Sea ice concentration from merged passive microwave and optical observations

Introduction and Motivation


Sea ice concentration (SIC) defines the percentage of sea ice within a certain area, typically a grid cell of a data product. More intuitively, it tells you where you have how much ice (in area, not in volume). This is of interest for a multitude of disciplines: physicists are interested in sea ice-ocean-atmosphere heat and momentum fluxes. Biologists are interested in photosynthesis rates. Chemists are interested in CO2 uptake. All these processes (and numerous others) are governed by the SIC in polar regions. Outside of natural sciences, shipping companies are interested in the accessibility of North-East and North-West passages. Ship captains are interested in getting a broad overview of the sea ice situation around their ship. The local population in e.g. the Canadian Arctic needs to know where to find polynyas or where to navigate safely on the sea ice with their snow scooters.

 

Background and Methods

Background

Different methods to retrieve the SIC have been developed since the use of satellites for large-scale monitoring of the earth system started in the early 1970s. Currently, two of them are used in this project: Thermal infrared measurements by NASA's  MODerate resolution Imaging Spectrometer MODIS and passive microwave measurements by JAXA's Advanced Microwave Scanning Radiometer 2 AMSR2. They exploit different parts of the electromagnetic spectrum. Thermal infrared data are acquired at 11 and 12 µm, while passive microwave data are measured at 89 GHz.

Thermal infrared SIC

For the thermal infrared SIC, we implement a method by Drüe and Heinemann (2004) to retrieve SIC from ice surface temperatures (thermal infrared data). Those are at a, for our purposes, sufficiently high resolution of 1 km but are only available under cloud-free conditions. A reference temperature for 100% SIC is derived based on the temperature distribution in a 48x48km window. The freezing point of sea water is taken as reference temperature for 0% SIC. Between these two reference temperatures, we interpolate linearly to retrieve SIC. Details are given in Drüe and Heinemann (2004).

Passive microwave SIC

We use our ASI (Artist Sea Ice algorithm) SIC product. It is based on the polarisation ratio at 89 Ghz. details are given in Spreen et al. (2008) and Kaleschke et al. (2001). We do not use the daily product which is publicly available, but data from single overflights to get the smallest possible timelag to the MODIS thermal infrared data.

Merged SIC

The two SIC datasets are merged with passive microwave SIC at 5 km resolution in the following way: One array of MODIS measurements spans 2030x1354 km. We split into boxes of 5x5km, corresponding to one passive microwave gridcell. Now, we assume that the mean of the passive microwave SIC in this cell is correct, but due to the coarser resolution features at spatial scales beneath 5km are not resolved. The thermal infrared SIC, on the other hand, resolve features down to 1km spatial resolution, but the magnitude is sometimes found to be too low. Therefore, we adapt the thermal infrared SIC within this 5x5km box such that they match the mean of the passive microwave SIC. Cloud gaps are filled by the passive microwave data alone. This way, we get a spatially consistent dataset without gaps which benefits from the high resolution of the thermal infrared data where they are available. 

Results

We have used the merged SIC dataset to study a polynya which opened North of Greenland in February and March 2018. Here, the merged SIC (top left), thermal infrared SIC (MODIS, top right) and the passive microwave SIC (ASI, bottom left) are compared. Our group's lead area fraction product (LAF) is used as independent reference product (bottom right). Several findings can be discussed based on this plot.

First, we note that the magnitude of the MODIS SIC is often smaller than that of the other passive microwave and the LAF SIC. This is probably due to the ice thickness distribution: The reference ice temperature is derived based on the local temperature anomaly. If thick sea ice present in the surrounding 48x48 km, the reference temperature will be close to the surface temperature. The surface temperature of thinner ice is still influenced by heat flux from the ocean below, so that 100% of thin ice have a higher surface temperature than 100% of thick ice. Therefore, they appear as smaller SIC. This is mitigated by the merging procedure because the merged SIC adopt the magnitude of the passive microwave SIC, which is not influenced by thin ice if it is thicker than 10 cm.

Second, we note that the merged product resolves much more leads than the passive microwave product. This is not a deficiency of the passive microwave product, but simply due to the higher spatial resolution of the thermal infrared data. Still, it demonstrates the benefit of using thermal infrared and passive microwave SIC together. The LAF product is the only one to actually show open water. It is based on a binary ice/water classification at 80m resolution which is downsampled to 1 km resolution. 

Third, we note that over the polynya at the right part of the image the thermal infrared and LAF-based SIC are higher than the passive microwave SIC. This can be explained as follows: When the ice is pushed away from the coast and open water is exposed to the cold atmosphere, it freezes quickly. However, due to the offshore drift, it does not grow as one homogeneous layer, but is broken up, so that frazil and grease ice form. These change the backscatter and the thermal signature, so that the thermal infrared and the LAF-bsaed SIC show comparably high SIC. The polarisation difference which the passive microwave SIC are based on is not that strongly affected in the early freeze up phase, so that the passive microwave SIC are still close to 0.

 

 

References and contact

  1. Drüe, C. and G. Heinemann, G. (2004): High-resolution maps of the sea-ice concentration from MODIS satellite data, Geophysical Research Letters, Wiley Online Library
  2. Spreen, G., L. Kaleschke and G. Heygster (2008): Sea ice remote sensing using AMSR-E 89 GHz channels J. Geophys. Res.,vol. 113, C02S03, doi:10.1029/2005JC003384
  3. Kaleschke, L., C. Lüpkes et al. (2001): SSM/I Sea ice Remote Sensing for Mesoscale Ocean-Atmosphere Interaction Analysis, Canadian Journal of Remote Sensing, vol. 27

MODIS data (cloudmask, geolocation, reflectances) have been retrieved from the website ladsweb.nascom.nasa.gov/search/. MODIS ice surface temperatures have been retrieved from ftp://n5eil01u.ecs.nsidc.org/pub/MOSA/MYD29.006

This project is worked on by Valentin Ludwig. Questions regarding this topic can be adressed to vludwig(at)uni-bremen.de