the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Estimating seasonal bulk density of level sea ice using the data derived from in situ and ICESat-2 synergistic observations during MOSAiC
Abstract. Satellite retrievals of Arctic sea ice thickness typically assume fixed values of sea ice bulk density (IBD), overlooking its seasonal evolution and spatial heterogeneity, which are influenced by factors such as the age, deformation, brine, and air inclusions of the sea ice. This study investigates the seasonal variability of IBD during the Arctic freezing season from October to April, across the Distributed Network (DN) scale of the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition. To estimate IBD, we combined sea ice and snow observations from ice mass balance buoys, snow pits, repeated transects, and ice cores, together with high-resolution along-track freeboard data obtained from airborne laser scanning (ALS) and the Ice, Cloud, and land Elevation Satellite-2 (ICESat-2). Assuming hydrostatic equilibrium, IBDs were determined for the level ice components of the MOSAiC ice floes, which consisted predominantly of second-year ice (SYI). Our results revealed significant seasonal variability of the IBD with two main phases during the MOSAiC freezing season at scales of DN (~ 50 km), L-sites (~ 25 km), and Main Coring Site (MCS, ~ 50 m). Throughout the freezing season, the mean IBD estimated at the DN scale (910 ± 7 kg m−3) was close to that of the SYI cores at the MCS (912 ± 2 kg m−3), highlighting the SYI-dominated regional ice properties. We also identified that sea ice freeboard, along with the ratios of ice freeboard to total freeboard or ice freeboard to thickness, are critical indicators to determine IBD at the scale of tens of kilometers. We have therefore developed parameterizations for IBD that are expected to be applicable throughout the freezing season for the SYI region, which is also the ice type that currently dominates the central Arctic Ocean. The proposed parameterizations have the potential to optimize basin-scale IBD estimation and improve satellite-derived sea ice thickness.
- Preprint
(9694 KB) - Metadata XML
-
Supplement
(1646 KB) - BibTeX
- EndNote
Status: open (until 06 Nov 2024)
-
RC1: 'Comment on egusphere-2024-2821', Anonymous Referee #1, 02 Oct 2024
reply
The study titled “Estimating seasonal bulk density of level sea ice using the data derived from in situ and ICESat-2 synergistic observations during MOSAiC” integrates observations of snow and ice thickness from 15 ice mass balance buoys, along with snow freeboard data from ICESat-2, airborne laser altimeters, and snow density data to estimate bulk ice density for a drifting ice floe and its surroundings at various scales. It also compares these estimates with manual observations of ice density and proposes several parameterizations, primarily using ice freeboard and thickness as key parameters.
Parametrization of ice density is a valuable task, and the study provides new data and implements known concepts to derive density estimates from large scale observations. The methods are clearly described, yet some details do not allow a reproducibility. The overview of remote methods is accurate enough, while an overview of in situ measurements is quite limited. In addition, there are a lot of inconsistencies related to the description of representative ice types and ice and snow thicknesses, related to a limited usage of existing in situ observations during the described expedition.
While the study offers valuable insights into the upscaling of sea ice freeboard data during the ice drift, where measurements were limited to individual floes, there are some methodological challenges that should be considered. For instance, the ice density estimates rely on a methodology with certain limitations. Some of these issues were addressed by manually adjusting the freeboard measurements, although the rationale and approach behind this adjustment could be more clearly explained. While these adjustments resulted in more realistic density estimates, there remains significant variability in the data, similar to the previously observed seasonal fluctuations in ice density.
Given this level of variability and the presence of an adjustment parameter that heavily influences both the density values and their seasonality, it is challenging to recommend a robust parameterization based on the current data. Additionally, the assumption that the adjustment is time-independent may need further justification. As a result, it becomes difficult to fully assess the quality of the density estimates due to both the high variability and the manual adjustments made during the analysis.General comments
A parameterization of sea-ice density can indeed be a valuable tool for both modelers and observers. However, I have some concerns regarding the potential usefulness of the parameterization presented here for such applications. Firstly, it is based on a methodology with significant uncertainties, which in many cases seem to exceed the actual seasonal variability of sea-ice density. Secondly, the current parameterization focuses on correlations between density and ice freeboard or thickness, without addressing important physical factors such as region, ice salinity, ice temperature, or water temperature. Since ice freeboard is inherently linked to its density, this creates a positive feedback loop. While snow may also play a role, it is less likely to directly influence ice density. Low ice freeboard can result from both high ice density and reduced thickness, complicating the interpretation. Similarly, the relationship between ice thickness and density involves multiple factors, including the region, ice age, and type, in addition to seasonal variations. Suggesting that only very thin ice has lower density seems to oversimplify the complexity of these processes. Additionally, it may be less useful to rely on ice thickness as an input parameter, as this is typically the primary goal of many remote sensing estimates rather than an available input variable.
Another significant concern regarding the methodology of this study involves the so-called “spatial scale adjustment.” First, the issue with buoys is not solely about spatial coverage or the number of buoys deployed. Rather, it arises from the fact that the buoys were installed in ice that may not be representative of the broader area. Secondly, such a substantial adjustment requires a more detailed explanation than a brief mention with a reference to supplementary materials. The adjustment of 7 cm to the freeboard, which accounts for roughly a third of the average total freeboard during the observation period, is quite significant. According to the methodology, ice freeboard is calculated as the difference between ICESat-2 snow freeboard and buoy snow thickness. Since the buoy snow thickness remains unchanged, applying a 7 cm adjustment equates to adjusting the ice freeboard by 7 cm, which corresponds to approximately 65 cm of ice thickness. This can be shown considering the initial buoy ice thickness was around 100 cm, while the modal ice thickness from electromagnetic sounding was between 40-50 cm, giving a similar difference in ice thickness, as the adjustment equivalent. Such a considerable adjustment to the interpretation of ice thickness, which ultimately impacts the density estimates, warrants a more thorough explanation of both its value and physical meaning. Additionally, it is perplexing that instead of adjusting the relatively uncertain estimates of snow and ice thickness from the buoys, the ICESat-2 and airborne laser scanner data were adjusted instead and presented in this adjusted form. This 7 cm adjustment was applied across the entire seven-month observation period, but it is not clear why this adjustment was considered time-independent. If the buoys were installed in areas with unrepresentatively thick ice, it seems unlikely that this bias in ice thickness would remain constant throughout the ice growth season. Numerous observations indicate that by the end of the cold season, initially thinner ice can grow more rapidly due to thinner snow cover, making the differences between thick and thin ice less distinguishable. Furthermore, another issue arises from the fact that buoys were installed in areas with thicker ice, which is known to correlate strongly with snow thickness. Even toward the end of the season, ice with greater surface roughness tends to accumulate more snow compared to smoother, younger ice. This raises questions about the representativeness of the snow thickness measurements from the buoys. These considerations lead to a broader question regarding the interpretation of the seasonal changes presented in the study. How might these changes be affected by the unrepresentativeness of the buoy measurements?
A third concern relates to the interpretation of in situ density measurements. Based on the abstract from the coring dataset by Oggier et al. (2023), it appears that the sea-ice density was measured at constant laboratory temperatures rather than at in situ conditions. However, this is (1) not explicitly mentioned in the study, and (2) there is no description of how density was recalculated to reflect in situ temperatures, if such a correction was applied. If this adjustment was not made, any discussion of how brine volume evolution affects sea-ice density becomes less meaningful. Furthermore, the interpretation of the measured density shows a larger standard deviation from actual weighing than from the combination of ICESat-2 freeboard data and buoy snow and ice thickness. This seems to contradict previous estimates of uncertainties in density measurements (Hutchings et al., 2015; Pustogvar and Kulyakhtin, 2016). Specifically, weighing has consistently been found to be one of the most precise methods, while density estimates based on ice freeboard and thickness measurements tend to have standard deviations several times higher—by an order of magnitude—according to Hutchings et al. (2015), even when the thickness and freeboard were measured in situ. Given this prior evidence, it is somewhat surprising that the reported standard deviations suggest the opposite pattern in this study. A clearer explanation of these discrepancies would clarify the interpretation of these measurements.
Specific commentsLine 21: The phrasing suggests that this study considers deformations, brine, and gas inclusions in its estimates of ice density, but this does not appear to be accurate. It would be helpful to clarify this point. Impact of deformed ice, brine and air inclusions were not analysed in this study.
Line 23: The term "Distributed Network scale" might not be clear to readers unfamiliar with the MOSAiC Distributed Network. Could you specify the scale or provide additional context to make it more accessible and inclusive?
Line 24: The statement regarding ice density estimation is not entirely accurate. Ice density was calculated separately from a combination of buoy ice thickness and altimetry freeboard data and from ice coring data. It might be helpful to differentiate these methods more clearly.
Line 28: The assertion that MOSAiC ice floes predominantly consisted of second-year ice needs further support. (1) Several studies suggest otherwise, and (2) the term "predominantly" is vague in terms of areal fraction. Additionally, it would be useful to clarify what is meant by "MOSAiC ice floes"—how many floes are being referenced, and which ones specifically?
Line 32: The ratio of ice freeboard to ice thickness cannot serve as a direct measure of ice density since they are mathematically correlated. Density is typically used to convert ice freeboard to ice thickness. Meanwhile, such relationship may be complicated by the presence of deep snow or absence of accurate snow observations. The latter is directly related to the results of this study. Please clarify the interpretation here.
Line 34: The statement that second-year ice (SYI) dominates the central Arctic Ocean contradicts existing studies, such as Tschudi et al. (2019, 10.5194/tc-14-1519-2020), which show that first-year ice (FYI) is the primary ice type in the Arctic. Could you provide more context or evidence to support this claim?
Line 59: The referenced study from Timco and Weeks (2010) offer an overview of sea-ice properties but do not draw any specific conclusions about the heterogeneity or seasonal evolution of ice density. They only presented ice density at various temperatures (assuming zero gas volume) and discussed a broad range of density measurements without the kind of conclusions mentioned in your study. Could you clarify this?
Line 63: The claim that sea ice density influences permeability, meltwater infiltration, and biogeochemistry is quite bold. The cited studies may show correlations, but it seems inaccurate to assert that density directly defines these processes. Permeability, for example, is largely governed by ice age, microstructure, and brine volume. Could this section be reframed to reflect the complexity of these relationships? Also, the term porosity is used in unconventional and inconsistent way. In most cases it is used as equivalent to gas volume fraction (line 440), while typically it should combine both gas and brine volume fractions. Therefore, please explain how gas volume is influencing permeability.
Line 66: If by winter you mean a period from December to February, this is not accurate statement. Most referenced measurements were collected in winter and spring.
Line 71: This (small errors of weighing) was known way before the referenced publication.
Line 90: The robustness of the parameterizations is limited by the assumption of hydrostatic equilibrium, which has been shown to be less precise (Hutchings et al., 2015). For instance, Jutila et al. (2022) propose density values above 950 kg/m³ for relatively thin ice, but such values are unrealistic, as they have not been validated by in situ measurements. How do you account for these limitations?
Line 101: Most of laser scans were significantly smaller than the linear scale of tens of kilometres (Hutter et al., 10.1038/s41597-023-02565-6). Could you clarify the scale?
Line 132: The exclusion of melt ponds from the sampling is not straightforward, as refrozen melt ponds are a substantial part of the ice cover, covering around 20-30% area. And their smooth surface also influence the amount of snow which may accumulate above such ice type. Additionally, surface conditions are not the primary issue when it comes to hydrostatic balance. The representativeness of the ice thickness at the buoy sites is more important, and this is not adequately discussed. How does excluding ridges and deformed ice improve representativeness?
Line 140: You state that 27% of the buoys were in FYI and 73% in SYI, but Kortum et al. (2024, 10.5194/tc-18-2207-2024) showed equal fractions of FYI and SYI based on freeboard measurements. How was this potential unrepresentativeness of buoy sites handled? Moreover, SYI is not typically associated with ice deformation when thinner FYI is present, which may influence evolution of its areal fraction.
Line 141: First, does this influence your results? Probably not. Also, exceptionally large ice growth should be quantified of important. Illustration shows that platelet ice presumably contributed to 10-20 cm ice growth which later disappeared. Is it relevant for this study?
Line 173: There's an inconsistency here—earlier, you stated that 27% FYI was representative in an SYI-dominated area, yet here, 43% FYI is also considered representative. Meanwhile, Kortum et al. estimated this fraction as 50%. Could you clarify this discrepancy?
Line 179: It would be more robust to demonstrate the representativeness of the selected sites by comparing their ice thickness or freeboard with large-scale observations, rather than just stating they were selected for representativeness.
Line 182: Please clarify what is meant by "ice porosity." What does it include?
Line 183: According to Oggier et al. (2023), salinity was measured directly from the sample after weighing, not interpolated from another core. Also, could you explain how brine volume and density were estimated for the measured temperatures?
Line 189: The statement that some cores were longer and might come from rafted ice, leading to substantially different densities, is not supported by the data from Oggier et al. (2023). Please provide evidence for this assertion.
Line 268: The ice pack does not consist solely of level ice. Please rephrase this statement, as ridges and deformed ice make up a substantial portion of the ice cover, as shown in transect observations. Also, how can you be certain that modal snow freeboard represents the same level ice described at the buoy sites? A simple comparison of ice thickness distribution or mean, modal and median values could be also useful. What if ice consists of both FYI and SYI level ice, with a bimodal thickness distribution, which is not seen in snow freeboard data?
Line 287: What are "MOSAiC ice floes, type II"? Could you provide more information on this classification? If you refer to the types of uncertainties, then they are described after this part of the manuscript.
Line 291: The 7 cm adjustment is a significant point, yet it is only briefly mentioned. Could you explain in more detail how this value was estimated and why it was applied as a time-independent adjustment throughout the seven-month period?
Line 294: How is this possible if you only use a single source of snow thickness from buoys? Here you attempt to match ice freeboard calculated as difference between ICESat-2 snow freeboard and snow thickness from buoys. You do not have data about large-scale snow thickness (line 314). Therefore, you lift the snow and ice freeboards together which leads to increased estimate of ice thickness which fits better the initial buoy ice thickness. From these estimates of ice freeboard and thickness you estimate the ice density. Therefore, it is not clear what do you mean by the snow depth difference between satellite and buoys sites, as second one is unknown. Similarly, this is an important assumption which cannot even be generalized for future campaigns. It deserved not to be described in supplementary materials.
Line 315: While seawater density might not be the primary source of uncertainty, why not use directly measured seawater salinity and temperature from the expedition? Why was the chosen reference used instead?
Line 317: Again, more evidence is needed to support the claim that the MOSAiC Distributed Network (DN) was dominated by SYI. How this is possible when the measured by ICESat-2 snow freeboard in autumn was close to 10 cm? Was it representative for SYI, especially SYI similar to ice where buoys were installed? Because buoys had ice freeboard around 10 cm and snow thickness of 13 cm, giving the snow freeboard way above 10 cm.
Line 318: Krumpen et al. (2020, 10.5194/tc-14-2173-2020) report that the modal ice thickness in L-sites was only 40-50 cm, significantly thinner than at buoy sites. Why is this discrepancy not addressed?
Line 359: The ice thickness range of 0.6–2.0 m is stated as representative for level ice, yet Krumpen reports much thinner modal thicknesses. Early transects from Itkin et al. (2023,10.1525/elementa.2022.00048) also mainly show modal ice thickness around 0.5 m. Could you provide clarification on this?
Line 370: Arbitrary selection of coring sites does not necessarily make them representative in terms of ice thickness. This should be better supported.
Line 375: You obtain level ice freeboard using modal values of snow freeboard. Why is a different approach applied here to the transect snow thickness? Itkin et al. (2023) showed a snow thickness distribution with initial modal snow thickness around 10 cm. Similarly, the initial snow thickness above level ice in their paper is also close to 10 cm. Therefore, what is the meaning of the initial average 24 cm from transects?
Line 377: It appears that specific conditions at the coring sites resulted in shifting the ICESat-2 freeboard upwards. However, ICESat-2 and ALS initially showed freeboards close to 10-15 cm, suggesting the buoys were in areas with unrepresentative snow thickness. Could this be clarified?
Line 384: This is a misinterpretation of the study from Itkin et al. (2023) You claim that buoy snow depth was close to transects. Yet, in the Fig. 9 of Itkin et al. (2023) the shown snow depth above level ice is below 10 cm.
Line 437: You mention that the DN is dominated by SYI, but later state that MOSAiC ice floes were predominantly FYI. This contradicts previous observations (e.g., Guo et al., 10.5194/tc-17-1279-2023; Kortum et al.). Could you clarify what exactly is being referred to as "MOSAiC ice floes"?
Line 437: Your estimate of ice density includes 7 cm freeboard adjustment; how can you discuss these minor differences between measured FYI and SYI density considering buoy method uncertainty close to 50 kg/m³? Moreover, there is not any clear difference in winter FYI and SYI density based on weighing. How can potential difference in several kg/m³ can support whether FYI or SYI was dominating the area of investigations? You are saying that 910 kg/m³ is closer to 912 kg/m³, while 908 is closer to 905 kg/m³? A difference in 2 kg/m³ is beyond measuring uncertainties and spatial variability even for the best performing methods to measure ice density. No such claims can be made. Finally, it is not correct to average FYI density for the whole period of observation, as it includes a period of lower densities in autumn. If you exclude those lower values from the average value of 905 kg/m³, would you be able to see any measurable difference between two ice types?
Line 443: in line 437 you claim that DN is dominated by SYI, while here you claim that MOSAiC ice floes were predominantly FYI. What are MOSAiC ice floes? This contradicts many other observations (Guo et al., Kortum et al.). MOSAiC Central Observatory was one of the thickest floes in the area with the largest SYI fraction, unlike surroundings (Krumpen et al., 2020), which is opposite to the presented claims.
Line 446: What is the basis of the claim of substantial sub-weekly ice density variability. This is a clear artefact of spatial variability of ICESat-2 tracks and corresponding freeboards which has nearly nothing to do with ice density changes.
Line 448: You are saying that FYI had newly formed SYI fractions at the ice bottom? What does it mean?
Line 451: First, which layers are you referring to? These layers may contain more brine than which other layers? For FYI, how is it possible, if ice is gradually getting colder and less saline upon ice growth? The brine volume is easy to estimate, and it was estimated in several publications, there is no need for such speculations. And, following the modelling study from Griewank and Notz (2008), brine volume is decreasing in Autumn, contributing to decreasing density if gas is not considered. And if you discuss SYI, there were no new layer under SYI in autumn, only in winter, which can be clearly seen in Fig. 4a. Only since cold winter month, SYI with larger snow and ice thickness may have a substantial thickness gain.
Line 461: What does it mean? There is a clear relationship between temperature, salinity, and brine volume which explain why saltier FYI has a stronger vertical gradient of brine volume. Also, most of SYI formed under the remaining 0.5 m at the surface is a new ice, with its physical properties close to FYI. Therefore, there is no surprise that FYI and SYI are eventually becoming more similar. Homogenization is a too broad description of these processes.
Line 462: What is the evidence for these claims? You do not present ice salinities in this study. But if you check FYI bulk salinity from Oggier et al. (2023), it has decreasing trend very similar to the broadly used parametrization from Kovacs (1996, 10.21236/ada312027). Also, please explain in more detail the suggested process behind the initial density decrease. If you consider only brine, desalination leads to decrease of brine volume and corresponding density decrease, that is correct. But it cannot be significant only due to brine, the absolute values are less than 5 kg/m³. One can estimate brine volumes from the measured salinity and temperature and draw the same conclusion. Yet, the observed density change is much larger, and it was increasing, not decreasing (Fig. 5a).
Line 466: This is an example of overfitting. Density estimates from ICESat-2 has a large spread of around 30 kg/m³. The increasing trend in autumn is barely based on a single data point in Fig. 5a.
Line 498: Can you explain more specifically what is the mentioned scale-related density variability? You cover different areas in DN and L-site scales, but there is no clear difference between them following Fig. 5a. Also, the mean values are nearly identical, but this would not be possible without in situ measurements, providing less uncertain estimates. And what supports the well-known importance of ice density for ice thickness retrieval but based purely on your measurements? How do you validate thickness estimates?
Line 520: Repeating this statement doesn’t make it more evident. Without scale adjustment, the density estimate from ICESat-2, and buoys not only show an opposite decreasing seasonal trend to in situ measurements, but they also give unrealistic density values above 950 kg/m³. Therefore, it is not clear why this difference in correlation should be compared between different methods.
Line 548: Previously, it was mentioned that seasonality of ice density is significant. Yet, none of the suggested parameters are related to seasons. Ice thickness has a large range from thin ice in Atlantic sector to thick ice in Pacific sector during the same sampling time. Therefore, ice freeboard cannot be an accurate factor related to the seasonality of ice density.
Line 580: What is the basis of this claim? A typical modal ice thickness is around 1.7 m (Sumata et al., 2023) which includes both FYI and SYI. Why FYI cannot have 0.8-1.8 m thickness, only SYI? Also, there is a substantial difference in ice thicknesses wit thinner FYI in the start of Transpolar Drift and older ice closer to Canadian coast, where ice thickness is also typically larger than 0.8-1.8 m.
Line 627: If you claim that L-site ice was FYI and SYI, while DN was mainly SYI, why then snow freeboard from ICESat-2 and ALS showed such a great agreement despite ALS covering only a relatively small area around MOSAiC ice floe? Also, if most of ice was SYI, why ice density seasonality is close to the FYI in situ measurements, not SYI with no seasonality?
Line 675: The study from Kovacs does not properly describe methodology of density measurements. You describe it as “theoretical equation”, what does it mean? There is a few better documented ice density reports, why this specific study was used?
Line 680: What is the basis for such claims? There are in situ measurements of ice density for relatively thin ice (20-30 cm), and those values are mostly around 900-910 kg/m³ (10.1016/0165-232X(95)00007-X), which is close to MOSAiC values. Also, MOSAiC ice was quite thin, around 40 cm. Sampling ice thinner than 30 cm is typically considered unsafe, and there is a very short time when ice is that thin in Central Arctic, as thin ice growth exponentially faster.
Line 682: Similarly, why a method based on substantial adjustments to be able to get realistic estimates of ice density can support limitations of more accurate method, based on measurements of ice density? What exactly was done to support this statement?
Line 697: The was not any seasonality in SYI density estimates. A strong seasonality and corresponding dependence of ice density from its thickness was only observed for FYI. Therefore, why the parametrization based on FYI measurements can optimize retrieval of SYI? This should be better explained.
Line 707: This is generally surprising that SYI is treated as a separate ice type from FYI. Maybe it could be reasonable many years ago, but for MOSAiC observations with SYI having low-salinity remnant layer below 40-80 cm (Oggier et al., 2023), most of SYI thickness by the end of winter is FYI grown under SYI, which may have density values like pure FYI. Since all such data is available, it is worth discussing. In brief, MOSAiC SYI was much closer to FYI than SYI sampled by Jutila et al., with at least two times higher thicknesses.
Line 734: You previously mentioned that without scale adjustments your estimates would be 50 kg/m³, therefore you should explain how this bias should be removed for future analysis prior to providing parametrization based on adjusted estimates.
Line 738: You haven’t provided a strong proof that the provided parametrization with a strong seasonality in autumn is related to SYI, not FYI. Observations suggest that this is different.
Line 742: It is generally looks impossible that estimates from hydrostatic balance have smaller standard deviations than measurements from weighing. First, even in situ measurements of freeboard and thickness give much higher deviations of density as described by Hutchings et al. (10.3189/2015AoG69A814) Second, density of separate layers of ice should not be compared with bulk density.
Line 743: This is one of the most inconsistent claims throughout the whole paper. Can you say specifically what exactly was dominated by FYI and SYI and provide some evidence for this.
Line 748: The assertion that warm air intrusions substantially influenced future changes in ice density evolution is significant but not discussed earlier in the text. Could this be elaborated?
Line 756: The criticism of in situ density measurements requires more analysis. How can 15 buoys at the fixed sites capture a broader picture better than around 40 coring events at slightly different sites? Additionally, how does the smaller uncertainty from weighing lead to challenges, as this seems counterintuitive?
Citation: https://doi.org/10.5194/egusphere-2024-2821-RC1
Data sets
Scale-related bulk density of level sea ice during the MOSAiC freezing season (October 2019 to April 2020) Yi Zhou et al. https://zenodo.org/records/13690816
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
127 | 21 | 5 | 153 | 18 | 1 | 1 |
- HTML: 127
- PDF: 21
- XML: 5
- Total: 153
- Supplement: 18
- BibTeX: 1
- EndNote: 1
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1