Projekt climateprediction.net składa się z trzech odrębnych eksperymentów - pierwszy do zbadania używanego modelu, drugi do sprawdzenia jak modele odtwarzają klimat z przeszłości i trzeci, by w końcu przedstawić prognozę klimatu dla XXI wieku. Każdy z modeli, które dystrybuujemy będzie używany dla wszystkich trzech eksperymentów. Każdy rozprowadzony model jest niepowtarzalny i różni się od wszystkich innych na trzy sposoby: warunki początkowe w jakich jest uruchomiony, atrybutów, które wymuszają na nim jeden konkretny stanu klimatu i parametry, które tworzą rzeczywisty model.

Każdy model klimatu musi posiadać pewną liczbę przybliżeń, zwanych parametryzacją. Aby dowiedzieć się więcej na ten temat, kliknij tutaj. Zasadniczo oznacza to, że w modelu są pewne liczby, które biorą pod uwagę pewne stałe wartości, ale wartość ta nie jest znana na pewno i zakres wartości może również być prawdopodobny. Eksperymenty zbada wpływ na modelowany klimat 20-tu najbardziej źle rozumianych parametrów modelu - takich jak stosunek między liczbą kropel w chmurze a ilością faktycznych opadów (aby zobaczyć te parametry kliknij tutaj). Jest możliwe, że niektóre kombinacje parametrów mogą odtwarzać klimat z przeszłości bardzo dobrze, ale podają znacznie odmienne prognozy na to co może się wydarzyć w przyszłości. Niektóre kombinacje parametrów nie będą działały, dostarczając całkowicie nierealny klimat (na przykład Ziemia zamarza lub oscyluje między bardzo gorącym i bardzo zimnym klimatem co kilka lat) i prawdopodobnie psując model. Nie jest możliwe aby poinformować z wyprzedzeniem jakie będą kombinacje.

Niektóre rzeczy, które uważasz, że nie są częścią klimatu wywierają ogromny wpływ na zmiany klimatyczne - takie jak wybuchy wulkanów (po Pinatubo który wybuchł w 1991 popiół miał wpływ na klimat przez kilka lat), aktywność słoneczna i oczywiście skład atmosfery. Te zjawiska nazywamy mechanizmem wymuszenia. Jeśli się zmieniają, wtedy wymuszają zmiany klimatyczne.

"Trzepot skrzydeł motyla w Brazylii może spowodować tornado w Teksasie". Ten słynny cytat dotyczy faktu, że bardzo niewielkie różnice w tym, co dzieje się na świecie teraz może mieć ogromny wpływ na to, co stanie się w przyszłości. Ponieważ nie możemy mieć doskonałej wiedzy o tym, co dzieje się teraz (do skali poszczególnych motyli), oznacza to więc, że do przedstawienia pełnej prognozy wszystkiego co może się wydarzyć w przyszłości, musimy wziąć pod uwagę wszystko, co może się dziać teraz. Aby to zrobić, musimy użyć zakresu warunków początkowych dla naszych modeli kiedy rozpoczynamy prognozowanie klimatu.

 

 

 

Eksperyment	Cel	Metodologia
1 Czułość modelu na parametry.	Określenie odpowiednich zakresów parametrów.	Każda symulacja obejmuje 3 etapy: Kalibracja (15 lat) Przedprzemysłowe CO2 (15 lat) Podójne CO2 run (15 lat)
2 Symulacja lat 1920-2080	Ocena umiejętności modelu poprzez porównanie z klimatem z przeszłości. Prawdopodobieństwo przewidywania przyszłego klimatu.	Uruchomienie modelu z warunkami początkowymi, wymuszeniami i parametrami.

 

---

Ten eksperyment ma bardziej na celu uczenie się, w jaki sposób model reaguje na zmiany warunków początkowych i parametrów, niż na zdobywanie wiedzy o odtwarzaniu klimatu Ziemi. Z tego powodu używany model ma udoskonaloną atmosferę, ale uproszczony ocean (jednowarstwowy, "płyta" oceanu). Oznacza to, że niektóre elementy systemu klimatycznego (takie jak prądy oceaniczne, oscylacje El Nino) nie są odzwierciedlane, ale model przebiega dużo szybciej i dużo więcej obliczeń może być zakończone.

Wiedza uzyskana z tego eksperymentu o sposobie w jaki model reaguje na zmiany parametrów zostanie wykorzystana do opracowania kolejnych faz eksperymentu climateprediction.net - kombinacji parametrów które oczywiście nie działają, można uniknąć.

Eksperyment składa się z 3 etapów, a każdy model, który jest rozpowszechniany uzupełnia wszystkie 3 etapy w unikalny zestaw warunków początkowych:

Kalibracja (etap 1)

Faza 1 jest fazą kalibracji eksperymentu. W tej fazie temperatura powierzchni oceanu jest niezmienna. Ruch lub strumień ciepła, do lub z oceanów potrzebny do utrzymania oceanu w stałej temperaturze jest obliczany. Jest to proste rozwiązanie, aby mieć bardzo prosty model oceanu, który nie może przechowywać ciepła w sposób prawdziwy, głęboki. Złożony ocean może. Etap obejmuje lata 1810-1825.

Przedprzemysłowe CO₂ (etap 2)

Jest to etap kontroli. Wymaga to uruchomienia modelu na 15 lat z poziomem CO₂ w modelu atmosfery utrzymywanym na stałym poziomie sprzed rewolucji przemysłowej, 282ppm. W przeciwieństwie do etapu 1, tutaj temperatura powierzchni oceanu może się różnić w zależności od tego jak dużo energii ocean odbiera i wysyła. Można jednak śmiało założyć, że ilość ciepła krążąca w oceanie jest taka sama jak w etapie 1, więc ciepło strumieni obliczane w etapie 1, jest stosowane. O ile atmosfera zaczyna robić coś zupełnie innego, a bilans energetyczny na szczycie atmosfery zmienia się, to temperatura całej atmosfery powinna pozostać bez zmian. Jeśli jest to przypadek, średnia temperatura powierzchni na świecie powinna być w przybliżeniu stała i nie zmieniać się znacząco z roku na rok lub mieć tendencję do bardzo różnych temperatur, i mówimy, że model jest stabilny. Etap obejmuje lata 1825-1840.

Podwójne CO₂ (etap 3)

Na tym etapie poziom emisji gazów cieplarnianych jest podwojony, a model jest realizowany przez kolejne 15 lat. W dobrym modelu, atmosfera powinna dostosować się do tej zmiany, i uregulować w nowym stabilnym stanie równowagi (który może być taki sam, cieplejszy lub chłodniejszy). Etap obejmuje lata 2050-2065.

Wyniki będą wskazywać, jakie kombinacje parametrów działają (w zakresie wytwarzania atmosfery, która zachowuje się w podobny sposób do rzeczywistości i nie zamraża się lub gotuje, albo tworzy epoki lodowcowe w czasie kilku lat). Będziemy zatem mieć możliwość wykorzystania wyników aby dokonać wyboru parametrów w głównym eksperymencie.

Porównując etapy pojedynczego i podwójnego CO₂, możemy obliczyć wrażliwość klimatu w modelach - jest to różnica pomiędzy średnią temperaturą powierzchni na świecie w modelu z przedprzemysłowym CO2 a tym ze zdwojoną ilością CO2. Jest to użyteczny wskaźnik pokazujący jak zachowuje się model klimatu, choć jest nieco sztuczny, jako że wartość dwutlenku węgla w atmosferze nie jest stała dla 15 lat, ale zmienia się w sposób ciągły.

[Uwaga: eksperyment uruchomiony we wrześniu 2003 roku..]

W rozszerzeniu do tego eksperymentu, wprowadzona została 4 faza: Thermohaline Circulation Experiment na kilka miesięcy od czerwca 2004.

We wrześniu 2005 r. uruchomiono 5 fazę: Sulphur Cycle Experiment, która była realizowana przez kilka miesięcy.

Drugi eksperyment będzie używał pełnego, połączonego modelu atmosfery i oceanu. Oznacza to, że ocean jest w stanie odpowiedzieć na więcej zmian w atmosferze niż w eksperymencie 1, dając nam pełniejszą symulację klimatu. Eksperyment będzie używał:

• Kombinacji parametrów, które zostały zidentyfikowane w eksperymencie 1, tj. takich, które działają stabilne i tworzą trwały klimat

• Zakresu warunków początkowych takich samych, jak używanych w eksperymencie 1.

• Eksperyment będzie posiadał wymuszenia poprzez obserwacje CO₂, wybuchy wulkaniczne itp. z lat 1920-2000 i zakres możliwych scenariuszy, co może się zdarzyć w ciągu najbliższych 100 lat.

Korzystając z każdego modelu do produkcji retrognozy (określenie stanu pogody w przeszłości na podstawie statystycznego prawdopodobieństwa - przyp. tłum.) lat 1920-2000, a następnie porównując prognozy z tym co się właściwie stało, będziemy mieli obraz jak dobre są nasze modele - czy większość z nich dobrze odtwarza co się właściwie stało? To również pozwala nam określić "rangi" modeli w zależności od tego jak dobrze działają. Wszystkie modele będą również wykorzystywane do produkcji prognoz na przyszłość - do 2080 roku. Kiedy ten eksperyment się zakończy, będziemy mieli szereg prognoz klimatu dla XXI wieku.

Istnieje problem czasu w odniesieniu do oceanu. Pojemność cieplna oceanów jest tysiące razy większa niż atmosfery, dlatego osiągnięcie stanu równowagi przez ocean trwa znacznie dłużej niż atmosfery. Dlatego długa faza "rozkręcania się" jest wymagana przed możliwym przeprowadzeniem eksperymentów z wykorzystaniem modeli oceanu. Jeśli nie zostanie to zrobione, ocean może wciąż dostosowywać się do rozpoczęcia warunków eksperymentu, a model będzie już przetwarzał narzucone warunki początkowe i stan równowagi wystąpi dopiero na koniec eksperymentu.

W ramach przygotowań do uruchomienia w połączeniu, sprawdziliśmy i porównaliśmy wyniki z różnych (w rozdzielczości i topografii) modeli i metod "rozkręcania się", przed podjęciem decyzji o ostatecznym projekcie. Posiadamy również wybrane parametry fizyczne i ich zakresy, które zakłócają modele oceanu. Stworzyliśmy maski obejmujące odpowiednie baseny oceaniczne i inne konkretne obszary zainteresowania, w celu wymuszenia zmienności w czasie diagnostyki oceanu, takich jak ciepło i zawartość wody słodkiej. Stworzyliśmy również korektę obszarów strumieni, które zostaną wykorzystane w fazie "rozkręcania się" w celu utrzymania realistycznego klimatu powierzchni oceanu.

Połączony model działa asynchronicznie, co oznacza, że najpierw przez jakiś czas działa model atmosfery a potem trwa model oceanu. W tym eksperymencie poszczególne składniki działają na zmianę przez jeden dzień.

[Uwaga: eksperyment uruchomiony w lutym 2006 roku.]

---

Ten eksperyment to nic innego jak kontynuacja eksperymentu 2, z tym że zamiast obserwacji mechanizmów wymuszania użyjemy wachlarzu możliwych scenariuszy, co może się zdarzyć w ciągu najbliższych 100 lat - w zakresie emisji gazów cieplarnianych, wybuchów wulkanów, aktywności słonecznej itp.

Kiedy ten eksperyment się zakończy, będziemy mieli szereg prognoz klimatu dla XXI wieku. Ostatnim etapem jest określenie "wagi" każdego modelu w zależności od pozycji w rankingu w eksperymencie 2 - tak, na przykład, jeśli model, który działał bardzo dobrze w eksperymencie 2 i przewiduje ocieplenie o 2 stopnie, i inny który działał źle w eksperymencie 2 przewiduje ocieplenie o 10 stopni, to wierzymy pierwszemu.

Wreszcie, mamy nadzieję, że wytworzymy najbardziej kompletne probabilistyczne prognozy klimatu na następne stulecie.

[Uwaga: eksperyment uruchomiony w lutym 2006 roku.]

click here to read about the experimental strategy in more detail.

Translation by kempler (BOINC@Poland)

Eskperyment Obiegu Siarki

Informacje poniżej przedstawiają różne rodzaje modeli klimatycznych używanych przez CPDN, w tym kilka które pojawią się w przyszłości.

W pełni połączone modele

Był to standardowy model Met. Office jeszcze kilka lat temu i jest nadal aktywnie wykorzystywany do badań zmian klimatycznych. Kolejne wersje tego modelu są planowane przez CPDN.

• Atmosfera: rozdzielczość 2,5 x 3,75 stopni szerokości i długości geograficznej, 19 pionowych poziomów (znany jako N48; porównywalna rozdzielczość do T42). 30 minutowy takt dla dynamiki, 3 godziny dla transferu promieniowania.
• Ocean: rozdzielczość 1,25 x 1,25 stopni szerokości i długości geograficznej, 20 pionowych poziomów, 1 godzinny takt.

Obieg siarki i węgla, dynamiczna roślinność itp. są opcjonalne.

HadCM3 z obniżoną rozdzielczością oceanu. Został on wykorzystany do eksperymentu BBC Climate Change (w tym wersji CPDN) i eksperymentu Geoinżynierii. Wersja CPDN używa dwóch rodzajów zmian topografii oceanu, obie bez Islandii.

• Atmosfera: tak jak HadCM3.
• Ocean: 2.5 x 3.75 stopni, 20 poziomów, 1 godzinny takt.

Szybki, o małej rozdzielczości wariant HadCM3. Ten model zaplanowano dla Eksperymentu Millennium.

• Atmosfera: 5,0 x 7,5 stopni, 11 poziomów.
• Ocean: 2,5 x 3,75 stopni, 20 poziomów, 12 godzinny takt, brak Islandii.

Model atmosferyczny w połączeniu z prostym oceanem

HadCM3 ale z 1 warstwą termodynamiczną oceanu (płyta oceanu). Eksperymenty CPDN Slab, Obiegu Siarki, i mid-Holocene używały/używają tego modelu.

• Ocean: 2.5 x 3.75 lub 1.25 x 1.25 stopni.

Modele atmosfery

Atmosferyczny element HadCM3 z ustalonymi temperaturami powierzchni morza. Nigdy nie używany przez CPDN jako samodzielny model w standardowej rozdzielczości 2,5 x 3,75 stopni.

HadAM3, ale w rozdzielczości N144 (1,25 x 0,83 stopni rozdzielczość, 30 poziomów), z 10 minutowym taktem dla dynamiki, używany w eksperymencie Seasonal Attribution Project.

Taki jak HadAM3, ale z rozdzielczością 1.25 x 1.875 stopni i poprawioną fizyką. Stosowany do sprawdzania i przypisania eksperymentu i jest planowany do eksperymentu śledzenia burz.

Wysokiej rozdzielczości, regionalnie zmienny HadAM3 z poprawioną fizyką. Będzie on służył do modelu regionalnego w PRECIS.

Ma rozdzielczość 0,44 x 0,44 stopni z obracającymi się biegunami aby osiągnąć wyższą dokładność. Rozdzielczość 50 km x 50 km w 19 poziomach. Używany jest również podwójny wariant rozdzielczości czyli 0,22 x 0,22 stopni.

Model hybrydowy

HadRM3, kierowany przez globalny model, taki jak HadCM3, HadCM3L lub HadAM3P. CPDN planuje wykorzystanie HadCM3L i kierowany przez HadRM3.

Podziękowania: William Ingram; Neil Massey, Myles Allen and Hiro Yamazaki.

Translation by kempler (BOINC@Poland)

This experiment runs simulations of perturbed physics versions of a regional climate model running inside a global climate model.

• This experiment is now in the alpha test stage.

Dynamical models of the global climate system, such as the model used in the main climateprediction.net experiment, divide the atmosphere and ocean into blocks and simulate the transfer of energy, mass, moisture, and other properties between those blocks. Earth is very big so these models need to use a large number of blocks. The blocks used in the atmospheric component of the main climateprediction.net experiment model are about 300km across (more precisely, 3.75 degrees longitude by 2.5 degrees latitude) with 19 blocks stacked on top of each other, while those in the highest resolution climate model currently being run on a supercomputer are about 100km across. Getting them smaller is difficult: halving their size in each dimension means slowing down the model by a factor of 16 (because the time step also has to be shorter for smaller blocks). But this is much larger than the resolution needed to figure out what has happened or may happen in your area. Also, keep in mind that useful information from these models is only produced at much larger scales (i.e. when they stop looking blocky). Is there any way around this?

Maybe there is. One idea is to run a much higher resolution climate model over a smaller area than the globe, covering a few million square kilometres. At each time step this "regional climate model" takes the weather at its edges (including the surface) from what a global model says it is. The regional model then simulates all the weather going on inside this domain using its much higher resolution.

This sounds like a neat idea, but there are still many things we do not know about regional modelling. This experiment addresses the question of how important uncertainty in parameters in the regional model are for simulating the regional climate. Essentially, it is the main climateprediction.net experiment but using a regional climate model.

This experiment will run both the global and regional models together. The global model will be HadCM3L, identical to what is run in the main climateprediction.net experiment. Like the main experiment, values of uncertain parameters in the model will differ across simulations. The regional model will be HadRM3, used by the UK Met Office's PRECIS programme. Values of uncertain parameters in the regional model will also be altered across simulations.

This experiment will focus on two regions: one set of simulations will have the regional model placed over western North America, while the other set will have the regional model placed over southern Africa.

For the first time, climateprediction.net will perform regional climate modeling for western North America. Regional modeling provides better spatial detail, which is critically important in mountainous regions. By producing thousands of simulated model futures, this regional experiment will for the first time provide detailed probabilistic answers to key questions about aspects of climate change of great societal relevance that go beyond changes in annual mean temperature and precipitation: frost days, measures of heat waves, number of consecutive dry days, extreme daily precipitation, wind speed, extreme wind events, snowpack, and coastal upwelling, to name a few. Changes in these quantities could affect agriculture, energy demand, human health, coastal ecosystems, flood risk, water supply, and many more aspects of economic and environmental values.

The domain covered by the regional model for the Western North American experiment is shown in this image (showing topography in metres):

The resolution is about 24km. Note how individual mountain ranges are visible here; in standard global climate models these mountain ranges are merged into a blob. Thus, the higher resolution of the regional model should lead to a more accurate representation of how the mountains affect the weather here.

The domain covered by the regional model for the Southern African experiment is shown in this image:

Because this region is bigger than the Western North American region, resolution here will be lower, about 49km. Southern Africa contains a number of mountainous regions. Also, weather in the tropical areas is dominated by small thunderstorms, rather than the large weather systems typical in mid latitudes. So we expect the increased resolution of the regional model to be important here too.

This project addresses the two key limitations on climate change science in southern Africa:

• a constrained capacity to undertake robust exploration of the range of possible future climate at a regional scale;
• the limited understanding of the complex interaction of key physical processes governing the regional climate response to anthropogenic forcing.

Richard Jones, Simon Wilson (UK Met Office)
Myles Allen, Tolu Aina, Milo Thurston, Hiro Yamazaki (University of Oxford)
Philip Mote, Eric Salathé, Valérie Dulière (University of Washington)
Bruce Hewitson, Dáithí Stone, Chris Jack (University of Cape Town)

This project is funded by Microsoft Research.

The following parameters are varied in the climateprediction.net experiment:

vf1 Ice fall speed through clouds – important for the development of clouds and determining type (rain, sleet, hail, snow) and amount of precipitation

ct This relates how quickly cloud droplets convert to rain.

rhcrit ‘critical relative humidity’ relates the grid box scale atmospheric humidity to the amount of cloud in that grid box

cw_land, cw_sea This relates how much water there is in a cloud to when it starts raining, which is dependent on the condensation nuclei concentration – the more condensation nuclei there are (bits of dust, salt etc. in the atmosphere on which raindrops can form) the smaller the raindrops.

entcoef This parameter determines how rapidly a convective cloud (imagine a plume rising over a power station, or a bit thunder cloud) mixes in clear air from around it.

eacf Empirically adjusted cloud fraction This calculates how much cloud cover there will be when the air is saturated.

dtheta This has to do with the initial state of the atmosphere – what it looks like when the model starts in 1810.

ice_size This gives an effective radius for ice crystals in clouds – i.e. what radius would they have if they were perfectly spherical. It is important in the radiation scheme, to calculate how much incoming or outgoing radiation is reflected etc.

i_st_ice_sw, i_cnv_ice_sw, i_st_ice_lw, i_cnv_ice_lw These parameters all allow for non-spherical ice particles in the radiation scheme.

asym_lambda This has to do with how rapidly air mixes by turbulence in the boundary layer (the layer of the atmosphere closest to the Earth).

G₀ This has to do with the fact that the ability of turbulence to mix air varies with how stable the air is – the more stable the air, the less turbulent mixing you get.

z0fsea This parameter governs the transfer of momentum and energy between tropical oceans and the air (wind) above them.

charnock This parameter governs the transfer of momentum and energy between seas and the air (wind) above them.

r_layers This is related to the number and size of plant roots in the soil – and, consequently, to how water is taken up from the soil and into the atmosphere by plant transpiration.

eddydiff This parameter governs the diffusion of heat from the slab ocean to ice, where there is sea-ice present in the model.

start_level_gwdrag Gravity waves are waves in the atmosphere for which gravity is the restoring force – think of air passing over a mountain, it is forced upwards over the mountain, and then gravity will pull it back down, resulting in an oscillation (you often see clouds form downstream of mountains as a result). The air particles oscillating in these waves tend to lose energy because of friction (drag), and this energy manifests itself as heat. This parameter determines the lowest model level on which gravity wave drag is applied

kay_gwave kay_lee_gwdrag These parameters govern the way that gravity waves are formed as air interacts with surface features, such as mountains.

Alpham, dtice These have to do with the fact that the albedo (reflectivity) of sea ice varies with temperature.

diff_coeff, diff_exp Diffusion coefficients and exponents govern how quickly something spreads through the material it is in – so, for example, if you put a drop of oil dyed purple into a beaker of un-dyed oil, how rapidly the dyed oil mixes with the oil around it until all the beaker has the same colour. Diffusion refers to mixing due to the random motion of particles, rather than turbulent mixing which happens when there are actual vortexes mixing things (which would happen if you stirred the beaker with a spoon). In the case of the atmosphere, the horizontal diffusion coefficient and exponent determines the rate of diffusion of heat from a warm air mass to a cold one.

diff_coeff_q, diff_exp_q These diffusion parameters determine the rate at which water vapour diffuses from a very humid air mass to a relatively dry one.

In addition, in experiment 2 we vary:

anthsca The scaling factor given to the anthropogenic sulphur dioxide emissions to allow for the range of uncertainty in emissions.

isopyc Models the effects of mixing of water along surfaces of constant density in the oceans.

mllam, mld Mixes the top ~100m of the ocean.

vdiff Coefficient of vertical diffusion of temperature and salinity in the oceans.

vvisc Parameterises the friction between the different vertical layers in the oceans.

More information on experimental parameters can be found on the data portal.

What was the climateprediction.net Seasonal Attribution Experiment?

Why look at attributing extreme weather events in this way?

How can I help?

What climate model are you using and what is 'BOINC'?

Who are we?

What was the climateprediction.net Seasonal Attribution Experiment?

The climateprediction.net Seasonal Attribution Project uses computing time donated by the general public to run state-of-the-art high-resolution model simulations of the world's climate. These simulations are used to determine the extent to which the risk of occurrence of extreme weather events is attributable to human-induced climate change.

We focus on extreme weather events that occur on a seasonal timescale, and in our current project we focus specifically on the United Kingdom floods of Autumn 2000 which occurred during the wettest autumn ever recorded, causing widespread damage and an estimated insured loss of £ 1.3 billion.

• Half of the climate model simulations we run are of the Autumn 2000 period, specifically including within them the effects of human-induced climate change caused by the emission of greenhouse gases. We call these the "Industrial Autumn 2000" simulations.
• The other half will simulate a representation of the the Autumn 2000 climate had there not been any human-induced emissions of greenhouse gases over the last century. We call these the "Non-Industrial Autumn 2000" simulations.
• By then comparing the results of these Industrial and Non-industrial simulated climates, and recording the occurrence of floods like that of Autumn 2000 in each of them, we can determine how the frequency of occurrence (or 'risk') of such a flood has changed, and therefore how much risk is attributable to human-induced emissions of greenhouse gases over the last century.

There is also an article with more details about the science behind this project.

We are also collaborating with other research groups who are interested in using our simulations to perform similar attribution studies, for snowmelt in western North America, and heatwaves in South Africa and India.

Back to top ^

Why look at attributing extreme weather events in this way?

Recent extreme weather events, having large societal and economic impacts have prompted the debate about effects of human activity on the world's climate. One way to answer this question is to compare the world's current climate with what it would look like had it not been for human activity - similar to the way an epidemiologist might compare samples of smokers and non-smokers to attribute the effects of smoking to lung cancer. The problem is that we cannot observe what a climate without the presence of human activity looks like, since we can only observe the current state of the climate. Hence we have to resort to simulations of such a climate using state-of-the-art climate models.

Furthermore, small differences in how we initialize these simulations can have a significant impact on the end results - reflecting the fact that we do not have completely perfect models and that in the real world small differences in what is going on now can have significant effects on what happens in the future (to quote a famous statement, 'The flap of a butterfly's wings in Brazil can set off a tornado in Texas', the so called butterfly effect). To account for this uncertainty we will run a large number of simulations of each climate - again, similar to how an epidimiologist might study a large number of patients to have confidence in his results. Also, because the Autumn 2000 flood event was in itself extreme this further necessitates running a large number of simulations before we might expect to reproduce an event of that magnitude. We are currently aiming to complete about 10,000 simulations for each of the Industrial and Non-Industrial Autumn 2000 climates.

We focus particularly on the United Kingdom Autumn 2000 floods because, aside from causing widespread damage, they occurred during the wettest autumn since records began in 1766. Thus we might expect any signal of such an event in the simulations to be more prominent and easier to detect than if we had investigated less severe and/or shorter lived weather phenomena. Also, it is only relatively recently that models have been developed with the required resolution to sufficiently capture the storms and weather patterns associated with the Autumn 2000 floods.

Attributing the risk of extreme weather events to climate change in this way also has very interesting implications for liability for such events.

Back to top ^

How can I help?

We aim to complete around 10,000 climate model simulations of each of the Autumn 2000 climates. Running all these simulations, however, is very computationally expensive because they are so many and because the model is of high spatial resolution - it's beyond our own resources. This where your can help and participate in our project!. We invite you to download and run one of these simulations on your own computer and so help in determining the risk of occurrence of the United Kingdom Autumn 2000 floods attributable to human-induced climate change.

The simulation will run automatically as a background process on your computer when you switch your computer on, and you can schedule when you want it to run. It should not affect any other tasks you use your computer for. It will simulate the world's weather for a one year period from April 2000 - March 2001, so as to capture important atmospheric conditions surrounding and including the Autumn 2000 period. As it runs, you can watch the weather patterns on your, unique, version of the world evolve. A single simulation run continuously takes approximately 3-4 weeks on a computer with a Pentium 4 processor. Once completed, the results are sent back to us via the Internet, but you do not need to be continuously connected to the Internet whilst the simulation is in progress otherwise, and you will be able to see a summary of your results on this web site.

Back to top ^

What climate model are you using and what is 'BOINC'?

For our simulations we use the state-of-the-art HadAM3-N144 climate model, which was developed at the Hadley Centre for Climate Prediction and Research. It is a high spatial resolution version of the standard Unified Model used by the UK Met Office (learn more about climate modelling and the Unified Model). At the earth's surface it has a horizontal resolution of approximately 100km² at mid-latitudes enabling it to reasonably capture the storms and weather patterns associated with the Autumn 2000 floods flooding. Specifically, in the simulations we record daily surface temperature, precipitation, 500hPa geopotential height and surface winds over the Atlantic-European region. Monthly temperature, precipitation, 500hPa geopotential height, mean sea-level pressure and soil moisture over the entire globe is also output to assess larger scale weather systems. Furthermore, daily temperatures and precipitation over the Northwest US, India and South Africa are recorded for related attribution projects investigating snowmelt and heatwaves.

To distribute and manage the running of simulations on participating computers we use 'BOINC' (Berkeley Open Infrastructure for Network Computing) software. BOINC allows the user to control the run behaviour of a simulation on their computer. This includes allowing you to schedule when you want to run the simulation, how much of your computers resources to dedicate to the project, selecting a screen-saver mode, and interactive visualisation of your simulation.

Back to top ^

Who are we?

The main researchers on this project are Pardeep Pall, Myles Allen and Dáithí Stone, based at the Atmospheric, Oceanic and Planetary Physics department of Oxford University. We work closely with the climateprediction.net project team and are extremely grateful for their contribution to this project. This work is funded by WWF International and a UK Natural Environment Research Council studentship.

Back to top ^

This experiment uses a model of the atmosphere and land surface only, with sea surface conditions, atmospheric greenhouse gas concentrations, and other external factors imposed on the model. This allows us to study the range of weather possible given these conditions.

Recent extreme weather events include the flooding in the UK of Autumn 2000 and Summer 2007 and the European heat-wave of 2006. The severity of these events highlight the need for an improved capability to determine whether the chance of such events are influenced by human-induced climate change, associated with raised levels of atmospheric greenhouse gases in the atmosphere.

When these events occur, the public currently receive, via the media, a series of unfounded and conflicting claims and counter claims as to whether climate change has had an influence on the extreme weather.

This experiment aims to provide a scientific method of attributing the chance of extreme weather events to increases in greenhouse gas emissions.

The previous climateprediction.net Seasonal Attribution project pioneered a method of attributing the risk of an extreme weather event occurring to the increase in greenhouse gas concentrations in the atmosphere. This is done by computing very large ensembles of climate models under two different climate scenarios. The first scenario has the atmosphere resembling, as closely as possible, the observed atmospheric conditions in the period when an extreme weather event occurred. The second scenario has an atmosphere that is an estimate of what the conditions would have been if no man-made greenhouse gases had been emitted, in effect that the industrial revolution had never happened. By comparing the results from the two ensembles, the attributable risk of the event occurring can be calculated. This project will follow the method of the Seasonal Attribution project but expand it to a general framework for attribution studies. A key component of this framework is the validation of the climate model to be used.

Validation of the model involves comparing the statistics of the model output with the statistics of the observed atmosphere. Observed datasets and reanalysis products, such as the ECMWF ERA-40 reanalysis data set, are used to compare against. We are interested in the performance of the model in predicting rainfall, especially prolonged and heavy precipitation extremes in Europe. We are also interested in the ability of the model to predict the storm track and modes of atmospheric variability, such as the North Atlantic Oscillation, the Scandanavian Pattern and blocking activity. Comparison datasets are calculated from the observed and reanalysis data, and corresponding datasets are calculated from the ensemble output.

We also investigate the ability of the model to simulate the frequency of occurence of precipitation extremes over a range of timescales from 1-day to 90-day events. If necessary and appropriate, an empirical adjustment based on model output statistics is computed and used to adjust model-simulated fields to be consistent with observations.This correction can then be used in future attribution studies.

The model used is the Hadley Centre HadAM3P model. This is related to the model used in the Seasonal Attribution project and the regional model used in PRECIS. It is a full atmosphere general circulation model (GCM), with prescribed sea surface temperatures (SST) and sea ice concentrations (SI).

The resolution is N96: 192 grid boxes in the longitude and 145 grid boxes in the latitude. This equates to 1.875° x 1.25° and a grid box size of 208 km x 139 km at the equator.

Other changes from the Seasonal Attribution project include an interactive Sulphur Cycle, changes to the physics which improve the general atmospheric flow, an improved land surface scheme and improvements to the parameterisation of clouds and precipitation. These improvements have a beneficial impact on the statistics of the model we wish to study in this experiment, in that they are closer to observed values, and make up for the loss in resolution.

The experimental set up is what is know as a time slice experiment. We are running the simulations from 1959-2000. This period is broken up into 2 year periods, which overlap, e.g. there is a period from 1961-1963 and a period from 1962-1964. This allows us to examine how the model diverts from its initial starting conditions and removes any bias that may be caused by the model "spinning up" into a stable state.

There are no physics perturbations in this experiment. Each model run consists of the initial starting conditions, a perturbation that is applied to the initial conditions and the forcings for the period that is to be run. The forcings include the sea surface temperatures (SST), sea ice (SI) concentrations, emissions and constants for the sulphur cycle, ozone, natural volcanic emissions, solar variability and greenhouse gas emissions.

• The initial conditions are taken from a long run of the model from 1959-2000, which was performed on University of Oxford computing facilities.
• Sea surface temperatures (SST) and sea ice fractions (SI) are taken from HadISST a global dataset of observed SST and SI from the Hadley Centre.
• Ozone is taken from observations up to 1991, then the IPCC SRES A2 scenario is used to the year 2000, as observations are not available in the format needed by the model.
• Natural volcanic emissions are from the Sato et al. dataset.
• Solar variability is taken from the Solanki and Krivova dataset. Greenhouse gas emissions are taken from the IPCC SRES A1B scenario.
• The initial condition perturbations are created by computing next day differences in the 3D field of potential temperature from the long run. There are 1741 of these perturbations. Along with 39 starting conditions, this creates an ensemble of almost 70,000 members!

The Validation Experiment should provide us with an evaluation of the model and, if necessary and appropriate, information on possible calibration of the model's output. If it passes this test, the we will proceed to an experiment to look at the degree to which anthropogenic emissions of greenhouse gases contributed to the risk of the summer 2007 floods in the UK.

This experiment will have a similar setup to the Seasonal Attribution project. A set of simulations of 2006-2007 will be run as in the Validation Experiment. Then, another set will be run for the same period, but in a hypothetical climate in which humans had never emitted greenhouse gases. This will mean both reducing the greenhouse gas concentrations in the model and also decreasing the ocean temperatures and sea coverage according to various estimates of the contribution of the greenhouse gas emissions. The implementation will closely follow that used in the Seasonal Attribution project.

Further extension is planned for the detailed simulation of the historical climate using the surface conditions that are extracted from pre-industrial coupled model simulations, performed at a coarse numerical resolution (eg. Millennium Experiment).

Greenhouse gas concentrations in the atmosphere are well measured for the recent past, and so we do not, in this experiment, need to investigate any possible uncertainty. For the future, we are using 4 possible scenarios, the first two of which are taken from the 2001 Intergovernmental Panel on Climate Change report.

• SRES scenario A1B Formal definition
• SRES scenario B1 Formal definition
• Stabilisation at 400ppm - this assumes that legislation or alternative means cap emissions so that the concentration of CO₂ in the atmosphere remains constant at 400ppm.
• Sequestration at 2050 - this radical scenario assumes that in 2050 mechanisms are put in place to remove CO₂ from the atmosphere.

Past

For the past (1920-2000), we have used 4 data sets of observations of the solar index (a measure of the amount of energy the Earth receives from the Sun). There is a reasonable amount of variation between these data sets, which are all based on observations. In case all of these substantially underestimate the actual trend in solar index, (in which case it could be argued that a large part of the observed warming in the second half of the 20th century might be caused by the Sun) we have arbitrarily created a 5th data set by doubling the trend in solar index in the Lean, Beer and Bradley data set. /p>

In this figure, the x axis shows dimensionless units related to the solar constant.

In all the data sets, you can see the 11 year solar cycle, caused by a regular variation in sunspot activity on the Sun.

Why do the data sets all join at the start of the period, rather than at the end, when you would have thought observations were best? In fact, it would be have been better to do the latter, but, as we are starting all our simulations from a 'spin up' we have to be careful that there is not a sudden jump in a forcing at the join. In fact, it is the trend (or change over the period we are looking at) rather than the absolute value that really matters, so an offset at the start doesn't matter. Reassuringly, all the data sets are at the same point in an 11 year solar cycle in 1920.

Future

As no-one knows how the Sun's energy output will vary over the next 80 years, we have created 3 scenarios - either the solar index will carry on increasing at the same rate it has increased over the past 80 years, or it will decrease at the same rate, or it will neither increase nor decrease. It is a reasonable assumption that reality will lie somewhere in between these cases.

Past

Only volcanic eruptions large enough to force dust up into the relatively stable stratosphere have a significant effect on the world's climate. Pinatubo, which erupted in 1992, cooled the Earth noticeably for about 2 years. Again, there is a reasonable amount of uncertainty in observations of volcanic emissions in the past - particularly in the pre-satellite era. For the past 80 years, we have created 5 data sets based on the Sato and Amman observations of volcanic aerosol in the stratosphere. This data is divided into 4 latitude bands of equal area - 90ºS-30ºS, 30ºS to the equator, the equator to 30ºN, 30ºN to 90ºN.

Future

For the future, we have created 10 possible scenarios, as we have, of course, no idea what volcanoes may erupt where. One scenario simply repeats the recent past according to the Sato (2002) data set. Two more are based on observations of the preceding 80 years, based on the Sato and Ammann data sets. The remaining 7 are subsets of observations of 1400-1960, based on a data set constructed by Crowley.

Tropospheric and stratospheric ozone values are set according to observations, which are well constrained for the recent past. For the future, we use two scenarios - one is the IPCC B1 scenario (Formal definition), the other comes from the Hadley Centre, which predicts recovery of the ozone hole by about 2025.

Ammann CM et al, "A monthly and latitudinally varying volcanic forcing dataset in simulations of 20th century climate", GRL, 2003

Sato 2003 - http://www.giss.nasa.gov/data/strataer/.
Sato, M., J.E. Hansen, M.P. McCormick, and J.B. Pollack 1993. Stratospheric aerosol optical depth, 1850-1990. J. Geophys. Res. 98, 22987-22994

SK Solanki & NA Krivova Can solar variability explain global warming since 1970? , J. Geophys. Res., 108, (2003)

J Lean, J Beer & R Bradley RECONSTRUCTION OF SOLAR IRRADIANCE SINCE 1610 - IMPLICATIONS FOR CLIMATE-CHANGE (1995, Geophys. Res. Letters, 22, 3195-3198) extended to 1997 (Lean, pers. comm., 1998)

DV Hoyt & KH Schatten, A DISCUSSION OF PLAUSIBLE SOLAR IRRADIANCE VARIATIONS, 1700-1992 , J. Geophys. Res., 98, 18895-18906 , (1993)

Lockwood & Foster

5000 runs have successfully completed and returned results to climateprediction.net!

Below are a few examples of what they did - the normal and the decidedly abnormal. This is very much raw, unprocessed data!

We will continue developing this page as we process the data coming back to us....

Thank you to everyone who is making this experiment possible!

click here to see:

A stable, 'normal' run

An unstable run

A warm, wet outlier

A cold, dry outlier

and here are some examples of things that can be seen using the new visualisation package which is being tested at the moment and should be available to everyone before the end of the year:

 

A cold, damp week in London, December 1828

---

 

 

Most of the runs we are getting back at the moment look fairly similar to this 74958.

 

The left plot shows global mean temperature (the average temperature on the surface of the world), and the right plot the global mean precipitation (rain, snow etc.) for the same model run.

The blue part (1810-1825) is the 'spin-up' phase where the model was settling down - if this wasn't a straight line, the model run would probably have crashed. The green line (1825-1840) is the control part of the experiment, with pre-industrial carbon dioxide levels. We say that the model is 'stable' if this line is also approximately straight - nothing should be forcing the climate to change. The red line (2050-2065) is where the carbon dioxide levels have been doubled. The difference between this and the green line is what we're interested in. In this case, there is approximately a 3°C temperature increase, and a 5% increase in precipitation worldwide.

Why does the temperature rise when carbon dioxide is doubled? Click here to read more about the Greenhouse Effect.

Why does the precipitation rise when carbon dioxide is doubled? As the temperature of the air increases, for a fixed amount of water vapour in that air the relative humidity will decrease, so warm air can hold more water vapour before it saturates than cold air. However, this doesn’t determine how fast moisture circulates through the water cycle (you can read more about this here - follow the links to 'weather' and then 'water cycle'). The intensity of the water cycle is controlled at a global level by how fast water can condense rather than by how much water vapour there is in the atmosphere. As water vapour condenses to form clouds, it releases latent heat. If nothing removed this heat, the air would warm up and would be able to hold more moisture, so the condensation would stop. What actually happens is that the atmosphere gets rid of this heat, mostly in the form of longwave radiation. As the atmosphere warms up, outgoing longwave radiation increases (click here to read more about this) which allows more cloud droplets to form and so the whole water cycle intensifies.

Why does the precipitation fall initially when carbon dioxide is doubled? Initially, the carbon dioxide insulates the atmosphere, trapping longwave radiation. The amount of longwave radiation lost to space falls. Less cloud droplets can form because the atmosphere cannot get rid of the energy released by condensation fast enough.
Eventually, in most models, the temperature of the Earth increases and, as it does so, the outgoing longwave radiation increases again, compensating the direct insulating effect of increasing carbon dioxide. This makes future precipitation changes so much more uncertain than temperature changes: in some models, there might even be a net reduction in rainfall following a doubling of carbon dioxide for much more than a year or two.
You can read more about this in the Allen, Ingram and Stainforth Nature paper which is on our publications page.

---

Several of the models are 'unstable', that is, the set of starting conditions and parameters supplied to the model meant that, even though it didn't do anything bizarre in the spin-up phase, it failed to settle at a given temperature in the control (green) phase. Experiments 40017 and 91121 are examples of this.

 

 91249 is an example of a stable experiment that went warmer and wetter than most in the doubled carbon dioxide phase.

---

 40015 is an example of a stable experiment that went colder and drier in the doubled carbon dioxide phase.

This graph shows temperature (solid) and precipitation (dashed) over London, U.K. for 11-18th December, 1828 in one model. If you follow the temperature along from the start, you can see it rising during the day, then starting to fall ... but at 9pm that evening (you can't tell that from the graph) temperatures start to rise again... strange, until you see a huge spike in precipitation - so the rise in temperature was due to the warmer air behind a warm front. Another, wetter frontal system comes through a couple of days later, but the whole week the temperatures hover around freezing, so you can imagine it being grey and cold, and certainly very icy ....

The climateprediction.net project comprises three separate experiments (details are given below).

All the experiments have three main parts:

1. initial condition ensembles
2. perturbed physics ensembles
3. forcings

Initial condition ensembles involve the same model, with the same forcings, run from variety of different start dates. Because the climate system is chaotic, tiny changes in things such as temperatures, winds, and humidity in one place can lead to very different paths for the system as a whole. We can work around this by setting off several runs started with slightly different starting conditions, and then look at the evolution of the group as a whole. This is similar to what they do in weather forecasting.

Perturbed physics ensembles form the main scientific focus of the whole project. Modern climate models do a good job of simulating many large-scale features of present-day climate. However, these models contain large numbers of adjustable parameters which are known, individually, to have a significant impact on simulated climate. While many of these are well constrained by observations, there are many which are subject to considerable uncertainty. We do not know the extent to which different choices of parameter-settings or schemes may provide equally realistic simulations of 20th century climate but different forecast for the 21st century. The most thorough way to investigate this uncertainty is to run a massive ensemble experiment in which each relevant parameter combination is investigated. Thus the perturbed physics ensemble is the central feature of the climateprediction.net project. You can read more about the rationale for this experiment by following the other links in the menu bar on the left.

Forcings are the things which drive the climate system. The chaotic variability we target in the initial condition ensemble is due to factors internal to the climate system, while things such as solar variability, sulphate (volcanic, etc) forcing and greenhouse gases are treated as external to the climate system. We call them forcings because they force the system from the outside: if these things change, we expect the climate system to respond. [If the sun puts out more energy, we would expect the Earth to heat up, for instance.]

The climateprediction.net strategy to achieve our ultimate objective of predicting potential future climate states is based on the following three steps:

Step	Goal	Methodology
Experiment 0 Pilot experiment to check feasibility of distributed computing [tell me more]	Identify what physics of the HadSM3 model provide the best / an acceptable simulation of present-day climate with prescribed present-day ocean surface temperature.	To run the atmosphere-slab ocean model with different physics to check the model simulation of present day climate. This experiment corresponds to the spin-up run of experiment 1.
Experiment 1 Explore model sensitivity to physical parameters [tell me more]	Identify suitable parameters and ranges by use of the HadSM3 model.	To realise an ensemble of simulation with perturbations made to a number of paramters. Each simulation includes 3 steps: Model spin-up (15yrs) Model control (15yrs) Double CO2 run (15yrs)
Experiment 2 Explore model sensitivity to initial conditions, historical forcings [tell me more]	Assess predictive model skill by making a probabilistic hindcast of the past climate by use of the HadCM3 model.	Make an ensemble of hindcast simulations for the period 1920-2000 by perturbing the initial conditions, and running a range of historical forcings. Compare model outputs with observation to assess how well the model performs.
Experiment 3 The climateprediction.net project to explore potential future climate state [tell me more]	Make a probabilistic forecast of the future climate by use of the HadCM3 model.	Make an ensemble of forecast simulations for the period 2000-2080 period by perturbing the physics of the model, running an initial condition ensemble and running a range of possible future solar, sulphate and greenhouse gas futures.

This experiment uses an atmosphere-slab ocean model (see expt 1) to simulate the present day climate. Parameters will be varied to explore the sensitivity. The main objective is to check the feasibility of distributed computing and get your feedback (learning by doing).

[Note: this has already been completed.]

This experiment will be slightly simpler than the atmosphere-ocean global circulation model (AO-GCM) which will be used in the main climateprediction.net release. It will still have a full atmosphere but the representation of the ocean will be simplified into just a single layer. This is known in the trade as a "slab" model. It is not so useful in forecasting future climate because the "ocean" can only respond in very limited ways. However, it is very useful for investigating the sensitivity of the model to changes in the atmospheric parameters. Because it doesn't have a full ocean model it responds much more quickly to changes in factors such as levels of greenhouse gases. The knowledge we gain from this experiment about model sensitivity will be used to design properly the climateprediction.net experiment. By perturbing parameters which control the models physical processes (such as cloud formation) it is possible to see different realisations of climate change.

As in the main experiment everybody's model will be unique because each will have a different combinations of parameters. We will be asking you to carry out 3 separate phases with each combination of parameters.

Calibration (phase 1)

It involves running the model for 15 years (which should take about 3 weeks on an up-to-date computer) to make sure the atmosphere and the slab ocean are "in balance" (they are in a stable, equilibrium state where large scale effects such as global mean temperature, do not change substantially from year to year). This experiment will provide some forcing fields to the 2 other phases.

Control run (phase 2)

Standard run where the levels of greenhouse gases in the model atmosphere are kept at pre-industrial levels.

Double CO₂ (phase 3)

Double CO₂ run with levels of greenhouse gases about twice what they are in the control phase. The results will give an indication of how sensitive the climate is to the parameters we vary. We will therefore be able to use the results to guide our choice of parameters in the main experiment.

In experiment 1, we have an initial condition ensemble, a pertubed physics ensemble, and a forcing ensemble that includes pre-industrial levels of CO2, and 2xCO2. This is to help us find the climate sensitivity of the models we are studying. [The IPCC defines climate sensitivity as "equilibrium climate sensitivity is defined as the change in global mean temperature, T2x, that results when the climate system, or a climate model, attains a new equilibrium with the forcing change F2x resulting from a doubling of the atmospheric CO2 concentration."]

[Note: This is the experiment launched in 2003.]

The second experiment will use the full atmosphere-ocean GCM (the coupled model). This means the ocean is able to respond much more dynamically than in experiment 1, giving us a more complete simulation of the climate. The basic parts of the experiment are:

1. an initial condition ensemble;
2. a perturbed physics ensemble (the interesting and viable models from experiment 1); and
3. historical forcings (from the period 1920-2000).

The initial condition ensemble will be similar to those used in parts 1 and 3- see the descriptions of experiment 1 for details.

The perturbed physics ensemble will comprise those models that are viable, stable climate models, that span the interesting regions of parameter space we have identified in experiment 1. We will be looking to include as many of these as possible, and will really only be throwing out those models which have gone unstable because of the choice of parameter sets. We imagine that there will be regions of parameter space (families of closesly related models with similar parameter settings) that are viable (it's theoretically possible that some will even be better than the standard model!), and regions that are not. We will find those regions in experiment 1 and use the good ones in experiment 2.

For our forcings in this part of the experiment we will use the data from the climate rercord from 1920-2000. We will start a bunch of experiments in 1920 and force them with historical data for fifty years. This process is called a hindcast: it's like a forecast, only you know the outcome. We know what happened 1920-2000, but it's still a challenge for the model to do a good job of simulating it. We can use the models' performance in simulating the past to see how good they are as fully-coupled models simulating recent and present climate. If they're good at that, we'll use them for predicting the future, too.

[Note: Launched in 2005.]

The prediction experiment. As in the second experiment we will be using the fully coupled model. Using the models that have managed to do a fairly good job of simulating the historical 1920-2000 climate, we run an ensemble prediction of the period 2000-2080. There are three parts to the experiment:

1. an initial condition ensemble;
2. a perturbed physics ensemble (the survivors from experiment 2); and
3. a future forcings ensemble.

The initial condition ensemble will be similar to those used in parts 1 and 2 - see the descriptions of experiment 1 for details.

The perturbed physics ensemble will comprise those models that score well on the climate prediction index. This index (developed in conjunction with the Hadley Centre) gives a measure of how well a model fits observations. Models that score well have performed well in experiment 2, in which they have simulated the second half of the twentieth century (see above). Models that score badly will be discarded (or down-weighted) for experiment 3.

The future forcings ensemble is necessary because we don't know what the sun or the volcanoes are going to do over the next fifty years. We also don't know how levels of greenhouse gases are going to change over that period. So we're going to run a large number of different possible futures, in which we vary solar, sulphate and greenhouse forcing, to span what we hope will be the likely range.

Experiment 3 will thus comprise the best models we have that are consistent with observations, running predictions that span likely behaviour in the three major climate forcings. Each model and each future scenario will be started off from a range of different initial conditions to check that the results we get back are not simply idiosyncratic functions of the choice of start date.

And the result of all this will be, we hope, the world's best guess at a probabilistic climate forecast for 2080.