Topics

1

Weather / Very short duration rainfall is less extreme (mainly statistically flat)

« on: September 26, 2021, 1128 UTC »

After a sabbatical I am back analyzing the weather.   This analysis is about short duration rainfalls that cause flash floods.  It's a very simple analysis looks at the maximum rainfall for a one hour duration through six hours duration.  The trend is the slope of the linear fit for the highest rainfall of the year.  I also do the trend for each month, although the monthly trend is interesting for reasons other than flooding.  The flash flood is really based on the maximum rain for the year.  I try to get 70 years of data, but you will see on some individual stations there can be a lot of missing or junk data even with a reasonable threshold for file size.

There's a consistent pattern which shows up with a few exceptions at individual stations.  The pattern is that the longer the duration, the more likely that there's a less negative (or positive) trend in max rainfall.  It's most noticeable in October.  It's possible that warmer oceans are lingering into that month causing higher extreme rainfalls or perhaps late season hurricane remnants since the upward trends seem to be clustered in the eastern US.

I simply follow the data, but  I ask myself why would extreme hourly rainfall be trending down in light of global warming which increases atmospheric moisture in general?  My guess, so far, is that the dynamics have changed, perhaps with less severe cold fronts there are less extreme rainfalls.

Results, code and compressed (processed) data: https://followthedata.dev/wx/rfhourly/

2

Weather / Rainfall trends DC and C-Ville

« on: December 06, 2020, 1700 UTC »

I added up days of rainfall for DCA and Charlottesville.  There is an increase in days of measureable rain (a hundredth or more) but a slight decrease in "real" rain (a tenth or more).  That decrease is probably negligible.  That plot is here:
Because of that slight divergence, I got interested in the trend in each bin of rainfall from very light to very heavy.  Unfortunately the daily rainfall totals don't indicate the timing so there's no way to know if 0.2 one day and 0.8 the next day is really one rainfall of 1.0 falling before and after midnight.  So I double counted the days and trended that.  Double counting creates a few trend artifacts.   Finally I did the seven day totals (not double counted).  The seven day plots show an increase in seven day rain amounts of about 2 inches.  For our area that's a relatively wet week.  More plots and links to the code here: https://virtualcoinclub.com/wx/rftrends/  The README has instructions to download the data for any location although you have to search a bit to find longer complete records.

3

Weather / Atlantic hurricane rapid intensification trends

« on: November 24, 2020, 1817 UTC »

Continuing my exploration of the HURDAT2 dataset, I've calculated trends for the rapid intensification of hurricanes in the Atlantic basin. The claim is that warmer waters from global warming are leading to more episodes of RI.  I believe that claim is true for some categories of RI although there is a lot of natural variation.  However the vanilla claim of RI, which is defined as strengthening of 30 knots or more in 24 hours, is very common, about 60% of all hurricanes have at least one 24 hour period with >= 30 knots of strengthening.  That 60% number is flat over the past 30 years:

However if you consider all tropical storms, not just hurricanes, there's a pronounced drop in RI:

I believe that is due to the detection and naming of more tropical storms.  There's an insinuation in some news stories that the glut of named storms this year is also due to global warming.  But they have to pick one thing to blame on global warming: more rapid intensification or more named storms, and then show the evidence.  I believe that Wilma (2005) and Felix (2007) are potential evidence of extreme RI that might be increasing although it is very sporadic. Finally I also calculated and plotted rapid weakening. 
Not surprisingly a number of the strongest storms weakened rapidly.  I explicitly excluded weakening over land since that is obvious and not what I am looking for.  I assume the trend in rapid weakening should be flat.  Storms that head into high latitudes can weaken rapidly but there should be no trend in high latitude movement over time.  Instead I see a slight uptick in rapid weakening over time.  The descriptions of those storms sometimes include the phase "unexpected collapse". Graphs and link to code at https://virtualcoinclub.com/wx/ri/  The page includes a table generated from the data of all the rapidly intensifying storms since 1950.

4

Weather / Western US Dry Season

« on: November 13, 2020, 0004 UTC »

I wrote a litle code to read the rainfall and high temperature for some stations in California and Oregon with the goal of seeing if the dry season has lengthened.  In many cases it has: https://virtualcoinclub.com/wx/dryseason/  Most notably in Santa Barbara, although the code needs a little work.  I think the missing years are because there was not enough rain adding up to trigger the start of the dry season (no rain in the fall is ok, just means dry season extends to day 365).  Often global warming will be blamed for a longer dry season.  I think that's possible and in future work I'll look at the intensity of the dryness based on high temperature.  But some of the cause of the earlier dry season in Sacramento could be the draining of the delta.  There were once huge wetlands in interior California.

In some cases in Oregon the dry season lasts longer, e.g. Ashland.

5

Weather / Summer extreme temps warming out west and cooling in the east

« on: October 24, 2020, 1625 UTC »

Summers are warming in the west when measured by extreme temperatures (highest temperature for the month).  Don't want average temperatures since temperatures at night are not an issue.  Can't use average temperatures anyway due to the time-of-day observation bias issue that makes it look like temperatures are dropping  when they are actually rising.  In a nutshell, the observers used to go out and set the min/max thermometer manually every day in the early evening.  But that meant that the max for the next day may in fact be the max for the current day if the next day is cooler.  Then switching to morning resets the min would be duplicated if the next day's min was warmer.  Finally switching to electronic min/max there would be no bias but unfortunately the new data would be incompatable with the old data.  So instead I simply find the max temperature for each month.   Thus the time-of-day issue is moot.
Hotter days out west are a problem, creating for example higher rates of evapotranspiration and soil moisture decrease.  That's the main reason that the average drought (which starts and ends naturally) is more severe than the average drought years ago.  Up to 4F rise per century (with a lot of natural variation) in Utah:

In the east it's a different story.  Summer extremes are dropping.  You would not know that from reading the news, but their sources are usually ASOS sensors at airports.  Those run hot especially in hot weather.  And as we know they are next to the runway.  Here's a rural Maine station with a 5F drop per century in July:

Here are all the stations I did with a link to the data and code https://virtualcoinclub.com/wx/temp/. The data came from here: https://www.ncdc.noaa.gov/cdo-web/search  I did not cherry pick any data.  I simply sorted by oldest starting date and looked for stations that continue to the present day with 98% or more data (although I prefer 99% or 100%).  The first station I found with that criteria with the list sorted by oldest starting date was the one I downloaded and plotted.

6

Weather / Atlantic hurricanes are slowing down at landfall

« on: October 12, 2020, 0206 UTC »

There are a couple papers on slowing hurricanes.  One is about Atlantic hurricanes https://www.nature.com/articles/s41612-019-0074-8 and seems reasonable by my analysis.  They compute the mean rate of motion and show a drop in the slowest 5% of storms (fig 1b) and corresponding changes in direction (fig 1d).  However I do not believe the first part of the title "Hurricane stalling along the North American coast ..." is quite accurate.  I believe that stalling, where steering currents break down, and the hurricane drifts, then stops or almost stops, then drifts in another direction, is a rare weather event.  I believe those weather events are determined by natural cycles.  The slowing is real though, in the 4-6 mph hurricanes, and the reasons in the paper are plausible.

The analysis I did uses the same HURDAT2 data they used.  However I did not use HURDAT2 landfall markings because I do not believe those are consistently applied.  In particular landfalls at non-6-hour intervals are captured in newer but not older data.  Instead i used the NOAA GLOBE data which contains elevation data in which any elevation greater than -500 meters is non-ocean.  The gist of my algorithm is finding all  landfalls for each storm and then finding the slowest.  In the case of Claudette the second landfall was slowest.  For Wilma the first of three landfalls was the slowest. Dorian had only one point of landfall (7 meters).  Thus I will miss some landfalls on very small islands.  But those misses will be consistent through the record and not affect the trend other than possibly lowering its statistical validity a small amount.

The code is linked on the webpage: https://virtualcoinclub.com/wx/slow/   Here are the 70 year trends  The 0-3 mph landfalls ("stalled") do not have a statistically significant trend.  There is too much natural variation in those types of weather patterns.  But the 4-6 mph landfalls have increased  significantly.   The 0-6 landfalls as a whole have risen from 40% to 50% of total landfalls.  That means hurricanes and tropical storms in the Atlantic basin are generally slowing down at landfall.

7

Weather / Atlantic hurricanes corrected for improved detection

« on: September 30, 2020, 0052 UTC »

I read the HURDAT database with some python and looked for hurricanes that came close to land.  To get figure out land, I used NOAA GLOBE which is linked on the page.  The NOAA GLOBE data is a set of simple raster files, binary files with nothing but 16 bit integers representing meters of elevation for every square km of the globe  (roughly square km for lower latitudes, varies at higher latitudes).  If the meters of elevation is -500, then that's ocean.  The code to read in the raster files and determine elevation  for any lat/lon is very simple using the rasterio library.  I think installing that sucked in GDAL which is a pretty large library, but I seem to have both on my mac now.

Anyway once I figured out how to determine if hurricanes came close to land, or not, then I assume that  percentage should be constant.  There's no reason for that percentage to change.  I used it to correct the tropical storm and hurricane numbers in the HURDAT database.  Web page with explanation, source code and instructions is here: https://virtualcoinclub.com/wx/count/count.html    There's a very striking increase in very strong hurricanes (high end cat 4 and cat 5):

Also tropical storms are increasing.  Other hurricanes are flat.  Some of the TS increase may be policy changes.  I can't correct for those, only presumed observation changes based on the heuristic above.  If you want to recreate the graphics you'll need to install matplotlib.  It's a very nice package for simple bar graphs, line graphs and scatterplots.

8

Weather / Hurricane Teddy

« on: September 17, 2020, 0234 UTC »

Any Mainers?  Remember when the euro was the only model predicting Sandy's left turn?  Now the euro is an outlier again predicting  Teddy will go left into eastern Maine: