days when the MJO was in phases 2 and 6, all supported by corresponding anomalies in surface wind and temperature. During July, similar variability in sea ice concentration was found in the North Atlantic sector during MJO phases 2 and 6 and Siberian sector during MJO phases 1 and 5, also supported by corresponding anomalies in surface wind. (3) The MJO modulates Arctic sea ice regionally, often resulting in dipole-shaped patterns of variability between anomaly centers. These results provide an important first look at intraseasonal variability of sea ice in the Arctic.     

Interannual variability of sea‐ice cover in Hudson bay,Baffin bay and the Labrador sea

Abstract

The spatial and temporal relationships between subarctic Canadian sea‐ice cover and atmospheric forcing are investigated by analysing sea‐ice concentration, sea‐level pressure and surface air temperature data from 1953 to 1988. The sea‐ice anomalies in Hudson Bay, Baffin Bay and the Labrador Sea are found to be related to the North Atlantic Oscillation (NAO) and the Southern Oscillation (SO). Through a spatial Student's i‐test and a Monte Carlo simulation, it is found that sea‐ice cover in both Hudson Bay and the Baffin Bay‐Labrador Sea region responds to a Low/Wet episode of the SO (defined as the period when the SO index becomes negative) mainly in summer. In this case, the sea‐ice cover has a large positive anomaly that starts in summer and continues through to autumn. The ice anomaly is attributed to the negative anomalies in the regional surface air temperature record during the summer and autumn when the Low/Wet episode is developing. During strong winter westerly wind events of the NAO, the Baffin Bay‐Labrador Sea ice cover in winter and spring has a positive anomaly due to the associated negative anomaly in surface air temperature. During the years in which strong westerly NAO and Low/Wet SO events occur simultaneously (as in 1972/73 and 1982/83), the sea ice is found to have large positive anomalies in the study region; in particular, such anomalies occurred for a major portion of one of the two years. A spectral analysis shows that sea‐ice fluctuations in the Baffin Bay‐Labrador Sea region respond to the SO and surface air temperature at about 1.7‐, 5‐ and 10‐year periods. In addition, a noticeable sea‐ice change was found (i.e. more polynyas occurred) around the time of the so‐called “climate jump” during the early 1960s. Data on ice thickness and on ice‐melt dates from Hudson Bay are also used to verify some of the above findings.     

A note on brine layer spacing of first‐year sea ice

Abstract

Brine layer spacing has been measured in a core sample taken 19 January 1978 from Eclipse Sound, Baffin Island, Canada. Observations on snow and ice conditions and a record of air temperatures for the entire growth season allowed correlation of the brine layer spacing with the growth rate of the sea ice. Growth rate is related to climatology, and the vertical brine layer spacing profile in the ice provides a record of previous weather conditions. It is suggested that the spacing is inversely proportional to the growth rate, and could also be dependent on crystallographic orientation. The spacing decreased rapidly with depth near the bottom of the core sample, and this is not compatible with a general relation between spacing and growth rate. Before a definitive statement can be made, cores from a variety of locations, grown in a range of meteorological conditions, will have to be studied.     

Impact of the winter North Atlantic Oscillation (NAO) on the Western Pacific (WP) pattern in the following winter through Arctic sea ice and ENSO: part I—observational evidence

Climate Dynamics - On the basis of a 51-year statistical analysis of reanalysis data, we propose for the first time that the positive phase of the Western Pacific (WP) pattern in the winter is...     

Simulation of sea ice and ocean variability in the arctic during 1955–2002 with an intermediate complexity model

Abstract

A number of recent sea‐ice and ocean changes in the Arctic and subarctic regions are simulated using the global University of Victoria (UVic) Earth System Climate Model version 2.6. This is an intermediate complexity model which includes a three‐dimensional ocean model (MOM 2.2), an energy‐moisture balance model for the atmosphere with heat and moisture transport, and a dynamic‐thermodynamic sea‐ice model with elastic‐viscous‐plastic rheology. The model is first spun up for 1800 years with monthly wind stress forcing derived from the National Centers for Environmental Prediction (NCEP) climatology winds and a pre‐industrial atmospheric CO₂ concentration of 280 ppm. After a second spin‐up for the period 1800–1947 with daily climatology winds‐tress forcing, and a linearly increasing atmospheric CO₂ concentration, the model is run with interannually varying wind stresses for the period 1948–2002 with an average forcing interval of 2.5 days and an exponentially increasing atmospheric CO₂ concentration varying from 315 to 365 ppm. However, the analysis of the model output is only carried out for the years 1955–2002.

The simulated maximum and minimum sea‐ice areas for the Arctic are within 6% of the observed climatologies for the years 1978–2001. The model output also shows a small downward trend in sea‐ice extent, which, however, is smaller than has been observed during the past few decades. In addition, the model simulates a decrease in sea‐ice thickness in the SCICEX (SCientific ICe EXpeditions) measurement area in the central Arctic that is consistent with, but smaller than, that observed from submarine sonar profiling data.

The observed variability and magnitude of the export of sea ice through Fram Strait is quite well captured in the simulation. The change in correlation between the North Atlantic Oscillation (NAO) index and the sea‐ice export around 1977 as found in a data study by Hilmer and Jung (2000) is also reproduced. Within the Arctic basin the model simulates well the patterns and the timing of the two major regimes of wind‐forced sea‐ice drift circulation (cyclonic and anticyclonic) as found earlier by Proshutinsky and Johnson (1997). The influence of variations in the Fram Strait ice export on the strength of the North Atlantic thermohaline circulation and surface air temperature are also determined. In particular, it is shown that 3–4 years after a large ice export, the maximum meridional overturning streamfunction decreases by more than 10%.

The temperature and salinity increase at depths of 200–300 m, as observed in the eastern Arctic by Morison et al. (1998), between the USS Pargo cruise in 1993 and the Environmental Working Group (EWG) Joint USRussian Arctic Atlas climatology for the years 1948–87, are just visible in the model simulation. The increases are more noticeable, however, when the ocean model data are averaged over the pentade 1995–2000 and compared with model data averaged over the pentade 1955–60. The fact that these, and some of the other modelled changes, are smaller than the observed changes can likely be attributed to the relatively coarse resolution of the UVic Earth System Climate Model (3.6°E‐W and 1.8°N‐S). Nevertheless, the fact that the model captures qualitatively many of the recent sea‐ice and ocean changes in the Arctic suggests that it can be successfully used to investigate other Arctic‐North Atlantic Ocean climate interactions during past and future eras.     

Climate tendencies of changes in multiyear sea ice thickness in the Arctic Basin (1970–2005)

The area integral of the sea ice thickness in the Arctic Basin is estimated from the measurements of sea ice surface fluctuations at drift-ice stations. The 1970–1990 linear trend is indicative of an approximately 10-cm reduction in the average sea ice thickness over the entire Arctic Basin, which makes 3% of the average ice thickness (about 3 m). Seasonal changes made 40 cm. The amplitude of variations of the average ice thickness in that period is 20 cm with a period of changes of approximately 6–8 years. The observations were interrupted during 1991–2003 and then resumed in 2004. During 1990–2005, the old ice thickness over the entire Arctic Basin decreased, on average, by 110 cm.     

Comparison of airborne electromagnetic ice thickness data with NOAA/AVHRR and ERS‐1/SAR images

Abstract

Snow‐plus‐ice thickness and surface‐ice roughness data collected by a helicopter‐towed sensor package was used to identify surface‐ice properties in March 1992 AVHRR and SAR images for the land‐fast and mobile pack ice off the northern coast of Newfoundland. The sensor package consisted of an electromagnetic induction sensor and laser profilometer. Observed snow depths and ice thicknesses verified that snow‐plus‐ice thickness over undeformed ice can be obtained to an accuracy of ±10 cm. Snow‐plus‐ice thickness and surface roughness data for flight sections covering several hundred kilometres indicated the change in pack ice properties seen in images from thin, smooth coastal ice and open water conditions to thick, rough consolidated offshore pack ice. Ice charts covering the same area showed similar variations in ice conditions based on AVHRR and fixed‐wing reconnaissance data. In the ERS‐1 SAR image, low backscattering coefficients were associated with large, smooth coastal floes interspersed with areas of high backscatter indicating the presence of waves in open water areas. Backscattering coefficients were higher in the rubble areas near the inshore edge of the pack ice than in the interior of the pack ice itself. Distinguishing ice types on the basis of tone alone in SAR imagery was found to be problematic; however in combination with other remotely sensed data such as AVHRR data, SAR data will become more useful in distinguishing ice types.     

Change and variability in sea ice during the 2007–2008 Canadian International Polar Year program

In this paper we describe sea ice change and variability during the Canadian International Polar Year (IPY) program and examine several regional and hemispheric causes of this change. In a companion paper (Barber et al., Climate Change 2012) we present an overview of the consequences of this observed change and variability on ecosystem function, climatically relevant gas exchange, habitats of primary and apex predators, and impacts on northern peoples. Sea ice-themed research projects within the fourth IPY were designed to be among the most diverse international science programs. They greatly enhanced the exchange of Inuit knowledge and scientific ideas across nations and disciplines. This interdisciplinary and cultural exchange helped to explain and communicate the impacts of a transition of the Arctic Ocean and ecosystem to a seasonally ice-free state, the commensurate replacement of perennial with annual sea ice types and the causes and consequences of this globally significant metamorphosis. This paper presents a synthesis of scientific sea ice research and traditional knowledge results from Canadian-led IPY projects between 2007 and 2009. In particular, a summary of sea ice trends, basin-wide and regional, is presented in conjunction with Inuit knowledge of sea ice, gathered from communities in northern Canada. We focus on the recent observed changes in sea ice and discuss some of the causes of this change including atmospheric and oceanic forcing of both dynamic and thermodynamic forcing on the ice. Pertinent results include: 1) In the Amundsen Gulf, at the western end of the Northwest Passage, open water persists longer than normal and winter sea ice is thinner and more mobile. 2) Large areas of summer sea ice are becoming heavily decayed during summer and can be broken up by long-period waves being generated in the now extensive open water areas of the Chukchi Sea. 3) Cyclones play an important role in flaw leads??regions of open water between pack ice and land-fast ice. They delay the formation of new ice and the growth of multi-year ice. 4) Feedbacks involving the increased period of open water, long-period wave generation, increased open-ocean roughness, and the precipitation of autumn snow are all partially responsible for the observed reduction in multiyear sea ice. 5) The atmosphere is observed as remaining generally stable throughout the winter, preventing vertical entrainment of moisture above the surface.     

Sea‐ice interaction and the stability of the thermohaline circulation

Abstract

The role of sea‐ice in affecting the stability and long‐term variability of the oceanic thermohaline circulation (THC) is studied in this paper. The emphasis is placed on studying how sea‐ice might affect the stability and the long‐term variability of the THC through modulations of the surface heat and freshwater fluxes. A simple box model is analyzed to elucidate qualitatively the distinct physical meanings of these two processes. The analytical solution of this simple model indicates that, for the long timescales considered here, the thermal insulation stabilizes the THC while the freshwater feedback increases the effective inertia of the coupled ice‐ocean system. Sea‐ice insulation lessens the negative feedback between heat flux and the SST, and therefore, allows the SST to play a greater role in counteracting changes of the THC and high latitude salinity field. The freshwater feedback effectively links the surface heat flux to a freshwater reservoir, and thus, increases the effective inertia of the coupled ocean‐ice system. A two‐dimensional ocean model coupled with a thermodynamic sea‐ice model is used to estimate quantitatively the magnitudes of these two feedbacks. The numerical experiments involve the model's responses both to initial anomalies and to changes of forcing fields. For the free response cases (model responses to initial anomalies without changing the forcing fields), the model shows that the decay rate of an initial anomaly is greater when sea‐ice is included. For small perturbations the thermal insulation effect dominates over the freshwater feedback. The latter becomes increasingly more important for larger perturbations. In response to a change of external forcing, the presence of sea‐ice reduces the magnitude and the pace of the model's response. The numerical results are qualitatively consistent with the analytical solution of the box model.     

On the relationship between arctic sea‐ice anomalies and fluctuations in Northern Canadian air temperature and river discharge

Abstract

A lagged cross‐correlation analysis of climatic data from the period 1953–1984 was carried out for three regions of Northern Canada (Beaufort Sea, Hudson Bay, Baffin Bay/Labrador Sea) to determine the relationships between sea‐ice anomalies and surface air temperature and river discharge anomalies. Significant negative correlations at the 95% level were found between sea‐ice and temperature anomalies. A significant correlation at the 95% level was found between sea‐ice and river discharge anomalies in only one of two subregions studied.     

Spatial and temporal characteristics of minimum temperature in winter in China during 1961–2010 from NCEP/NCAR reanalysis

Based on the surface 2?m monthly minimum temperature from the National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis dataset, the spatial and temporal characteristics of winter minimum temperature during 1961–2010 have been analyzed in China. Results showed that the minimum temperature in China has a significant increasing rate of 0.25° per decade calculated by the Mann–Kendall statistical test, which is consistent with the global warming trend. Empirical orthogonal function (EOF) analysis reveals that there are three main patterns that can explain more than 57.6% of the total variance of the winter minimum temperature. The EOF1, EOF2, and EOF3 account for 34.8%, 13.5%, and 13.5% of the total inter-annual variance, respectively. The EOF1, EOF2, and EOF3 patterns are synchronous in northern China, central China, and on the Tibetan Plateau. There exist a decrease trend in the corresponding time coefficients of EOF1 and EOF2 and an increase trend in that of EOF3 since the 1960s. Both the corresponding time coefficients of EOF1 and EOF2 have significant positive correlations with the 500?hPa geopotential heights of the Arctic region and negative correlations in the regions lower than 40°N, while a significant positive correlation is found between the corresponding time coefficients of EOF3 and 500?hPa geopotential heights in the low latitudes. This suggests that rapid warming occurs in northern China and on the Tibetan Plateau, while the weakest trend locates in southeast China. Thus, warming in winter is more pronounced at higher altitudes and latitudes. These patterns are tightly connected with the atmospheric circulation.