Tuesday, August 22, 2017

North Pacific Blog Post

I added a new post on the Alaska "Blob Tracker" blog today, with a brief discussion of a new paper published in the Journal of Climate.  Thanks to Brian Brettschneider for pointing out the paper to me.


Sunday, August 20, 2017

Climate Normals in Changing Environment

Hi, Rick T. here. One of the things that interests me is how people adjust to a changing climate. Anecdotally, it was vaguely humorous to me last winter to see how quickly many people have incorporated three consecutive mild winters into a perception of a "new normal". This was underlying the many comments I heard about how cold the winter of 2016-17 was in Alaska. Of course, through the multi-decade lens, it wasn't notably cold for the winter (through parts of the state were, by any measure, cold in March). So that got me to thinking: given that many climate variables in Alaska are changing, how can we provide estimates of "normal" and associated variability that takes into account the ongoing changes?

One approach I've been toying with to make these kinds of estimates is with the use of quantile regression. Quantile regression is something of cousin to the more familiar least-squares regression, but is computationally more tedious, so was not much utilized until the advent of modern computing. Nowadays, it's trivially simple to use on the kinds of climate datasets that I mostly work with, that is, point-based time series. So the first question you ask: what is a quantile? A quantile is, to quote Wikipedia, "…cutpoints dividing the range of a probability distribution into contiguous intervals…". Quantiles can have any value between zero and one. So, the 0.5 quantile divides a distribution into two equal sizes: half the values are above and half the values are below. You've heard of this: it's better known as the median. A quantile of 0.843 divides a distribution into two parts: the quantile is the value of the distribution for which 84.3% of the distribution is below and 15.7% above. Quantile regression is a method to estimate the quantile values of a dataset when one variable is (possibly) dependent on one or more other variables. The second question you ask: why would you want to use quantile regression? There are a couple of reasons. First quantile regression is not nearly as sensitive to outliers as ordinary linear regression, which in effect models the mean. Secondly, and most significantly for my purposes here, quantile regression allows us to generate estimates of not only the central values of a distribution, e.g. mean or median, but also allows for estimates of how other aspects of the distribution are (possibly) changing.

As an example of this approach, below is a plot of some climate data that you are probably familiar with: spring breakup dates of Tanana River at Nenana (for this version I've used  "fractional dates" which incorporate the time of breakup, which does not matter to this analysis). There is no statistically significant trend through into the 1960s, so I construct the quantile regression to have zero slope in this time period. The purple line is the segmented median (0.50 quantile) date of breakup, which in this case we're looking at the dependence of breakup date on the year (i.e. the trend). The green-shaded area represents the area between the 0.333 and 0.666 quantiles. So, this plot should partition the breakup dates into three (roughly) equal categories: one-third below the green shading (significantly early break-ups), one-third inside the green shading (near normal) and one-third above (significantly later than normal). From this, it's easy to see that break-up dates during the first days in May in the mid-20th century were solidly in the "significantly earlier than normal" category, but the same dates are now in the "significantly later than normal" category.
Below is another example. Here I've plotted the Alaska-wide January through March average temperature from the NCEI Climate Divisions data set. In this case there is no strong evidence for a change in the regression slope that would be better fit with a segmented analysis. In this plot, the purple line is again the regression of the median (0.50 quantile), but the shaded area in this case represents one standard deviation (if the season average temperatures are normally distributed) either side of the mean (approximated by the 0.159 and 0.841 quantiles). You'll notice that the median and +1 standard deviation estimates have increased more than 3°F since 1925. However, the -1 standard deviation estimate has not changed at all. This suggests that late winter temperatures have become more variable: "cold" late winters are about as cold as they were 90 plus years ago, but the warmest late winters are now significantly warmer than back in the Roaring Twenties. How can that be?

Well, in part it's a feature of my analysis. The estimated slope of the 0.159 quantile (the bottom of the shaded area) is about the same as the median. However, at the 90% confidence level, the 0.159 quantile estimate crosses zero (for all you P-value fans, in this case this is the same as saying there is insufficient support to reject the null hypothesis of "no trend"). The 90% confidence estimate does not cross zero for the median or the 0.841 quantile. My convention is: if there is not robust statistical support for a non-zero trend, plot it as zero. More important than any convention, is there something interesting going on physically? I would suggest that yes there is. The late winter season has seen no long term change in the larger regional scale cryosphere variables, i.e. late winter sea ice extent in the Bering Sea shows lots of inter-annual variability, but no trend; snow cover extent is near the seasonal maximum with no trend at high latitudes. This means that given the appropriate weather pattern it can still be cold. Since cyrosphere changes are evidently not at play, ocean temperatures and increasing greenhouse gas forcing are the obvious suspects that would support increased warmth but at this point still allow the cold "tail" to hang on.

The quantile regression I've presented here allows us to make reasonable estimates of the current  distribution of some climate variables in the face of change. This simple linear approach is not likely to be sufficient in the future. For instance, in looking at the Tanana at Nenana breakup dates, I suspect that we are starting to (or will be soon) butt up against astronomical constraints on how early breakup can be given expected terrestrial climate forcing in the next century; e.g. a solar noon sun angle of 30ยบ above the horizon (Nenana on April 1) can only do so much heating. In that scenario, well need to employ non-linear techniques. But that's a topic for another day.

Updated to respond to Richard's comments and questions of Aug 21.
Here's a plot of the quantile regresion slope at 0.05 increments and the associated confidence intervals (90% level) for the Alaska statewide late winter (JFM) temperatures (data plotted above). In this case both the tails show higher spread in the confidence intervals than most of the middle, which I would expect. One wonders though what's going on at the 0.60 and 0.65 quantiles.
Here is some data with more a problematic structure. This is over a century of first autumn freeze dates at the Experiment Farm at UAF. I've included the segmented median and the "near normal" category (0.333 to 0.666 quantiles):
Here the "problem" is the cluster of very late dates between 2001 to 2011. Below, the quantile regression slope and confidence levels seem reasonable until the very high end. Notice the spread of the 0.95 is lower than others above the 75th percentile. I don't think this is realistic, and must be due to that cluster of very late (top ten) dates.
If we push it out even further and make it even more fine grained (quantiles 0.02 to 0.98 every 0.01)  more artifacts emerge, such as the occasional spikes in the bounds, and then the impossibly small confidence interval above the 95th percentile. For me the moral of this story is that it's important to do this exploratory review first, especially if the focus is in the far extremes of the distributions, where potentially other tools are better suited.   

Thursday, August 17, 2017

Summer Wanes

Summer is waning quickly now in Alaska's interior, and some cooler temperatures are finally showing up.  There have been no freezes in the Fairbanks area yet, but the first freeze occurred this morning at Chicken (29°F).  This is the 3rd latest first freeze in the 21 years of data from Chicken; the record latest was on August 21.

Similarly, the Chalkyitsik RAWS saw its first freeze this morning (32°F); this is also the 3rd latest on record (19 years of data; the record latest is August 22).

Last year at about this time I commented on the persistent warmth at the Goldstream Valley Bottom (Ester 5NE) coop site near Fairbanks, and it's been a similar story this summer.  From July 1 through August 15, the lowest temperature was 38°F, compared to 40°F in the same period last year.  In every other summer in this site's 20-year history, the temperature dropped to at least 34°F in this period, and indeed the average date for the first freeze is August 2.  The unusual warmth has been very persistent in the last few weeks.

Looking back at summer conditions across the entire state, the highest reliable temperature measurement was 94°F at the CRN site southeast of Tok, although a more remarkable heat wave occurred just 12 days ago at Skagway, when the temperature rose to 93°F - an all-time record for the site.  These were the highest temperatures in Alaska since June 2013, when Talkeetna smashed its all-time heat record with an astonishing 96°F.

Wednesday, August 16, 2017

Minchumina Follow-Up

Yesterday I posted what I thought was a bit of a mystery regarding solar radiation and temperature data from the Lake Minchumina RAWS, but within just a few minutes reader Gary pointed to a possible solution: increasing shade from vegetation that may have grown up right next to the RAWS instruments.  Here's a 2004 photo from the Western Regional Climate Center website (click to enlarge):

As Gary noted, the photo faces approximately east, so the tree growing up on the right side appears to be roughly southwest of what look like the thermometer and pyranometer in the middle of the arm.  Obviously if this and other vegetation hasn't been controlled in the 10+ years since the photo was taken, then it may have provided increasing amounts of shade over the instruments in recent years; and this would explain the reduction in both solar radiation and warm bias.

Interestingly the hourly solar radiation data support the idea that shading has developed from objects to the south and southwest.  The chart below shows the mean hourly solar radiation (units of langleys) during May on a kind of polar plot; the distance away from the center indicates the radiation amount in each hour, and the angle from the vertical corresponds to the average position (azimuth) of the sun in that hour.  So over the course of the day the solar radiation starts small in the east, increases as the sun moves towards the south, and decreases as the sun goes west.  The blue line shows the averages for 2009-2013 and the red line is for 2015-2017.

The plot makes clear that the reduction in sunshine is fairly small in the morning until about 11am in May, but then it appears that the shading effect is pronounced by around 1-3pm, when the sun is just west of south.  This is nicely consistent with the apparent location of vegetation in the photo.

The charts below show similar results for June, July, and August.  Interestingly the month of June is the only month in which there appears to be no shading from the southeast, i.e. around 9am-noon, and this makes sense if we consider that the sun rises highest in the sky near the solstice; so whatever vegetation has grown up to the southeast, it's apparently not yet high enough to cause shading in June.

In conclusion, I think the problem is just about solved - it looks like the Minchumina radiation data have been seriously affected by shading in recent years, and this has also altered the temperature bias relative to the nearby airport thermometer.  Final confirmation will await a site visit: anyone want to take a field trip?

Tuesday, August 15, 2017

Increasing Clouds at Minchumina?

Some weeks ago when I was exploring the artificial warming reported by RAWS thermometers on sunny days, I noted a remarkable reduction in the RAWS warm bias at Lake Minchumina over the past several years.  The chart I showed earlier is reproduced below, with the solid lines indicating the difference in monthly means of daily high temperature (RAWS minus airport AWOS).  In 2009-2012 the RAWS high temperatures averaged about 4-5°F warmer in May through July, but in 2015 and 2016 the difference was only around 2°F.

In the previous post I suggested that the systematic trend towards a smaller warm bias could be related to increasing cloudiness; if solar radiation has been lower in recent summers, then there would be less artificial warming of the RAWS thermometer.

To explore this hypothesis, I looked at solar radiation data from the Lake Minchumina RAWS for May through August - see the chart below.  To minimize issues related to missing data, I first calculated the mean solar radiation for each hour of the day in a given year and month, and then I took the 7am-7pm mean while requiring fewer than 10% missing data points to arrive at a valid monthly number.

The results are rather striking, with a pronounced drop-off in solar radiation since 2014-2015 at the Lake Minchumina RAWS.  Broadly speaking the decrease in reported solar radiation corresponds to the diminution of the RAWS warm bias, so this seems to be physically consistent.  But is it possible that the solar radiation can have dropped off so significantly and in such a sustained manner in each month from May through August?  This would imply a really significant shift in the summer climate and I'm inclined to be skeptical: a more likely explanation, perhaps, might be that a systematic error has been developing in the solar radiation measurements.  For example, perhaps the sensor has somehow become progressively more obscured or inefficient.

We might have a lot more confidence in the Minchumina RAWS data if other nearby RAWS sites showed a similar reduction in solar radiation.  The map below shows the area we're dealing with.

Looking first at data from Fairbanks, there is little evidence of a sustained increase in summer cloudiness and in fact the overall trend in solar radiation is slightly upward since 2007.  But of course Fairbanks is a long way from Lake Minchumina.

The data from the Telida RAWS (southwest of Minchumina and at similar elevation) do suggest a downward trend in solar insolation in each of the summer months - see below.  The magnitude of the trend is less than at Minchumina, and in particular June and July of this year were comparable to earlier years, which is very different from the situation reported at Minchumina.  So the Telida data seem only mildly supportive of the Minchumina trend.

Looking farther to the southwest, the Farewell RAWS has also reported a modest decrease in summer sunshine over the past 10 years, but again the change in the past few years is nowhere near as dramatic as at Minchumina.

The McKinley River and Wein Lake RAWS also show some indications of diminished solar radiation, but the trends are not pronounced and consistent across all months as at Minchumina.  So again, this is not conclusive.

What about other independent sources of information on cloudiness and humidity?  The airport AWOS at Lake Minchumina has reported cloud coverage at sub-hourly intervals for some years, so this should be of some value - see below.  No significant trend is evident, but we must bear in mind that the AWOS ceilometer can't detect high-altitude cloudiness; and moreover the bi-annual oscillation in the cloud cover is very odd and more than a little suspicious, so I'm not sure the data are trustworthy.

Finally, a quick look at reanalysis data suggests that, contrary to expectations, relative humidity has been lower than the long-term normal over our area of interest in recent years.  Of course, the NCEP global reanalysis is incapable of reproducing the local flow patterns around the Alaska Range and so its broad analysis may not correspond to local trends, but nevertheless it does not support the idea of increased summer cloudiness.

In conclusion, the situation at Lake Minchumina unfortunately remains a puzzle.  The significant reported decrease in solar radiation is quite consistent with the dramatic reduction in the RAWS warm bias, suggesting that recent summers have been consistently and significantly more cloudy than earlier summers; but data from other sites and sources give a mixed message as to whether we should believe the Minchumina trend.  Judging from the multi-site consensus of the RAWS data, it does seem that summers have been relatively cloudy of late in the upper Kuskokwim valley, but that may be the only conclusion we can make.

Saturday, August 12, 2017

The Fairbanks Flood of 1967: The Rainfall

Hi, Rick T. here. You don't have to live in Fairbanks very long to hear about the great flood of August 1967. Even for a community long accustomed to significant flooding, this was extreme. Something like 95% of the city was flooded, causing millions of dollars in damage. Rasmussen Library at UAF has posted a number of videos from the flood on Alaska Film Archives YouTube channel. The NWS Fairbanks Forecast Office and the Alaska-Pacific River Forecaster Center have also put together a very nice online storybook with a short description of the meteorology and hydrology and lots of photos. There has also been work by the NWS, the City of Fairbanks and FNSB Borough to survey and put up high water mark signs around town, and many of these have been installed in the past couple weeks, such as this one downtown on the north bank of the Chena River.

In this post, I want to look at the rainfall that lead to this flooding from a climate perspective. The rainfall August 11-13, 1967 stands out as the highest of record for daily and multi-day totals. The only rainfall records this event does not hold are short duration records, which are all thunderstorm related. To set the stage, here is the background: the second half of July 1967 saw well normal rainfall: 3.07" fell between July 16 and 31. That's still the fifth highest "second half of July" total. However, that was followed by a dry start to August: only 0.02" of rain fell during the first week of the month. But then the skies opened. Below is a plot of the hourly and cumulative rainfall for the week of August 8-15, 1967 (data extracted from Fairbanks August 1967 Local Climatological Data). More than half an inch of rain fell on the 9th. This was followed by about 36 hours with very little rain. Starting late in the afternoon on the 11th, moderate rain fell without much of a break until the morning of the 13th, though light rain continued to dribble on into the 15th before the fire hose finally shut down.
The following totals were recorded at the Airport, all of which still stand as the highest of record:
  • 24 hour: 3.44" 11pm AKST August 11 to 11pm AKST August 12
  • 36 hour: 4.40" 4pm AKST August 11 to 4am AKST August 13 
  • 48 hour: 4.76" 3pm AKST August 11 to 3pm AKST August 13
  • Single calendar day: 3.42" August 12  
  • Two consecutive calendar days: 4.29" August 11-12
  • Three consecutive  calendar days: 4.98" August 11-13
So beyond "highest of record", what's the climatological context? Was this a one-in-a-million event, or is there some reasonable likelihood it will be broken?  To answer this I've compiled annual extremes of a few of these parameters and then fitted a generalized extreme value (GEV) distribution. If that's greek, no worries: GEV is a standard technique for analyzing extreme event frequency and generating estimates of return periods.

Maximum 24-hour precipitation (not necessarily calendar day) has been recorded in Fairbanks since the Weather Bureau office opened in the summer of 1929. For two and three day totals, I've included the cooperative data from the Ag Experiment Station starting with 1915, when daily precipitation began to be regularly recorded. In the graphic below I show an example of the annual times series, in this case the maximum 24 hour precipitation (upper left) and then the GEV analysis for the annual  maximum 24 hour precip (upper right) and annual maximum two and three day consecutive days (bottom row). The red line shows the fitted return period, while the open circles are the observations (which are plotted in rank order).  I should point out that there is no significant trend in any of the annual values.
So what are the return periods for the precipitation amounts that occurred in August 1967?

With  87 years of data:
  • 3.44" in 24 hours is expected to occur on average once in 203 years
With 102 years of data:
  • 4.29" in two consecutive days is expected to occur on average once in 269 years
  • 4.98"  in three consecutive days is expected to occur on average once in 352 years
I'm actually not a fan of return periods. It's not technically wrong, but many people would look at those numbers and say "gee, I'll never see that." In fact, the return periods are just an alternate way of expressing the probability of occurrence. So if I tell you that there is 13% chance that Fairbanks Airport will receive precipitation totaling 3.44" or more in 24 hours hours once in the next 30 years, that's equivalent to saying the return period is 203 years, but the take-home message is different. Now 13% in 30 years is not high (and that makes the dubious assumption that there is no change in extreme precipitation events in a warming world), but it is hardly unthinkably rare.

The GEV analysis I've presented here differs somewhat from that published in the 2012 NOAA Atlas 14 primarily in that I used a longer period of record (for extremes analysis, the longer the better), and, as near as I can tell, the maximum 24 hour precipitation (as distinct from calendar day) was not used in compiling that work.

The maximum 24 hour precipitation for August 1967 is listed as 3.42" in the August 1967 Local Climatlogical Data publication and has been carried through ever since. In fact the hourly precipitation data in the same publication shows that the correct amount is 3.44", 11pm on the 11th to 11pm on the 12th.

Thursday, August 3, 2017

Extent of Humid July

In the previous post a question arose as to whether the very high humidity observations from Fairbanks airport in July were accurate or if perhaps the sensor might be malfunctioning to some extent.  To shed a bit more light on this, I looked at other sites around the interior and also - thanks to a suggestion from Rick T. - looked at the surface dewpoint reported twice per day on the Fairbanks upper-air sounding.

The sounding data clearly support the record July humidity reported by the ASOS instrument - see the chart below.  Last year saw the most humid July on record at the Fairbanks upper-air site (adjacent to the airport), and this July was even more humid.

Looking at hourly data from Fort Wainwright and Eielson AFB, the same thing is observed; both of these sites also saw a record high monthly mean dewpoint.  This is a record for any calendar month, not just for July; and at all three sites the previous record was in July 2016.

I also pulled out the historical data for several other sites across the interior to see how far afield the record moist airmass extended.  The charts below show the July mean dewpoint for 12 sites divided broadly into eastern and western groupings, with the record high values indicated by markers.  Among the "eastern" sites, Nenana and Northway observed record humidity in July 2017, although the 1973-74 data from Delta Junction and the 1962 record from Fort Yukon look highly suspect, and if we take these out then the record also occurred in 2017 at these sites.

Farther to the west, July 1998 was the most humid July (and calendar month) on record at Galena, Indian Mountain, and Bettles, and July 2004 was the most humid month at McGrath, but a new record was set last month at Minchumina and Tanana.

In conclusion, the data suggest that record humidity occurred last month at least throughout the Tanana River valley, as new calendar month record dewpoints were observed at every site I looked at from Northway down to Tanana (assuming the 1974 Delta Junction data is wrong). 

The July 500mb and MSLP patterns (see below) do not show an amplified pattern over Alaska, but the modest upper-level ridge over northern and western Alaska was persistent and prevented cool, dry air from reaching the interior from the north.  Daily minimum temperatures were almost entirely above normal in Fairbanks as shown in the chart below.


It seems to have been the stagnant weather pattern, more than anything, that allowed humidity to pool over the Tanana River valley, although the widespread above-normal sea surface temperatures (and therefore enhanced evaporation) surrounding Alaska presumably played a role.  Here's a map of recent SST anomalies, as shown by Rick in his recent climate briefing.