Data Description

Yearly maps of early melt (earliest observed melt conditions), melt (melt conditions observed from this point until freeze), early freeze (earliest observed freeze conditions), and freeze (freeze conditions observed from this point on) for the surface of sea ice derived from SSM/I data. Reference:

Markus, T., J. C. Stroeve, and J. Miller (2009), Recent changes in Arctic sea ice melt onset, freezeup, and melt season length, J. Geophys. Res., 114, C12024, doi:10.1029/2009JC005436

Product Updated, 3/07/2019

This update adds melt maps for 2018.

Product Updated, 2/26/2018

This update adds melt maps for 2017. Added NetCDF versions of the datafiles and a python routine to read them. Updated README.

Product Updated, 4/18/2017

This update adds melt maps for 2015 and 2016. 2016 uses F17 for the first part of the year and then switches to F18. The intercalibration between F17 and F18 looks good so at the moment no bias adjustment is done. We will continue to monitor the performance of F18.

Fixed a bug that didn't reinitialize the default freeze value between pixels which meant that in a low number of pixels there was no new freeze value determined and the algorithm used the prior result. This fix made the version 730s across all sensors.


Product Updated, 7/2/2015

In early October 2011, the AMSR-E instrument developed a fault and further data was unavailable on the archive. The previous source data for this routine, SSM/I F13, stopped providing data at the end of 2007 and at that time the routine had been adapted to AMSR-E and maps were processed from 2003 through 2010. The routine has now been adapted to use SSM/IS F17 as a data source for microwave brightness temperatures with the version “701s”, and the 2011 maps were processed.

In February 2014, we investigated anomalous behavior of the routine in coastal areas on the early freeze and freeze maps. We discovered that the bias correction used for F17 was misreporting ice concentration values during the melt period, and that was resulting in the poor freeze-up values. The bias was adjusted and the datset was reprocessed for 2007 on with version "702s".

The following figure shows average melt onset, freeze onset, and average melt duration in days for the entire Arctic.  


figure displaying trends in melt onset, freeze onset, and melt season length


These figures show the trends above on a single plot, with the space between the curves (the melt season length) in just blue and in a colormap to show the trend in melt season length due to warming.

a plot showing the increasing melt season length as a function of time 

a plot showing the increasing melt season length as a function of time as shown by a color trend

a plot showing the increasing Arctic freeze onset anomaly as a function of time

Data Format

Readme for melt / freeze products

4 files:

YYYY - year
VVVV - version, current set is 730s.

Integer arrays, 304 x 448, big endian (windows and mac will need to byteswap). Values are the day of year where the event described in the filename occurs (in 'melt', a value of 175 means melt conditions are present on June 24 in a non-leap year).

1 file per year:
YYYY - year
VVVV - version, currently 730s.

Four datafields: earlymelt, melt, earlyfreeze, freeze

There is a sample routine on our site called that reads and plots the data.

Data Flags:

open water
valid data

All negative values represent pixels where melt or freeze could not be determined; the different negative values correspond to different criteria in the algorithm and should not be considered for analysis. The earliest melt possible due to the algorithm is day 75. Data is no longer masked for the polar "hole" in the coverage, and users should note that the hole changes size with satellite transitions.


Regional Summary Files:




XXXX - product (early melt | melt | freeze | late freeze | inner diff (freeze - melt) | freeze anomaly)


These are text tables that give the average value per year (rows) of the named product per region. There has been some quality control on the pixels included in the averages. The first row gives the column labels, which are:


col - region code - region

0 - YYYY - Year

1 - 2 - Sea of Oktosk      

2 - 3 - Bering Sea

3 - 4 - Hudson Bay

4 - 6 - Baffin Bay

6 - 7 - Greenland Sea

7 - 8 - Barent Sea

8 - 9 - Kara Sea

9 - 10 - Laptev Sea

10 - 11 - E. Siberian Sea

11 - 12 - Chukchi Sea

12 - 13 - Beaufort Sea

13 - 14 - Canadian Archipelago

14 - 15 - Central Arctic

15 - 0 - Entire Arctic (all the above regions)

Updated on 7/02/2015

Removed the AMSR-E products from this distribution in favor of SSM/I / SMMR trend only, since AMSR-E stopped providing data. Updated 2007+ to improved bias adjustment, version 702s.

Updated on 4/18/2017

Updated through 2016. Version changed to 730s across all the instruments as a bug in the default ice freeze value for some cases was corrected.


Contact info:

README last updated 6/12/2019

Data Files

  • - zip, 3.44 MB: Melt and freeze files for 1979-2018, zipped
  • melt_update_2019.tar - tar, 42.11 MB: Melt and freeze files for 1979-2018, tarred
  • - zip, 4.12 MB: NetCDF version of the dataset, 1979-2018, zipped
  • melt_update_2019_ncdf.tar - tar, 41.79 MB: NetCDF version of the dataset, 1979-2019, tarred
  • - zip, 93.63 kB: Melt and freeze files for 2018 only, zipped
  • 2018_melt.tar - tar, 1.05 MB: Melt and freeze files for 2018 only, tarred
  • - nc, 1.04 MB: The NCDF file for the 2018 melt / freeze season only.
  • - py, 1.82 kB: Python 3 routine that shows how to read and plot the data.
  • - zip, 15.23 kB: Regional summary files for melt / freeze / early melt / late freeze / inner diff (freeze - melt, the length of the melt season) / freeze anomaly, organized by year (rows) and Arctic regions (cols).