Thursday 19 September 2019

European Meteorological Society Meeting highlights on station data quality and communication #EMS2019

Last week I was at the Annual Meeting of the European Meteorological Society in Copenhagen, Denmark. Here are the highlights for station data (quality) and communication.

Warming in Svalbard

Øyvind Nordli and colleagues estimated the warming on the Arctic island of Svalbard/Spitsbergen; see figure below. They use the linear red line to estimate the total warming and claim 3.8°C of warming. I would say it warmed a whooping 6°C (11°F). The graph already mostly shows that such a linear trend based estimate will underestimate the total warming.

The monthly data was already published in 2014. At that time I would have called it 5°C of warming; recent years were very warm.

They put a lot of work in the homogenization; even made modern parallel measurements to estimate the effect of past relocations of the station. The next, almost published, paper is on the daily data, so that we can study changes in the number of growing, freezing or melting days.

Warming in the tropical hot spot

There is a small region high up in the air in the tropics that is dear to many climate "skeptics", the tropical hot spot. It is one of the coldest places on Earth which warms strongly when the world is warming (for any reason). Because some observations do not show as much warming there, climate "skeptics" have declared this region to be the arbiter of climate truth, these observations and satellite estimates to the be best we have and most informative for the changes of our climate.

The warming for a GISS model equilibrium run for a 2% increase in solar forcing showing a maximum around 20N to 20S around 300mb (10 km).

Back to reality, it is really hard to make good measurements of such a cold place starting at such a tropically warm place. The thermometer needs to be reliable over about 100°C of range. That is a lot. It is not that easy to launch a weather balloon up to such heights and colds; the balloon will expand enormously. While the countries making these measurements are among the poorest on Earth.

What I had not realized is how few weather balloon make it to such heights. A poster by Souleymane Sy showed this; see Figure below. For trend estimates the sharp drop off above the pressure level of 300mb is especially very worrying. Changes in this drop off level due to changes in equipment can easily lead to changes in the estimated temperature. There is a part of the tropical hot spot below 300mb; that would be the part I would prioritize in trend estimates.

Number of radiosonde stations recording at least a given percentage of temperature and relative humidity monthly data at mandatory pressure levels since 1978 to present time for the Tropics (20° North to 20° South).

Weather forecasts in America and Europe

Communication at the EMS mostly means presenting the daily TV weather forecasts. There was a lovely difference between American and European presenters. The Americans were explaining how to dumb down your forecast as much as possible. A study found that most high school students in Alabama could not find their county on a map of Alabama; so the advice is to put a city name on every number on the map. The Europeans presented their educational work.

Our Irish friends had made three one-hour shows about the weather on consecutive days between 7 and 8pm when normally the soaps are running; light information in a botanical garden with a a small audience.

German weather presenter Karsten Schwanke got a price for his educational weather forecasts, which add information on climate change; for example in case of Dorian show the increase in the sea surface temperature. For Schwanke providing context is the main task of TV weather, the local numbers are available from a weather app.

Karsten Schwanke explains the relationship between the jet stream, wild fires and the drought in Europe. In German.

An increasing problem is fake weather predictions. Amateurs who can make a decent map are often seen as reliable sources, which can be dangerous in case of severe weather.

American weather caster Jay Trobec reported that it is common to have weather information three times during a news block, before, in the middle and at the end. In Europe you just get weather at the end. In America the weather is live, a presenter explaining everyone should leave the disaster area they went to to make this live broadcast. In Europe typically reported and the weather shown in videos. Trobec stated that during severe weather people watch TV rather than use the internet.

Live hurricane weather. :-)

The difference is likely that there is not that much severe weather in Europe, you normally watch the weather to see if you have to take an umbrella with you, rarely to see whether your house will soon be destroyed. Live weather would be looking at a weather presenter slowly getting wet in the drizzle. In addition, European public media have an educational mandate, they are paid by the public to make society better, while in America media is commercial and will do whatever makes money.

In the harbor of Copenhagen is the famous little mermaid. Tourists ships went to see it, had to keep quite a distance and could only show her back. Typically the boats only waited a few seconds because there was nothing to see. But due to commercial pressure they had to have the little mermaid on their tour schedule. They follow demand, whether the outcome is good or not.

Short hits communication

  • When asked what 30% probability of rain means for a weather prediction most people gave the wrong answer: that 30% of the region would experience rain. The formally correct answer is that 30% of the cases this prediction is made you will experience rain. To be fair to the people, I often explain the need to give such a percentage by saying that in case of showers we cannot tell whether it rains in Bonn or Cologne. I feel this is quite common explanation and the main effect. The German weather service is working on providing more detailed probabilistic information to weather brigades. That seems to be appreciated (and they answered the question mostly right).
  • Amanda Ruggeri won the journalism award for her story on sea level rise in Miami, which was reviewed by ClimateFeedback who found its scientific credibility to be "very high". Recommended read.
  • EUMETSAT operates the European satellites once in space. They also make MOOCs ([[Massive Open Online  Courses]]). They have one on the oceans and one on the atmosphere. They are a great way to introduce these topics to new people and in future they plan to do more live. 
  • Climate change is seen as the Top Global Threat according to global polling by the Pew Institute. In 2018 67 percent of the world sees climate change as a major threat to their country.  
  • During a Q&A someone remarked that it would be good to talk about the history of climatology more because people are spreading the rumor that climatology is a new field of science trying to make it sound less solid.
  • In case I have any Finnish speaking readers, Finland has a two-yearly bulletin on weather and climate, recently revamped.
  • Copernicus has a "new" journal on statistical climatology, ideally suited for homogenization studies: Advances in Statistical Climatology, Meteorology and Oceanography (ASCMO). It does not have an Impact Factor yet, but seeing the editorial team and reading a few articles it is clearly a serious journal and likely will get one soon. It is worth building up such a journal to have an outlet for statistical/methodological studies on climate. We already published there once; post upcoming.
  • Did you know about STATMOS, an American Research Network for Statistical Methods for Atmospheric and Oceanic Sciences?

Short hits observations

  • I had seen people use measurements of cosmic rays to estimate the soil moisture between the surface and the probe, but it was new to me to use it to measure the amount of snow on top of a glacier.
  • Michal Zak of the Czech Hydrometeorological Institute and colleagues had an interesting way to estimate how urban a station is. They computed the absolute day to day differences of the maximum and of the minimum temperature and subtracted them from each other. If the maximum temperature varies more a station is likely urban, if the minimum varies more it is likely rural. For Prague and its surrounding the differences between stations were not particularly large and smaller than its seasonal cycle, but it could be a useful check. This could also be a measure that could help one to selected climatologically similar pairs of stations in relative statistical homogenization.
  • The Homogenization Seminar in Budapest will be from 18 to 21 of May 2020. Announcements will follow, e.g., on the homogenization list. (I should write less mails to the homogenization list; at EMS someone asked to be added to the homogenization newsletter.) 
  • Carla Mateus studied Data Rescue (DARE) as a scientific problem. By creating one really high quality transcribed dataset as a benchmark, she studied how accurately various groups transcribed historical observations. Volunteers of the Irish meteorological society were an order of magnitude more accurate (0.3% errors) than students (3.3%). Great talk.
  • Our colleagues from Catalonia studied the influence of the time of observation. Manual observations tend to be made at 8am, while automatic measurements often use a normal calendar day. This naturally mattered most for the minimum temperature. With statistical homogenization the small breaks are hard to find, to formulate it diplomatically.
  • Monika Lakato has ambitious plans to study changes in hourly precipitation in Hungary motivated by increases in rain intensity (precipitation amount on rainy days).
  • Peter Domonkos studied how well network-wide trends are corrected in the new MULTITEST benchmark dataset (the presentation as pptx file). He found that his method (ACMANTv4) was able to reduce this error by about 30% and others were worse. It would be interesting to study what is different in the MULTITEST dataset or this analysis because the results of Williams et al. (2012) are much more optimistic; here 50 to 90% of the trend error is removed for similarly dense networks.
  • ACMANTv4 is on GitHub and about to be published. Some colleagues already used it. 

Meteorological Glossaries

Miloslav Müller gave a talk on the new Slovak meteorological glossary, listing many other glossaries. So I now have a bookmark folder full of glossaries.
To finish with a great audience comment on the last day, not directly weather related: "In Russian education everything is explained, you do not have to remember or study." I loved that expression. That is the reason I studied physics, I also loved biology, but you have to remember so much and my memory is very poor for random stuff like names of organisms. When you understand something, you (I?) automatically remember it, it does not even feel like learning.

Related reading

The IPCC underestimates global warming. This post explains why using linear regression underestimates total warming

Annual Meeting of the European Meteorological Society