Breaking the Ice

Before getting to the main topic of today’s post, I’ll mention that there’s a terrific story from a Belgian who attended the training session to become a presenter of Al Gore’s slide show about global warming: http://www.worldchanging.com/archives/005588.html. Quite a good read! The cavalry is coming — and we’re it.

A few days ago I posted about ice-out (day of breakup of the ice) on Moosehead Lake. I pointed out that it’s far too easy to get the wrong idea from a cursory look at data, that in fact the data presented did not show signs of global warming, but that more up-to-date data actually does. When I went hunting for ice-out data for Moosehead Lake, I actually found data for 29 lakes in New England. So I’d like to discuss what the data show.

One caveat should be mentioned: temperature isn’t the only factor affecting ice-out date. Precipitation, for example, also has an effect; a heavy rain can lead to earlier ice-out simply because of the effect of adding a lot of running water. Nonetheless, temperature is a big factor in ice-out date, so trends in ice-out can be used as a useful clue to trends in temperature.

One way to get an idea about the overall change is to do a linear fit to look for a trend (probably the most popular method for trend analysis). This doesn’t mean that the data follow a linear trend, or that the trend goes up to the present! But it does give a clue about the overall change. None of the lakes shows a positive overall trend with statistical significance, i.e., none of them shows signs of ice-out occuring later in the year. But 17 of the 29 lakes show a statistically significant negative trend; ice-out is occuring earlier (which we would expect if the region is warming).

More relevant is a trend analysis of the recent period. So I repeated the analysis for data from 1970 to 2005. This is much more indicative of the actual response of New England lakes to recent global warming, but at the same time, it greatly reduces the likelihood of detecting trends with statistical significance, simply because there’s so much less data to work with. In spite of this handicap, the results are impressive: 13 of 29 New England lakes showed a statistically significant trend over the time span 1970 to 2005. Again, all of the trends are negative, i.e., ice-out is occuring earlier rather than later.

With 29 data sets, the natural thing to do is to combine them — to average over all lakes — in order to increase the statistical reliability and to get a picture of the behavior over the entire New England region. But you can’t just average the data for all 29 New England lakes, because each data set covers a different time span; the earliest (Sebago Lake) starts in 1807, the latest (Squa Pan) doesn’t begin until 1930. Also, some of the lakes show ice-out consistently earlier or later than others (they cover a wide region of New England), and it’s good to get them on the same “scale.” So I began by defining ice-out anomaly as the difference between the ice-out date for a given lake, and the average for that same lake over the “reference period” 1940 to 2005. Negative values mean earlier-than-average ice-out (warmer conditions) while positive values indicate later-than-average ice-out. Now I can safely average the anomaly values, to get an idea of the overall trend throughout New England.

Average ice-out anomaly for all 29 New England lakes does show a statistically significant trend. For the entire time span, the trend is for ice-out to occur earlier by 8.8 days/century — but this isn’t really indicative of the impact of recent global warming. For the time span 1970 to 2005, the trend is also statistically significant and does indicate the impact of recent warming. What’s most impressive is that the rate of change of ice-out is a whopping 23 days/century! If this trend continues, then by the year 2100 ice-out on New England lakes will, on average, happen 22 days earlier than it’s happening today.

The sad fact is that the trend is likely to continue; in fact it’s likely to happen even faster. Global warming is expected to happen (slightly) faster in the future, as we continue to dump greenhouse gases into the atmosphere at an alarming rate. It’s not inconceivable that by 2100, a large number of New England lakes simply won’t freeze over at all.

A few of the lakes showed even more rapid changes. Lake Pontoosuc shows, since 1970, a trend toward earlier ice-out of 49 days/century. Sebago lake shows the strongest trend of all; ostensibly it’s more than 100 days/century, but the trend is to a certain degree undefined because for four of the years in question, Sebago lake didn’t freeze over.

Does this prove global warming? Of course not. New England is only one small region of the globe. But it’s evidence which doesn’t depend on super-high-tech instruments (like a thermometer) that this region has warmed significantly over the last 35 years. If you study regional temperature trends, you’ll discover that New England hasn’t been warming as fast as other areas of the U.S., or as fast as the entire planet.

The natural thing to do would be to study ice-out dates for lakes distributed around the world. Maybe some day I’ll get around to that, but for now I’m preoccupied blogging, investigating the cause of the ice ages, and of course doing that little thing called work.