SEARCH

SEARCH BY CITATION

Keywords:

  • sea ice;
  • interannual variability;
  • climatology

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

[1] The Canadian Ice Service Digital Archive (CISDA) is a compilation of weekly ice charts covering Canadian waters from the early 1960s to present. The main sources of uncertainty in the database are reviewed and the data are validated for use in climate studies before trends and variability in summer averaged sea ice cover are investigated. These data revealed that between 1968 and 2008, summer sea ice cover has decreased by 11.3% ± 2.6% decade−1 in Hudson Bay, 2.9% ± 1.2% decade−1 in the Canadian Arctic Archipelago (CAA), 8.9% ± 3.1% decade−1 in Baffin Bay, and 5.2% ± 2.4% decade−1 in the Beaufort Sea with no significant reductions in multiyear ice. Reductions in sea ice cover are linked to increases in early summer surface air temperature (SAT); significant increases in SAT were observed in every season and they are consistently greater than the pan-Arctic change by up to ∼0.2°C decade−1. Within the CAA and Baffin Bay, the El Niño-Southern Oscillation index correlates well with multiyear ice coverage (positive) and first-year ice coverage (negative) suggesting that El Niño episodes precede summers with more multiyear ice and less first-year ice. Extending the trend calculations back to 1960 along the major shipping routes revealed significant decreases in summer sea ice coverage ranging between 11% and 15% decade−1 along the route through Hudson Bay and 6% and 10% decade−1 along the southern route of the Northwest Passage, the latter is linked to increases in SAT. Between 1960 and 2008, no significant trends were found along the northern western Parry Channel route of the Northwest Passage.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

[2] The Arctic's floating sea ice cover is a key component of the global climate system. Not only does it regulate the transfer of moisture, heat and momentum between the atmosphere and ocean but the positive sea ice-albedo feedback plays a lead role in the polar amplification of climate change [Intergovernmental Panel on Climate Change, 2007]. Between 1979 and 2006, pan-Arctic reductions in the annual average Arctic sea ice extent estimated from the passive microwave satellite record were on the order of 3% to 4% decade−1, and seasonally, the largest reductions have occurred in summer at a rate of 6.2% decade−1 [Parkinson and Cavalieri, 2008]. These hemispheric wide reductions in Arctic sea ice have been linked to changes in the atmosphere [e.g., Serreze et al., 2007], ocean [e.g., Maslowski et al., 2001] as well the phase and strength of large-scale climate oscillations namely the Arctic Oscillation (AO) [e.g., Rigor and Wallace, 2004] and the El Niño-Southern Oscillation (ENSO) [e.g., Liu et al., 2004]. Regionally, approximately 15% of the total summer Arctic sea ice cover is in Canadian waters and within these waters; Parkinson and Cavalieri [2008] report decrease in sea ice extent by 19.5% decade−1 in Hudson Bay, 1.2% decade−1 in Canadian Arctic Archipelago (CAA), and 16.0% decade−1 in Baffin Bay/Labrador Sea.

[3] As the sea ice cover recedes, fundamental changes in the Arctic ecosystem are expected and it is also anticipated that commercial use of Arctic waters will increase dramatically, particularly in summer [Hassol, 2004]. Specific to Canada, speculation about a navigable Northwest Passage (NWP) through the CAA and the potential security, economic and environmental implications is a reoccurring topic in the global media. Although a temporary sea ice-free route through the western Parry Channel region of the NWP opened in late August 2007, the consensus is that sea ice conditions in this region will remain hazardous during a transition to an ice-free Arctic Ocean [Melling, 2002; Howell et al., 2008, 2009]. Reductions in sea ice cover in the Hudson Bay region, however, are among the greatest in the circumpolar Arctic and the region has only traces of ship hazardous multiyear ice (MYI). As a result, the most likely scenario to become reality in the near future is the proposed Arctic Bridge, a Russia-to-North America shipping route between Murmansk, Russia and Churchill, Manitoba.

[4] General Circulation Models (GCM) are used to predict future sea ice conditions under different greenhouse gas scenarios. The latest estimates from the IPCC AR4 models predict an ice-free Arctic as early as 2050 under the business as usual scenario and Stroeve et al. [2007] caution this might be a conservative estimate. While GCMs can reproduce sea ice dynamic and thermodynamic processes in the Arctic Ocean and peripheral seas, they do not resolve many regions in the Canadian Arctic particularly the narrow waterways through the CAA. Sou and Flato [2009] developed a regional CAA model and found that a completely ice-free CAA is unlikely by 2050. To reconcile differences between model predictions and to assist in future model development, the main causes of interannual variability in summer sea ice in the Canadian Arctic needs to be better understood.

[5] A consistent passive microwave satellite record of Arctic sea ice cover begins in 1979 and this data has been used to document both annual and seasonal changes in Arctic sea ice. Unfortunately, this data set is known to have problems resolving sea ice conditions within the majority of the Canadian Arctic [Agnew and Howell, 2003] and trend analysis and variability studies at this time are limited to 30 years. The Canadian Ice Service Digital Archive (CISDA), a record of weekly ice charts, provides more detailed sea ice information and extends Canadian sea ice conditions as far back as 1960. In this analysis we make use of the CISDA to (1) report on trends in summer sea ice cover in the Canadian Arctic; (2) explore the potential influence of surface air temperature (SAT) anomalies, the ENSO, and the AO on sea ice trends and interannual variability; and (3) discuss long-term trends and variability specifically along the major shipping routes. Linkages to the AO and ENSO are explored since there is evidence in the literature suggesting these large-scale climate oscillations may influence interannual sea ice variability in Canada. There is strong evidence in the literature of both ENSO and AO signatures in the SAT field over Canada [e.g., Shabbar, 2006; Liu et al., 2004]. With respect to sea ice, coincident strong El Niño and positive NAO (AO) events have been linked to heavy ice conditions in Hudson Bay and the Labrador Sea [Mysak et al., 1996] and Kinnard et al. [2006] relate the first principal component in monthly ice concentration in Canadian waters with the ENSO, where ENSO leads by 3 months.

2. Canadian Ice Service Digital Archive

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

2.1. Overview

[6] Sea ice forms every winter in northern Canadian waters. By mid-January waters of the western Arctic, eastern Arctic, Hudson Bay and Labrador Sea (Figure 1) are completely ice covered and the only interannual variability in winter sea ice coverage is the southern extent of the ice edge in the Labrador Sea. During the summer months sea ice does not completely melt in most regions and old ice, ice that has survived at least one summer season, can be present as far south as the coast of Labrador. Shipping in northern waters is only possible from mid-June to late October, the period between spring breakup and fall freezeup, during which ice charts are produced and released to the public by the Canadian Ice Service of Environment Canada.

image

Figure 1. CIS regional ice chart boundaries.

Download figure to PowerPoint

[7] The Canadian Ice Service Digital Archive (CISDA) is a compilation of all the weekly ice charts produced by the CIS. The majority of the archive is composed of the regional ice charts that are produced weekly for four separate regions in Canada (Figure 1): the western Arctic (1968 to present), the eastern Arctic (1968 to present), Hudson Bay (1971 to present), and the east coast (1968 to present). In general, the regional ice charts are only produced during the shipping seasons, however, production of monthly winter charts in the eastern and western Arctic began in 1980. Prior to 1998 ice charts were produced on paper and these have been digitized and combined with the post-1998 charts that are in digital format to create the CISDA-regional charts data set (CISDA-R).

[8] Between 1958 and 1974 weekly historical charts were produced by CIS at the end of the operational season at set 7 day (or occasionally 8 day) intervals specifically for climatology. This allowed for the use of additional data, particularly remotely sensed data and aerial and shipping reconnaissance data that were not available in near real time. The northern historical charts, which cover the area of the eastern and western Arctic (Figure 1), were produced for the summer months only. The southern historical charts, which include Hudson Bay (Figure 1), were produced year-round. These charts have recently been digitized creating the CISDA-historical charts data set (CISDA-H).

[9] A third set of Arctic ice charts for the summers of 1961–1979 are being prepared for digitization (CISDA-P). These charts were compiled for the Polar Continental Shelf Project (PCSP) and are based on ice reconnaissance flights designed to gather climatological rather than operational ice information [Lindsay, 1971, 1976, 1981]. The PCSP charts were produced biweekly at the end of the season and cover the eastern and western Arctic (Figure 1). In addition to aerial surveys, the historical and regional ice charts along with all remotely sensed and ground observation data were used in their production. It is anticipated that including the digitized PCSP charts (CISDA-P) in CISDA will improve the reliability of sea ice information, particularly, in the remote areas of the western high Arctic and along the northwest coast of the Canadian Arctic Archipelago.

[10] The regional, historical, and PCSP ice charts represent an integration of remotely sensed information, surface observations, airborne and ship reports, operational model results, along with the expertise of experienced ice forecasters. To complete chart areas where no up-to-date information existed, the ice forecaster consulted past ice and meteorological conditions and drew upon their expertise and experience to predict the present ice position and characteristics. This process is called nowcasting. In many sea ice regions in Canada, such as the western Queen Elizabeth Islands or Hudson Bay, where ice conditions are often stable, persistence was frequently the basis of nowcasting. In very dynamic regions such as the southern Beaufort Sea, where information collected many hours or days in the past cannot be expected to represent the current conditions, nowcasting is difficult, and requires the knowledge of an experienced ice forecaster.

[11] Sea ice information in CISDA is represented using the World Meteorological Organization (WMO) “egg code.” Each individual egg attribute (total concentration, partial concentration, stage of development and form of ice (floe size)) is stored in a separate column (field) in the database. The codes used on the charts have changed over the years. During digitization, the original codes were captured and converted to the egg code if necessary. The regional charts data set (CISDA-R) is updated in near real time and is available to the public as polygonal data in Arc/Info GIS E00 format. A gridded data set is also available on request as dbf files (for the individual chart regions) or as a merged grid point file on a 0.25° grid in the Network Common Data Format (NetCDF). In addition, various climate products based on CISDA-R such as climatic ice atlases [Canadian Ice Service (CIS), 2002] and departure from normal products are accessible online (www.ice-glaces.ec.gc.ca). The historical and PCSP chart data sets (CISDA-H/CISDA-R) will be available to the public in the near future.

2.2. Climatology

[12] A climatic atlas [CIS, 2002] for sea ice in northern Canadian waters was published by CIS based on the regional ice charts (CISDA-R) for 1971–2000. Products from this atlas are used to provide a general overview of the summer ice regime in Canada. Reference can be made to Figure 4 for the geographic location of specific sea ice regions in Canada. The median ice concentration for selected weeks throughout the summer season is shown in Figure 2. The first regions to clear in spring are the southeastern Beaufort Sea and northern Baffin Bay (Figure 2a). In the southern Beaufort Sea flaw leads develop year-round [e.g., Barber and Hanesiak, 2004; Carmack et al., 2004] and in northern Baffin Bay the North Water Polynya forms every year [Dunbar, 1969]. Increased absorption of solar radiation in these open water areas in spring accelerates melt. By mid-September when ice concentrations are at the annual minimum, Hudson Bay, Hudson Strait, the north Labrador Sea and the central region of the western Arctic waterway are completely ice free (Figure 2d). Ice concentrations remain high throughout the summer season in Nares Strait, the western high Arctic, Kane Basin, McClintock channel and along the north coast of the CAA. New ice starts to form in October and by mid-November the Arctic is completely ice covered. By January, Hudson Bay and Hudson Strait are completely ice covered and the ice edge in the Labrador Sea reaches its most southern limit in March.

image

Figure 2. Median ice concentration for all ice, 1971–2000 [CIS, 2002].

Download figure to PowerPoint

[13] Figure 3 shows the median ice concentration and frequency of presence of old ice for selected weeks throughout the summer season. When ice concentrations are at their annual minimum in mid-September, sea ice in the Beaufort Sea and throughout the western CAA is predominantly MYI. Sea ice that forms during the winter (first-year ice) and survives the summer season is reclassified as second-year ice on 1 October. In this paper, any ice that is not first-year ice is classified as old ice or multiyear ice, the two terms are used interchangeably. The western and eastern high Arctic regions are both a source of MYI and sink for MYI from the Arctic Ocean. As a sink, MYI from the Arctic Ocean is continuously forced up again the islands and MYI is advected by winds and current into the CAA [Melling, 2002] where over the course of a few summers it melts. As a source, cooler summer temperatures and high ice concentrations throughout the summer facilitate the in situ formation of MYI. The southern regions of the CAA are predominantly a sink for MYI although there is some in situ formation in the western Parry Channel and M'Clintock Channel [Howell et al., 2009]. Notable increases in median MYI concentrations between mid-September and mid-October in, for example, western Parry Channel, M'Clintock Channel and the western high Arctic and Victoria Strait (Figures 3c and 3e), are due to the promotion of FYI to MYI on 1 October. Foxe Basin and the Baffin Inlets are atypical southern regions, MYI in these regions is primarily from in situ growth and there is little ice flux in and out of these regions. In Hudson Bay and Baffin Bay there is no in situ growth of MYI since the sea ice clears every summer. The source for MYI in northern Hudson Bay is Foxe Basin and MYI enters Baffin Bay from eastern Parry Channel and Kane Basin [Tang et al., 2004].

image

Figure 3. (a, c, and e) Median ice concentration for old ice, 1971–2000, and (b, d, and f) frequency of presence of old ice, 1971–2000 [CIS, 2002].

Download figure to PowerPoint

[14] The CIS has recently established sea ice regime regions (CISIRR) to be used with the CISDA. The regions (Figure 4) reflect unique ice regimes defined by predominant ice conditions and regional forcings such ocean currents, synoptic climatology and bathymetry. In defining CISIRR, careful attention was given to previous schemes developed by ice and ocean research scientists [Crocker and Carrieres, 2000a; Melling, 2002], operational ice forecasters [Jeffers et al., 2001] and operational users of ice information (B. Goreman, personal communication, 2006). The smallest CISIRR regions are in the CAA where the narrow passages between islands forces very different ice regimes. However, these regions are not too small for the resolution of the CISDA, the smallest CISIRR is 2407 km2 (Figure 4, region 18 in the western high Arctic) which is much larger than the coarsest resolution source data (Table A4). For a detailed explanation of the rationale behind the boundaries see CIS [2007b].

image

Figure 4. Regions and subregions of the Canadian Arctic (CISIRR). Subregion numbers correspond with Table 1.

Download figure to PowerPoint

2.3. Statistical Analysis in the Context of Known and Suspected Errors

[15] The type and detail of information in CISDA, particularly floe size and stage of development, far exceeds what is attainable from a single satellite source. However, because the ice charts are an integration of ice information from a variety of sources, the greatest shortcoming of the data set with respect to statistical analysis is that errors in estimates are nonuniform in both space and time. In Appendix A of this paper the early government reports [CIS, 2007a, 2007b; Crocker and Carrieres, 2000a, 2000b] that describe the database are synthesized; the main sources of uncertainty in the database are reviewed and discussed along with estimates of their contribution to the overall error in ice position, ice concentration and stage of development. In this section, the rationale behind the choice of sea ice parameter and time period used for the statistical analysis in this study is first presented. Next, the time series are checked for evidence of systematic biases that could be introduced from known and suspected biases in CISDA (see Appendix A).

2.3.1. Sea Ice Parameter

[16] The choice of an ice parameter for statistical analysis is not trivial. The lack of continuity in the data set has the potential to degrade data quality. Temporal changes in any specific ice parameter could be due to changes in the quality of the input data or interpretative procedures and not to real changes in the state of the ice environment. The parameters used in this study are the average sea ice coverage for all ice types combined (all ice coverage (AIC)), multiyear ice (multiyear ice coverage (MYIC)), and first-year ice (first-year ice coverage (FYIC)) in each CISSIR region calculated using

  • equation image

Awater is the total sea surface area and Aice is the total sea ice area. Aice can be subdivided into FYI or MYI and N is any number of weeks that define a season of interest. The summer season is of interest and for the eastern and western Arctic (Figure 1) it is chosen as the 17 week period from 25 June to 15 October. For Hudson Bay (Figure 1) it is chosen as the 23 week period from 25 June to 19 November. These season definitions have been adopted by CIS because they maximize the use of the regional charts which are only created weekly in the Arctic during the summer shipping season [CIS, 2007a]. For any given region there is a single value for ice coverage that defines the sea ice severity during the summer shipping season.

[17] A seasonal parameter for the CISIRR overcomes random errors in the data set (see Appendix A) and minimizes the influence of gross errors in a single ice chart. However, the impact of spatial and temporal discontinuity in the data on statistical analysis, particularly trends, needs to be addressed. This is handled in two steps. First, a subjective assessment of data quality in each CISIRR region is used to determine the most appropriate data and time period for statistical analysis. Second, the potential impact of known and suspected systematic bias in the data set on statistical analysis is assessed.

2.3.2. Addressing Nonhomogeneity in Data Quality

[18] Quality Index (QI) scores (see Appendix A) were used to categorize and assess the data quality of the AIC and MYIC time series to determine an appropriate start year for statistical analysis. Data over a time period classified as “not considered for trends” or “not considered” are unreliable for trend analysis because observations were too scarce at the beginning of the record. A computed trend in this circumstance could be caused by a change in data quality over time, not real environmental change. QI scores are based on a detailed analysis of the AIC and MYIC time series in each region with particular attention to homogeneity in the time series. Regions in the fair, good and excellent categories are considered of high enough quality for any statistical analysis.

[19] The data quality in AIC/MYIC for each region derived from the CISDA regional ice charts over the 1968–2007 period is shown in Figure 5. In general, data quality is higher in the operational regions. The AIC/MYIC time series in the northern regions of the CAA and along the northern coast of the CAA are not reliable for trend analysis. If the regional charts are replaced by the historical charts for the overlapping period of 1968–1974, the data quality increases for these remote regions allowing for trend analysis (Figure 5). Figure 5 (bottom) shows the data quality in AIC only for the time series from the blended data sets taken back to 1960. It is apparent that trend analysis beginning in 1960 for AIC is only reliable along the main shipping routes. In summary, we have used a combination of the regional ice charts and the historical ice charts (1968–1974) in this study in order to have the best possible estimate of ice conditions and for the 1960–2008 period we have only calculated trends along the main shipping routes.

image

Figure 5. (top) Qualitative quality index scores in each subregion for AIC/MYIC for the regional ice charts, 1968–2008; (middle) AIC/MYIC for the blended historical and regional ice charts, 1968–2008; and (bottom) AIC for the blended historical and regional ice charts, 1960–2008.

Download figure to PowerPoint

2.3.3. Addressing Systematic Bias

[20] There are two known systematic bias in the CISDA (see Appendix A): (1) underestimation of low concentrations of ice occurring with all satellite imaging sensors and (2) overestimation of total and multiyear ice concentration in remote areas. The influence of systematic underestimation of low ice concentrations on AIC is explored by comparing AIC with AIC calculated by excluding regions with sea ice concentrations less than 2/10 (AIC > 2/10). Correlations between the two AIC parameters are all greater than 0.99 and trend analysis repeated using AIC > 2/10 revealed no detectable difference in the results. It is therefore assumed that the bias in estimates of low ice concentrations within the CISDA has no significant impact on interannual variability or trends in AIC. The bias in old ice associated with nowcasting in the early years in remote areas of the high Arctic is not formally addressed but it is noted and taken into consideration in the interpretation of the results.

[21] There are two likely sources of systematic bias in the CISDA (see Appendix A): (1) the switch to the egg code in 1983 and (2) the operational introduction of RADARSAT-1 in 1996. To test for potential shifts in CISIRR AIC/MYIC that may have been introduced by changing chart techniques (i.e., switch to egg code) and source data over time (i.e., RADARSAT), a comparison is made between CISDA and three other sea ice data sets: sea ice concentration estimates from Scanning Multichannel Microwave Radiometer (SMMR)-Special Sensor Microwave/Imager (SSM/I) passive microwave sensors using the NASA Team algorithm [Cavalieri et al., 2008], MYI concentration estimates from a neural network analysis of SSMR and SSMI/I data [Belchansky et al., 2004], and sea ice concentration from the Hadley Centre data set (HadISST2.1) [Walsh, 1978; Knight, 1984; Rayner et al., 2003]. The sea ice algorithm used to estimate ice concentration from the passive microwave data is sensor specific in order to homogenize data between sensors [Cavalieri et al., 2008]. The data is available as monthly averages on a 25 km degree grid from 1979 to present from the National Snow and Ice Data Center. Estimates of MYI concentration are also on a 25 km degree grid and are available as monthly averages for January, February, and March from the International Arctic Research Centre. The Hadley data for Arctic sea ice concentration is a compilation of the Walsh [1978] data set, the National Ice Center (NIC) ice charts [Knight, 1984] and the passive microwave record [Cavalieri et al., 2008]; it is available as monthly averages on a 1° grid at the British Atmospheric Data Center. Although the Hadley record of sea ice concentration is a blended data set, considerable effort has been made to homogenize the data which includes a correction for surface melt effects on retrievals from satellite microwave-based estimates.

[22] For the comparison to be valid not only must the baseline data sets have minimal time-varying bias, but they must be independent of CISDA. The passive microwave data was rarely used in chart preparation due to the high error in ice concentration estimates (4/10; see Table A3); when it was used it was only used to delineate the ice edge in the absence of any other satellite or ground observations. Post-1979, the main data source in the Hadley data set is the passive microwave record [Rayner et al., 2003] and pre-1979 the main data sources are the NIC ice charts and the Walsh data set. Although the CIS ice charts are not a data source in either the NIC ice charts or the Walsh data set, they likely share common source data prior to 1979. For this reason the Hadley data set is only used to check for bias in the early 1980s when the passive microwave record is too short to be used.

[23] For comparison, the CISIRR subregions are grouped into four large regions: southern Beaufort Sea region (Figure 4, Beaufort Sea), Canadian Arctic Archipelago region (Figure 4, western Arctic waterway, western Parry Channel, M'Clintock Channel, Franklin, western high Arctic, eastern high Arctic, Baffin Inlets, Foxe Basin, and Kane Basin), Baffin Bay region (Figure 4, Baffin Bay), and Hudson Bay region (Figure 4, Hudson Bay, Hudson Strait, Davis Strait, and north Labrador Sea). AIC and MYIC is calculated from each data set for each of these four regions. For consistency, all data is first interpolated onto a common 1° grid and seasonally averaged. The summer shipping season in all data sets is chosen as the July-August-September (JAS) average for the three Arctic regions and the July-August-September-October (JASO) average for the Hudson Bay region.

[24] The results for AIC are shown in Figure 6. In general, there is good agreement in interannual variability between CISDA, the Hadley Centre data set (HAD) and the passive microwave record (SSMI) and for all regions correlation coefficients range between 0.9 and 0.95. The Hadley data set is only used in the Beaufort Sea because before 1979 observations were scarce in all other regions [Rayner et al., 2003]. The offset between CISDA and SSMI (Figure 6) is a result of the consistent underestimation of low ice concentrations derived from the passive microwave sensors and has been investigated by Agnew and Howell [2003]. The Rodionov [2004] regime shift detector is used to test for significant shifts in the (CIS-SSMI)/SSMI and the (CIS-HAD)/HAD metrics. It is assumed that a significant shift in this metric would be due to a bias in the CISDA data that is not present in the passive microwave or Hadley data sets. The regime shift detector requires no a priori hypothesis about the timing of a shift and here it is run with a Hubert weight parameter of 2, no prewhitening and cutoffs ranging from 5 to 15 years. No significant shift was detected in the CAA, Baffin Bay or Hudson Bay regions. A significant shift was detected in 1998 in the Beaufort Sea region for (CIS-SSMI)/SSMI which comes shortly after the introduction of RADARSAT-1 in 1996; the mean percent difference before 1998 is ∼40% and after 1998 it increase to ∼70%. We believe that this shift is more likely due to an increase in the passive microwave bias in underestimating the presence of sea ice in regions with low ice concentrations or heavily decayed ice and less likely due to an increase in observable low ice concentration areas from RADARSAT-1. Of the four regions, the Beaufort Sea has experienced the greatest increase in the length of the summer melt season [Stroeve et al., 2006] while maintaining considerably high ice concentrations in summer due to the large fraction of MYI. As the length of the melt season increases, ice concentrations decrease and the fraction of ponded ice increases. In the Beaufort Sea, 1998 was the warmest year on record [Atkinson et al., 2006] and with the exception of 2008 it was also the lightest ice year on record. Further supporting this argument is the fact that no significant shift was detected in any other region. In particular, no shift was detected in Hudson Bay where low ice concentrations are prominent during breakup and where historically there was less direct observation of the ice compared to the Beaufort Sea. It is expected that if there is a bias in summer AIC due to the introduction of RADARSAT-1, it would be greatest in Hudson Bay. As a final check, the two most likely influential changes, the switch to the egg code in 1983 and the introduction of RADARSAT-1 in 1996, are tested using a simple difference of means test on the adjacent periods [e.g., Hare and Mantua, 2000]. The 1996 shift is tested using (CIS-HAD)/HAD and (CIS-SSMI)/SSMI; the 1983 shift is tested using (CIS-HAD)/HAD. The difference in mean was not significant to the 95% confidence interval for either period in the metrics tested. In conclusion, no evidence of a time-varying bias was detected in the AIC time series.

image

Figure 6. Summer all ice coverage, 1980–2005, for the Beaufort Sea, Canadian Arctic Archipelago, Baffin Bay, and Hudson Bay regions. The time series are generated from the CISDA (CIS), Hadley data set (HAD), and passive microwave record (SSMI).

Download figure to PowerPoint

[25] The results for MYI are shown in Figure 7. For MYI, the CISDA is compared against the Belchansky et al. [2004] data set that contains monthly estimates of MYI ice concentration for January, February, and March. The comparison was made in all three months, but because the results were similar in each month only February is shown and discussed here. There is a strong agreement between the two MYIC time series (Figure 7) and the correlation is high (r = 0.75). The Rodionov regime shift detector algorithm was also applied to the (CIS-SSMI)/SSMI metric with no prewhitening and cutoff lengths ranging from 5 to 15. No significant shifts were detected at the 95% confidence level. A difference of means test was used to check for a significant shift in 1996 with the introduction of RADARAT-1 and again no significant change was detected at the 95% confidence level. As in the AIC time series, no evidence of a time-varying bias was detected in the MYIC time series.

image

Figure 7. February multiyear ice coverage, 1980–2005, for the Beaufort Sea, Canadian Arctic Archipelago, and Baffin Bay regions combined. The time series are generated from the CISDA (CIS) and the passive microwave record (SSMI).

Download figure to PowerPoint

3. Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

[26] Trends were calculated by fitting least squares lines to the data. Estimates of the standard deviation of the slope are used to quantify error and statistical significance is determined by testing the null hypothesis of a zero slope using a two tailed t test at a confidence level (i.e., the p value) of 95% (p value < = 0.05). As a second test of significance, a Monte Carlo simulation with 10,000 repetitions is used to calculate new confidence bounds for each time series; this is done to address the potential inflation of the true number of degrees of freedom in each time series due to autocorrelation [e.g., Ebisuzaki, 1997]. The time series for each simulation is first detrended, then, the residuals are smoothed until the lag 1 autocorrelation coefficient equals the true lag 1 autocorrelation coefficient of the residuals in the original data, and finally, the simulated trend is added back to the simulated residual time series. The 2.5 and 97.5 percentiles of the 10,000 simulated trends give the 95% confidence bounds. The Monte Carlo approach makes no assumption about the distribution of the data; however, it does assume a first-order autoregressive (AR1) process. The partial autocorrelation function was calculated for each time series and the results are consistent with an AR1 process. Correlation between variables is tested using the Pearson's correlation coefficient (r). Statistical significance of r is determined using a two-tailed t test at a confidence level (i.e., the p value) of 95% (p value < = 0.05), the effective number of degrees of freedom is calculated using Neffective = N(1 − rxry)/(1 + rxry), in which N is the sample size and rx and ry are the lag 1 autocorrelation coefficients of the two time series. Based on the analysis in section 2.3 any correlation calculated using the data start in 1966 and trends are calculated starting in 1968. Along the major shipping routes where data quality is adequate for trend analysis back to 1960 (Figure 5) trends are recalculated and compared to the 1968 results.

4. Results and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

4.1. Trends in AIC and MYIC (1979–2008 and 1968–2008)

[27] Passive microwave derived estimates of sea ice concentration is the longest record from a single data source and it is the most widely used for studying long-term change in Arctic sea ice. As a first step, trends in AIC calculated from CISDA are compared to this data set. This is done to place estimates from CISDA in the context of the literature and although the regions and the time period being compared are not an exact match, it is expected that any large differences between the two data sets will be detected. The most recent published results from passive microwave data report significant decreasing trends in extent and area coverage in each season and in all months for the 28 year period 1979–2006 [Parkinson and Cavalieri, 2008]. The greatest decrease is in summer, where the reported decrease in summer ice area is 7.8% decade−1. In Canadian waters, significant decreases in summer ice area reported by Parkinson and Cavalieri [2008] are 20% ± 6% decade−1 in Hudson Bay, 18% ± 5.2% decade−1 in Baffin Bay and 6.9% ± 3.3% decade−1 in the CAA. In comparison, significant trends in summer AIC calculated from CISDA for similar regions between 1979 and 2008, are −15.7% ± 3.9% decade−1 in the Hudson Bay region, −16.0% ± 4.1% decade−1 in the Baffin Bay region and −4.4% ± 2.0% decade−1 in the CAA region. Trends calculated from the CISDA and the passive microwave records do not differ by more than the margin of error in the estimates. The time series in summer AIC (Figure 8) beginning in 1979 are also consistent with Parkinson and Cavalieri [2008]. The time series for the CAA and southern Beaufort Sea regions exhibits considerable interannual variability and are characterized by an extreme minimum in 1998 and most recently a minima in 2008 in the southern Beaufort Sea and 2007 in the CAA. The time series for Hudson Bay and Baffin Bay are characterized by a shift in the mid- to late 1990s. They do not share the extreme minima in 1998 and in both time series, the last decade is marked by consistent light ice conditions and an apparent reduction in interannual variability.

image

Figure 8. Time series of summer all ice coverage for the Beaufort Sea, Canadian Arctic Archipelago, Baffin Bay, and Hudson Bay regions from 1968 to 2008.

Download figure to PowerPoint

[28] Trends in summer AIC from 1968 to 2008 for the grouped CISIRR are listed in Table 1 and the time series for each of the grouped CISIRR are shown in Figure 8. Taken back to 1968, trends in AIC calculated from CISDA for each grouped region drop to −11.3% ± 2.6% decade−1 in the Hudson Bay region, −8.9% ± 3.1% decade−1 in the Baffin Bay region, −2.9% ± 1.2% decade−1 in the CAA region, and −5.2% ± 2.4% decade−1 in the Beaufort Sea region (Table 1). In almost all CISIRR, the magnitude of the trends decrease if 1968 is used as the start year compared to 1979.

Table 1. Slopes of Lines of Linear Least Squares Fit Through AIC and MYIC for 1968–2008a
CIS Regions and SubregionsAIC 1968–2008MYIC 1968–2008
102 km2S% decade−1102 km2S% decade−1
  • a

    Units are 102 km2 and % decade−1 with estimated standard deviations. Significance (S) is an indication of which trend estimates are statistically significant at the 95% and 99% confidence levels. Regions and subregions consistent with Figure 1 are shown in bold italics and italics, respectively. Numbers in parentheses correspond to those used in Figure 4. Trends that are statistically significant are shown in bold.

Beaufort Sea region−22.0 ± 10.095−5.2 ± 2.4−12.5 ± 8.2 −4.6 ± 3.0
   Alaska (1)−14.3 ± 3.799−10.9 ± 2.8−10.9 ± 3.199−16.4 ± 4.7
   Mackenzie (2)−2.7 ± 2.9 −5.4 ± 5.6−0.4 ± 1.6 −1.9 ± 8.8
   Banks (3)0.9 ± 1.5 4.4 ± 2.31.23 ± 1.1 12.3 ± 10.4
   Prince Albert (4)−0.1 ± 1.3 −0.1 ± 2.11.63 ± 1.3 3.5 ± 2.9
   Canada Basin (5)−5.7 ± 2.195−3.6 ± 1.3−4.2 ± 2.6 −3.2 ± 2.0
Canadian Arctic Archipelago region−2.3 ± 1.095−2.9 ± 1.2−23.4 ± 12.3 −4.1 ± 2.1
   Western high Arctic−0.2 ± 3.3 −1.5 ± 1.0−2.3 ± 1.7 −1.9 ± 1.4
      Kellett-Crozier (11)−0.2 ± 0.1 −2.4 ± 1.5−0.3 ± 0.3 −6.3 ± 4.8
      Balantyne-Hazen (12)−0.2 ± 0.195−0.8 ± 0.3−0.8 ± 0.495−2.9 ± 1.4
      Gustof Adolf (13)−0.2 ± 0.195−0.5 ± 0.2−0.4 ± 0.3 −1.2 ± 1.1
      Byam Martin (14)−3.3 ± 2.695−1.9 ± 1.50.3 ± 0.2 8.3 ± 5.7
      KCI West (15)−0.2 ± 0.2 −0.8 ± 0.7−0.2 ± 0.4 −1.4 ± 2.3
      Peary-Sverdrup (16)−0.4 ± 0.295−1.4 ± 0.5−1.2 ± 0.595−4.1 ± 1.6
      KCI East (17)−0.2 ± 0.1 −1.7 ± 1.0−0.3 ± 0.2 −3.9 ± 3.4
      McDougall (18)−0.1 ± 0.0 −3.8 ± 2.00.0 ± 0.5 5.3 ± 9.0
      Penny (19)0.1 ± 0.1 2.5 ± 2.00.4 ± 0.19929.6 ± 8.1
      Wellington (20)−0.1 ± 0.2 −2.2 ± 3.10.2 ± 0.2 42.6 ± 24.1
   Eastern high Arctic−2.1 ± 0.5 −3.7 ± 1.4−3.9 ± 4.5 −2.4 ± 4.5
      Jones Sound (2)−0.7 ± 0.395−5.0 ± 2.20.0 ± 0.2 3.6 ± 9.4
      Eureka (11)−0.7 ± 0.395−3.9 ± 1.5−0.3 ± 0.4 −4.9 ± 6.4
      Norwegian Bay (12)−0.8 ± 0.495−2.9 ± 1.4−0.1 ± 0.4 −1.1 ± 4.3
   Western Parry Channel−2.5 ± 1.595−2.0 ± 1.2−0.6 ± 2.5 −0.9 ± 3.8
      M'Clure (1)−0.9 ± 0.8 −2.3 ± 1.9−0.7 ± 1.1 −2.9 ± 4.2
      Viscount Melville (2)−1.3 ± 0.8 −1.7 ± 1.00.2 ± 1.6 0.4 ± 4.0
      West Barrow (3)−0.2 ± 0.1 −3.9 ± 2.00.0 ± 0.1 −1.8 ± 8.5
   Eastern Parry Channel−0.5 ± 0.7 −3.8 ± 5.10.1 ± 0.2 6.8 ± 10.9
      Lancaster/East Barrow (1)−0.5 ± 0.6 −3.7 ± 5.00.1 ± 0.2 7.2 ± 11.3
      Lancaster Mouth (2)−0.5 ± 0.9 −4.1 ± 7.60.0 ± 0.0 4.6 ± 13.8
   Western Arctic waterway−4.0 ± 2.2 −6.7 ± 3.80.1 ± 0.4 4.0 ± 10.0
      Amundsen (1)−1.4 ± 0.9 −8.6 ± 5.70.1 ± 0.1 47.6 ± 31.1
      Coronation Maud (2)−2.4 ± 1.2 −6.5 ± 3.4−0.2 ± 0.2 −7.7 ± 10.9
      Amundsen Mouth (3)−0.3 ± 0.7 −3.6 ± 8.70.2 ± 0.3 15.1 ± 22.8
   M'Clintock Channel−0.1 ± 0.6 −2.0 ± 1.3−1.1 ± 1.3 −3.8 ± 4.8
   Franklin−1.3 ± 1.0 −3.3 ± 2.7−0.8 ± 1.0 −6.7 ± 8.3
      Larsen Victoria (1)−0.8 ± 0.7 −3.2 ± 2.7−0.7 ± 0.8 −7.7 ± 3.1
      Peel Sound (2)−0.5 ± 0.4 −3.7 ± 2.9−0.1 ± 0.3 −2.7 ± 10.4
   Baffin Inlets−5.8 ± 1.699−6.8 ± 1.9−5.4 ± 1.499−16.9 ± 4.1
      Prince Regent-Boothia (12)−5.1 ± 1.495−6.6 ± 1.9−5.4 ± 1.499−17.0 ± 4.3
      Admiralty (13)−0.4 ± 0.195−8.2 ± 2.9−0.0 ± 0.1 −13.3 ± 10.5
      Pond Inlet (14)−0.3 ± 0.195−7.5 ± 3.1−0.0 ± 0.0 −6.0 ± 13.4
   Foxe Basin−7.4 ± 1.995−8.9 ± 2.3−0.4 ± 0.295−20.6 ± 9.6
   Kane Basin−1.0 ± 0.495−2.8 ± 1.1−0.3 ± 0.5 −1.5 ± 3.2
BAFFIN BAY REGION−15.9 ± 5.595−8.9 ± 3.1−1.6 ± 1.8 10.7 ± 12.4
Baffin Bay region−7.0 ± 2.995−9.5 ± 3.90.5 ± 0.2 13.0 ± 19.0
   East Baffin Bay (1)−8.7 ± 2.099−11.1 ± 2.6−0.0 ± 0.1 −0.2 ± 13.6
   West Baffin Bay (2)−0.3 ± 1.5 −1.0 ± 5.31.1 ± 0.1 16.7 ± 10.2
   Northwest Baffin Bay (3)−37.1 ± 8.695−11.3 ± 2.6   
Hudson Bay region−16.5 ± 5.099−10.4 ± 3.1   
   Northwest Hudson Bay (1)−6.7 ± 2.195−13.6 ± 4.3   
   Central Hudson Bay (2)−5.8 ± 2.695−7.5 ± 3.4   
   East Hudson Bay (3)−1.7 ± 1.0 −11.5 ± 6.8   
   Hudson Bay Narrows (4)−2.4 ± 0.7795−13.2 ± 4.4   
Hudson Strait−5.0 ± 1.099−16.0 ± 3.4   
Davis Strait−6.6 ± 1.899−14.2 ± 3.9   
North Labrador Sea−1.6 ± 0.499−17.8 ± 4.8   

[29] The spatial distribution of trends in AIC, beginning in 1968 and calculated for the CISIRR, are shown in Figure 9 (top) and are listed in Table 1. The greatest decreases are in the Hudson Bay region where the results for every CISIRR except east Hudson Bay indicate significant decreases between 7.5% ± 3.4% decade−1 and −17.8% ± 4.8% decade−1 (Figure 9 and Table 1). Ice loss is also high in Baffin Bay where significant decreases in AIC range from 9.5% ± 3.9% decade−1 to 11.1% ± 2.6% decade−1 and are found in all regions except northwest Baffin Bay (Figure 9). Within the Beaufort Sea, only two regions, Alaska and the Canadian Basin, show significant decreases at 10.9% ± 2.8% decade−1 and 3.6% ± 1.3% decade−1, respectively (Figure 9 and Table 1). Although not statistically significant, AIC has actually increased in the Banks region (Table 1). This is consistent with Kwok [2008] who showed that from 2003 to 2007 the predominant motion of the Arctic Ocean pack ice in the summer months was into the Banks region. Within the CAA CISIRR there are significant (but mild) reductions in AIC in several of the high Arctic regions (Figure 9). In general, reductions are greater in the western high Arctic, they range from 0.8% ± 0.3% decade−1 to 1.9% ± 1.5% decade−1, than in the eastern high Arctic where they range from 2% ± 2.9% decade−1 to 5% ± 2.9% decade−1 (Table 1). This is to be expected as the eastern CAA breaks up earlier than the west where some regions do not break up at all during the summer [Melling, 2002]. Interestingly, there have been no significant reductions in summer AIC in any CISIRR along the southern or northern Northwest Passage routes. The greatest decreases within the CAA region are in Foxe Basin and the Baffin Inlets where reductions by 8.9% ± 2.3% decade−1 and 8.2% ± 2.9% decade−1, respectively, are on the same order of magnitude as losses in Hudson Bay and Baffin Bay (Figure 9 and Table 1).

image

Figure 9. Trends in summer (top) AIC and (bottom) MYIC from 1968 to 2008; units are % decade−1. Only trends significant to the 95% confidence level are shown.

Download figure to PowerPoint

[30] Trends in MYIC from 1968 to 2008 for the grouped and individual CISIRR are listed in Table 1 and show no significant change. Within the CAA significant trends were identified in very few CISIRR regions (Figure 9, bottom). The largest reductions are in Foxe Basin and Prince Regent Sound-Gulf of Boothia in the Baffin Inlets, 20.6% ± 9.6% decade−1 and 17.0% ± 4.3% decade−1, respectively. Only slight decreases in MYIC are identified in the western high Arctic and an increase in MYIC is identified in the Penny Strait region of the western high Arctic (Table 1 and Figure 9). The lack of significant decreases in MYIC within most regions of the CAA can be attributed to the CAA being a sink as well as a source for MYI; reductions in in situ MYI formation have been shown to be balanced by increases in import from the Arctic Ocean [Howell et al., 2009]. The only other CISIRR region where a reduction in MYIC is found is in the Alaska region of the Beaufort Sea where the magnitude of the trend is −16.4% ± 4.7% decade−1.

[31] Sea ice of the Canadian Arctic has decreased considerably since 1968 and in general, the greatest decreases are in the seasonal first-year ice regimes. It is possible that these reductions are, in part, a result of a thermodynamic response of the ice to a general warming over the arctic [e.g., Serreze et al., 2007]. However, trends in MYI were found to be considerably less than in all ice, highlighting the importance of sea ice dynamics, particularly within the CAA where the dominant source of MYI is likely changing. To better understand these trends, we first quantify the relationship between variability in sea ice and regional surface temperature. In section 4.2, we explore the relationship between variability in sea ice and the two dominant modes of climate variability in the Northern Hemisphere, the AO, and the ENSO.

4.2. Sea Ice Variability and Surface Air Temperature

[32] Temperature data are from the NCEP/NCAR reanalysis project [Kalnay et al., 1996]. The NCEP data set is commonly used in sea ice studies [e.g., Rigor et al., 2002; Liu et al., 2004] and it is chosen because it extends back beyond the satellite era and has data over the oceans. Given the coarse grid resolution, 2.5° grid, monthly mean surface air temperature was averaged over three regions (Figure 10). It has been noted that NCEP air temperatures are consistently too cold over the Canadian Arctic [Sou and Flato, 2009]. To ensure the reanalysis data represents interannual variations in air temperature, it is compared to monthly in situ surface observations from weather stations. A consistent record of monthly mean air temperature is available for 15 stations in the Canadian Arctic from 1966 to present (Figure 10). The data is quality controlled and is freely available from the Environment Canada National Climate Data and Information Archive (www.climate.weatheroffice.ec.gc.ca). Monthly mean station temperature data are averaged in each region and then correlated with the monthly regional averaged NCEP surface air temperature. There is good correlation between the station data and the NCEP data, correlation coefficients in each region, for each month, exceed 0.8 and generally exceed 0.9. Seasonally, the correlation is lower during the summer months.

image

Figure 10. Regions used to create area averages of seasonal NCEP surface air temperature and the location of the Environment Canada stations used for validation. From west to east the stations are Barrow, Mold Bay, Sachs Harbour, Tutuyuktuk, Cambridge Bay, Kugluktuk, Resolute Bay, Churchill, Eureka, Clyde River, Hall Beach, Coral Harbour Pond Inlet, Inukjuaq, Iqaluit, and Kuujjuak.

Download figure to PowerPoint

[33] Significant trends in seasonal averages of SAT for each of the three regions (Figure 10) over the 1966–2007 period are plotted in Figure 11 and are compared with trends in pan-Arctic (>60N) SAT. The 5 month averages are used to match the length of summer sea ice season. Pan-Arctic there has been significant increases in SAT in all seasons with the greatest increases (∼0.6°C decade−1) occurring during winter and the smallest changes occurring during the summer months (∼0.3°C decade−1) (Figure 11). In the Canadian Arctic significant changes are observed in every season and are greater by up to 0.2°C decade−1 than pan-Arctic change (Figure 11). The trends in SAT in the Canadian Arctic are consistently greater than pan-Arctic wide yet, they follow the same general pattern indicative of surface-based Arctic amplification [Serreze et al., 2009] with the greatest change occurring during autumn/winter and the least during summer. The smallest trends in air temperature are observed in Hudson Bay. In this region, although there is little variability between seasons and no significant trends in winter, the greatest increases occur in autumn (Figure 11).

image

Figure 11. Trends in regional and pan-Arctic NCEP surface air temperature by season 1966–2007; trends are expressed in °C decade−1 and only trends significant at least to the 95% confidence interval using the standard f test are shown. The data is averaged over 5 month periods and is plotted centered on the middle month; the regions correspond to those shown in Figure 10.

Download figure to PowerPoint

[34] Has the warming in SAT caused the observed decreases in summer AIC? This is a difficult question to answer, there are many feedbacks in the system and changes in both SAT and AIC could be in response to changes in another state variable in the climate system. To explore the relationship between SAT and AIC, the data are first detrended. This is done to ensure that any correlation between the two variables stems from joint interannual variability and not a common trend. The first lag is chosen so that there is no overlap between SAT and sea ice; this removes the potential for any correlation to be as a result of a response in SAT to changes in sea ice. AIC in each region is correlated with the preceding spring (FMAMJ, JFMAM), winter (DJFMA, NDJFM, ONDJF), and fall surface air temperatures (SONDJ, ASOND). Again 5 month averages are used to match the length of the summer sea ice season. Since there is little memory in the atmosphere and since there is little ice motion in most CISIRR regions between late fall and early spring, we suggest that increases/decreases in SAT change ice thickness and ice cover and that it is this sea ice anomaly that is later reflected in summer AIC.

[35] The maximum coefficient of determination between SAT and AIC is shown in Figure 12 along with the season when the strength of the relationship is at its maximum. The correlation between AIC and SAT is statistically significant in most CISIRR; the exceptions are Nares Strait, the western Beaufort Sea, Baffin Bay and the central CAA (Figure 12, top). In almost all subregions the maximum correlation is with spring SAT, when air temperature leads by approximately 4–5 months (Figure 12, bottom). In Hudson Strait, Davis Strait, the north Labrador Sea and the eastern Beaufort Sea, correlations are strongest with SAT in winter while in the western Beaufort Sea region the correlation is strongest with SAT in fall (Figure 12). The strongest relationships between summer AIC and SAT occur in the Hudson Bay region and Foxe Basin where at least 30% of the variance in summer AIC is linked to spring SAT. In the majority of the CISIRR between 10% and 30% of the interannual variability in summer AIC can be linked to preseason SAT. The correlation coefficients ranged between −0.32 to −0.76 with an average value of −0.46. The strength of the relationship increased if the trends were left in the time series (average r = −0.51) and if the there was no lag (average r = −0.57). The range and strength of the lagged relationship to SAT is comparable to other studies. For example, Francis et al. [2005] explored the relative contribution of winds, radiation fluxes, and advected heat on the position of the ice edge in the six peripheral Arctic seas, with sea ice lagging by 80 days. The forcing parameters explained anywhere from a few percent to approximately 40% of the variability in ice retreat.

image

Figure 12. Correlation between summer AIC and 5 month averaged SAT, correlations are expressed as the (top) percentage of variance explained and (bottom) season when the correlation with air temperature is maximized.

Download figure to PowerPoint

[36] To test whether or not the recent decreases in summer AIC are linked to increases in presummer SAT, the statistical significance of trends in the residual AIC time series are evaluated after the influence of SAT is removed using linear regression. With the exception of Hudson Strait, Davis Strait and the north Labrador Sea no significant trends were detected in the residuals. This result suggests that decreasing trends in AIC are linked to increasing trends in presummer SAT. This is not the case for Hudson Strait, Davis Strait and the north Labrador Sea where the correlations with regional SAT are greatest in winter (Figure 12), the season where there has been no significant increase in temperature (Figure 11). And this is also not the case for Baffin Bay and Nares Strait, the only regions with statistically significant decreasing trends in AIC but no statistically significant correlation with antecedent SAT. The observed decreases in these two regions are likely due to changes in ice dynamics.

4.3. Sea Ice Variability and the AO/ENSO

[37] Correlation coefficients between the NDJFM averaged ENSO index (NINO3.4 data from http://www.esrl.noaa.gov) and summer AIC, summer MYIC and summer FYIC in each CISIRR are shown in Figure 13. Few statistically significant correlations emerge with AIC, however, correlations with MYIC in Baffin Bay and with both MYIC and FYIC in the CAA which range in magnitude from 0.32 to 0.64, are comparable to the strength of the SAT correlations.

image

Figure 13. Correlation coefficients between the NDJFM-averaged ENSO index and summer (top) AIC, (middle) MYIC, and (bottom) FYIC.

Download figure to PowerPoint

[38] The opposing sign of the MYIC and FYIC correlations explains the lack of statistically significant correlations with AIC (Figure 13) and suggests a combination of thermodynamic and dynamic processes. In any CISIRR, an increase in MYI is due to either in situ formation (i.e., FYI survival) or advection from another region. Baffin Bay is mostly open water during the summer, there is never any in situ MYI formation and there is always room for MYI import. The main source of MYI in this region is sea ice that is advected from the Arctic Ocean through Nares Strait [Tang et al., 2004; Kwok, 2005]. An ice arch controls the flux of sea ice through Nares Strait and this flux is shutdown by the onset of landfast conditions. Results from Samelson et al. [2006] suggest that the formation of the blocking landfast ice is controlled by both winds and air temperature. Within the CAA, Howell et al. [2009] showed that a longer melt season reduces the summer concentration of FYI; this creates larger open water areas which facilitates the import of MYI. In both Baffin Bay and the CAA, dynamic import and by extension summer MYIC is to some extent controlled by temperature. Warmer temperatures either cause reductions in FYI areas in the CAA or they weaken the blocking landfast ice in Nares Strait.

[39] Is there an ENSO signature in SAT that might explain the correlations between ENSO and MYIC/FYIC shown in Figure 13? A weak relationship between the ENSO index and spring SAT in the western Arctic is found but no statistically significant correlations were found between ENSO and SAT in the eastern Arctic or Hudson Bay regions in any season. However, composite plots of summer SAT for the main La Nina and El Niño episodes do show opposing anomalies in SAT over the Canadian Arctic. The El Niño SAT composite shown in Figure 14c for June-August shows anomalously warm temperatures over the whole high Arctic that range from 0.3°C to over 1.1°C. At first glance the SAT composite (Figure 14c) is not consistent with the literature as both Shabbar [2006] and Liu et al. [2004] reported anomalously cold surface air temperatures over the Canadian Arctic during El Niño events. The difference between this study and the two published studies is season and lead time. Shabbar [2006] looked at composites of winter SAT associated with El Niño events and Liu et al. [2004] correlated the nino3.4 index with Arctic SAT with ENSO leading by one month. In this study, anomalous warm temperatures over the Canadian Arctic are in summer with El Niño leading by approximately 6 months.

image

Figure 14. Composite summer (JJASO) (a) 500 mb geopotential height (Z500) and (b) sea level pressure and (c) composite midsummer (JJA) surface air temperature departure from normal (1971–2000) for the 1965–1966, 1972–1973, 1982–1983, 1986–1987, 1991–1992, and 1997–1998 winter El Niño events. Dark (light) shading denotes statistical significance at the 95% (90%) level.

Download figure to PowerPoint

[40] Composites of summer sea level pressure (SLP) and 500 mb geopotential heights (Z500) anomalies for the main El Niño events show significant changes in the atmospheric circulation over the Arctic. The anomaly pattern in Z500 (Figure 14a) is characterized by a center of low height anomalies over Alaska that reaches across the Chukchi and East Siberian Seas, between two centers of high height anomalies over the eastern Canadian arctic and eastern Russia. A similar pattern is seen in sea level pressure (Figure 14b) and the associated anomalous southeasterly flow from northeastern Canada will bring warm continental air into the western Canadian Arctic. We suggest that the SLP and Z500 anomaly patterns could be a summer signature of El Niño that results in warmer SAT over the Canadian Arctic Archipelago.

[41] The preceding winter (JFM) index of the AO (data from http://www.cdc.noaa.gov) was correlated with AIC and MYIC in each CISIRR. Of all the CISIRR, only two statistically significant correlations emerged, they are for AIC in western Baffin Bay (r = −0.38) and MYIC in northern Baffin Bay (r = 0.39). Since so few relationships were found this is not explored further.

4.4. Arctic Bridge and Northwest Passage: Trends in AIC 1960–2008

[42] Strong reductions in sea ice cover found in response to increased SAT, raises many questions concerning the future of navigation through the Canadian Arctic. Data is available in CISDA-H for the 1960s and it is shown in Figure 5 that the data is of high enough quality for trend analysis along the major shipping routes. Extending the CISDA time series back to 1960 made little difference to the magnitude (or significance) of trends along the Arctic Bridge; reductions still range between 11% and 15% decade−1 (Figure 15). Along the NWP routes, trends in AIC beginning in 1960 indicate significant decreases along the southern route that range from 4% ± 1.9% decade−1 to 9.2% ± 4.1% decade−1. These regions are found to be more strongly correlated with SAT than the northern route through western Parry Channel (detail not shown in Figure 12) where no significant trends were found beginning in 1968 (Figure 9) or 1960 (Figure 15).

image

Figure 15. Trends in summer AIC from 1960 to 2008; units are % decade−1. Only trends significant to the 95% confidence level are shown.

Download figure to PowerPoint

[43] Due to changing ice conditions, the Arctic Bridge and the NWP route through the western Arctic waterway are likely to become major shipping routes in the future. Although no trends were detected in ice conditions along the northern NWP route through western Parry Channel, the fact that this route was navigable in 2007 fueled expectations. The 2007 opening of the northern route is the first recorded opening in the satellite record, the CISDA-R record and the CISDA-H record. However, there is a record of this route opening in the CISDA-P record. Figure 16 is the PCSP ice chart for the period of 26 August to 5 September 1962. There is a wide ice-free route through western Parry Channel (Figure 16, blue is open water).

image

Figure 16. Polar Continental Shelf Project ice chart for the period between 26 August and 5 September 1962. Blue is open water, purple is multiyear ice, and orange is first-year and/or second-year ice [Lindsay, 1971].

Download figure to PowerPoint

[44] In the context of climate change, the obvious question is: are the conditions that caused the opening in 1962 different than the conditions that caused the opening in 2007? Howell et al. [2009] show that the opening in 2007 was caused by a combination of anomalously warm temperatures, a preconditioned MYI cover that was more susceptible to melt and southerly winds that prevented MYI from the northern CAA from flowing into western Parry Channel. These factors all held for the 1962 opening. Figure 17 compares summer SAT anomalies in 1962 with 2007, it is clear that 1962 was warmer in the western Canadian arctic compared to 2007 and warmer in the eastern arctic compared to climatology (not shown). There is evidence of a preconditioned ice cover; breakup in the summer of 1961 was more advanced than in normal years [Lindsay, 1971]. Evidence of southerly flow is seen in the ice chart (Figure 16) where mobile ice in western Parry Channel is piled up south of Melville Island. If the prevailing winds were from the north, the ice would have drifted south blocking the open water route. These results suggest that almost 45 years ago similar conditions setup a scenario whereby an open water route formed along the northern route of the NWP.

image

Figure 17. Difference in summer (JAS) surface air temperature departures from normal (1971–2000) over the Canadian Archipelago for 1962 minus 2007.

Download figure to PowerPoint

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

[45] Trends in sea ice in the Canadian Arctic, from 1968 to 2008, were documented using the CISDA. Over this 40 year period, AIC has decreased by 11.3% ± 2.6% decade−1 in the Hudson Bay region, 2.9% ± 1.2% decade−1 in the Canadian Arctic Archipelago region, 8.9% ± 3.1% decade−1 in the Baffin Bay region, and 5.2% ± 2.4% decade−1 in the Beaufort Sea region. Regionally, the greatest decreases in AIC have occurred in the seasonal ice regimes. Few significant reductions in MYIC were revealed highlighting the importance of MYI import from the Arctic Ocean. Results from the comparison of CISDA against the Hadley and the passive microwave records of sea ice concentration revealed no bias in the CISDA data set. This result in combination with an in depth review of the data quality in AIC in each region is evidence that trend estimates are real and are not an artifact of changing data sources over time.

[46] Potential causes of trends and interannual variability in summer AIC were explored. Antecedent SAT was found to explain between 10% and 58% of the interannual variability in summer AIC in almost every region and after removing the linearly regressed effect of SAT on summer AIC, no significant trends were detected in the residuals. These results suggest a universal thermodynamic component to interannual variability and the long-term trend in AIC forced by spring SAT. ENSO was found to explain between 10% and 40% of the interannual variability in MYIC and FYIC in Baffin Bay and the central CAA. We suggest the likely reason, that to our knowledge has not been identified in the literature, is the opposing response of MYI and FYI to positive SAT anomalies over the Canadian Arctic in the summer following strong winter El Niño events where decreases in FYI allows for increased MYI import. The summer SAT anomalies are associated with changes in the atmospheric circulation where anomalous southeasterly winds likely advected warm continental air into the Canadian arctic. This relationship could be used for forecasting summer MYI conditions along the Northwest Passage.

[47] Results from this study provide evidence to support the growing expectation of the potential for increasing SAT to reduce summer sea ice cover and facilitate navigation through the Canadian Arctic in the near future. In particular, trends and variability in summer AIC along the Arctic Bridge and the western Arctic waterway route through the Northwest Passage was linked to SAT. The lack of statistically significant trends in MYIC (1968–2008) and AIC (1968–2008 and 1960–2008) along the western Parry Channel route of the Northwest Passage supports the notion that MYI concentrations will remain high in this region as long as advection from the Arctic Ocean through the northern CAA continues. Furthermore, this region of the NWP was navigable in 1962 and it is shown that the atmospheric conditions that forced the 2007 clearing were similar to 1962.

Appendix A:: Canadian Ice Service Digital Archive

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

[48] This appendix is a synthesis of the official CISDA documentation and early government reports that describe the database [CIS, 2007a, 2007b; Crocker and Carrieres, 2000a, 2000b]. All of the material relevant for using the data in climate studies is summarized.

A1. Source, Type, and Estimate of Error

[49] There are three main sources of error in the ice charts: chart preparation techniques and procedures, source data, and nowcasting. In this section, each source of error is discussed and its contribution to the overall error in ice position, ice concentration and stage of development is estimated. Errors resulting from the process of digitizing the paper charts (prior to 1998) are also examined. An attempt is made in these discussions to categorize errors as random or systematic.

A1.1. Chart Preparation

[50] The four major technological advancements in chart preparation (Table A1, italics) all increased the data quality. In 1975, the switch from teletype to fax began to increase the timeliness and overall usefulness of satellite imagery in the preparation of the regional charts. In 1979, forecasters began using the epidiascope, a projection system that allowed one image to be projected onto another. The epidiascope allowed slightly more accurate transfer of ice information than the previous manual techniques. In 1983, the switch from the ratio code to the egg code allowed for more detailed reporting of the stage of ice development. The ratio code contains a single category for first-year ice (30–200 cm), whereas the egg code contains four different first-year ice subcategories. In situations where detailed thickness information is available, the ratio code data (pre-1983) would be less accurate. This probably did not result in an immediate increase in accuracy in 1983 but permitted a gradual improvement as detection technology improved. The introduction of the egg code did not affect ice concentration because it is reported in tenths in both the ratio code and egg code. The preratio code in use prior to 1966 contained the same basic information as the ratio code. However, because the code and conventions for reporting partial concentrations and stage of development were different from the ratio code only the total concentration was digitized for the pre-1966 historical charts (the partial concentration information was recorded as a text string). The code used on the PCSP charts prior to 1966 was similar but not identical to the 1960–1965 historical charts. Finally in 1995 the Ice Service Integrated System (ISIS) was installed. This fully integrated, GIS-based system replaced the older IDIAS system. The ISIS operational interface is easier to use, has increased functionality and is more stable.

Table A1. The Six Time Periods That Have Been Identified as Being Significant in the Data Quality of the CIS Regional Chartsa
Time PeriodSource Data Technique DescriptionChart Preparation Technique Description
1968–1974Infrequent availability of satellite data (NOAA VHRR); intense shipping and airborne observations in shipping areas. 
1975–1977Increasing availability of near-real-time satellite data (NOAA VHRR, Landsat MSS); intense shipping and airborne observations in shipping areas.Introduction of fax technology in 1975 for data transmission.
1978–1982Availability of near-real-time satellite data (NOAA AVHRR, Landsat, Nimbus-7 SMMR); introduction of airborne SLAR in addition to airborne observations; intense shipping and airborne observations in shipping areas.Introduction of the epidiascope in 1979.
1983–1990Availability of near-real-time satellite data (NOAA AVHRR, Landsat MSS/TM, Nimbus-7 SMMR, limited operational use of SSM/I); airborne SLAR in addition to airborne observations; intense shipping and airborne observations in shipping areas.Introduction of the “egg code” for reporting ice conditions.
1991–1995Availability of real-time satellite data (NOAA AVHRR, Landsat MSS/TM, limited operational use of ERS-1 and SSM/I); introduction of airborne SAR in addition to airborne SLAR and observations; intense shipping and airborne observations in shipping areas. 
1996–2005Availability of near-real-time satellite data (NOAA AVHRR, RADARSAT-1, limited operational use of ERS-2); airborne SAR/SLAR in addition to airborne observations; intense shipping and airborne observations in shipping areas.Introduction of advanced computer software for the generation of ice charts (IDIAS and ISIS).

[51] Table A2 summarizes the estimated positional errors associated with various components in the chart preparation process. The maximum error ranges from ±5 km to ±7.5 km and is associated with earlier technologies. It is believed that any error associated with chart preparation is random because ice interpretation has always been and continues to be carried out manually; it is unlikely that a systematic bias was introduced with any of the technological advancements. It is acknowledged that errors can be introduced by the limitations of human observers who must estimate distances and lengths and mentally integrate varied ice characteristics over large areas. Although these gross errors are impossible to quantify, ongoing effort is made to correct these errors manually as they become apparent.

Table A2. Estimated Positional Errors Attributed to Various Components of Chart Preparationa
SourcePosition (±km)
Compilation and drawing (digital)1
Compilation and drawing (epidiascope)5
Compilation and drawing (manual)7.5
Generalization1
Deformation of materials (paper)1.5
Digitization1
Data transmission (digital)0
Data transmission (fax)5
Data transmission (code)5
A1.2. Source Data

[52] The many sources of ice information used to generate the ice charts have varying limitations. In general, the accuracy of the information is a function of sensor resolution, distance from the target and the nature of the ice feature being observed. Table A3 lists the estimated errors in ice position, ice concentration and stage of development associated with the main sources of ice information. A complete list of data sources can be found in Table A4.

Table A3. Typical Estimated Observation Errors Associated With the Principal Sources of Ice Informationa
SourcePosition (±km)Total Concentrationb (tenths)Stage of Developmentc (% confidence)Floe Sized (% confidence)
  • a
  • b

    The accuracy of total concentration estimates made from satellite data is highly dependent on the concentration. Many sensors, including RADARSAT-1 are poor at showing low ice concentrations. Concentration data derived from SSM/I imagery has been shown to systematically miss low concentrations of ice (<3/10ths).

  • c

    A 75% confidence in stage of development indicates the ice was properly characterized every 3 out of 4 times.

  • d

    The accuracy of floe size information is highly dependent on the floe sizes. Small floes cannot be resolved on satellite imagery. Therefore, the estimates in for floe size indicate the “ability to determine floe size above the sensor resolution.”

  • e

    Visual observations are strongly influenced by range. Near to the flight path or shore station, observations are more accurate than at large distances from the observer.

  • f

    The positional accuracy of ship reports decreases with increasing distance from the coast.

RADARSAT-10.50.57590
ERS-1/ERS-20.2516590
NOAA AVHRR (digital)225575
NOAA AVHRR (predigital)2045575
NOAA VHRR2025575
SSM/I12.54NA75
LANDSAT0.517590
Aircraft SLAR (GPS navigation)0.515090
Aircraft SLAR (INS navigation)215090
Aircraft SAR0.250.58095
Aircraft Visual4 (GPS navigation)0.518595
Aircraft Visual (INS navigation)218595
Helicopter Visual (GPS)0.518595
Helicopter Visual (pre-GPS)118595
Shoree0.548580
Ship (GPS)0.528590
Shipf (pre-GPS)228590
Table A4. Source Data Used in Chart Preparationa
 PlatformDates in OperationWavelengthGround ResolutionSwath Width
  • a

    Crocker [2002].

Satellite sensors     
   SR (scanning radiometer)NOAA 1–5January 1970 to March 1979visible, thermal IR3.2–8.0 km2900 km
   VHRR (very high resolution radiometer)NOAA 2–5October 1972 to March 1979visible, thermal IR1.0–1.9 km25,800 km
   MSS (multispectral scanner)LANDSAT 1–7July 1972 to presentvisible, thermal IR∼80 m185 km
   AVHRR (advanced very high resolution radiometer)TIROS-N NOAA 6–11 NOAA “next”October 1978 to presentvisible, near IR, thermal IR1.1 km2580 km
   SSMR (scanning multichannel microwave radiometer)NIMBUS-7October 1978 to July 1988passive microwave20–80 km783 km
   TM (thematic mapper)LANDSAT 4–7July 1982 to presentvisible30 m185 km
   SSM/I (Special Sensor Microwave Imager)DMSPJune 1987 to presentpassive microwave25 km1400 km
   SAR (synthetic aperture radar)ERS-1July 1991 to June 1996 30 m80–100 km
 ERS-2April 1995 to present 30 m80–100 km
 RADARSAT-1November 1995 to present 9–100 m45–500 km
 RADARSAT-2December 2007 to present 9–100 m45–500 km
Airborne sensors     
   VisualDC-4; Lockheed Electra; Dash-7; challenger; CCG helicopter and others1968 to presentvisible∼10–100 km∼10–30 km
   SLAR (side-looking airborne radar)Lockheed Electra, Dash-71978 to presentX band∼50–300m100 km
   SAR (synthetic aperture radar)Challenger1990 to 1995X band30 m, transmitted as 100 m50 km
Surface observations     
   VisualCCG, marine vessels, radio and DEW line stations, lighthouses and others1960 to presentvisiblevariablevariable

[53] There is no known bias in determining ice position from either satellite imagery or visual observation. The positional accuracy of information from different sensors ranges from about ±20 km for predigital NOAA AVHRR and NOAA VHRR to ±0.25 km for aircraft synthetic aperture radar (SAR) and ERS-1/ERS-2 SAR. Predigital NOAA AVHRR and NOAA VHRR provided valuable information in stable ice regimes in remote areas, however, they were not frequently used in shipping areas and were not available in real time for use in the regional chart analysis. In the historical charts (1960–1974) they were used to improve the continuity in the analysis in shipping areas. Further investigation into the source data for a particular region could increase the position accuracy and allow for the study of more detailed ice features.

[54] Ice concentration is relatively easy to observe visually and a number of remote sensing tools have been shown to provide accurate total concentration data. The highest error is 4/10 and is associated with predigital AVHRR, the DMPS Special Sensor Microwave Imager (SSM/I), the Scanning Multichannel Microwave Radiometer (SSMR) and shore-based observations. Shore-based observations make up a very small proportion of the source information and were used mainly to verify events such as harbor breakup and clearing. Predigital AVHRR as noted above was used in shipping areas only to extend the aerial or temporal coverage of the more accurate data sources and in more stable regimes in remote areas to monitor the formation of polynas and leads and to monitor the progress of breakup in years when ice in these areas became mobile. Ice analysts were and are very reluctant to use satellite based passive microwave data to determine ice concentration during the melt season. The high error associated with SSM/I occurs because the imager does not resolve low ice concentrations, very thin ice and heavily ponded ice [Agnew and Howell, 2003]. SSM/I was and is generally used only to aid the identification of the main ice edge and as an extension to other more reliable sources. All sensors are subject to systematic underestimation of low concentrations of ice; the extent of the error depends on the resolution of the sensor and floe size. However, this type of error is reduced in the ice charts data set because trained analysts compensate for the sensor underestimates when preparing the charts.

[55] There are currently no sensors in operational use that can distinguish ice thickness to the same resolution as the stage of development (ice type) category contained in the egg code. It is usually possible to distinguish multiyear ice from first-year ice, and young and new ice (less than about 30 cm in thickness) from first-year ice. Data from RADARSAT-1 SAR, the most widely used satellite data source since its launch in November 1995, has by far the highest confidence level at 75%. The ice analyst infers ice type from the tone, texture and shapes in the SAR imagery. For example, in most cases, multiyear ice is easily identified because it appears bright in the imagery; the high backscatter is a result of volume scattering from large air inclusions in the ice. Beyond these broad classifications, the accuracy of the thickness information relies heavily on aerial and surface observations with an estimated confidence level of 85%. Multiyear ice and the different forms of young and new ice can be readily distinguished by the trained observer. Ship-based observations are probably the most accurate in terms of stage of development because the observer is close to the ice and can often observe the ice floes on edge as the vessel transits through the ice.

A1.3. Nowcasting

[56] Estimated position, ice concentration and stage of development errors due to nowcasting are shown in Table A5. In general, the shorter the nowcast period, the smaller is the associated error. Position errors range from 5 km to 25 km and concentration errors range from 1/10 to 3/10 for nowcasts ranging from a few hours to 2 days. The stage of development accuracy is listed as being constant because this characteristic changes very slowly once the ice has reached the first-year stage. However, nowcasting errors in stage of development are significantly higher when ice is first forming. Once the ice has formed its growth is predictable using simple freezing degree day models.

Table A5. Typical Estimated Nowcast Error for a Range of Nowcast Durationa
SourcePosition (km)Total Concentraion (tenths)Stage of Developmentb (% confidence)Floe Sizeb (% confidence)
  • a

    CIS [2007b]. These errors represent a broad range of nowcasting from anticipating the change in ice edge of breakup conditions from a reconnaissance flight or satellite image that is a day old in shipping regions where the ice is very dynamic to confirming the lack of change in vast areas of consolidated ice or open water.

  • b

    The error estimates indicate the level of confidence with respect to the reported stage of development and floe size codes associated with nowcast (i.e., assuming the initial values were correct). For example, a value of 95% indicates that changes in the stage of development are correct 19 out of 20 over the nowcast period.

6–12 h (temporal adjustment)519595
24 h (no data for 1 day)1029590
>48 h (no data for 2+ days)2539585

[57] In general, these errors are assumed to be random because the nowcasting process is manual. However, there is a known systematic bias in multiyear ice estimations (and to a smaller degree total concentration estimates) associated with nowcasting. Before satellite data were available in near real time, in the absence of direct observation, it was common practice for ice analysts to assume that any ice in the high Arctic was multiyear ice. If there was any doubt, the practice was to overestimate the stage of development. This bias is significantly reduced in the historical charts but is still present in the remote areas in the early years.

A2. Space and Time Dependence of Errors

[58] The greatest shortcoming of the data set is the changing error with space and time. The changes in chart preparation technique, source data, and nowcasting over time are illustrated by the time line in Table A1. With respect to time series homogeneity, the change from the ratio code to the egg code in 1982–1983 and the introduction of RADARSAT-1 in 1996 are the most important developments.

[59] The accuracy of an ice chart varies regionally and the overall accuracy of a chart is highly dependent on the proportion of the chart derived from each data source. Within a single ice chart, the highest-quality data are found near the shipping routes and communities; this holds true even during the remote sensing era. Figure A1 lists the percentage of each chart derived from direct observations, satellite sensors and nowcasting. Surface observations have played a small role in overall chart production in all regions. The number of observations and therefore the proportion of the charts covered by surface observations have decreased steadily over the 30 year period from 2% to 1%. It should be noted that surface observations are still important for verification of the information derived from remotely sensed data and their importance in overall chart preparation is therefore much greater than what is inferred from the total area coverage. Aerial reconnaissance increased as navigation and radar systems became available then decreased in the mid-1990s when the availability of RADARSAT-1 data reduced reliance on aerial observations. Although satellite observations were available as early as 1968 they were of limited use operationally until the introduction of facsimile technology in the mid-1970s. The most significant event in terms of satellite coverage occurred in 1996 when RADARSAT-1 became operational.

image

Figure A1. Proportion of historical and regional charts covered by observational information and nowcasts: Arctic regions. Solid arrows indicate coverage was relatively stable; dashed arrows indicate the values increased or decreased in an approximately linear manner. For historical charts 1960–1974 and regional charts thereafter. Bracketed values are for the regional charts during the overlap period [CIS, 2007b]. Superscript 1 indicates nowcasts >1 day. Superscript 2 indicates typical nowcast duration.

Download figure to PowerPoint

A3. Quality Assurance and a Data Quality Index

[60] Quality assurance was carried during chart production, postchart review, chart digitization and during the preparation of the Sea Ice Climatic Atlas [CIS, 2002]. Very few errors were found due to the transfer of chart codes to the digital database, however, a number of errors were identified in the interpretation of chart information prior to digitizing. Gross errors have been corrected and these quality assurance efforts are ongoing as updates are made to the database.

A3.1. Caveats for Statistical Use

[61] An important result of the experience gained during this process is a list of caveats to guide statistical use of the database. Specific caveats for consideration in statistical analyses of the datasets [CIS, 2007b] are as follows:

[62] 1. The dataset, for the most part, consists of weekly observations over an active ice season, but missing charts do exist with increasing frequency towards both the beginning and end of the active season.

[63] 2. Reports of trace amounts of ice appear throughout the dataset, but a decrease in the number of these trace reports per polygon is seen in 1982 as the “ratio code,” which allowed for multiple traces per polygon was replaced with the “egg code,” which allows for only 1 trace of “thick” ice and 1 trace of “thin” ice.

[64] 3. The encoding of areas of “fast ice” has not always been consistent for the eastern Arctic and western Arctic particularly during the “ratio code” years which makes it difficult to differentiate between “fast ice” and “drift ice.”

[65] 4. Up to and including 1996, charts for adjacent regions were not produced on the same day and this can result in differences in ice parameters between charts in those areas where charts overlap.

[66] 5. Some inconsistencies have been noted in the continuity of amounts of old ice (multiyear/second-year/old) from the end of the summer melt season to the beginning of the next season. This may be reasonable in areas where ice motion still occurs during the winter season, e.g., the Beaufort Sea, yet unreasonable in areas where little or no ice motion occurs, e.g., the Canadian Arctic Archipelago.

[67] 6. Some inconsistencies have been noted in the differentiation between multiyear and second-year ice. It is recommended that studies involving multiyear ice should consider multiyear, second-year, and old ice as one ice type.

[68] 7. Areas of “no data” exist in the dataset. At times these areas can be large and their effect on statistics may need to be considered.

[69] 8. It has been noted that the use of “multiple codes” can affect statistics, particularly the statistics on old ice. A “multiple code” is an area on a chart where two or more ice codes were placed with no line of delineation between them. Upon digitization, an approximate midpoint between multiple codes was drawn. In general, the effect will be minimal due to the similarities in ice conditions between the codes. However, areas of low concentrations of old ice have been affected, most notably in Baffin Bay.

[70] 9. Concentration of “icebergs” 1/10 or greater is sometimes included in the total concentration. This should be considered for studies pertaining only to sea ice.

[71] 10. The encoding of “9+” for total concentration can also affect values. Numerical attributes in the CISDA consider this to be 9.7/10 or 97% total ice cover. This differs from the sum of the partial concentrations, which in this case would be 10/10 or 100%. This applies to all regions and all years in the CIS Digital Charts Database.

[72] 11. Changes in the digital base map can affect statistics, but their impact is considered minor.

[73] 12. Changes in chart extents throughout the dataset can have a marked effect on values. Care must be taken to ensure usage of a consistent area throughout the dataset for the analysis.

A3.2. Quality Index

[74] A qualitative quality index (QI) was developed to portray the variability in data quality over space and time using the ice regime regions defined by CIS (CISIRR) and described in section 2.1. The QI is intended as a tool to assist users of the database to assess the applicability of the data to a particular problem. The QI was developed based on extensive interviews with those most familiar with the database and was peer reviewed by CIS staff. The development process was iterative and the final result represents a general consensus among this group of experts.

[75] The period of record was broken down into eras, each representing consistency in source data and preparation methods (Table A1). The sources of ice information were grouped into four categories: Shipboard observations, Airborne observations, Airborne SAR and SLAR (side-looking airborne radar), and Satellite observations (optical, infrared, SAR). For each data era and for each CISIRR subregion each of the four data sources was assigned a qualitative score from 0 to 5 where 0 is poor confidence 3 is average confidence and 5 is high confidence. This quality score takes into account the quality of the source data, for example predigital AVHRR versus RADARSAT-1, and the availability over space and time of the information. These basic scores therefore account for the data source and the degree of nowcasting required in each subregion for each time period. The scores are then weighted for their contribution to the ice chart in each era and for the number years in the era. The final scores for each subregion and for each era are listed in Table A6. These scores can be combined to get a QI value for each subregion for any period of interest

  • equation image

where ωsubregion is the score for the subregion, i is a specified time period from 1 to 6, j is a specified ice reconnaissance data set from 1 to 4, xj is the raw score applied to ice reconnaissance data set j for time period I, αj is the weight for the contribution of ice reconnaissance data set j to the regional ice chart and βi is the weight for the contribution of the time period i to the entire time period.

Table A6. Quality Index Scores by Period for Each Region for the Blended Historical and Regional Ice Chart Dataseta
CISIRRHistoricalRegional
1960–19651966–19671968–19701971–19741968–19701971–19741975–19771978–19821983–19901991–19951996–20062007–2008
Beaufort Sea (1)2.52.73.43.63.13.13.83.83.74.04.94.9
Beaufort Sea (2)2.52.73.43.63.13.13.83.83.74.04.94.9
Beaufort Sea (3)1.72.22.52.72.22.22.92.92.93.54.94.9
Beaufort Sea (4)0.80.91.82.01.51.52.52.52.53.04.74.7
Beaufort Sea (5)0.20.31.11.30.80.82.52.52.53.04.54.5
Arctic Ocean periphery (12)0.20.31.21.40.80.82.12.12.12.53.53.5
Arctic Ocean periphery (13)0.20.31.21.40.80.82.12.12.12.53.53.5
Western high Arctic (11)0.80.91.41.40.91.02.12.12.12.54.44.4
Western high Arctic (12)0.80.91.41.40.91.02.12.12.12.54.44.4
Western high Arctic (13)0.80.91.41.40.90.82.12.12.22.54.44.4
Western high Arctic (14)1.61.71.82.41.61.82.72.42.22.54.44.4
Western high Arctic (15)0.80.91.41.90.91.02.12.12.12.54.54.5
Western high Arctic (16)0.80.91.41.90.91.02.12.12.12.54.44.4
Western high Arctic (17)0.80.91.31.61.01.02.22.22.22.54.44.4
Western high Arctic (18)2.42.83.03.22.62.63.53.63.54.04.74.7
Western high Arctic (19)1.01.11.61.81.21.22.32.32.32.64.54.5
Western high Arctic (20)2.22.62.83.02.42.43.23.23.23.94.84.8
Eastern high Arctic (2)1.21.31.51.71.21.22.32.32.32.64.54.5
Eastern high Arctic (11)1.01.11.31.71.21.22.32.32.32.64.54.5
Eastern high Arctic (12)1.01.11.51.71.21.22.32.32.32.64.54.5
Western Parry Channel (1)1.41.51.61.81.11.22.42.22.22.54.54.5
Western Parry Channel (2)1.81.92.22.41.91.92.52.42.53.04.54.5
Western Parry Channel (3)3.23.73.94.13.63.64.34.44.24.55.05.0
Eastern Parry Channel (1)3.23.73.94.13.63.64.34.44.24.55.05.0
Eastern Parry Channel (2)3.23.73.94.13.63.64.34.44.24.55.05.0
Western Arctic waterway (1)2.63.13.43.63.13.13.83.83.74.04.94.9
Western Arctic waterway (2)2.02.32.42.62.12.12.82.62.73.14.74.7
Western Arctic waterway (3)2.63.13.43.63.13.13.83.83.74.04.94.9
McClintock Channel1.41.51.61.81.31.32.42.22.22.54.44.4
Franklin (1)2.02.32.93.12.62.63.33.23.23.64.84.8
Franklin (2)1.42.32.42.62.12.12.82.62.73.14.74.7
Baffin Inlets (12)1.21.31.51.71.21.22.32.32.32.64.54.5
Baffin Inlets (13)1.21.51.71.91.41.42.01.82.02.64.54.5
Baffin Inlets (14)1.01.11.31.51.01.02.22.22.22.54.44.4
Foxe Basin1.21.31.51.71.21.22.32.32.32.64.54.5
Kane Basin0.80.91.41.90.91.02.12.12.12.54.44.4
Baffin Bay (1)2.63.13.43.63.13.13.83.83.74.04.94.9
Baffin Bay (2)2.22.32.32.52.02.02.52.52.53.04.54.5
Baffin Bay (3)2.63.13.43.63.13.13.83.83.74.04.94.9
Hudson Bay (1)2.62.93.23.42.92.93.53.63.54.55.05.0
Hudson Bay (2)1.21.33.23.42.92.93.53.63.54.55.05.0
Hudson Bay (3)1.21.31.51.71.21.22.32.32.32.64.54.5
Hudson Bay (4)2.63.23.23.42.92.93.53.63.54.55.05.0
Hudson Strait2.63.23.23.42.92.93.53.63.54.55.05.0
Davis Strait2.63.43.43.63.13.13.83.83.74.04.94.9
North Labrador Sea1.21.31.51.71.21.22.32.32.32.64.54.5

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

[76] The authors wish to thank John Walsh, Roger DeAbreu, Chris Petrich, David Huard, two anonymous reviewers whose suggestions and comments greatly improved this manuscript. We also wish to thank all of the many people at the Canadian Ice Service who have contributed to the CISDA. Support for this project was provided by Environment Canada (Tivy, McCourt, Chagnon, Crocker, Carrieres), the Natural Sciences and Engineering Research Council of Canada (Howell, Yackel), and the International Arctic Research Center (Tivy).

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Canadian Ice Service Digital Archive
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. Conclusions
  8. Appendix A:: Canadian Ice Service Digital Archive
  9. Acknowledgments
  10. References
  11. Supporting Information
FilenameFormatSizeDescription
jgrc11576-sup-0001-t01.txtplain text document5KTab-delimited Table 1.
jgrc11576-sup-0002-taA01.txtplain text document2KTab-delimited Table A1.
jgrc11576-sup-0003-taA02.txtplain text document0KTab-delimited Table A2.
jgrc11576-sup-0004-taA03.txtplain text document2KTab-delimited Table A3.
jgrc11576-sup-0005-taA04.txtplain text document2KTab-delimited Table A4.
jgrc11576-sup-0006-taA05.txtplain text document1KTab-delimited Table A5.
jgrc11576-sup-0007-taA06.txtplain text document3KTab-delimited Table A6.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.