User:Soap/Climate data essay

There is currently no reliable source for climate data either online or offline; every site that I can find has at least some extremely suspect data. Examples below. If some of this data seems believable to you, note that in some cases the data is only off by 3 or 4°F, so it may not seem suspect ... but while it's reasonably common to find one station reporting a few degrees above or below its surroundings on a given day due to random fluctuations, it will never persist on every single day and become part of the station's climatology.

How does this happen? There are many ways. For example, clearly flawed data such as 140F in Alaska may be omitted from the list of daily record highs but still used to calculate the normals.

Background
UPDATE! I havent added much to this page lately, but Im still here! I think I found another malfunctioning weather station. THis time, it doesnt even have the usual "These values are not official NWS readings" disclaimer. Not that I can see anyway. Or maybe North Conway is about to open up a tropical resort next to Mt. Washington where palm trees and pineapples sprout out of the ground when every town around them is getting six inches of snow.

Eastport
Also it looks like the old Eastport water temperature data is gone now so you'll just have my word alone to go on. The new graph begins seemingly at the time the station was repaired. Although evidence of the problem can be seen at the tail end of the old data right before the cutoff begins. The text file that was the source of this graph is also diapered up with "999°C" readings now.

Update
Weatherbase is so far wrong that I'd propose blacklisting it at this point if not for the fact that I believe it would lead to frustration among new editors.

Here is just one of over a hundred examples of data on Weatherbase that is not merely wrong, but so obviously wrong that with a moment's thought even a layman with little knowledge of climate should realize something is wrong. [http://www.weatherbase.com/weather/weather.php3?s=336940&cityname=Silverpeak-Nevada-United-States-of-America&units=us 96F average high temperature .... at 12000 feet elevation!]

Examples of bad data
This section includes only data recorded in archives, not instantaneous weather station malfunctions like the day that Abingdon, Maryland recorded an afternoon temperature of –∞°F.

Weatherbase
Many of these readings are for Maine because I live here and check local weather more often. I'm sure people could find as many errors per unit of data for everywhere else. '''UPDATE: I have not added to this page significantly since around 2015, but I can see that the problems are still there and added one more entry just now in July 2022. Why havent they improved? My guess is that commercial websites simply have no interest in whether their data is accurate or not.
 * Weatherbase
 * Bishop, California is somehow an island of ice in the middle of the desert despite not being on higher ground. FIXED!
 * Calling this ludicrous is too generous.
 * Puerto Cortés, a town along the west coast of Mexico, has an average annual temperature of only 32°F (a little bit colder than Duluth, Minnesota). Not exactly great weather for surfing, is it? The source of the problem is Weatherbase's long-standing bug of using "0°F" as a placeholder for incomplete data.  A few months of the year thus have their high temperatures entirely as 0°F, and since this makes the high temperature data for the full year also incomplete, the average annual high temperature becomes 0°F as well.
 * Kneeland, California's highest recorded temperature in March is somehow lower than its average high temperature.
 * Metlakatla, Alaska has winters 40°F colder than its immediate surroundings, for no apparent reason. Compare Annette which is just a few miles away.
 * Cape Girardeau, Missouri has bad temperature data including a few months where the average high temperature is higher than the record high temperature and some other months where the average mean temperature is higher than the average high temperature. Also the site claims that Cape Girardeau has an average high temperature, averaged over all months, of 4°F (colder than Barrow, Alaska).
 * Sutton, Vermont is another station with similar situation: average highs that are higher than the record highs, an impossibly warm average December high, and an annual average monthly high of 4°F.
 * Slubice, Poland looking more like a town in the high desert of California, at least in terms of temperature.
 * weatherbase showing subfreezing temperatures off the coast of California in June (I know there's a cold current there, but this is IMPOSSIBLE.) FIXED!
 * weatherbase claims 68°F in Antarctica in the middle of winter.
 * weatherbase has data for a nonexistent town in Maine. (Further investigation based on the latitude and longitude given shows that it's likely to be Portsmouth, New Hampshire).
 * 91°F in northern Alaska in December. (Other weird data is there as well, such as December being warmer than September and May being warmer than July ... the page says it was 21 years of records but the only way I could see these temperatures being believable is if it was really just a record of one very erratic year.)
 * UPDATE: Interestingly, they seem to have replaced the bad data with other data. Now instead of December being warmer than September, December is the coldest month of all.  Unusual, but if this is in fact just a one-year record despite saying that it's 21, it might in fact be accurate.  However the reading of 91°F in December remains in the table.  This may be a data recognition error, as 91 is also the highest temperature recorded in June, so perhaps their data scraping program mistook the December field for the annual field.

Note that this section has the most examples of errors because it has the most data ... so I don't believe that weatherbase is the worst of the bunch, by any means. Most of the cities listed above don't even exist in the data sources listed below.
 * It's never been warmer than 85 degrees in Bangor, Maine? Or colder than 0? Come on.
 * Moqiang, an island of cold in China's Shanxi province. The place has an elevation of 3199 feet, but this is much too little to result in such extremely cold weather.
 * Berbera, Somalia is a manually operated station that often only takes one reading per day, but nevertheless is recorded as a high and low for the day by weatherbase.com.
 * Chambéry, France has identical values for "high temperature" and "average temperature" ... so which one is it?
 * Godbout, Quebec has 881 cm (347 inches) of rain every year. FIXED!
 * Freeport Illinois somehow has winters 10°F warmer than surrounding areas.
 * 9°F in Wales in May. (Nearby stations show temperatures around 30°F).
 * 11°F in Scotland in June.
 * There's no such place as Plzen, Germany (though there is a Plzen in the Czech Republic).
 * Cold winters in Roosevelt, Utah. FIXED!
 * Cairo, IL ... not really bad weather data, but the geographical coordinates for this city are quite a bit off ... Google Maps places these coordinates just north of the Tennessee/Mississippi border. UPDATE: Now the link takes you to "New Canaan, Mississippi", seemingly with the same weather data that had been reported for Cairo, IL.  And meanwhile Weatherbase no longer has data for Cairo, IL.
 * Anomalous data in San Marino, seemingly lagged one month behind nearby cities such as *Perugia and others here.
 * Wrong hemisphere.

'''It would appear that many of the errors involve substituting one row of data for another, or one column for another. For example, the surprisingly narrow range of temperatures in Bangor Maine seems to be due to their data scraping program mistaking NOAA's "highest mean maximum" and the corresponding minimum for the record highs and lows.'''

WXUG

 * Weather Underground (aka wunderground, WXUG)
 * Weather Underground climate data for much of New England shows various stations with surprisingly cool temperatures in summer, even giving inland stations lower temperatures than the coast ... I personally suspect that the cause of this problem is an entire year of data with 32°F for the entire year somehow snuck into the data and was never discovered by whoever was responsible for uploading the data. Note that the anomaly only appears on stations for which the data goes back to 1996.  I emailed the Help desk of wunderground.com but never got a response.  Since the conversion to the new layout, the data hasn't gotten better, and in fact now there are stations in New England with ridiculously high annual average rainfall totals, such as 150 inches, with some months averaging over 20 inches (and not in winter, so it can't be explained as snow.)
 * Pierre SD hits 0°F in July 1989 because they were using a thermometer that only had 2 digits. The data nevertheless is recorded by WXUG as a new record low for the month and enters the city's climate data, giving them a monthly average high of 68°F (21 degrees below average).
 * This North Conway, New Hampshire data may not seem suspicious at first, even to locals, but comparison with nearby towns leads me to believe the station is malfunctioning. 40F average high as of Jan 11, but nearby Plymouth at a similar elevation has merely 24F.

weather.com
(now united with WXUG)


 * Weather.com
 * Baltimore is an island of heat, with average summer temperatures exceeding anything you'll find along the rest of the East Coast (or even inland) until you reach Georgia. This heat anomaly is too much to be explained by the station being located in the deep city, which anyhow usually only makes a noticeable difference for night temperatures.  All other climate data sources that I can find show Baltimore with temperatures similar to Washington DC. FIXED!

EC

 * Environment Canada
 * Environment Canada (EC) has claimed that Winnipeg has the coldest winters of any city in the world over 600,000 in population, yet there are cities such as Ulan Bator which are both larger than Winnipeg and colder in winter. The page containing this claim has now been removed from EC's website, yet as of March 17 2012 is still being used on Wikipedia. (see this link).  Note that the other cities which qualify are all in Russia and China.
 * They also claim that it was 88°F on March 29, 1926 in Beaver Creek, BC.
 * Humidex of 23 in January in Eureka, Nunavut. Apparently someone dropped the minus sign.

NOAA

 * eielson visitor center, a mtn station in alaska (3700 ft), is warmer than Miami in Aug 2005, including recording 140°F in both july and august. base link here.
 * Here is a graph of ocean temperatures off the coast of Eastport, Maine, showing them to be far warmer than the surrounding areas, and even warmer than the water off the coast of Northern Florida. Bad readings begin in early October.
 * Update: Eastport bathers are now enjoying a tide of 112°F water on the beach as of February 4th 2010, down from a high of 127°F on the previous day. (text format)
 * Further update: 151°F on February 24 and no sign of a repair anytime soon.
 * Update as of March 24, 2010: Finally repaired. Temperatures maxed out at 164°F early in the morning, also hitting an all-time record low of —183°F a few hours later.


 * Tsunami warning for Connecticut sent by Alaskan weather station.
 * 111.47 feet of snow fall in 24h on October 19, 2015. The original data gives it as 1337 inches. Perhaps it was not snow but sleet ?

Eielson
If this data were verified, I would move here:

This station is not listed on their map of Alaska, and I only stumbled upon it by chance clicking around an unlabeled map, leading me to believe that they know the data is wrong but provide it nonetheless for those who are curious. Fahrenheit/Celsius conversion errors cannot explain such phenomena as March being colder on average than February, December being often warmer than November, and the coldest month on record being March 2007. More obvious errors exist, as well ... DRI provides year-by-year climate data showing that the 140°F above is part of a string of similarly unbelievable readings that went on for months.

It is possible that the many repeated values like -41F are just this station's way of filling in where the station was down for a period of time. I have no idea if there was an original correct -41F which echoed throughout the year or if even that was incorrect.

Though I thought this data is so far off the scale that it should be obvious to even a casual viewer, I understand that laypeople may not see at a glance what's so obvious. Just as we once had Aleknagik with the climate of the Aleutians, the Eielson data (without the extremes) made it onto Weatherbase and someone added it to Wikipedia. The original IP editor who added it likely lived in Jakarta, Indonesia and so I could certainly see how they'd not know the difference between "cold" and "very cold". But Im a bit disappointed that we let the bad climate data sit in the article for five and a half years.

Anaktuvuk Pass
But the -11F in June for Anaktuvuk Pass does seem to be real. So says a source which also rejected -17F in June and similar temps in Aug and September.

Paper sources
The Times Books World Weather Guide  has had questionable data that seems to match no local weather station but in the paper edition Im familiar with there was no way to prove or disprove the given data. One example is an anomalously warm climate for Monaco, but this climate data now appears on Wikipedia alongside a somewhat cooler one. It is possible that the warmer readings are from a microclimate.
 * Paper sources
 * Just in case you thought only Internet sources had these problems, the World Book Encyclopedia (a children's encyclopedia) mixed up daily and monthly highs and lows, and has maps drawn with no attention to topography, as if tall mountains and deep valleys had identical climates, and Rand McNally road atlases have had typos such as —5 instead of (probably) 5 for the December low in Anchorage.  The Cairo   IL problem may have originated from an old encyclopedia.

Twitter
Claims of a very anomalous high temperature in Beijing in Feb 2021 have been circulating on Twitter, but so far I have seen no mention of it   outside social media.

Other info
It bothers me that the record low temperatures for most US states are in such perfect round numbers: Alaska −80°F, Montana –70°F, Minnesota –60°F, Maine/NH/VT –50°F. Why not –79 or –81? And these are with modern equipment mostly, its not just a guy with a graph saying "Eh, this marker's pretty fat, I can make -42 look like -50". Maine was too embarrassed to be left with a -48F record low while puny little NH and VT both had exactly -50.00000000F so they put up thermometers in the mountains and said "Hey look! We found a lake in the very northern corner of the state where the temperature was exactly -50F last night, so now we can just be buddies in the -50 Club!" They didnt go lower because they already tried that in the 1990s when they started screaming "Allagash -55F!!!" and people believed it until someone from Massachusetts got tired of it and peeled the fake numbers off the thermometer.

Warsaw, Missouri climate data is a bit suspicious. Never been colder than –8F in January, but yet they managed to get all the way down to –40F (exactly –40! not a bit less) in February. Then the record high for July is 118F, far far higher than the other summer months.
 * NOTE, this appears to have been due to the extremes data being a mix of two sources. One source was likely for a very short record period, and did not have much to talk about, whereas the other included at least the all-time record high and the all-time record low.  The climate chart has since been replaced with something more harmonious, though with no source.  WxBase resembles neither the old Wikipedia chart nor the new one.
 * Comparing it to nearby stations, I'd say the 118F high and the —40F low are likely legitimate, though remarkable.

Links
Old but still relevant discussions on various Wikipedia discussion pages:
 * Administrators%27_noticeboard/IncidentArchive831
 * Wikipedia_talk:WikiProject_Cities/Archive_9
 * User_talk:CambridgeBayWeather/Archive28
 * User_talk:Soap/Archive_4