Friday, August 15, 2014

Long-Term Precipitation Trends – Part I: Alaska

** Updated 8/17 with charts reflecting percentage changes instead of raw number changes **

Continuing with the theme of precipitation trends, I wanted to look at the rate of change of various precipitation thresholds over the last number of years to see if they are more or less frequent. To evaluate the rate of change, I looked at 4 different precipitation thresholds over the course of the Fairbanks climate record (good precipitation records extends back to 1915). Figure 1 shows the annual number of days for four different precipitation classes in Fairbanks (>= 0.01", >= 0.05", >= 0.25", and >= 0.50"). The only one of the trend lines that was significant at the 95% level (p<=0.05) was the number of days per year with at least 0.01" of precipitation. All the other thresholds were close to the even line but none showed a statistically significant trend (positive or negative). End of story, right?

Figure 1. Precipitation frequency (days per year) at Fairbanks for four different precipitation classes from 1915 to 2013. They are: 1) at least 0.01", 2) at least 0.05", 3) at least 0.25", and 4) at least 0.50".

A dangerous practice in the field of climatology is to make sweeping generalizations by looking at a single station's data. So, let's look to see how things have changed across a wide swath of Alaska. Unfortunately there are not many stations with a record as long as Fairbanks. We can increase the sample size to 21 stations when we set the begin date at 1951 instead of 1915. Figure 2 shows the stations we will use to look at the statewide trend in precipitation events.

Figure 2. Stations (21) used to assess long-term precipitation frequencies.

Looking at the statewide data, I was originally looking at "heavy" events so the categories are a little different. Actually, I was looking for national trends so the criteria were set somewhat higher than would ordinarily be appropriate for Alaska. In any event, the categories are: 1) at least 0.05", 2) at least 0.50", 3) at least 1.00", and 4) at least 2.00". That makes for a little bit of an apples to oranges comparison with Figure 1 – but let's progress nonetheless. Figures 3 through 6 below show the annual percentage difference of all stations from the 1951-2013 station average. To eliminate the effect of stations with large values overwhelming stations with low values, percentage differences were used instead of raw numbers. For example, a station that averages 50 days with 0.05" from 1951-2013 has a 10% increase if a year recorded 5 extra days with 0.05". A station that average 20 days with 0.05" from 1951-2013 has a 25% increase if a year recorded 5 extra days with 0.05". If you combine the raw numbers from those two stations, they would sow a 15% increase in the number of days with 0.05". However, comparing the percentages yields an increase of 17.5%. Therefore, comparing the annual percentage deviation is more meaningful since stations with larger averages would ordinarily drown out stations with low averages.

The number of days with >= 0.05" and >= 0.50" are noticeably increasing at a statistically significant rate (95% level) in Alaska. The >= 1.00" and >= 2.00" give mixed signals and neither trend is statistically significant. This is due in large part to small numbers and a relatively few stations responsible for most of the deviations.

Therefore, I feel comfortable stating that the number of small precipitation events and the number of moderately-sized precipitation events have definitely increased over the last 60 years here in Alaska. This is not so surprising given the fact that warmer air can hold more moisture and warmer air at the surface with constant temperature aloft (not analyzed) would lead to lapse rate instability. As for the frequency of 1" and 2" precipitation events, they are too infrequent to make judgments on.

Figure 3. Statewide average of percentage change from 1951-2013 average annual number of days with at least 0.05" of precipitation.

Figure 4. Statewide average of percentage change from 1951-2013 average annual number of days with at least 0.50" of precipitation.

Figure 5. Statewide average of percentage change from 1951-2013 average annual number of days with at least 1.00" of precipitation.

Figure 6. Statewide average of percentage change from 1951-2013 average annual number of days with at least 2.00" of precipitation.

1. I would argue that a straight line fit us not appropriate in any of these charts. For the Fairbanks chart there is a clear step at 1927. And the other graphs have clear steps at 1977 during the Pacific regime change. A fit to the two segments of the step for each graph would yield insignificant trends. And I agree that the bottom two graphs have to little data to make a good comparison.

I'm going to guess that the interior has seen relatively little precipitation increase but the southcentral and panhandlers has seen a lot by storm tracks changing via the 1977 shift. Thus only bigger storms get in and give us more rain when it does rain. The changed storm tracks have also given us milder winter temps since 1977. This is how I'm starting to see all of these changes.

1. Thanks for the comment Eric. One of the problems with linear trends, as you indicated, is that they ignore smaller scale patterns. That being said, the longer the time period we look at, the more macro-scale patterns tend to come through. I am not so sure that the pattern shifts in Figures 1, 3, 4, 5, and 6 look like the PDO shifts that are apparent in temperature plots (not shown). I'll run a regression for the pre- and post-1977 data for Fairbanks and report back on significance.

2. http://jisao.washington.edu/pdo/
http://jisao.washington.edu/pdo/PDO.latest

Here's the PDO indices since 1900. Digest and compare with Brian's very interesting trends. Whether there was a permanent change of regime in 1977 is debatable, given the unknown trends for the PDO prior to the scale presented (see Kaplan cited for some earlier info).

In the scheme of things this time frame is a very limited point of data in the earth's climate since the last Glacial maximum 20-25K years ago. We need to look at least at that overall period to truly note the impact of a century's climate.

The PDO is only one factor affecting our climate in Alaska.

Gary

3. Am not sure this applies directly to precipitation, but I find this type of visual presentation fascinating as it may relate to this summer's weather. Select the modes for effect:

http://www.worldclimateservice.com/thickness/

Gary