SEARCH

SEARCH BY CITATION

Keywords:

  • sea ice;
  • Canadian Arctic Archipelago;
  • QuikSCAT;
  • Northwest Passage;
  • remote sensing

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[1] Estimates of annual sea ice melt onset, freeze onset, and melt duration are made within the Canadian Arctic Archipelago (CAA) using SeaWinds/QuikSCAT data from 2000 to 2007. The average date of melt onset occurred on day 150, the average freeze onset occurred on day 266, and the average number of days of melt was 116. Melt onset occurred first, and freeze onset occurred last within the Amundsen, Western Arctic Waterway, and Eastern Parry Channel regions, whereas the reverse occurred in the Queen Elizabeth Islands (QEI) and the M'Clure and Viscount-Melville regions. Multiyear sea ice (MYI) increases occurred from 2000 to 2004 because of dynamic import and first-year sea ice (FYI) being promoted to MYI, but this replenishment virtually stopped from 2005 to 2007, coincident with longer melt seasons. Only after two consecutive long melt seasons (2005–2006) and almost no replenishment were regions to the south of the QEI cleared of MYI. We argue that this is because MYI must slowly ablate on the underside while in transit within the CAA from the small oceanic heat flux and can therefore survive for several years in southern regions without replenishment. Net positive dynamic MYI import into the CAA was observed in 2007 following MYI removal during 2005–2006. Longer melt seasons will continue to reduce the inventory of FYI in the CAA following the melt season. Longer melt seasons within the CAA will likely not reduce MYI dynamic import, but it remains to be seen whether or not this MYI will be able to survive longer melt seasons as it migrates to the southern regions.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[2] Since 1979, sea ice in the Arctic Ocean has experienced considerable decreases during the summertime [Serreze et al., 2007] that have been linked to longer melt seasons [Stroeve et al., 2006]. Simulations from most global climate models predict continued and potentially rapid decreases in sea ice [Holland et al., 2006; Zhang and Walsh, 2006]. However, compared to observational records, climate model simulations may actually be far too conservative at predicting a summertime sea ice free Arctic [Stroeve et al., 2007] as exemplified by the rapid reduction of Arctic sea ice in 2007 [Nghiem et al., 2007; Stroeve et al., 2008]. The year 2007 also marked extreme record low sea ice conditions in the Arctic but 1998 still represents record low sea ice conditions within the Canadian Arctic Archipelago (CAA) [Atkinson et al., 2006].

[3] The CAA is composed of a collection of islands and channels that face the Arctic Ocean to the north (Figure 1). The narrow channels of CAA contain a mix of seasonal first-year sea ice (FYI) and perennial or multiyear sea ice (MYI) such that the spatial resolution (1.4° × 1.4°) of even the latest state-of-the-art climate models (e.g., Community Climate System Model, version 3; [Collins et al., 2006]) still has difficulty resolving the sea ice melt processes within these subresolution channels. The islands interspersed among the waterways of the CAA further complicate model projections because of the vastly different and variable specific heat capacities of snow covered land, snow covered sea ice, and water; particularly during the winter to spring transition. Snowmelt timing from the margins of the islands is assumed to be much earlier than the adjacent sea ice cover but its impact on sea ice melt onset timing is unknown. Regional melt onset (MO) and freeze onset (FO) information are important for improving climate models simulations and within the CAA this information is particularly critical. Even as climate model spatial resolution increases, problems with the CAA's intricate islands and channels will still remain. Additionally, the effect of a longer melt season may not necessarily imply reductions in MYI because large-scale sea ice dynamics continuously force MYI from the Arctic Ocean up against the CAA [Agnew et al., 2001; Melling, 2002].

image

Figure 1. Map of the Canadian Arctic Archipelago and its subregions. The locations of the QuikSCAT time series algorithm test sites are also shown.

Download figure to PowerPoint

[4] The exchange of MYI between the Arctic Ocean and the CAA is normally blocked by thick landfast MYI barriers. However, Melling [2002] suggested the effect of a longer melt season in the CAA might weaken these barriers allowing more MYI to enter the CAA (Figure 1). MYI is thicker than seasonal FYI and if the CAA is subject to increased amounts of MYI, this will impact the seasonal thermodynamic and dynamic processes operating in the CAA. The narrow channels in the CAA also help shield MYI from the wind and warm currents that are particularly influential on the melting of MYI in the Arctic Ocean. MYI exchanges between the CAA and Arctic Ocean from 1997 to 2002 find that the Queen Elizabeth Islands (QEI) exhibited a net import whereas the M'Clure Strait exhibited a net export [Kwok, 2006]. From 2000 to 2004 the CAA exhibited increases in MYI from in situ growth as well as dynamic import from both the QEI and M'Clure Strait [Howell et al., 2006, 2008]. Alt et al. [2006] provide evidence for the weakening of the MYI barriers within the QEI from 1998 to 2005. Increasing surface air temperatures (SAT) within the CAA were linked to reductions in the amount of MYI formed in situ from 1982 to 2004 but dynamic import has compensated [Howell et al., 2008]. Clearly, exploring the links between MYI and melt season processes within the CAA is important, especially the MYI in situ contribution which has received less attention. In addition MYI is the most significant barrier to successful navigation of the Northwest Passage.

[5] Remote sensing, especially in the microwave region of the electromagnetic spectrum, has proven particularly useful for the monitoring and study of sea ice melt processes. Estimates of sea ice MO conditions over broad-scale areas have primarily relied on passive microwave data from the Scanning Multichannel Microwave Radiometer (SMMR) and Special Sensor Microwave Imagery (SSM/I) [e.g., Serreze et al., 1993; Smith, 1998a; Drobot and Anderson, 2001; Belchansky et al., 2004a; Stroeve et al., 2006]. Unfortunately, very few studies have focused on the CAA [Howell et al., 2006] because of the intricate mix of sea ice types and significant land contamination of the coarse resolution passive microwave pixels within the narrow channels. Studies of FO and melt duration (MD) are limited [Smith, 1998a, 1998b; Belchansky et al., 2004b; Stroeve et al., 2006] and for the most part have excluded the CAA. However, SeaWinds/QuikSCAT (QuikSCAT) Scatterometer Image Reconstruction (SIR) active microwave data has large areal coverage and high spatiotemporal resolution making it ideally suited for monitoring and mapping sea ice melt dynamics within the CAA [Howell et al., 2006].

[6] The objective of this study was to extend the QuikSCAT melt algorithm developed by Howell et al. [2006] to include estimates of FO and thereby derive MD in order to report on changes in melt season parameters and sea ice conditions within the CAA from 2000 to 2007. We begin with a description of the extended algorithm followed by a discussion of the resulting regional spatiotemporal distribution of MO, FO, and MD. We then explore the links between changes in the melt parameters to changing sea ice conditions (with particular emphasis on MYI) within the CAA during this time period. Specifically, we addressed two important interrelated questions: (1) Can MYI within the CAA survive longer melt seasons? And (2) will MYI in the CAA continue to decrease?

2. Data

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

2.1. Seawinds/QuikSCAT (QuikSCAT)

[7] QuikSCAT data for the period 2000–2007 was obtained from the NASA Scatterometer Climate Record Pathfinder (SCP) project. The SeaWinds scatterometer on board the QuikSCAT satellite is a dual-polarized real aperture radar operating at 13.4 GHz (Ku-band). QuikSCAT provides normalized cross-section backscatter values at fixed incident angles of 46° (HH) and 54.1° (VV) over a swath width of 1800 km with twice daily temporal resolution (i.e., daily ascending and descending passes). QuikSCAT data is available in two image products, eggs and slices, at Scatterometer Image Reconstruction (SIR) enhanced and nonenhanced grid resolution. The spatial resolution for the nonenhanced grid products is 11.125 km for slices and 22.25 km for eggs while the spatial resolution of SIR enhanced is ∼8–10 km and ∼4 km for eggs and slices respectively [Long, 2000]. The SIR products take advantage of the spatial overlapping backscatter measurements taken at different times, thus increasing the sampling density in order to increase the spatial resolution [Long et al., 1993]. However, limitations imposed on the SIR data are the sampling density, nulls introduced by the aperture function(s), the acceptable noise level, and the temporal stability of the study area [Long et al., 1993; Early and Long, 2001]. Howell et al. [2005] evaluated QuikSCAT SIR data for monitoring sea ice thermodynamic phase changes within the CAA. They found that SIR egg combined product is more appropriate because SIR slice combined product produced noisier results despite sea ice conditions within the CAA being temporally stable for most of the year. As a result we selected the QuikSCAT SIR enhanced combined pass egg image product (HH polarization) because it maximizes spatial resolution while minimizing noise for monitoring sea ice melt parameters within the CAA.

2.2. Extended AVHRR Polar Pathfinder (APP-x)

[8] The extended advanced very high resolution radiometer (AVHRR) Polar Pathfinder (APP-x; [Wang and Key, 2005]) climate data (only available until 2004) was used to validate the QuikSCAT-detected melt season parameters. Twice daily averages of SAT, incoming shortwave radiation (K[DOWNWARDS ARROW]), outgoing shortwave (K[UPWARDS ARROW]), incoming longwave (L[DOWNWARDS ARROW]), and outgoing longwave (L[UPWARDS ARROW]) data were obtained from the APP-x data set. Net all-wave radiation (Q*) was subsequently determined as the incoming minus the outgoing fluxes. The APP-x data is provided at a 25 km spatial resolution and has been validated with in situ data from the Surface Heat Budget of the Arctic Ocean (SHEBA) field experiment [Maslanik et al., 2001], Collaborative Interdisciplinary Cryospheric Experiment (C-ICE), and the Canadian Arctic Shelf Exchange Study (CASES) [Howell et al., 2006]. The APP-x data set likely provides the best available high-latitude satellite-based estimates of surface energy balance parameters with adequate areal extent and spatial resolution to represent the CAA and complement QuikSCAT-detected melt stages. However, Key et al. [1997] point out that there is still some uncertainty in satellite-derived surface radiation parameters at high latitudes. The root-mean-square error (RMSE) for SAT, K[DOWNWARDS ARROW], K[UPWARDS ARROW], L[DOWNWARDS ARROW], and L[UPWARDS ARROW] are calculated at 1.98 K, 34.4 W m−2, 22.4 W m−2, 26.6 W m−2, and 9.4 W m−2, respectively [Wang and Key, 2005].

2.3. Digital Ice Charts

[9] Regional digital ice charts provided by the Canadian Ice Service (CIS) were used to determine both FYI and MYI conditions during the QuikSCAT-detected melt season from 2000 to 2007. Digital ice charts are provided on a weekly basis by the CIS from 1968 to present (Canadian Ice Service Digital Archive Documentation Series data are available at http://ice.ec.gc.ca/IA_DOC/cisads_no_001_e.pdf). The regional digital ice charts are derived weekly, from the integration of data from a variety of sources including surface observations and aerial and satellite reconnaissance (with the primary source being RADARSAT-1 since 1996), and represent the best estimate of ice conditions on the basis of all available information at the time (Canadian Ice Service Digital Archive Documentation Series data are available at http://ice.ec.gc.ca/IA_DOC/cisads_no_001_e.pdf). Note that the period covered in this study is well within the RADARSAT-1 era. Agnew and Howell [2003] compared the CIS digital ice charts to passive microwave concentration estimates from the National Aeronautics and Space Administration (NASA) Team algorithm [Cavalieri et al., 1999]. They found the CIS digital ice charts to be a more accurate representation of ice coverage within the CAA compared to the NASA Team algorithm which can underestimate ice coverage between −7.1% and −32.6% during shoulder seasons [Agnew and Howell, 2003].

[10] The CIS regional ice charts were also used to estimate changes in the amount of FYI promoted to MYI at the end of the melt season and the amount of MYI dynamically imported into regions of the CAA. Previous studies have also used the CIS ice charts to demonstrate that the regional icescape of the CAA contains an inventory of FYI following the melt season and MYI has been dynamically imported from the Arctic Ocean [Melling, 2002; Alt et al., 2006; Howell et al., 2008]. By definition FYI is promoted to MYI on 1 October and therefore, the last week of September usually represents the amount of MYI that survived the summer melt season. The CAA is virtually immobile and landfast on 1 April and MYI dynamic import within the CAA can be estimated by taking the difference between MYI coverage on the last week of September from coverage on 1 April. Positive values represent MYI import and negative values represent MYI than has been exported or melted. These are not exact values and some uncertainty must be acknowledged. During the melt season some MYI within the CAA can be lost because of melt or export across the M'Clure Strait, Amundsen, Jones Sound, and/or Lancaster Sound boundaries. MYI export across the QEI boundaries is negligible and is almost an anomalous occurrence [Atkinson et al., 2006; Alt et al., 2006] as the Arctic Ocean pack ice only permits southward drift from the QEI during the melt season [Marko, 1977; Melling, 2002; Alt et al., 2006] As a result, positive values determined by this method represent the minimum amount of MYI imported into the CAA. This is viewed as a conservative estimate because if there is export of imported MYI across the M'Clure Strait, Jones Sound, or Lancaster Sound boundaries and/or a considerable amount of the imported MYI melted during the season, then MYI import into the CAA will be greater. MYI can also enter and move throughout the CAA as long as landfast conditions are not established and therefore, small changes in MYI can occur during the early winter months outside of the melt season [Alt et al., 2006].

[11] An approximate estimate of the amount of FYI promoted to MYI within the CAA can be calculated by taking the difference in MYI coverage across 1 October (i.e., MYI coverage the week after, minus MYI coverage the week prior). Positive values represent aging or dynamic import and negative values represent export or melting. The first uncertainty with this estimate is that MYI dynamic import or export can occur during this 1-week time window and has the potential to over or underestimate. Slight overestimation could result from QEI given the persistence of the Arctic Ocean pack ice which continually permits MYI import as long as landfast conditions are not established. More ice tends to be exported as opposed to being imported via the M'Clure Strait, Amundsen, Jones Sound and/or Lancaster Sound [Melling, 2002; Kwok, 2006] that would result in an underestimation of the amount of FYI promoted to MYI. Even with these potential uncertainties, ice motion is very slow and intermittent within the CAA and it is unlikely that considerable import/export will occur within this 1-week time window to significantly affect the net overall estimate of FYI promoted to MYI. The second uncertainty is the FYI being promoted to MYI may not have originated within CAA. Melling [2002] has shown that there is an inventory of FYI within the CAA following the melt season and as previously mentioned, more ice has been found to be exported from the CAA than imported. Given these factors it is likely that only a small amount of FYI may have originated outside the CAA but this is not easily verified. Therefore, we acknowledge that our estimate of FYI promoted to MYI within the CAA only represents the inventory of FYI found on 1 October and not how much MYI the CAA actually produces (i.e., forms in situ).

3. Melt and Freeze Detection

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[12] Sigma nought (σ°) is the scattering coefficient used to measure the ratio of returned power to transmitted power of microwave energy usually expressed in decibels and is strongly linked to the dielectric properties of the material [Ulaby et al., 1986]. The dielectric constant (ɛ*) is the mathematical construct used to quantify the dielectric properties of the material and is given by the following equation:

  • equation image

where, ɛ′ is the ability of the material to allow incident electromagnetic energy across the interface (i.e., dielectric permittivity), ɛ″ is ability of the material to dissipate penetrated energy (i.e., dielectric loss), and j = (−1)0.5. Permittivity describes what happens to electromagnetic energy when it impinges upon a boundary and loss describes the electromagnetic loss once energy has penetrated the material. This total loss is a combination of the absorption loss (i.e., the transformation of energy into another form) and scattering loss (energy deflected to travel in directions other than incident) [Hallikanien and Winebrenner, 1992]. The presence of water in liquid phase within the snow and sea ice cover significantly affects the dielectric properties (and ultimately the microwave backscatter coefficient, σ°) of the snow and sea ice system during both the melting and freezing processes [Ulaby et al., 1986; Winebrenner et al., 1994, 1996; Smith, 1998a; Kwok, 2004]. Large changes in σ° from winter conditions can therefore be exploited to detect MO and FO dates, and the duration of the melt season can be estimated by taking the difference between the dates.

[13] Using the CIS digital ice charts, six FYI test sites in 2000 and six MYI test sites in 2004 were selected to illustrate the algorithm's applicability for detecting MO and FO (Figure 1). The lack of in situ observations of MO and FO dates makes validation difficult especially when estimates are required at regional or global scales. Variability in MO and FO dates is not uncommon between algorithms because of differences in methodology and/or the sensor used. With respect to our QuikSCAT algorithm, while the spatial resolution of QuikSCAT SIR is improved over passive microwave data, ambiguities in the σ° evolution with respect to varying ice types at sub pixel resolution can still be present. Following Howell et al. [2006] we only apply the algorithm to homogenous FYI σ° < −18 dB and MYI σ° > −11 dB pixels established during the winter seasons (i.e., April for MO and December for FO) and then apply a kriging interpolation routine to create continuous MO and FO surfaces. Also, dynamic ice movement may cause erroneous MO and FO detection near the boundary between the Arctic Ocean polar pack and the CAA for the M'Clure and Amundsen regions as well as between Baffin Bay and the Eastern Parry Channel. The small channel width and persistent landfast conditions make this problem negligible for the Arctic Ocean-QEI boundary.

3.1. Melt Onset

[14] As the snow-ice interface temperature on FYI approaches −5°C, the dielectric constant increases with increasing brine volume and snow basal layer grain size [Assur, 1958] which creates a substantial increase in volume scattering which then causes an increase to σ° [Barber and Nghiem, 1999]. For MYI, which has negligible salinity and is essentially brine free, the liquid water increase within the snow cover causes significant microwave absorption which masks volume scattering from the hummock ice layer, thus decreasing σ° [Winebrenner et al., 1994; Barber et al., 1995]. This upturn (FYI) and downturn (MYI) is clearly present at the test sites (Figures 2 and 3) and also corresponds to increases in SAT. The RMSE of APP-x SAT is 1.98 K which can result in SAT at MO ranging between −4.3°C and −7.7°C compared to in situ values [Howell et al., 2006]. In situ SAT values near −5°C for both FYI and MYI at MO within the CAA have been observed by Barber et al. [1995]. MO was estimated for FYI and MYI using a threshold of absolute change in σ° of greater than 2 dB from stable winter conditions. This threshold was based on numerous σ° temporal evolution sites within the CAA and is deemed to be a representative MO threshold within the CAA [Howell et al., 2005, 2006]. Changing the threshold by ±1 dB typically results in a change in the date of MO by only 1 or 2 days, which is small compared to the year-to-year variability in the date of MO.

image

Figure 2. Temporal evolution of QuikSCAT σ° observed at Peel Sound, Amundsen 1, Western Arctic Waterway 1, Western Artic Waterway 2, QEI 1, and QEI 2 FYI test sites in 2000. QuikSCAT σ°-detected melt onset using the 2 dB threshold is indicated by the vertical line, QuikSCAT σ°-detected freeze onset using the 8 dB is indicated by the second vertical line, and the gray curve is temperature (right scale). Test site locations are shown on Figure 1.

Download figure to PowerPoint

image

Figure 3. Temporal evolution of QuikSCAT σ° observed at Viscount-Melville, M’Clure, QEI 1, QEI 2, M’Clintock 1, and M’Clintock 2 MYI test sites in 2004. QuikSCAT σ°-detected melt onset using the 2 dB is indicated by the vertical line, QuikSCAT σ° freeze onset using the 5 dB threshold is indicated by the second vertical line, and the gray curve is temperature (right scale). Test site locations are shown on Figure 1.

Download figure to PowerPoint

3.2. Freeze Onset

[15] FO was estimated by reversing the chronology of MO detection to identify the significant upturn in σ° from an open water surface (i.e., formation of thin FYI) and/or melt ponded surface (i.e., MYI). For FYI at C-band, the high dielectric constant of ocean water causes variable scattering as a function of wind speed [Barber and Yackel, 1999] and when ice starts to form (i.e., water changes from a liquid to a solid) the scattering begins to stabilize and increase. This process is also observed at Ku-band (Figure 2). Unfortunately, in situ SAT data during FO is unavailable to evaluate APP-x during this time but decreasing trends in SAT are present for all FYI sites at QuikSCAT FO detection (Figure 2).

[16] The freezeup process of FYI is complex and variable, which poses some challenges for the detection of FO from σ° changes. These effects are particularly problematic because the FO algorithm works in a reverse chronology. For example, newly formed young ice can have a high σ° as a result of high dielectric frost flower structures on the surface or a low σ° as a result of specular surface scattering from grease ice that also reduces wind roughening and Bragg surface scattering of the ocean surface [Tucker et al., 1992; Grenfell et al., 1998]. The extensive growth of frost flowers may cause FO estimates to be detected a few days later than in the absence of frost flower growth because frost flower crystals appear very rough at the small Ku-band wavelength [Drinkwater and Crocker, 1988]. FO can also be detected late if wind-roughened leads or ice rafts and ridges are present, all of which act to increase σ°. On the basis of these potential influences, we evaluated σ° change thresholds from 4 dB to 10 dB in intervals of 2 dB in order to evaluate their sensitivity to FO dates and thus select a representative transition threshold for FO over FYI within the CAA.

[17] From Figures 4 and 5, all sites except QEI 1, Amundsen 2, and Jones Sound, the 4 dB threshold detected FO dates very late (Table 1) and was unable to skip over high σ° values caused by the presence of either frost flowers, ridges, and/or wind roughened leads. The 6 dB threshold performed slightly better but is still not able to skip over late season high σ° values and fall within the initial upturn region (Figures 4 and 5). As a result, FO was still detected too late within the QEI 2, Western Arctic Waterway 1, Western Arctic Waterway 2, Eastern Parry Channel, and Peel Sound using the 6 dB threshold (Table 1). The 8 dB and 10 dB thresholds seem to perform the best and for the most part fall during the upturn from open water conditions (Table 1 and Figures 4 and 5). No detection was possible using the 10 dB over Peel Sound and FO seems to occur too early in some regions (e.g., Western Arctic Waterway 2 and Jones Sound) which may be the result of open water wind roughening rather than FO. As a result, we selected the change threshold of 8 dB from winter conditions to represent FO detection. This threshold is high enough such that FO is more likely to fall during the upturn from an open water surface (Figures 4 and 5). Despite this high threshold, it is acknowledged that some FYI pixels may still exhibit false FO detection because of frost flowers, ridges, and/or wind roughened leads (e.g., Peel Sound) resulting in slightly longer melt durations.

image

Figure 4. Temporal evolution of QuikSCAT σ° observed at Peel Sound, Amundsen 1, Western Arctic Waterway 1, Western Artic Waterway 2, QEI 1, and QEI 2 FYI test sites in 2000 specifically during freeze onset. QuikSCAT σ°-detected freeze onset using the 8 dB threshold is indicated by the solid vertical line, and the gray curve is temperature (right scale). The dashed vertical lines are freeze onset dates corresponding to different backscatter thresholds. Test site locations are shown on Figure 1.

Download figure to PowerPoint

image

Figure 5. Same as Figure 4 except for Amundsen 2, Eastern Parry Channel, Prince Regent, and Jones Sound.

Download figure to PowerPoint

Table 1. Summary of Freeze Onset as Detected by QuikSCAT Over First-Year Ice and Multiyear Ice for Varying σ° Change Thresholdsa
RegionFreeze Onset Dates Based on Change Threshold Over FYI
10 dB8 dB6 dB4 dB
  • a

    Corresponding spatial locations are shown on Figure 1. FYI, first-year ice; MYI, multiyear ice.

Peel SoundNo Detection290295297
Amundsen 1290291297335
Amundsen 2280285291291
Western Arctic Waterway 1285285303317
Western Arctic Waterway 2275290296311
Queen Elizabeth Islands 1262262262263
Queen Elizabeth Islands 2261262285288
Eastern Parry Channel262268299301
Prince Regent277278278325
Jones Sound262278279279
 Freeze Onset Dates Based on Change Threshold Over MYI
Region9 dB7 dB5 dB3 dB
Viscount-Melville241247247247
M'ClureNo Detection243243243
Queen Elizabeth Islands 1234241243243
Queen Elizabeth Islands 2234234243243
M'Clintock 1258259259259
M'Clintock 2258268269272

[18] During the FO process over MYI, the transition from liquid water in melt ponds and the snowpack reduces surface scattering at C-Band and allows volume scattering from the hummocky surface to increase significantly [Winebrenner et al., 1996]. The same mechanisms also manifest at Ku-band (Figure 3). For MYI, FO is defined from a significant upturn in σ° from a newly refrozen melt ponded surface. Contrary to FYI, wind roughened leads and ridging do not present a significant problem for MYI where a very smooth transition to winter conditions is apparent (Figure 3). Smaller σ° change thresholds ranging from 3 dB to 9 dB in intervals of 2 dB were evaluated in order to select a representative transition threshold for FO over MYI. Less variability in FO dates as a result of changing σ° thresholds is present over MYI and only the 9 dB seems to be inappropriate for FO as evident by no detection within the M'Clure region (Table 1). We selected the 5 dB because, although it is similar to 3 dB, it is likely more representative than the 7 dB and 9 dB. For example, the 5 dB change threshold corresponds more to the down trend in SAT that is observable at all sites (Figure 3), whereas the 7 dB and 9 dB change threshold detects FO just prior to this downturn especially over the QEI 2 (not shown).

4. Results and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

4.1. Spatiotemporal Distribution of Melt Onset, Freeze Onset, and Melt Duration

[19] The date of MO, FO and MD length varied among and between regions of the CAA from 2000 to 2007. Within the CAA the average date of MO occurred on day 150, the average FO occurred on day 266, and 116 was the average number of days of melt (Table 2). On average, MD was the in shortest in the CAA at 103 days in 2004 and the longest in 2006 at 136 days (Table 2). Air temperature over Resolute Bay, Nunavut located in the central CAA (Figure 1) that has previously been used as a proxy indicator of the climate regime of the CAA [e.g., Falkingham et al., 2001; Atkinson et al., 2006] shows that 2004 experienced the coldest temperatures and 2006 experienced the warmest temperatures in almost every month from 2000 to 2007 (Table 3).

Table 2. Mean Year Day of Sea Ice Melt Onset, Freeze Onset, and Melt Duration in the Canadian Arctic Archipelago as Detected by the QuikSCAT σ° for 2000–2007a
YearMelt Onset (day)Freeze Onset (day)Melt Duration (days)
  • a

    Standard deviation is in parentheses.

2000153 (13.4)260 (20)107 (28.2)
2001148 (11.9)266 (19)118 (27.2)
2002152 (8.8)267 (14.3)115 (19.5)
2003152 (13.1)264 (20.1)112 (29.0)
2004154 (11.5)257 (16.3)103 (23.4)
2005145 (9.7)264 (16.4)119 (22.1)
2006141 (9.9)277 (21)136 (26)
2007154(12.6)271 (19.1)117(28.0)
Mean150(8.3)266 (14.3)116(21.6)
Table 3. Monthly Air Temperatures From the Environment Canada Meteorological Weather Station Located at Resolute Bay, Nunavut for 2000–2007a
Month20002001200220032004200520062007
Jan−33−31.5−30.6−27.4−33.9−28.4−30.3−31.5
Feb−32.2−31.3−34.1−34.4−33.1−31.6−26.3−30.5
Mar−25.1−28.7−30.6−29.7−34.2−29.3−25.2−34.2
Apr−21−22.7−23.6−21.8−22.1−22−19.4−19.9
May−12−11.1−10−9−11.2−9−6−13.5
Jun0.20.60.8−1−1.60.4−0.21.3
Jul4.44.82.552.23.94.47.4
Aug0.81.61.90.81.63.844.9
Sep−4.3−3.4−2.2−3.7−5.7−3.3−2−3.1
Oct−15−16.3−10.3−11−12−11.8−6−12.3
Nov−23.3−24.2−23−20.5−23.6−19.7−20.6−25.4
Dec−29.3−22.4−23.7−27.1−32.3−28.8−26.9−27.7

[20] Spatially, the average MO date occurred first and average FO date occurred last within the Amundsen, Western Arctic Waterway and Eastern Parry Channel regions (Figures 6 and 7) . The QEI regions were the last to experience MO (Figure 6) but the first to experience FO (Figure 7). As a result, the QEI experienced shorter MDs between 79 and 110 days, compared to longer MDs in the more southerly regions of the CAA from 2000 to 2007 (Figure 8). Both Smith [1998b] and Belchansky et al. [2004b] found that the shortest sea ice MDs in the Arctic occur immediately north of the QEI. The M'Clure and Viscount-Melville regions also experienced shorter average FO dates (Figure 7) as well as reduced MDs (Figure 8). Regions where reoccurring polynyas can form early in the season (i.e., Eastern Parry Channel, Amundsen, and Jones Sound) typically experienced longer MDs (Figure 8). These observations are in agreement with the results of Stroeve et al. [2006].

image

Figure 6. QuikSCAT σ°-detected melt onset in the Canadian Arctic Archipelago, 2000–2007. Legend is in year day.

Download figure to PowerPoint

image

Figure 7. QuikSCAT σ°-detected freeze onset in the Canadian Arctic Archipelago, 2000–2007. Legend is in year day.

Download figure to PowerPoint

image

Figure 8. QuikSCAT σ°-detected melt duration in the Canadian Arctic Archipelago, 2000–2007. Legend is in days.

Download figure to PowerPoint

[21] MD from 2000 to 2002 exhibited a similar spatial distribution but MD was shortest for the QEI in 2000 (Table 2 and Figure 8). The QEI region experienced longer MDs for 2001–2002 (91 to 99 days) compared to a shorter MD observed in 2000 (79 days). Although MO occurred later than most other years it is the early FO date (day 239) observed in 2000 that is largely responsible for the shorter MD. Alt et al. [2006] reported that FO occurred early in 2000 for the high-latitude regions of QEI because of northerly atmospheric flow off the Arctic Ocean.

[22] MD for 2003 was shorter than previous years within the QEI, M'Clure, Viscount-Melville, and Peel Sound regions at 94, 110, 107, and 121 days, respectively. While MO timing was fairly similar compared to previous years, FO dates are noticeably earlier (Figure 7). MO occurred later than average for most regions within the CAA for 2004 (Figure 6) and FO also occurred earlier than previous years (Figure 7). Temperatures were lower for 2004 compared to 2000–2003 for almost every month (Table 3).

[23] The year 2005 had the second longest MD within the CAA at 119 days (Table 2) and experienced a more homogenous spatial distribution for MO and MD compared to previous years (Figures 6 and 8) whereas the FO was early and similar to 2004 (Figure 7). From 1979 to 2005, Stroeve et al. [2006] found that in 2005 the Arctic experienced its longest melt season. The year 2006 had an even longer MD (136 days) compared to 2005 and temperatures during the melt season for 2006 were warmer than 2005, especially during the shoulder seasons within the CAA (Table 3).

[24] The year 2007 experienced a later than average MO date (154) and the second latest FO date resulting in a just slightly longer than average MD of 117 days. Temperatures were low during the start and end of the melt season but compared to previous years higher temperatures were found during the peak melt months of July and August (Table 3). Comiso et al. [2008] found anomalously warmer temperatures over the central Arctic for 2007 during these months.

[25] Overall the CAA has experienced considerable spatiotemporal variability in melt season parameters over the past 8 years. It has previously been observed that the CAA has experienced MYI increases from 2000 to 2004 followed by decreases in 2005–2006 [Howell et al., 2008]. MYI has different physical and thermodynamic properties than seasonal FYI [Maykut and Untersteiner, 1971] and we would expect changes in MYI to be reflected in melt season parameters. Moreover, despite the 2007 extreme record low sea ice in the Arctic, 1998 still represents record low sea ice conditions within the CAA. It therefore seems important to examine the linkages between melt season parameters in relation to MYI within the CAA.

4.2. Can MYI Within the Canadian Arctic Archipelago Survive Longer Melt Seasons?

[26] Figure 9 depicts the MD in relation to the observed MYI coverage on the last week of September (i.e., the amount that survived the melt season), the amount of FYI promoted to MYI, and the amount of MYI dynamically imported within the CAA from 2000 to 2007. Indeed, the CAA has accumulated increased amounts of MYI from both FYI promotion and dynamic import from 2000 to 2004. The increases are related to the recovery from record minimum MYI conditions in 1998 caused by an anomalous warming event within the CAA [Atkinson et al., 2006; Howell et al., 2008]. FYI to MYI promotion accounted for slightly more of the increases compared to MYI dynamic import (Figure 9).

image

Figure 9. Bar graph illustrating multiyear ice coverage on the last week of September, first-year ice promoted to multiyear ice, multiyear ice dynamically imported, and melt duration within the Canadian Arctic Archipelago from 2000 to 2007.

Download figure to PowerPoint

[27] Longer MDs are apparent from 2005 to 2007 and both MYI dynamic import and FYI promotion almost “shut down” but the resulting MYI decreases being a function of a longer MD are only strongly apparent in 2006 (Figure 9). The spatial distribution of MYI on the first week of April (i.e., immobile landfast conditions) and the last week of September are shown in Figures 10 and 11, respectively. On average, high concentrations of MYI are present in the QEI and the M'Clure regions that extend to Viscount-Melville and further south to the M'Cintock and Larsen Sound regions for 2005–2006, despite these regions experiencing longer MDs (Figures 8, 10, and 11). On the basis of these observations, it seems apparent that MYI can survive longer melt seasons even in the southern regions without replenishment. This raises the more compelling question of how this process is realized.

image

Figure 10. Spatial distribution of multiyear ice concentration in the Canadian Arctic Archipelago on the first week of April from 2000 to 2007. Legend is ice concentration in tenths.

Download figure to PowerPoint

image

Figure 11. Spatial distribution of multiyear ice concentration in the Canadian Arctic Archipelago on the last week of September from 2000 to 2007. Legend is ice concentration in tenths.

Download figure to PowerPoint

[28] The answer lies in the relationship between the increased thickness of MYI and the oceanic heat flux. Specifically, surface ablation from solar radiation cannot be the dominant heat source that ablates MYI in the CAA. Instead, it must slowly ablate from the underside, even as it slowly migrates to the southern regions of the CAA. Although only focused within the QEI region, Melling [2002] hypothesized this process by demonstrating that for MYI within the QEI the small 5–15 W m2 oceanic heat flux is responsible for ablating the MYI. MYI is very thick within the QEI (3.4 m on average) and its coverage (and concentration) remains high during the summer melt season. Therefore, the ocean is not likely to be warmed as much by solar radiation and accelerate the ablation process. Maykut and McPhee [1995] found that most of the ocean heat flux is solar energy deposited through leads (88%) and only 12% was transmitted through MYI less than 1.8 m thick. Higher-magnitude values were reported by Perovich [2005] that was attributed to a thinner (1.58 m) MYI cover. Thicker MYI results in less solar energy input to the sea ice–ocean system and this shallower energy input slows the positive sea ice–albedo feedback [Perovich, 2005]. As a result, even if MO occurs early within the CAA, atmospheric forcing from solar radiation is still likely insufficient to significantly ablate the MYI cover.

[29] Looking at the time series of MYI coverage in the QEI along with the APP-x derived Q* from 2000 to 2004 illustrates that MYI experiences only subtle decreases during the peak times of the melt season whereas most of the FYI completely ablates (Figure 12). Even the longer melt season in the QEI for 2002 resulted in minimal MYI losses during the melt season (Figures 8 and 12). An examination of MYI conditions from 2000 to 2004 within Viscount-Melville, M'Clintock, and Larsen Sound regions that are outside of the QEI reveals that a similar process occurs as MYI migrates southward. The aforementioned regions contained a mix of MYI and FYI with MYI eventually dominating the Viscount-Melville and M'Clintock icescape in 2003 (Figures 13 and 14). Within these regions MYI remains relatively stable but the FYI completely ablates and MYI increases typically occurred following FYI reductions (Figures 13, 14, and 15). While some of this MYI was FYI promoted to MYI a considerable portion has in fact slowly migrated from the QEI (Figures 13, 14, 15, and 16). This loss of MYI in the QEI and subsequent increase in Viscount-Melville for 2001 and 2002 (which also exhibited a long MD in the QEI) are evidence of this (Figures 12 and 13). The increases were not from the Arctic Ocean via the M'Clure Strait because Kwok [2006], using tracked RADARSAT-1 data, found that no MYI import occurred for 2001 and 2002. The southward migration of MYI from the QEI through the southern channels (i.e., M'Clintock, Peel Sound, and Baffin Inlet) of the CAA was particularly evident for 2003 and 2004 (Figure 16). Both the M'Clintock and Larsen Sound regions exhibited MYI increases and decreases when nonlandfast conditions persist, providing more evidence for this dynamic process (Figures 14 and 15). The ocean heat flux also depends on latitude such that a lower solar angle will increase the ocean heat flux [Krishfield and Perovich, 2005] and as a result greater melt is expected in regions south of the QEI, especially in Larsen Sound. Moreover, thinner MYI is expected as it migrates southward which results in more solar radiation transmitted through the ice, thus increasing the rate of under ice ablation. Nevertheless, MYI has slowly migrated to southerly regions of the CAA from 2000 to 2004 and has not completely ablated while in transit.

image

Figure 12. Time series of multiyear ice coverage, first-year ice coverage, and net all wave radiation (Q*) within the Queen Elizabeth Island region for 2000–2004. QuikSCAT σ°-detected melt onset is indicated by the first vertical dashed line, and freeze onset is indicated by the second vertical dashed line. The bold vertical dashed line indicates the date (1 October) when first-year ice is promoted to multiyear ice.

Download figure to PowerPoint

image

Figure 13. Same as Figure 12 except for the Viscount-Melville Sound region.

Download figure to PowerPoint

image

Figure 14. Same as Figure 12 except for the M’Clintock region.

Download figure to PowerPoint

image

Figure 15. Same as Figure 12 except for the Larsen Sound region.

Download figure to PowerPoint

image

Figure 16. Spatial distribution of multiyear ice dynamic movement within the Canadian Arctic Archipelago from 2000–2007. A negative value represents dynamic export or melt, and a positive value indicates dynamic import. Legend is concentration in tenths.

Download figure to PowerPoint

[30] More compelling evidence for the oceanic heat flux slowly ablating MYI within the CAA can be found by looking at MYI conditions following the longer melt season of 2005–2006. During this time period, regions of the CAA are not subject to significant replenishment from dynamic import and/or FYI promotion, yet considerable MYI was still present following the melt season (Figure 11). The year 2005 exhibited a long MD of 119 days (Table 2) but this was probably attributed more to the early MO because FO was similar to previous years, especially in regions of high concentrations of MYI (Figures 7 and 11). Specifically, temperatures were warmer during August and September for 2005 (Table 3) but the FO dates over the MYI regions were similar to 2004, which was a much colder year (Figures 7 and 11). This supports the notion that even with earlier MO dates the influence of solar radiation is reduced over MYI because early FO transitions (hence slower sea ice albedo feedback) still have occurred over the thicker MYI. MYI losses are apparent within the QEI for 2005 (Figure 17) and the Viscount-Melville, M'Clintock, and Larsen Sound regions also all experience losses of MYI but remnant amounts still remain following the melt season (Figures 18, 19, and 20). Slight MYI increases in the northern part of Viscount-Melville were from the QEI which exhibited some southward export of MYI (Figures 16, 17, and 18). Almost no positive dynamic import occurred in the M'Clintock and Larsen Sound regions (Figures 16, 19, and 20).

image

Figure 17. Time series of multiyear ice coverage within the Queen Elizabeth Island region for 2005–2007. QuikSCAT σ°-detected melt onset is indicated by the first vertical dashed line, and freeze onset is indicated by the second vertical dashed line. The bold vertical dashed line indicates the date (1 October) when first-year ice is promoted to multiyear ice.

Download figure to PowerPoint

image

Figure 18. Same as Figure 17 except for the Viscount Melville Sound region.

Download figure to PowerPoint

image

Figure 19. Same as Figure 17 except for the M’Clintock region.

Download figure to PowerPoint

image

Figure 20. Same as Figure 17 except for the Larsen Sound.

Download figure to PowerPoint

[31] In 2006, on average the CAA experienced both the latest mean FO date (277) and longest MD (136) of the years considered (Table 2). Spatially, the southerly regions contained appreciable amounts of MYI at the onset of the 2006 melt season (Figure 10) and MO began earlier in 2006 compared to 2005. The long MD of 2006, especially in regions south of the QEI, is a combination of both an early MO and later FO whereas the long MD in 2005 is primarily attributed to an earlier MO. The longer MD for 2006 was a function of warmer temperatures (Table 3) but was also influenced by MYI in the southern regions experiencing net ablation for over 2 years with almost zero replenishment. This reduced MYI concentration and thickness allowed surface heating to have more of an impact which eventually facilitated complete ablation following the 2006 melt season (Figure 11). Notice how the QEI experienced MYI increases for 2006, but very little MYI was exported to the southern regions (Figure 17). Subtle increases of MYI are observed in Viscount-Melville during 2006 that can be attributed to small export from the QEI and M'Clure regions (Figures 16 and 18). The slight MYI increases in M'Clintock and Larsen Sound regions during the season represent MYI drifting southward from Viscount-Melville (Figures 19 and 20). Examining the situation within the CAA for 2007 finds that it experienced a similar MD compared to 2005 at 117 days but 2007 began the melt season with small concentrations of MYI in the M'Clure, Viscount-Melville, and M'Clintock regions (Figure 10). Interestingly, although no net positive FYI to MYI promotion occurred, the CAA once again exhibited a positive net dynamic import of MYI for the first time since 2004 (Figure 9).

[32] Very little replenishment of MYI has occurred in the M'Clure, Viscount-Melville, M'Clintock, Larsen Sound, and Baffin Inlet regions of the CAA since 2004. However, the increased thickness of MYI only permitted its slow ablation from the underside, even as it migrated southward. Despite the longer melt seasons for 2005 and 2006 and little MYI replenishment, considerable amounts of MYI survived. Therefore MYI can survive long melt seasons even in the southern regions of the CAA. This now brings us to the question of whether or not MYI decreases are expected to continue within the CAA with the advent of longer melt seasons.

4.3. Will MYI in the CAA Continue to Decrease?

[33] The longer melt seasons of 2005–2007 within in the CAA seem to be responsible for reducing the amount of FYI promoted to MYI compared to 2000–2004 (Figure 9). Should this trend continue, MYI conditions in the CAA should reduce especially in the southern regions, however, MYI dynamic import is likely to play an important role. The regional icescape of the CAA was heavily congested with MYI at the onset of the 2005 melt season (Figure 10), thus restricting dynamic ice motion. Note that the spatial distribution of MYI at the start and end of the 2005 melt season changed very little (especially in the southern regions) (Figures 10 and 11). For southward drift from the QEI to occur there must be leeway in Viscount-Melville or in Larsen Sound and blockage in Viscount-Melville or Larsen Sound also restrict import from the Arctic Ocean via the M'Clure Strait [Marko, 1977]. No net dynamic import of MYI occurred for 2005 because Viscount-Melville and southern Larsen Sound were heavily congested with MYI, limiting free drift conditions thus preventing intrusions of MYI (Figure 10). This likely even prevented import from the Arctic Ocean into the QEI, as 2005 was the only year that the QEI exhibited a significant MYI deficit (Figures 12 and 17).

[34] The short MDs observed within the QEI restrict MYI mobility to only a few months a year and therefore it can take several years for MYI to migrate through the QEI into the southern regions of the CAA. Historically speaking (dating back to the early 1960s), following low-MYI years there has always been a 2–5 year period of MYI recovery within the CAA [Melling, 2002; Atkinson et al., 2006; Howell et al., 2008]. Large-scale sea ice dynamics continuously force MYI up against the QEI, and until this process stops, thick MYI from the Arctic Ocean can still flow through the regions of the QEI and then into more southern regions of the CAA. If the reduced MYI concentrations at the end of 2006 in the southern regions of the CAA can be interpreted as part of this process, then movement should begin to reinitialize within the next few years. Witness the QEI region exhibited MYI losses (mostly attributed to melt) in 2005 but considerable dynamic import occurred in 2006 (Figure 17). This suggests that the losses of 2005 provided room for MYI to enter the QEI in 2006. Subsequently in 2007 MYI was exported from the QEI and began to accumulate once again in the Viscount-Melville following the melt season (Figures 11 and 16) which contributed to a net dynamic MYI increase within the CAA (Figure 9). The cyclic nature of MYI within the CAA will therefore still facilitate its presence despite longer melt seasons.

5. Summary and Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[35] Estimates of MO, FO and MD from QuikSCAT were made annually within the CAA from 2000 to 2007. The mean date of MO occurred on day 150, the mean FO occurred on day 266, and 116 was the average number of days of melt. We have found considerable spatial variability in melt season parameters for the CAA from 2000 to 2007. On average, the more heavily concentrated MYI regions of the QEI experienced shorter MDs followed by the M'Clure, Viscount-Melville and M'Clintock regions. These MYI regions typically experienced earlier FO dates hence shorter melt seasons because MYI thickness results in less energy absorbed (hence, easier to extract) into the ocean–sea ice system. The seasonal FYI regions of the Amundsen, Western Arctic Waterway and Eastern Parry Channel regions experienced longer MDs.

[36] Changes in melt season parameters and MYI conditions within the CAA were also observed from 2000 to 2007. While MYI increases in the CAA occurred from 2000 to 2004 as a result of dynamic import and FYI promotion, this replenishment virtually stopped from 2005 to 2007 coincident with longer MDs. However, it took two consecutive, long melt seasons (2005–2006) to completely remove the MYI from the southern regions of the CAA. We argued that this must be because MYI in the southern regions of the CAA is subject to ablation predominantly from the small oceanic heat flux during summer melt season. It seems apparent that once MYI enters the channels south of the QEI, its may remain for several years following. This suggests that ship navigation through the Northwest Passage can still be compromised even in the absence of seasonal replenishment. It is also noteworthy to mention that very little FYI was promoted to MYI and/or was dynamically imported within the Northwest Passage since 2004. Hence, the low amounts of MYI in the Northwest Passage for 2006 and its temporary clearing in 2007 while significant, was perhaps not unexpected.

[37] The heavy MYI conditions within the CAA are undoubtedly attributed to persistence of MYI from either dynamic import and/or FYI promotion. Indeed a longer melt season will reduce the amount of FYI promoted to MYI, especially within the southern channels of the CAA, as shown in this study. A longer melt season will likely not reduce MYI dynamic import, but it may decrease the thickness of the MYI, promoting increased ablation where it can no longer survive for several years in the southern channels. However, one cannot discount Melling's [2002] suggestion that the weakening of the ice barriers within the QEI may increase MYI flushing events and the MYI that reaches these southern channels could actually be thicker and perhaps survive longer because of less seasonal ablation. The CAA exhibited net positive dynamic import in 2007 for the first time since 2004, and the CAA is conditioned to facilitate more dynamic import of MYI in the upcoming years. This will provide us with an opportunity to address this uncertainty.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[38] The authors wish to thank Steve McCourt (CIS) for providing the CIS digital ice chart data. This research was supported by the Natural Sciences and Engineering Research Council (NSERC) Post Doctoral Fellowship to S. Howell and NSERC Discovery grants to J. Yackel and C. Duguay. We also wish to thank Harry Stern, Bruno Tremblay, and the two anonymous reviewers who made this paper better.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Melt and Freeze Detection
  6. 4. Results and Discussion
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
jgrc11006-sup-0001-t01.txtplain text document1KTab-delimited Table 1.
jgrc11006-sup-0002-t02.txtplain text document1KTab-delimited Table 2.
jgrc11006-sup-0003-t03.txtplain text document1KTab-delimited Table 3.

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.