Something I’ve been interested in for the last several months is sea ice data. What makes it interesting is that as I understand it, models demonstrate the poles should be most sensitive to global warming leading the planet temp, especially in the Arctic. Recently I have been able to process the monthly and daily gridded arctic data as provided by NSIDC. The daily values allow a better analysis of trend than can be provided by the monthly data.
If you’re like me you recall the claims of fastest melt rate ever were made about 2007 , I fully believed them, because the graphs showed a much more negative value than in the previous 30 years as shown in Figure 1 below.
This effort was originally intended to investigate how bad the melt rate was in comparison to the natural variation, I didn’t get that far yet. Accessing and processing the gridded data was critical to the analysis, so I spent the time reading the literature and writing code. Having full access to the NSIDC data allows some interesting analysis, they do an excellent job on their site.
There are two primary algorithms used for processing ice data NasaTeam and Bootstrap. The descriptions of the data state the difference between the two is very small and the sets are interchangeable except that bootstrap is recommended for trend analysis in research publications. Bootstrap is only provided in monthly data format while NasaTeam is provided in both monthly and daily provided you’re willing to download over 1G of data, write code to process it, refit the land and missing data mask and sum the results. I am. Also, NasaTeam provides a near real time version of the polar ice data which has a different land mask and hasn’t been processed for missing data. This data isn’t as clean but I wanted to use it. I applied the same land mask as the rest of the series to insure that there was a consistent baseline for trend analysis. The missing data from Jan 2008 onward created noise in the series which I simply filtered out using a 7 day sliding window filter.
The mask looks like this Figure 2
The brown is land, black edges on land are coastline and light blue is the satellite data not measured. This mask is applied consistently through the entire data series. There was some question about masking on one of my other posts at WUWT where visually the land area seemed to change size, in the case of the NSIDC data they apply masks consistently except for the satellite hole and the near real time data.
The NasaTeam version of the arctic ice data looks like the plot below for 2009 (note the small size of the satellite data hole). This graph was created in R using the actual Nasa Team masks and data. I used the worst case land and polar masks to adjust the entire dataset to eliminate problems with consistency. Figure 3
Of course it’s an interesting picture, but what I wanted to know when I started this post was how bad was the worst melt rate in history and what is the actual melt area. In the plot below the arctic is losing sea ice at a rate of only 56K km^2/year. Of course sea ice area went up in the Antarctic during the same time frame though. Note the strong recovery in 08 of Figures 1 and 4, which actually exceeds values of most of the record, matching data back to 1980. Much of this is first year ice so the melt in 08 was expected to be a new record.
If you recall, in 2007 and 08 we were treated to headlines like this, which most of us accepted with a shrug.
I processed and analyzed the NasaTeam land area and missing data masks spending hours understanding different variances they list on their own website. After nearly everything I could find (except satellite transitions errors) was corrected (a different post) and corrections for variance in the measured pixel size, the final result in 30 day trends of arctic sea ice looks like the graph below (Figure 5). This graph is a derivative of the ice area plot. The maximum peaks and valleys represent the maximum rates of change in 30 day periods through the ice record.
Looking at this plot of the 30 day slopes of actual NASA gridded data, the maximum ice melt rate occurs in 1999 and in 2004 not in 2007. Surprisingly the maximum ice growth rates occur in 2007 and 2008, I don’t remember those headlines for some reason. Don’t forget when looking at the 2008 - 09 peak, the data is preliminary and hasn’t been through the same processing as the other data. From looking at the unprocessed data I doubt it will change much.
Certainly the 30 year arctic trend in ice area is downward, even the most committed global warming scientist has to admit this happens regularly in climate along with regular 30 year uptrends. The questions are, did we cause it or not, and was CO2 the instigating factor. The rapid recovery of ice levels has to have some meaning regarding the severity of the problem. This goes directly in the face of accelerated global warming and the doom and gloom scenarios promoted by our politicians and polyscienticians.
Why are my conclusions different from the news reported records? I think it’s likely due to the fact that the scientists used the monthly data which is processed using a weighted filter of the daily data that incorporates a longer time frame than a single month. This means their use of the monthly data to establish a monthly trend was in error and the real record down trends were actually set in 1999, 2003 and 1984. While the record uptrends were in 2007, 2008 and 1996.