|
|
|
COUPLED OCEAN-ATMOSPHERE PROCESSES AND EUROPEAN CLIMATE |
|
|
|
|
|
COAPEC (Coupled Ocean Atmosphere Processes and European Climate)Report of the Third Annual MeetingThe Third COAPEC Annual Meeting took place on Tuesday 2nd-Wednesday 3rd July, 2002, at Chamberlain Hall, University of Birmingham.
[top] Tuesday 2 JulyThe meeting opened with a review of the COAPEC thematic programme by the Science Co-ordinator, Helen Snaith (SOC). The programme is almost half way through its nominal 5-year existence and has completed its funding rounds. Three members of the core team (Bablu Sinha (SOC), Mat Collins and Pierre Philippe Mathieu (Reading)) have been in post since April 2000 and are now well into their respective research projects. A fourth member, Alan Iwi (RAL), has been in post since the beginning of October. [top] The first presentation was an invited talk by Richard Wood (Hadley Centre) who introduced the range of "COAPEC-related" research currently being undertaken at the Hadley Centre. The Hadley Centre has a range of models for use in coupled studies:
HadCM3, is used for a wide range of research projects.
HadCEM is a research "eddy permitting" (1/3° 40 vertical level ocean) version of HadCM3 run for a 150yr control then 80yr 2% CO2 scenario. The model shows better salinity characteristics in the Atlantic as the ocean model "corrects" errors in the atmospheric model evaporation. There is not much influence on global mean SST but in the increased CO2 run there is slightly faster warming than in HadCM3. GloSea is a version of HadCM3 with enhanced meridional resolution in the tropics, run as ensembles of 6 month forecasts (calibration to remove drift). Some of the results of the model are available from the MO seasonal forecasting web site (www.metoffice.com/weather/seasonal). The Met Office are starting a project, called HadCMIP, to compare the range of different models now coupled to HadAM3. The project will look at a range of areas and then develop teams to focus on specific model errors FAMOUS (Fast Met Office/UK Universities Simulation) has low resolution atmosphere and ocean, approximately half that of HadCM3, and runs up to ten times faster than HadCM3. FAMOUS is being developed in conjunction with Paul Valdes (Reading) primarily for use in projects requiring long spin-ups (e.g. palaeo or carbon cycle studies) or those requiring large ensembles (e.g. for studies of initial conditions). HadGEM is the next generation operational model, expected to go into operation during 2003/2004 with prototypes currently under test. The atmospheric model has a new dynamical core, physics and vegetation scheme and includes sulphur cycling, aerosols and link for chemical transport. The ocean model is 40 level 1x1°, rising to 13° in the tropics. Understanding gained from studies of HadCM3 can be fed into the development process, and the development team would be very grateful to hear about results from COAPEC science which could help HadGEM1 development. [top] The afternoon session started with a presentation by Steve George (Reading) on a 2nd round project looking at the effects of N Atlantic variability on European climate at seasonal timescales. Past studies using lagged SVD of SST and Z500 anomalies in models and observations seem to show some links, with maximum correlations at lags of approx 4 months, The approach is to look at ECMWF system II forecasts (HOPE ocean model coupled to the IFS atmosphere by the OASIS coupler run in burst forecasts of 40 members). This model has an inherent bias that it is not entirely obvious how to remove and one option is to look only at anomalies. Despite these biases, the model does show some areas of high predictability, e.g., skill in predicting the Pacific DJF SSTs. [top] The next presentation, by Simon Josey (SOC), gave an update on the SOC fluxes climatology. The original bias of 30 Wm-2 globally has been corrected using inverse analysis - scale adjustments to heat flux components constrained by ocean sections. The adjusted climatology now agrees with Ganchaud and Wunsch (2000) within error bars, although all estimates still diverge in the Southern Ocean. The report of this analysis, and the climatology, are available online. The next phase of the project is looking at improvements in the Long Wave and Short Wave flux parameterisations. The LW parameterisation in the N Atlantic has been improved using good quality measurements at sea and data recently collected in the Indian ocean should improve the situation there. The parameterisation may account for 10-12 Wm-2 in LW radiation. Compared to other climatologies the total fluxes may be similar but the sources may vary. [top] Alberto Naviera Garabato (UEA) presented work on data inversions using 1990s N Atlantic WOCE hydrographic sections, aimed at quantifying the THC. Work so far has concentrated on designing the inverse model. Various initialisation methods have been tried and additional constraints of zero mass transport across latitude bands applied. No totally satisfactory solutions have yet been found but preliminary solutions give the THC as ~20 Sv at 48N. Work is now focussing on trying to find a better inverse solution for (at least) high and low NAO conditions. Future work will involve looking at the same sections in HadCM3 data and calculating overturning rates directly and from the inverse model. This should give an indication of the skill of the inverse model at determining overturning from the observations. Comparisons will also be made of the observed and HadCM3 estimates of variability using the repeated AR7E&W and AR19 sections. [top] Helen Snaith then introduced the "New Observations" project: Argo floats in South Atlantic on behalf of Brian King (SOC). The main focus of the project is the sub-tropical gyre. In the South Atlantic there is a problem with the fresh water budget as estimates of the southward flux is too large, implying too much evaporation in the N. Atlantic. To try and address the lack of observations, 15 Argo floats will be deployed in the SE Atlantic, parked at 2000m, with a 10-day cycle. Data will be analyses using a repeat of the previous inverse analysis. [top] The session closed with a discussion on the use of observations and data requirements within COAPEC. The 1000-year HadCM3 data set being downloaded to BADC is now 66% complete, but there is a need to remove some overlapped fields. Some of the requested fields are missing (not saved) at the Hadley Centre. [top] The final session of the day was started with a brief introduction on to the CMIP (Coupled Model Intercomparison Project) programme by David Webb (SOC). As part of the programme, coupled model data are available via Lawrence Livermore National Laboratory. Currently there are not many people looking at the ocean components of coupled models, important for IPCC FAR. Anyone interested in the programme is encouraged to register interest through the web site. [top] Pierre-Philippe Mathieu (Reading) gave the first core team talk of meeting, reporting on his work on does ENSO influence Europe? Previous studies indicate an extension of the PNA into Europe and shifts in the storm track but there is still no real consensus in the literature. This study uses HadAM3 forced with observed SST for 1986-2000 and also using non-ENSO anomalies in the Atlantic whilst using ENSO anomalies in the Indo-Pacific. Looking at the 500mb height anomalies during El Niño events there is some reproduction of the observed anomalies but not in all cases. Atlantic SST anomalies can also influence European pressure but the response can be non-linear. There is an ENSO influence on Europe but there is also inter-event variability. [top] The next presentation was by Adrian Matthews (UEA) on Tropical Atlantic variability. Using SVD of SST anomalies (SSTA) and zonal wind stress anomalies in the tropical Atlantic, mode 1 represents 43% of the variance and shows both local and ENSO signals. The events are stratified into those where Atlantic and ENSO are important (AAE) and those where Atlantic, but not ENSO anomalies, (ANE) are significant. The ANE events result in a strong tripole signal in SST, but this is not so in AAE events. ANE events have a negative NAO component. HadAM3 experiments with tropical Atlantic SST anomalies capture tropical OLR anomalies and some MSLP and SAT response, similar to observations. The MSLP anomalies appear to be weak (a few mb) but the SAT anomalies are quite strong (2K). The impact back on ocean of these events is very variable. [top] Len Shaffrey (Reading) is looking at the partitioning of atmospheric and ocean heat transport in the N Atlantic. The Bjerknes compensation theory states that they must balance to give the same net transport. Oceanic interannual transports are dominated by Ekman effects, with links to the NAO. Atmospheric transports are more complicated. Atmospheric heat transports have been regressed against the scaled top 3 EOFs. Kelvin and Rossby wave adjustment in tropical Atlantic is observed following e.g. Northern Hemisphere forcing. The heat content anomalies propagate around a complicated path. Looking at decadal variability, there is some indication of Bjerknes compensation occurring, with THC implicated as the source. [top] Chris Old (Reading) is also looking at ocean heat anomalies, this time in the context of propagation of heat content anomalies in HadCM3. The aims of the project are to analyse HadCM3 control run data, assimilate observations into HadCM3 and determine the controlling physical processes. The COAPEC 100 year HadCM3 data show a cooling trend. There is some decadal variability in the data, with some NAO forcing apparent. There are some difficulties in closing the budget in the model, with discrepancies of ±0.5 PW. There is spatial structure in the residuals, perhaps associated with using monthly means i.e. high frequency signals may contribute to the budget. Using temperature class diagnostics, there is some strong temporal variability in the heat anomalies. Future work will involve continuing the present work and starting to assimilate hydrographic data into HadCM3. [top] The final presentation of the day was a Keynote talk by Laurent Terray (CERFACS) on SST forcing of low-frequency variability in the N Atlantic (the Presentation can be viewed online). Laurent started by proposing a paradigm for low frequency climate variability: Climate can be defined by a few inherent modes of variability with the addition of some weather noise, and variability can be thought of as shifts in the PDFs of the expansion coefficients of these modes. He then outlined a strategy for examining this paradigm by looking at a hierarchy of model experiments forced with observed and idealized SST patterns up to fully coupled simulations before presenting a number of different studies that have used this strategy to examine low frequency variability. The first study involved looking at a range of mechanisms that might be used to explain the observed co-variance between the winter NAO and summer N. Atlantic SST anomalies. The conclusion was that the persistent "horseshoe" pattern of SST anomalies could force an atmospheric stationary wave pattern with further feedbacks from atmospheric transient waves. The second study examined mechanisms for the observed correlation between autumn tropical Atlantic SST anomalies and an NAO like pattern in winter. Model simulations with the tropical Atlantic pattern give a significant mid-latitude response that, in turn, forces the tripole SST pattern. There is also some asymmetry between positive and negative forcing. He then gave a non-linear perspective on climate variability using clustering techniques. He defined 4 clusters associated with a NAO ±, a Gulf Stream mode and a ridge mode. The approach gives useful insight into the forcing by both tropical Atlantic and extra-tropical Atlantic beyond the usual linear thinking. The approach was further highlighted by a study of the "seasonal bridge mechanism" linking heating over the Caribbean and Sahel regions with a Rossby wave pattern. He turned finally to the role of the Tropical North Atlantic (TNA) region in decadal climate variability in the N. Atlantic. Analysis of century long SST forced experiments show that the TNA can force mid-latitude variability via a Rossby wave teleconnection and through a local enhanced Hadley Cell and corresponding increase in the jet stream and the baroclinicity. The mechanism was confirmed by 30-year experiments with idealized TNA forcing and shows that the forcing can be enhanced by non-linearities. In concluding he highlighted the need to better understand the variability of the important tropical Atlantic region (over both the ocean and the land) and to consider other possible factors such as the Indian Ocean, the Stratosphere and anthropogenic forcing. [top] Wednesday 3 JulyThe second day of the meeting was opened by Maarten Ambaum (Reading) who has been investigating the rôle of the stratosphere in the coupled system. The stratosphere can act as a long term "memory" via feedback of increased NAO leading to an enhanced stratospheric polar vortex causing low pressure over the arctic, further strengthening the NAO. Current research is investigating the effects of a high-resolution ocean model on at Intermediate GCM atmosphere to look at the length scales of SST anomalies, e.g. by reducing the width of the Gulf Stream. 750 day runs of the IGCM simulations with prescribed, variable width "Florida Current" give complicated responses is (and may not be statistically stable). He will then use linear wave theory to understand the model runs. [top] Evangelos Tyrlis (Reading) then introduced his PhD project that is using ECMWF ERS-40 data to look at the link between blocking and the NAO using a PV/theta blocking index. Preliminary results show that correlations are rather weak in winter but blocking is not as high as expected, possibly due to the area averaging used. [top] Mat Collins (Reading) then presented his work on Synthetic Analog Seasonal Forecasting. The concept is to select "analog" states, i.e.states that "match" the current situation in some sense, from runs of HadCM3 and then use these analogs to make ENSO forecasts. In this case, HadCM3 was sampled at TAO locations and temperature for Niño3 was used as a metric for similarity when scanning 2000 years of HadCM3 data. The system beats persistence and has some merit. [top] This research is being developed in a 3rd round project introduced by Myles Allen (Oxford). In the project, the climateprediction.com client will be adapted to do seasonal forecasting, improve ENSO forecasts and look at the problem of predicting European temperatures. [top] There followed two presentations introducing projects funded under the call for "User Driven Science" proposals, aimed at ridging the gap between scientific research and user application. The first of these, introduced by Sergio Pezzuli (Reading) looks at the use of climate forecasting in the UK Energy industry. The current systems in use take multi-regression of key points in the curve of daily demand vs temperature, precipitation etc. and have to include other diverse variables such as electricity costs and TV events. For seasonal demand, the only inputs are from climatology and non-climate parameters. The plan is to use Bayesian combinations of climatology and ensemble forecasts, to try and account for situations where ensemble variability is too low, and then to see if these forecasts can improve on the systems already used in the power industry. [top] The second User Driven Science project was then introduced by Glenn McGregor (Birmingham) who is planning on looking at the use of climate forecasts in seasonal Health sector planning (Presentation). Cold related deaths account for some 80,000 deaths per year in UK, the highest in Europe, and cold accounts for some 6.1 million hospital days. The current systems for planning include numerical weather prediction and tend to produce short timescale "fire-fighting" plans, whereas the aim is to move towards longer timescale "fire-prevention" planning. The project hopes to develop prototype climate-informed DST (Decision Support Tool). In order to provide a tool that is usable within the health Sector, a workshop will be held in order to understand the current decision making process, followed by diagnostic studies with evaluation. Throughout the process, communication between the climate researchers and the health sector is key. [top] This talk was followed by a brief discussion on user Driven Science in COAPEC. One of the main points raised in this discussion was that effective communication was always essential. Often there is fundamental misunderstanding between scientists and users due to a different interpretation of terminology. [top] Two related talks were presented either side of the morning break, the first by Dave Stainforth (Oxford) was an update on the status of the ClimatePrediction.com project. There have been a series of "behind the scenes" developments ready for public launch of the package including design of the visualisation packages and work on run allocation, data collection and experimental control. There has also been development on a related Open University, undergraduate level short course, possibly including experimental sub projects, and there are ongoing discussions about schools packages linked to the project. The Beta test is due for release in Aug/Sep 2002, with a public release expected in Nov/Dec. [top] The second ClimatePrediction.com talk was given by Nick Faull (Oxford) on his PhD project to look at perturbed physics ensembles with the ClimatePrediction.com models. In the HadCM3L (low resolution HadCM3) runs, the model will drift if your perturb parameters, unless each run has a long spin-up to re-calculate the flux corrections. In order to try and work around this, Nick has applied a correction to the flux correction calculated form the slab runs. This work is still in its early stages. [top] Chris Wilson (Liverpool) continued the session with a two part talk. Chris is investigating two topics: 1) ocean ventilation and 2) use of the coupled model FORTE. He concentrated on the latter topic during his talk. This work is stimulated by the work of Seager et al. (COAPEC 2nd Annual Meeting) and is concerned with the role of the ocean, as a source of local diabatic forcing in influencing the location and strength of the Northern Hemisphere mid-latitude storm tracks. After a brief description of the FORTE coupled model (Bablu's talk later), he showed plots and animations of the Eady growth rate, which picks out the storm tracks. He has performed a standard run and one with an aquaplanet (although orographic forcing of the atmosphere is retained). A comparison of these two runs indicates that ocean boundary heat transport (absent in the aquaplanet run) provides stronger storm tracks. Future work will address two topics: the effect of ocean resolution on heat transport by the boundary currents to the storm track region, and eddy effects; and the size of the ocean basins relative to the storm tracks - looking at the question of why the eastern Atlantic is warmer on average than the Pacific at the same latitude. [top] Andy Hogg (SOC), then described the development of a coupled QG model which explicitly models non-linear eddies and their effect in mid-latitudes. The advantages of such a model is that it is fast, eddy resolving and explicitly represents non-linear dynamics. Its disadvantages are that it uses reduced physics and has simplified bottom topography, and has not been done before, so there is a lack of experience to draw on. The atmospheric model has two layers, representing a 10km high atmosphere, and is connected to a mixed layer model at the surface. The ocean model has a mixed layer at its surface, connected to a two-layer ocean with a total depth of 4km. Andy showed animations of atmospheric surface temperature, lower layer pressure, and ocean temperature from the model. Stressing that the results are still preliminary, he showed atmospheric and ocean kinetic energy spectra from various model runs where the strength of the coupling between the mixed layer and atmosphere, and the effectiveness of geostrophic advection of temperature in the ocean, were modified. From the evidence of the differences in the KE spectra, he suggested that coupling acts to damp variability in the atmosphere at all frequencies except for a single peak in the KE spectra at a period of about 1 year, which he is still investigating [top] Further model developments were presented by Scott Osprey, (RAL) Scott is conducting sensitivity analyses using adjoint techniques. Quoting AW Robertson, who highlighted the 'red' nature of climatic timeseries, he suggested that the leading EOFs of variability have large spatial scales and long decorrelation times. This can be used to correct model drift in a scale-dependent way, rather than just relaxing to surface temperature and salinity climatology as is currently done with most ocean models, and which is not scale-dependent. In order to add scale-dependence to the relaxation, large-scale patterns (EOFs) are obtained from the first 100yrs of the HadCM3 control run (annual cycle removed) and their decorrelation times are worked out from autocorrelation of their principal component (PC) timeseries. Generally speaking for the first 12 PCs, the decorrelation time (e-folding time) reduces as the EOFs become higher order, although not as fast as hoped. The implementation of the scale dependent relaxation was done using the HOPE model by projecting the HOPE SST anomalies on to the HadCM3 EOFS. The amplitude of each EOF was then damped as per the relaxation time calculated previously. The residual field (i.e. residual after projecting onto EOFs 1-12) was damped strongly. [top] The final Keynote Talk was presented by Rainer Bleck, (Los Alamos National Laboratory) on the sensitivity of the oceanic overturning circulation in a coupled climate model to global warming Describing the ocean Meridional Overturning Circulation (MOC) as 'a capricious gear in the global climate mechanism', Rainer started by showing different climate-model predictions of THC evolution under global warming scenarios as published in the latest IPCC report. He emphasised the wide range of forecasts from no change to thermohaline circulation collapse. He then discussed a schematic due to W. Schmitz (2002) showing his view on observed water mass transport and transformation in the Atlantic Ocean. In particular the formation of North Atlantic Deep Water (~70% of the volume of oceanic water) occurs through overflow and entrainment processes over sills in the Greenland-Iceland-Norwegian Seas. As dense water flows into the North Atlantic, the overflow expands due to turbulence, mixing and Kelvin-Helmholtz instabilities. The degree of mixing is dependent on the speed of the overflow. The problem with level co-ordinates is that they tend to represent the overflow process as a horizontal plume followed by convection, which is an unrealistic way of representing the mixing processes. This is one motivation for using density as a co-ordinate to replace depth. This enforces physical constraints on the numerics. In particular, vertical advection smears out stratification in level models and lateral mixing should work along density surfaces. These processes are better represented in density co-ordinates. The main disadvantage of density co-ordinates is that they do not well represent turbulent mixing in the absence of stratification. This is where hybrid co-ordinates are introduced. These convert density surfaces continuously to level surfaces to prevent density surfaces outcropping (and consequent zero layer thicknesses). Rainer demonstrated that the method works well at the Brazil-Falklands Current confluence - a region of strong eddy activity and therefore a tough test of the numerics. Rainer then explained some details about the model: it uses a 'Pan-Am' (spherical bipolar) grid which removes the North Pole singularity problem by splitting it into two. This improves the resolution in the Arctic basins. Outside the Arctic there is a Mercator grid. The model is run without flux corrections and no separate spinups of ocean and atmosphere are required. Rainer described the results from two simulations, a control run starting from Levitus data, and a 2xCO2 run for 70 years, with a 1% increase in atmospheric CO2 concentration per year. The doubled CO2 experiment resulted in a 2K rise in globally averaged surface temperature. Rainer went on to describe the oceanic overturning in density co-ordinates, showing Atlantic, Pacific and Indian Ocean overturning separately as well as total overturning. The Atlantic overturning amounted to 20Sv, with slightly excessive overturning in the Pacific and too much bottom water in the Southern Hemisphere. The doubled CO2 experiment showed no reduction in overturning strength, unlike most level models but similar to ECHAM4, which is an isopycnic-coordinate model, providing food for thought. Rainer concluded by showing two diagnostics that he and co-workers have developed, which have proved to be illuminating. The first is called 'streamline bundling' and consists of separately summing adjacent northwards (and southwards) transport at each latitude, instead of the traditional zonally averaged value. This allows one to reduce the dimensionality of the dataset and also to look at regional processes (e.g. surface flow into the Labrador Sea and return flow at depth). The second diagnostic was diapycnal displacements. Comparison between the control and greenhouse gas run showed some differences between the two runs mainly at high southern latitudes. A further important difference, not related to the new diagnostics, was that the atmospheric low-pressure cell over Iceland was strengthened in the doubled CO2 run. Rainer concluded by saying that overall the coupled model gave good agreement with Schmitz's schematic. [top] The penultimate session of the meeting was opened by core team member Alan Iwi (RAL) presenting his work on coupled modelling on a PC cluster. This talk showed some validation results from various configurations of the Met Office Unified Model on a Beowulf cluster. The machine in question consists of 16 dual-processor Pentium 3 nodes, with Myrinet 2000 networking, and Linux operating system. The Intel Fortran Compiler was used in the runs shown. Speed benchmarks showed that 64-bit code on this machine runs at similar speed to the Manchester T3E; slightly faster than the T3E on 8 processors, slightly slower on 16 or 32 processors (i.e. scalability is slightly poorer). 32-bit code runs about twice as fast as 64-bit. HadCM3L (3.75° x 2.5° atmos, 3.75° x 2.° ocean) is about twice as fast as HadCM3 (same atmos, 1.25° x 1.25° ocean). Output was shown from integrations of HadAM3, HadCM3L and HadCM3 on the cluster, at 32- and 64-bit precision, and from comparative runs on the T3E (except for HadCM3L). This included: timeseries plots of global mean temperature and precipitation, as were annual mean upper-air temperature and wind fields, and ocean mixed layer depth and barotropic streamfunction. Results from the cluster at 64-bit were generally comparable with the T3E. However, in comparisons between 32- and 64-bit, significant differences of substantial magnitude were found in upper-air fields in those configurations that included an ocean model (even in HadCM3L where examination of surface temperatures showed little difference), and in the ocean fields themselves. An equation solver in the ocean convection scheme had been found not to converge at 32-bit; a workaround for this had been implemented, but similar undetected problems probably still exist at 32-bit. It is planned to continue the validation of the model on the Beowulf cluster.
This will involve investigating model-model differences between platforms in more detail, trying to identify the causes of any major differences found, and validating model climate against observational data.
(Field significance testing will be used for more rigorous statistical evaluation of difference fields.)
[top] Alex Megann (SOC) then presented the developments in CHIME (Coupled Hadley Isopycnic Model Experiment): an experiment to examine the influence of the ocean vertical coordinate on a climate model.
The model under test uses the HadAM3 atmosphere and ice models, coupled to the HYCOM ocean model, which has a hybrid isopycnic-depth vertical coordinate.
The ocean grid is identical to that of HadCM3 S of 55N, but has a bipolar grid in the Arctic (with poles at 55N, 70E and 110W, i.e. over land).
[top] Further model development, this time of a fast coupled model, was presented by the last of the core team Bablu Sinha (SOC). Coupling of HadAM3 to HYCOM (via the OASIS coupler) is in progress, but drifts still need to be checked. This talk concentrated on the FORTE model, which has coupled the IGCM3 spectral atmospheric model (T21 resolution, with 22 layers, incorporating a dynamical model, humidity, clouds, radiation, convection, a soil scheme) to the MOMA finite difference ocean model (15 layers, forced by winds and surface heat/salinity fluxes, incorporating free surface and a simple mixing convection scheme) via the OASIS coupler. The coupler runs as a separate process, enabling the parallelised ocean model and the computationally cheap atmosphere model, running on a single PC model, to communicate while running independently. The coupler does interpolation of fields and synchronisation of the models. Periodic coupling can also be used. The model has not yet reached equilibrium (as shown by time series of temperature and salinity). However, the temperature field shows realistic features of the ocean circulation. The salinity fields show the Pacific is too salty (which is under investigation), and this excess salinity leads to some spurious Pacific overturning. The atmospheric fields have also been evaluated and zonal wind stress and surface temperatures both compare reasonably with climatology. The overturning streamfunction has good similarities with HadCM3 and shows some decadal variations, so this model can be used to investigate variability on these timescales. [top] Developments of an even faster coupled model were then presented by Bob Marsh (SOC). The "GOLDSTEIN" model, a 3-D ocean model, coupled to a 2-D energy moisture balance model of the atmosphere, is designed to look at the role of sea-ice in interannual - decadal variability, as part of Paul Valdes' second round project. The model has internally computed heat and freshwater fluxes, prescribed winds, zonal mean albedo, and atmospheric heat and moisture diffusivityand implicit net atmospheric water vapour transport from Atlantic to Pacific. The model is run on coarse resolution (36x36) global grid with realistic geometry but smoothed topography and 8 ocean levels. Spatially variable ocean drag is applied, and coupling is performed every ocean and every 5 atmosphere timesteps The mean climate looks OK after a 2000 year spinup (run in only 10-12 hours real time!). An ice model has been incorporated. It is now planned to introduce physical refinements, an alternative grid for better Arctic representation, seasonality in solar forcing with stochastic noise. [top] The final modelling talk of the meeting was given by Doug Cresswell (UCL) who presented an Arctic sea ice model, which it is planned to use to run an ensemble of perturbed physics and forcing, and to optimise using remote sensing observations. The model is based on CICE code developed at LANL and has multi-category, multilayer thermodynamics, elastic-viscous-plastic dynamics, 100km resolution and a mixed layer model. Forcing is from the POLES forcing dataset (with ECMWF fields later in the time series and SST & SSS from PHC climatology. Preliminary results show that the modelled ice thickness is too thin compared to remote-sensing observations by about 1m. It is hoped to improve this using higher resolution. In time series of regional mean modelled and observed sea ice, it is possible to reproduce the observations well by tuning parameter values; the values needed are within the justifiable range, but there is no physical a priori reason to choose the particular values. The ice thickness increases during the 1980s, and decreases again during the 1990s - in agreement with other modelling studies. Data from ENVISAT and other satellites to come in next few years, that have smaller polar islands than previous satellites, will give useful validation. The model shows that the length the of ice melt season is a strong controlling factor of ice thickness, with tight anti-correlation between ice thickness and length of melt season. The model also shows that variations in the absorbed shortwave radiation are much greater than variations in the incoming down-welling shortwave, implying the importance of the sea ice albedo feedback. [top] There followed a brief discussion of model development within COAPEC. Any researchers with models, or model runs, of use to the wider community were encouraged to put information on the models onto the COAPEC CAST site. [top] The final talk of the meeting was from Mark Saunders (UCL) on Seasonal predictability of the winter NAO. Two models have been developed using statistical methods for predicting the NAO, one based on lagged modes of SST variability, the other based on the snow extent during the preceding summer. The NAO is the dominant mode of northern hemisphere winter atmospheric variability (particularly over the north Atlantic). Skilful NAO seasonal forecasts would bring benefits, but have been largely unsuccessful so far. For each of the two statistical models, monthly sea-surface temperature and mean sea-level pressure fields from NCEP were used to examine the predictability of three different winter NAO indices (to ensure that the predictability is not sensitive the to particular choice of indices). The first model uses statistically significant, and spatio-temporally stable principal components, with NAO-linked modes of SST variability identified from lagged anomalies using 1, 3, and 5-month averages. A 5-month (June-Oct) average gives better skill than 3-month (Aug-Oct) or 1-month (Oct) average, according to the various measures of predictive skill that are used. The correlation skill from principal components analysis is typically from 0.5 to 0.6, although there were periods of higher (1970s and 1990s) and lower (1960s and 1980s) skill over the last 50 years. The suggested physical mechanism is the persistence of the SST anomaly patterns, which feed back to influence the NAO through an NAO / SST tripole association. The second model, based on the lag-correlation of the NAO with the snow-cover extent peak for the preceding summer (June-July or July-August depending on the NAO measure used) gave good autocorrelations between the DJF NAO index and the preceding July-August snow extent. These give on average ~0.63 skill, with similar skill values according to various skill measures. The suggested physical mechanism is a change in albedo due the to snow cover extent, giving rise to latitudinal temperature gradients which persist through to winter. [top] |
| Maintained by Helen Snaith | Last modified: July 06 2005 15:45 |
