- climateprediction.net: Europe’s Record Heat Directly Tied to Climate Change
18.12.2014 12:39 Uhr
As 2014 comes to a close, Europe is virtually certain to lock in its hottest year in more than 500 years, and according to research by three independent teams of climate scientists, including climateprediction.net, the record can be closely attributed to climate change.
The three groups, from the UK, the Netherlands and Australia, each using a different method, found that Europe should best its previous heat record set in 2007, and that setting that record has been made at least 35 to 80 times more likely by the manmade rise of greenhouse gases in the atmosphere.
“All the groups are concluding there is a very substantial influence of human caused climate change on increasing the chances of setting a new record in 2014,” said David Karoly, an atmospheric scientist at the Australian National University and one of the project researchers.
The three teams in this project — from the Royal Netherlands Meteorological Institute (KNMI), the University of Oxford, the University of Melbourne and the Australia National University — are working as part of a new initiative with Climate Central to examine extreme weather events and determine climate change’s role in them. In the past, extreme-event attribution took much longer to determine, but the new effort aims to establish any link between climate and extreme weather events much sooner after events happen.
Extreme heat events are one of the signature extremes expected in a warming world, according to climate model projections, be they a record hot summer in Australia or a record hot year for any part of the planet or the globe as a whole. And in general, the link between climate change and extreme heat is clearer and easier to discern than for other forms of extreme weather, such as extreme downpours or drought.
Different Methods, Same Conclusion
In this analysis of extreme heat, each team used different methodology and came up with slightly different numbers for the 2014 record projection and the likely role that climate change played, but all showed a heavy influence of manmade climate change on the recent heat records.
Globally, as well as in Europe, nine of the 10 hottest years ever recorded have all occurred since 2000. No annual cold record has been set in Europe since 1956 as warming is making hot years hotter and more frequent. No year has been record cold globally since 1911.
This year has been particularly warm, driven largely by extraordinary warming in large swaths of the Earth’s oceans.
Of course, the world doesn’t warm uniformly. In the U.S., for example, the eastern part of the country has had a fairly cool year while parts of the West are on pace for record temperatures. And many other parts of the planet have been baking, including Europe.
Nineteen European countries are on track for their hottest year on record: Austria, Belgium, Croatia, Czech Republic, Denmark, France, Germany, Hungary, Iceland, Italy, Luxembourg, The Netherlands, Norway, Poland, Serbia, Slovakia, Slovenia, and the United Kingdom. Sweden is projected to tie its record high.
Our team at the University of Oxford, led by Myles Allen and Friederike Otto, used thousands of iterations of regional climate models embedded within larger global models to examine more localized weather events.
For example, our data for Germany determined that what was once a 1 in 80-year heat event has now become a 1 in 7-year event, making it 10 times more likely due to global warming.
Using our volunteers’ computers for our weather@home project, we simulated possible European weather based on the observed global ocean temperatures. At the same time, we also simulated a 2014 where there is no human-influenced climate change. Comparing those two “worlds” we found that the 2014 European temperatures were much more likely in the world with climate change than the one without.
“It is important to highlight that Oxford’s result crucially depends on the 2014 global ocean temperatures. The same study using 2000-2011 conditions gives a different result although the anthropogenic warming is the roughly same in these years,” said Dr Friederike Otto.
“When looking at smaller regions in Europe, we notice that there is a higher variability of temperatures,” Karsten Haustein, who conducted the analysis, said. “For example, in central Europe we found that the probability of reaching the observed 2014 temperatures is about 40 times higher. In an even smaller region such as the UK, we found that the probability has increased by a factor of about 10.”
Research from the Netherlands
As a continent, Europe is set to break its record by 0.12°C, according to new estimates made by Geert Jan van Oldenborgh at KNMI, and explained in the Climate Indicator Bulletin released by the World Meteorological Organization’s Regional Climate Center for Europe and the Middle East.
Van Oldenborgh used both modern and early temperature records, as well as sources like tree rings, which can act as a proxy for very old temperatures, to observe Europe’s temperature records back to 1500 and determined that 2014 will almost certainly be the warmest year Europe has experienced during the past 500 years.
When looking at the record of temperatures over time, “you see that the world is getting warmer and hence Europe is getting warmer,” van Oldenborgh said.
Using those observational records, van Oldenborgh’s analysis concluded that global warming has made a temperature anomaly like the one observed in 2014 in Europe at least 80 times more likely.
At the beginning of the 20th century, before the global warming signal emerged from the noise of year-to-year variability, the chances of seeing a year as warm as this one were less than one in 10,000, van Oldenborgh said. “It would be almost impossible to reach these temperatures a century ago.”
An Australian team led by Karoly and Andrew King of the University of Melbourne answered a slightly different question, which was to ask how much more likely it was to break the 2007 record with warming than without. To do this they used the same model simulations used in the Intergovernmental Panel on Climate Change’s most recent report.
They found that the accumulation of greenhouse gases, which increases the chances of a record warm year every year they accumulate, made such an event 35 times more likely.
The differences in the numbers calculated by each group are a product of the slightly different questions each team asked, as well as the inherent uncertainty in these sorts of measurements, given how rare extreme events are, van Oldenborgh said. But for this heat event, “these numbers are very close given the uncertainties,” he said. That agreement across studies “gives us confidence in the result that global warming made a temperature this high in Europe much more likely.”
The results are also in line with previous work of Karoly’s that found that the record warm year Australia experienced in 2013 was “virtually impossible” without the influence of global warming, as he put it.
“What we’ve been finding is that there’s a very clear signal on human influences on increases in the likelihoods of record temperatures,” Karoly said.
Download a fact sheet about European Attribution for 2014.
- climateprediction.net: New publication on the experimental setup of weather@home
30.10.2014 11:48 Uhr
A paper detailing the model development, experimental setup and validation of the weather@home project has been published in the Quarterly Journal of the Royal Meteorological Society.
The paper is open access and is available to read here:
The paper explains the background to climateprediction.net and how the project is using regional climate modelling to answer questions about the attribution of extreme weather events – “Was the event caused by anthropogenic climate change?”
weather@home – development and validation of a very large ensemble modelling system for probabilistic event attribution
Neil Massey, Richard Jones, Friederike Otto, Tolu Aina, Simon Wilson, James Murphy, David Hassell, Hiro Yamazaki and Myles Allen
Quarterly Journal of the Royal Meteorological Society, DOI: 10.1002/qj.2455
- climateprediction.net: New publication about extreme summer rainfall in England and Wales
14.10.2014 14:38 Uhr
A new publication by our Science Coordinator, Dr Friederike Otto, has been published in Climatic Change about our recent research into extreme summer rainfall events in England and Wales as an example for probabilistic event attribution.
In the summer of 2007 England and Wales experienced very heavy flooding. Summer precipitation and subsequently flooding are harder to model than winter or autumn rainfall and this recent publication highlights this again. We look at different possible drivers of high precipitation in summer and do not find a conclusive signal apart for the July precipitation which might have been exacerbated due to anthropogenic forcing. However, the focus of the paper is not so much the attribution of the extreme precipitation to external climate drivers but the quantification of the uncertainties involved in such a study. The paper is part of a special issue dedicated to exactly these issues which will soon be published in full.
Read the full article, which is open source, here.
- Collatz Conjecture: Corrupt Index Caused Authentication Errors
05.03.2015 23:58 Uhr
Summary: An index on the user table in the database was corrupt and has been fixed. If you have issues or notice something that is out of sync somehow, please let me know. Details: The email address index on the user table was corrupted. This kept people from being able to authenticate when connecting. Ever heard that the definition of lunacy is doing the same thing over and over and expecting different results? If that is true, some of you need some serious counselling. ;-) When people couldn’t connect, they attempted to authenticate and BOINC’s logic is that if it can’t find you, you must be new. So, it created another record because the index used to look up your existing record was corrupt — or for some lunatics, another and another and another and another… That resulted in multiple user records with the same email address and because of that, I couldn’t just drop and re-create the corrupt index because there were now duplicate data values that needed to first be removed. So, after cleaning up hundreds of records each of which had to be manually evaluated to determine whether any other data was associated with the user record (since BOINC doesn’t have any data integrity between tables because it uses no foreign key relationships in the database it will allow me to delete records that do have data associated with them which means I have to be very careful when fixing things). Once they were all removed, I was able to rebuild the index and allow everyone to access the project again.
- Collatz Conjecture: Android 5, aka Lolipop or Android L not getting work
12.02.2015 00:50 Uhr
Android 5 was recently released but it is not currently working – mainly due to work fetch issues. We are working to identify and fix the issues as soon as possible.
- Collatz Conjecture: Prepping for Android 5
02.02.2015 21:00 Uhr
Because Android 5 requires PIE support, the android app has been split into two plan classes, one for Android 4.1 and later (a new app), and another for earlier versions (the existing app). Please let me know if you run into any issues.
- Collatz Conjecture: Database Is Back Online
21.01.2015 02:50 Uhr
During the nightly backup process, the database once again stopped responding and had to be restarted this morning. Rather than bring it back online as I have been doing, I spent the day trying to duplicate the issue, but to no avail. Before bringing it back online, the database tables were optimized and I checked to make sure the database is in sync with the BOINC file system.
- Constellation: Google Summer of Code 2015 is on!
21.02.2015 09:58 Uhr
The Google Summer of Code 2015 started and organization applied until today. So if you are a student and want to code during summer time for open-source projects, have a look on www.google-melange.com for organizations you would like to work for. Constellation had been part in 2013 and 2014 and it rocked! So we also applied for this year and you can find our projects on our ideas page. Stay tuned for more news!
- Constellation: Constellation is back online!
20.02.2015 22:36 Uhr
Good news everyone, we’re finally back online! We are sorry to have you experiencing a longer than intended server down-time. Due to the new Comet Trails app, we generated more than expected results, which caused some trouble with our system. So thank you very much for helping this great application and also benchmarking our server. We already tackled the main problems, but we still leave the server under monitoring. So please be patient with us, if the server is not yet running as it used to. By the way, it is still the same server we started with in 2011 plus some new storage. He served us well, but if you like us to have a new and reliable one, please be so kind and donate on our donation page. We will keep you informed here or on our “emergency channels“. And we will also have some other good news coming up. So, thank you for your patience again! Best regards, Andreas
- Constellation: TYCHO paper now published in Acta Astronautica! [online / Volume 108]
21.01.2015 13:14 Uhr
So honored! My TYCHO paper is now published in Acta Astronautica! I did that as my Diploma Thesis at the Institute of Space Systems at the University of Stuttgart and the paper was selected for Acta Astronautica during the International Astronautical Congress 2013 in Beijing. It’s a concept study about a communication relays satellite in Earth-Moon Libration point (EML-4) as a permanent infrastructure. So something similar Elon Musk will need for his SpaceX and Google “space internet” ;) You can get it for free until March 5, 2015 -> http://authors.elsevier.com/a/1QN6u_29e7yxw- or here Volume 108, March–April 2015, Pages 156–170
- DistributedDataMining: Recent problems
10.12.2014 22:19 Uhr
Hi there, recently, we experienced some unfortunate problems: 1. The server hardware is old and can barely deal with the increasing load. This frequently leads to an unresponsive server. Website visitors are mainly affected since the boinc clients deal with the problem by trying to reconnect later on. The reconnects increase the problem additionally. 2. After getting back from two weeks of vacation, I realized that parts of the dDM database crashed and had to be restored. As far as I can see, we suffered no loss of important data. Maybe, some changes of user profiles were not stored. 3. The amount of dDM data increases steadily and the corresponding backup takes much more time than in the early stages of the project. As a result, the announced maintenance periods are no longer sufficient. I will think about a new backup strategy. 4. Some users reported about validation problems. I guess, this was due to the issues mentioned above. Right now, the validation of WUs is almost finished. I apologize for any inconvenience caused by these problems. Cheers, Nico
- DistributedDataMining: Forum under spam attack
22.04.2014 16:01 Uhr
Hi there, the ddm forums are currently under heavy spam attack. I’ve just deleted the relevant posts. To prevent further spam, from now on users can only post new messages if they have at least one credit point. Cheers, Nico
- DistributedDataMining: dDM went offline
21.04.2013 12:46 Uhr
Hi there, this week, we had some serious hardware problems which forced us to switch over to a new server. Caused by hard disc failures some data was lost – but nothing important. Now, dDM is running on new hardware. We are still recovering data by replaying our backups. That’s the reason why, some WUs still finish with download errors. We are confident, that we will be fully operational at Tuesday. We apologize for any inconvenience. Cheers, Nico
- DistributedDataMining: Usage of 'ARM' technology for the dDM project
12.03.2013 18:12 Uhr
Hello, Mr. Schlitter, I really enjoy this project and would like to contribute using the CPUs on my mobile device(s). Projects such as PrimeGrid, Enigma, SubsetSum@home, OProject@home, WUProp@home, and others have successfully used ‘ARM’ technology, which I understand is open-source, and I have been crunching for these projects for about a month successfully. Is DDM currently compatible with this technology, or are there plans to allow crunching on such devices? Many of the projects listed above have noted a sharp increase in the number of users as well as amount of research that is completed. Are you able to make this work for your project as well? I would be honoured to test and/or participate in any way. Thanks again, StandardbredHorse
- DistributedDataMining: Team challenges in September
28.08.2012 08:04 Uhr
Hi there, recently, I got message that two teams are going to focus on the dDM Project in September. The first one is team BOINC@Poland the second one is team The Knights Who Say Ni. I am going to create some more WUs in order to satisfy the upcoming requests. Many thanks goes to both teams for their support of our research. Best regards, Nico
- DistributedDataMining: New application version 1.36 for Medical Data Analysis
09.07.2012 19:57 Uhr
Recently, I’ve released version 1.36 of our application for medical data analysis. The new version takes care of finding the location of java on windows clients. As a consequence, it shouldn’t be necessary to set the path variable manually. Best regards, Nico
- DistributedDataMining: New application version 1.04 for our Multi-Agent Simulation of Evolution
21.06.2012 14:27 Uhr
Today, I’ve released version 1.04 of our Multi-Agent Simulation of Evolution. As planned, we finished the tests of this new application. Consequently, the latest version is now available for all dDM members. The new version deals with some problems that were reported by our beta testers:
- The first version was responsible for frequent hard drive activities. This was caused by writing checkpoints and detailed log files during the test period. Now, we have reduced the frequency between checkpoint writing and the level of detail in the log files.
- From time to time, WUs finished with error code -148, which indicates that something went wrong while starting the WU. We investigated this problem and did some code changes. Right now, we are not sure, if this solves the problem. We release the new version anyway in order to get some more log information and test data.
I also reorganized the structure of our forum:
- I added a new section Multi-Agent Simulation of Evolution dedicated to this new research topic and moved some recent postings.
- I removed the CafÃ© section because it was hardly used in the past and didn’t provide any additional benefit.
Best regards, Nico
- DistributedDataMining: New Research Topic
23.03.2012 16:05 Uhr
Hi there, currently, we are in the final stage of releasing a new application which deals basically with the simulation of evolutionary processes. Using the Multi-Agent Simulation of Evolution we investigate the biological phenomenon of aposematism (also referred to as warning coloration). This term describes the evolutionary strategy of certain animal species to indicate their unpalatability/toxicity to potential predators by developing skin colors and patterns that can be easily perceived by them. Prominent examples of toxic animals with distinct warning coloration are poison dart frogs, coral snakes and fire salamanders. For tackling this interesting research challenge, we developed a distributed multi-agent model that simulates the dynamic interactions of predator and prey populations over time. By systematically testing different adaptation and learning strategies for the agents and exploring the parameter space of our simulation model using the computational power of the dDM project, we might be able to deepen the understanding of the aposematism phenomenon and the evolutionary paths leading to it. So far, dDM members won’t get WUs of this new application. Currently, we are finishing our final tests and distribute these WUs to preselected hosts only. Soon, I’ll send out these WUs to beta test members. After finishing our beta test the new application will be available for all members. Cheers, Nico
- DistributedDataMining: Website Restructuring
22.03.2012 18:58 Uhr
Hi there, as you may have seen, I’ve restructured the dDM website in order to gain clarity. The main changes are:
- I reworked the start site, which now provides a brief overview about the dDM project, its history and its objectives.
- The research challenges went to a separate page.
- I also created a new site which briefly reports about the dDM achievements.
- The menu structure was revised. All items related to the member account were moved the right.
- I also created four different color schemes, which can be freely chosen by the members (left bottom corner).
Your comments, suggestions and wishes are much appreciated. I am going to add some more minor web site features soon and would be happy to include some of your ideas. Furthermore, I’ll update and add some research related information. That’s it for now. Cheers, Nico
- DistributedDataMining: Download problems caused by hardware issues
26.02.2012 19:17 Uhr
Recently, we had some hard-disc issues and as a result the data files of our medical application are temporary not available. We are going to reconstruct the data using the latest backup. Since this will take some hours, just a few workunits will be available in the meantime. We do not expect any problems regarding the user data (e.g. credits) because our ddm database is not affected. Best regards Nico
- DistributedDataMining: Security Update to Drupal 6.24
08.02.2012 15:07 Uhr
Due to some reported security vulnerabilities I’ve updated the content management system of our dDM website to Drupal version 6.24. Please, report any problems related to the dDM website in our forum. Best regards Nico
- DistributedDataMining: Server Maintenance
20.01.2012 11:18 Uhr
Since we are moving the dDM-Server to new hardware the project website and all BOINC functions won’t be available on Sunday (2012/01/22). So far, we don’t expect any problems. There should be no need for any changes in the client configuration. Best regards Nico
- DistributedDataMining: Scientific contribution
09.01.2012 11:56 Uhr
As already announced in an earlier post, some new research results from our Medical Data Analysis application were presented at the 162nd Meeting of the Acoustical Society of America which was held in San Diego, CA from 31 Oct – 4 Nov 2011. Our poster contribution was titled “Identifying relevant analysis parameters for the classification of vocal fold dynamics” and received lots of positive feedback from conference attendees and initiated interesting and inspiring scientific discussions. In the presented work, we systematically investigated the influence of a set of control parameters on the classification accuracy of the automatic diagnosis system for vocal fold dynamics based on high-speed videos. The particular suitability of certain parameter combinations was revealed in this study, helping to further improve the practical application of our diagnostic framework. The poster was heavily based on the results that we got from the extensive experiments conducted with the massive help of the DistributedDataMining community. An abstract of this work can be found in the accompanying conference proceedings: J. Acoust. Soc. Am. Volume 130, Issue 4, pp. 2550-2550 (2011). Thanks a lot for your support! Best, Daniel
- DistributedDataMining: New WUs and MD5 download errors
11.12.2011 09:44 Uhr
Currently, I am generating new workunits for our medical application. It turned out, I was acting a bit careless, because as a consequence many members are getting MD5 download errors now. The reason is quite simple. During WU generation files are getting copied to the server. Sometimes these files are still existing because we use the some data files for multiple workunits. In these cases we just change the learning parameters or the experiment setup. Anway, overwriting these existing files leads to problems, because the old and the new file have different MD5 checksums and hence all WUs that are related to the old files error out with an MD5 download error. About 30.000 WUs are affected. I’ll take care of it but it might take a while to identify the failing WUs. There is a chance, that the dDM members error out the affected old WUs in the meantime. In that case, the new WUs should work fine. Lesson learned: Never generate new WUs if you are in a hurry – in the end it takes much more time and causes preventable trouble. I am sorry for the inconvenience! Best regards, Nico
- DistributedDataMining: Let dDM benefit from your Christmas shopping
07.12.2011 16:16 Uhr
Hi there, As you know, the dDM project does not get financial support by universities, research institutes or private commercial organizations. Thus, we depend on private help to keep the project and our research running. Since the dDM website became part of the Amazon Associates Program, the dDM project can benefit from your Christmas shopping at Amazons online shop. Amazon rewards dDM with up to 6% of any issued gift certificate and up to 10% of each sold store item. If you intend to buy some Christmas gifts at Amazon, please remember dDM and follow the URL provided on this page. The provided URL links to the closest Amazon store of your region and contains our dDM partner ID. In case the wrong store is chosen, please select the right one out of the given country list. The dDM project gets a reward for each item bought after using these links. As always, all Amazon rewards or PayPal Donations are used to pay the server rent, maintenance and internet traffic. The possible remain is used to finance our scientific publications or conference presentations. Perhaps, we can even buy some new hardware in order to replace the aged project server. Thank you in advance for all of your generous support! Best regards and have a nice Christmas season! Nico
- DistributedDataMining: New application version 1.35 for Medical Data Analysis
05.12.2011 13:22 Uhr
Recently, I’ve released version 1.35 of our medical data analysis application. I am happy to announce that this version solves the problem of never ending workunits. The new version is out there for almost a week and so far there aren’t any workunits that had to be killed manually by our ddm members. Best regards, Nico
- DistributedDataMining: New application version 1.34 for Medical Data Analysis
29.11.2011 12:00 Uhr
Today, I’ve released a new version for our medical application. Version 1.34 should be able to detect the java location automatically. Doing so, it shouldn’t be necessary to adapt the PATH variable manually. This was necessary because it seems that the path variable that we used until now was changed by an automatic windows update. As a consequence our wrapper couldn’t find java and all workunits failed. Some users might have received an email notification and the suggestion to install java even if java was already installed on their computers. I am sorry for the confusion and the inconvenience caused by the circumstances. The new version uses the variables JAVA_HOME, JRE_HOME and JDK_HOME, which are usually set during java installation. I did some tests and it worked fine here. Lets see how it works out there. Any feedback and error reports are welcome in the forum related to the medical application. Regards, Nico
- DistributedDataMining: Update on Medical Application
29.08.2011 12:38 Uhr
Hi there, Here’s a quick update on our medical application: the latest results look really good! Lots of interesting relationships between the analyzed input variables were revealed and relevant research questions could be answered. Parts of these results will be presented at a scientific conference in November. Expect more details on that soon! We’ve learned a lot from the results regarding the use of parameter optimization and feature selection – which now allows us to create much more efficient (and hopefully even more reliable) experiments. For example, we’re planning to apply sophisticated search methods from the field of evolutionary computation combined with a more statistically sound validation approach. For that purpose, we developed a new version of our application, which will run the new experiments much faster and will take over quite a lot of the data analysis steps that were done manually before. This new version is being tested extensively at the moment and will be available soon. But we’re also thinking about implementing completely new ideas and exploring different research directions. For example, one potential project will be focusing on agent-based modeling in the field of simulating the emergence of biological and social phenomena. Well, that’s it for now. Thanks again for your support! Best, Daniel
- DistributedDataMining: Pending credit
17.08.2011 10:32 Uhr
As you might have noticed we had some minor problems regarding pending credits in the past. Today, I’ve fixed this issue. The problem was caused by a safety mechanism that is uses to avoid benchmark cheating. The algorithm marks suspicious workunits for further inspection and the affected workunits were pending until I’ve checked them manually. Unfortunately, there was a slight error in the cheating detection heuristic and as a result much more workunits than necessary were marked. Today, I found the problem and corrected the code. At a first glance, there won’t be any more pending workunits.
- DistributedDataMining: New application version 5.01 for Time Series Analysis Application
16.08.2011 14:39 Uhr
Today, I’ve released version 5.01 of the Time Series Analysis Application. The new version uses the same wrapper technology that is already in use for the medical application. It overcomes several problems and decreases the fraction of failing workunits. So far, the new version supports 32&64 bit Linux systems only. Versions for Windows will be published soon. As always, comments and error reports are welcome in the forum.