MOU UK-Japan Climate Collaboration
The aim of this collaboration is to produce ground-breaking climate simulations on the Earth Simulator which could not be achieved elsewhere. A key scientific goal is to evaluate uncertainty due to structural limitations on current climate models, either due to limited resolution or omitted physical or dynamical processes, with particular focus on improving the representation of regional climate and its variability. This will be achieved through a partnership between the Earth Simulator Center, the NCAS-Climate (Reading, UK) and the Met Office Hadley Centre (Exeter, UK).
Completed 100 years of simulation with coupled models at the lowest and highest resolution in model matrix (150km atmosphere-100km ocean and 100km atmosphere-30km ocean). This has allowed us to investigate the role of model resolution on variability (particularly ENSO) on global and regional scales.
Completed 25 years of 100km and 60km atmosphere model integrations forced by AMIP2 SSTs and SSIs, to investigate the role of atmosphere resolution in processes such as monsoon, tropical cyclones, local rainfall and the diurnal cycle.
Completed 20 years of 60km atmosphere model forced by AMIP2 SSTs with +4K added, to investigate the impact of warming on tropical cyclones
4 x 10 year ensemble integrations of 60km model to test model sensitivity to model formulation and improved land surface representation
Several experimental integrations of 60km atmosphere coupled to 30km ocean, to examine interaction of extreme events such as tropical cyclones on ocean processes.
The growth characteristics of minimal perturbations (changes to temperature of ~10-14K) added to a model state are an important diagnostic of any climate or forecasting model. Growth rates significantly greater than the theoretically expected rates documented by Rosinski and Williamson (SIAM Journal on Scientific Computing, 1997) indicate coding errors in the model, and may lead to problems with model validation on different computer platforms. Work carried out by UJCC on the Earth Simulator has shown the perturbation growth rates in the Met Office Hadley Centre models to be much too high. In order to find the coding errors causing this problem, the UJCC team have developed a novel technique for isolating the errors, and have been able to fix several important bugs as a result. This package of work will be continued in the UK with assistance from the UJCC team.
Investigation of instabilities in the 60km model uncovered several problems related to filtering, diffusion and other choices – these are now significantly alleviated.
The simulation of the tropical Pacific climatology is much improved when using the higher resolution ocean model. Tropical instability waves act to converge heat at the equator and warm the sea surface temperatures, a process not well represented at low resolution. A paper is complete on this work, and is submitted to the Journal of Climate.
The El Nino/Southern Oscillation variability is poorly represented in the low resolution coupled model, but with higher atmosphere and ocean resolution the characteristics of ENSO are reasonably well simulated, both in terms of the local process, and the impact of ENSO on global climate. Analysis reveals that the termination process of ENSO, due to wind stress curl causing heat content anomalies in the Western North Pacific, is the primary cause of this difference, a mechanism also seen in observations but not in any previous Hadley Centre model. A realistic ENSO simulation is very important in order to study regional climate variability and change. A paper on this work will be completed in early 2008.
Initial analysis of tropical cyclones (TCs) in the range of atmosphere models suggest that their intensities are better represented as model resolution is increased to 60km. Mid-latitude storm tracks show some improvement at higher resolution, in particular over the North-Atlantic sector. A systematic storm-track intercomparison, using a sophisticated three-dimensional diagnostic approach (TRACK, Hodges et al.) covering all our model resolutions and including similar integrations from some of our Japanese collaborators are well underway. Comparison with other data sets, including Japanese very high resolution 5-10km models and re-analysis data (e.g. ERA-40) and observational case studies will indicate how realistically such storms are simulated. Multi-decadal integrations will allow an examination of how TCs impact on the large-scale mean climate at mid-latitudes.
Regional climate and its variability also depend on seasonal to interannual variability (e.g. PNA, NAO) and weather-type variability (e.g. blocking), so these statistics will be examined in the AMIP2 and coupled integrations.
The Madden-Julian Oscillation (MJO) is the dominant mode of intraseasonal variability in the tropics. Via its profound effects on the tropical circulation in the Eastern Hemisphere, the MJO has strong connections with a number of other tropical phenomena, including ENSO, the Indian Monsoon and Tropical Cyclones.
During the year, a report was completed on the characteristics of MJO across the UJCC model matrix, and also in an atmosphere-only version of the Japanese MIROC model. The coupled UJCC models all produce good simulations of the MJO, but have problems propagating MJO anomalies past the Maritime Continent. Although the tropical climate is found to improve significantly with increased resolution, the affect on MJO was found to be small, partly due to a persistent bias in surface winds in the Eastern Pacific. The MJO will also be examined in the 60km atmosphere coupled model, as well as in the climate change integrations, to uncover any significant impact.
An integral component of the construction of the high-resolution AGCM has been the gathering and compilation of high-resolution boundary conditions, which have mainly focused on surface albedo (from MODIS) and soil physical parameters. UKMO standard soil maps had not previously been updated and their resolution was more appropriate for coarse GCMs, like HadCM3. In order to properly intercept and store soil water, which has as input the high-resolution precipitation produced by the 60km AGCM, it was found that more modern soil maps and updated parameters were needed. IGBP data at a nominal resolution of 1km were used, which also provide information about soil heterogeneity, which is important for hydrology work. Initial results (see Precipitation and river outflow section) indicate that model sensitivity is primarily limited to simulated impacts (mainly runoff and vegetation NPP), but work is underway on the analysis of evapotranspiration, which could feed back onto model climatology.
A dynamic closure for convection is implemented in the convective parameterization of the HadGEM1 model. Vertical velocity is used as a controlling factor for the timescale over which deep convection consumes its potential energy. This can be interpreted as the probability for deep convective clouds to be initiated. Furthermore, it decouples physical and numerical constraints on the convective timescale. Simulations with this new closure show that the tropical diurnal cycle of precipitation is improved by a delay of one hour. Cloud top height and the time at which deep convection occurs both increase. This changes the radiation balance drastically, through a higher cloud albedo and less outgoing longwave radiation. Near the surface, modification of boundary layer humidity through compensating subsidence and downdraughts influences the consecutive evolution of convection. However, the incorrect model representation of the diurnal evolution from shallow to deep convection is not changed.
Initial analyses of precipitation characteristics made over a mountainous area, the European Alps, have shown that high-resolution atmospheric models have less frequent and more intense precipitation over land due to better-resolved convergence near the surface, caused by orography. The precipitation distribution is also more localized in space and time, especially in winter over mountain peaks and along coastlines. The higher resolution model shows a clear improvement in the representation of snowmelt timing, attributed to the higher elevation of snow cover. This has a direct impact on the river flow seasonal cycle, which is much closer to observations. A paper on this work will be completed in mid-2008.
Coupled climate models often suffer from large biases in regions adjacent to coastlines. Although these areas only occupy 0.5% of the global ocean, they are a key part of the carbon cycle and our food supply. Observations show that these regions are very sensitive to climate change. While our low resolution model shows significant but opposite-signed biases in summer and winter, the high resolution model follows the observations much more closely. Although a higher resolution in either the atmosphere or the ocean improves the simulation, both are clearly needed to give a good simulation throughout the year. The increased atmospheric resolution improves the processes that determine the radiation balance over these stratocumulus areas in summer, while the ocean resolution moderates the seasonal cycle through a stronger upwelling response which is important throughout the year.