Identifying sea ice plumes is one of the essential factors in obtaining sea ice thickness. 4 SPATIAL DISTRIBUTION OF SST ANOMALY (A) AUGUST, (B) SEPTEMBER, (C) OCTOBER,. C) SEA ICE THICKNESS IN NOVEMBER 2016 DERIVED FROM CRYOSAT-2.
Introduction
- The role of sea ice in the Earth climate system
- Arctic amplification
- Remote sensing of sea ice
- Sea ice altimetry
- Basic principle of radar altimetry
- Concept of SAR altimetry and CryoSat-2
- Overview of papers
- Research questions and hypotheses
Sea ice albedo measured with the Advanced Very High Resolution Radiometer (AVHRR), which provides seasonal albedo for Arctic sea ice over a large area (Lindsay and Rothrock, 1994). Snow depth is essential to estimate sea ice thickness using the hydrostatic assumption.
Introduction
Farrell et al., 2009) proposed a threshold-based method to distinguish wires from ice flakes using various parameters extracted from ICESat level 1b data such as gain, reflectivity, radiance and waveform characteristics. Ricker et al., 2014) used various waveform parameters such as PP, SSD, stack kurtosis, and sea ice concentration to distinguish wires from ice flakes.
Data
- MODIS
- Sea ice type
- Airborne Electromagnetics data
The type of sea ice was used as an input variable to calculate sea ice thickness from ice freeboard. AEM uses electrical conductivity differences between seawater and ice to measure sea ice thickness with an accuracy of ±0.1 m over flat ice (Haas et al., 2010).
Sea ice thickness estimation and machine learning algorithms for lead detection
- Sea ice thickness estimation
- Machine learning algorithms for lead detection
This figure illustrates that the reach point of the first maximum power (the open square) was found before the maximum power peak, and that the retracking point (the open circle) was determined between the reach point of the first maximum power and the threshold level. (the dotted line). The open square indicates the range point of the first maximum power peak determined by the peak detection algorithm.
Results and Discussion
- Characteristics of five parameters based on CryoSat-2 waveform
- Comparison of lead detection performance
- Spatial distribution of Arctic sea ice freeboard and thickness
However, because the proposed machine learning-based approaches for lead detection distinguish the ocean from sea ice and lead, the ocean was excluded in the SSHA. The annual variability of sea ice thickness was high in the MYI zones, compared to the FYI zones. The amount of freeboard and sea ice thickness has apparently decreased compared to the observed Arctic sea ice thickness from February to March 2012 based on CryoSat-2 data.
The sea ice freeboard and thickness maps for March 2013 from this study differed slightly from the results of Ricker et al. 8 Arctic sea ice freeboard from CryoSat-2 for March and April between 2011 and 2014 based on IC scheme. 9 Arctic sea ice thickness from CryoSat-2 for March and April between 2011 and 2014, based on the IC scheme.
2013) yielded similar performance, See5.0 sea ice thickness showed generally better performance than that obtained using the existing methods. To further increase the freeboard and sea ice thickness produced in this study, the depth of snow intrusion must be taken into account.
Conclusions
They revealed that the use of snow load (i.e., W99) resulted in greater uncertainty in the estimate of sea ice thickness than the use of mean density. Such changes need to be carefully considered for a more accurate estimate of sea ice thickness, which requires further investigation. 2014) analyzed the random and systematic uncertainties of Arctic sea ice thickness from CryoSat-2 using the partial derivation of equation 1 under the assumption of hydrostatic equilibrium.
They showed that random uncertainty affects the estimation of sea ice thickness less than systematic uncertainty caused by the choice of a retracking threshold and the unknown penetration level of the signals on snow. To remove systematic uncertainty caused by the choice of a retracking threshold, Kurtz et al. 2014) used a waveform fitting approach to retrieve sea ice freeboard. The lead detection models proposed in this study yielded higher accuracy than the existing approaches for lead detection, implying a possible reduction of the uncertainty caused by the second factor.
This suggests that sea ice thickness in other months such as May or June could be retrieved when additional reference samples from the months were combined with the existing data. The results also showed that Arctic sea ice freeboard and thickness consistently declined from the Canadian Archipelago in particular, but rebounded in 2014.
Introduction
Linear mixture analysis based on the assumption that the spectra measured by sensors for a pixel are a linear combination of the spectra for all components within the pixel (Keshava and Mustard, 2002) was first applied in the field of altimetry research in the Polar Region. from Chase. and Hoyer (1990). They estimated sea ice type and concentration using linear mixture analysis based on Geosat waveforms. However, Geosat with a relatively small number of bins and coarser spatial resolution is not sufficient to detect small plumes in the Arctic winter and spring seasons.
In this study, we applied the linear mixture analysis concept to waveforms from SIRAL, CryoSat-2, to identify leads and produce monthly pan-Arctic lead fractions from January to May and from October to December between 2011 and 2016. The detected leads were visually evaluated with MODIS images (with a resolution of 250 m) and compared with other threshold-based lead detection methods. The lead detection of waveform mixture analysis is not easily affected by the update of the CryoSat-2 baseline, which is new and different from previous threshold-based lead detection methods.
The main objectives of this study are to 1) develop a new lead detection method based on waveform mixture analysis, 2) identify recent pan-Arctic lead fractions.
Data
- Sea ice edge data
- Monthly lead fraction maps
Methods
- Waveform mixture analysis
- Endmember selection
- Calculation of sensitivity in a 10x10 km grid
A monthly lead fraction was derived by dividing the number of lead observations by the number of total observations within a 10 km grid in a month. We divided the Arctic Ocean into two parts, given the number of CryoSat-2 observations. While the spatial resolution of the lead fraction is 10 km below 75°N, the spatial resolution of the lead fraction is 50 km above 75°N, as the number of CryoSat-2 observations decreased rapidly below 75°N.
Since each grid has a different number of CryoSat-2 observations, a sensitivity analysis was performed with respect to the number of observations per lattice. Thirty (30) percent of the number of lead and ice observations in 10x10 km grids were repeatedly permuted (ie, the number of lead and ice observations in a grid changed randomly each time an iteration occurred) and the standard deviation of the resulting lead fractions through 50 iterations was calculated in grid. Since the number of CryoSat-2 observations around the coastline is small (5–10), the thirty percent of the number of lead and ice observations changes randomly, resulting in large variation of the lead fraction in a grid.
A random change of thirty percent in the number of lead and ice observations results in a small variation in the proportion of lead in the grid. Sensitivities in terms of the calculation of monthly lead fraction maps depend on the number of CryoSat-2 observations.
Results
- Performance of lead classification
- Spatiotemporal distribution of lead fraction maps
- Grid sensitive analysis in 10x10 km
In Fig. Areas around the Arctic Ocean coastline clearly show higher lead fraction due to the shear zone (i.e., an area of deformed sea ice along the coast, Serreze and Barry, 2005) and sea ice flow. In particular, high lead fraction was found around the Beaufort Sea during the spring season (MAM) due to the Beaufort Gyre, a wind-driven ocean current.
However, the lead fraction around the Chuckchi Sea was lower than the lead fraction around the Beaufort Sea in January to April (i.e. winter season) 2011 and 2016, with the exception of 2015. 5 Monthly lead fraction maps based on waveform mixture analysis in January to May, October to December between 2011 and 2013. 6 Monthly lead fraction maps based on waveform mixture analysis in January to May, October to December between 2014 and 2016.
This results in an increase in the sensitivity of calculating the monthly lead fraction around the Arctic Ocean coastline based on the small number of CryoSat-2 observations. 7 (a-d) the number of lead observations, (e-h) the number of ice observations, (i-l) the standard deviation of the results based on the sensitivity analysis of the lead fraction from January to April 2011.
Discussion
- Comparison of lead classification methods
- Comparison to other lead fraction maps
- Lead dynamics
- Novelty and limitations
Scene-based lead fraction maps better represent the linear characteristic of diversions and coastal polynya than altimeter-based lead fraction maps. Altimeter-based lead detection methods identified tracks between deformed and fragmented sea ice, causing a higher lead fraction in the Chukchi Sea in January 2011 (Figures 3.8g and j). However, scene-based lead fraction methods could not properly detect the lead fraction in the Chuckchi Sea, resulting in a lower lead fraction.
Nevertheless, altimeter-based lead fraction maps well documented the overall spatial distribution of lead, particularly high lead fractions in the shear zone. The grid sensitivity analysis should be taken into account when interpreting the lead fraction maps around the Arctic Ocean coastline derived by the proposed waveform mixture analysis. Scene-based lead fraction maps better represent shoreline and the intrinsic form of lead (Röhrs and Kaleschke, 2012; Willmes and Heinemann, 2016).
The increase and decrease in the proportion of lead is related to the change in sea ice thickness. The spatial resolution of the monthly lead fraction maps has been improved to 10 km, showing detailed high-latitude spatial distribution of lead in the Arctic.
Conclusions
- Introduction
- Data and Methods
- Sea ice concentration and thickness
- MERRA-2 reanalysis and SST data
- Results
- Anomalous warming event in mid-latitudes and Arctic
- Sea ice minimum extent in November 2016
- Discussion
- Comparison of previous anomaly phenomena
- Sea ice thickness in November 2016
- Conclusions
Sea ice concentration and CryoSat-2-derived sea ice thickness are used to represent sea ice condition. Sea ice thickness from Lee et al. 2016) applied at lower propagation correction was used in this study. November sea ice thickness anomaly in a base period to describe surface variation especially in the central Arctic.
In addition, warm anomaly in North America is moved further to the Arctic due to southerly winds (Fig. 4.3d). e-h) sea ice concentration in August to November 2016. SST may also influence anomalous Arctic warming and the minimum sea ice extent in November 2016. Key factors behind the record minimum sea ice extent in November 2016 are complex rather than one factor.
Implying influences on the November 2016 sea ice anomaly might be possible based on previous analyzes from the literature. The significant decline in sea ice extent in November 2016 has occurred compared to November last year.
Estimation of Arctic sea ice thickness from CryoSat-2 satellite data using machine learning-based lead detection. The ice mass balance buoy: a tool for measuring and attributing changes in the thickness of the Arctic sea ice cover.
