a. Summary

Improvement or the model taken in newly is developed for many processes of the stratosphere by the climate model (CCSR/NIES a model, existing) of the atmosphere, the ocean, and a land side which consists mainly of a physical process.

About improvement of an air model (AGCM), improvement of many processes of an inadequate middle atmosphere (the stratosphere and mesosphere) is aimed at by the present model. That is, while change of the radiation from the physical chemistry process and the sun of an ozone layer peculiar to a middle atmosphere influences each other mutually and causes change of a middle atmosphere by invasion of the artificial origin substance to the inside of a middle atmosphere, it clarifies the mechanism which combines with change of the lower layer troposhere and produces a climate change by model experiment. Moreover, the action of an internal gravity wave and it clarify influence affect atmospheric circulation by a super-high resolution air model.

This fiscal year mainly introduced the extension to the middle atmosphere of the integrated model which combination of the atmosphere, the ocean, land, a carbon cycle, atmospheric chemistry, and aerosol process completed, and a physical process component (a σ-p hybrid coordinate system, a new radiation code, and Hines parameterization) required in the stratosphere, and did version arrangement / unification work during the program of an integrated model. Moreover, since the radiation balance of an air upper bed changed with change of a radiation code and perpendicular resolution, the tuning about itinerant Buddhist distribution of the troposhere was started. About improvement in the speed of a calculation code, the code was improved so that the parallelization function in a node of an Earth Simulator could be used in parallel between the conventional MPI nodes. About research of the air internal gravity wave using high resolution AGCM, it carried out succeedingly for improvement of Hines non-orographic gravity wave resistance parameterization. Moreover, for presumption of momentum flux change of the gravity wave at the time of global warming, the result of a warming experiment of a symbiosis first division title and a Sumi group was analyzed, and it collected into the paper.

b. Research purpose

The purpose of this research is to understand better various process in the atmosphere connected with it to be development and improvement of a climate physics model. In order to simulate correctly the interaction process of the change and climate of atmospheric composition in a middle atmosphere, the temperature place of the air motion which governs transportation of an air minor constituent or aerosol, and the atmosphere important for photochemical reaction process especially needs to be appropriately reproducible.

Cooling by heating which the large-scale circulation in a middle atmosphere, the seasonal variation of a temperature place, and solar-ultraviolet-radiation absorption according to ozone in order to more often reproduce change every year bring about, and the infrared radiation to which the greenhouse effect gas make into representation releases carbon dioxide, methane ozone, and steam, i.e., radiation process, and various airwaves ranging from hundreds of m to a planet scale are consider need to be appropriately express in a model, respectively. In order to express radiation process and a small-scale airwave correctly, it is thought that horizontal / vertical resolution of a model must be to some extent high.

However, resolution sufficient required when conducting the warming prediction experiment included to the interaction with atmospheric composition change over a long period of time is not clear enough even now. It is large-scale calculation using an Earth Simulator, and it is scientifically meaningful to clarify dependability over model resolution in each process, and it is the main subject of this sub subject matter while it is indispensable for architecture of an integrated model.

In the overall plan, while setting development of the general circulation model used as the foundation of a final integrated model as a long-term target, development is performed and offer of the air model united with the needs of each subgroup (partial integrated model) is performed.

c. A research program, a method, a schedule

Improvement or the model taken in newly is developed for many processes of the stratosphere by the climate model (CCSR/NIES a model, existing) of the atmosphere, the ocean, and a land side which consists mainly of a physical process. Horizontal resolution which does not carry out the parameter rise of the internal gravity wave but as for which direct picking treats it in order to adopt correctly the temperature of the stratosphere and the mesosphere atmosphere, and the effect of the internal gravity wave which plays a big role in circulation and substance transportation It is necessary to do a numerical simulation by the model about 20 km and perpendicular thickness 100 m. This experiment will be conducted by 2nd, and while even a middle atmosphere is included, it enables it to include new parameterization in resolution atmospheric chemistry and a climate joint model in 3rd. The chemistry and the climate joint model containing a middle atmosphere cooperate also with development of sub subject matter (2) "warming and atmospheric composition interaction model", furthers development in parallel, and experiments by inside resolution models, such as a synergistic effect of ozone layer depletion and warming.

d. The research program in the Heisei 17 fiscal year

The upper bed of an integrated model is extended to the stratosphere, and it aims at making realistic especially seasonal advance of the atmospheric general circulation of important inside and high latitude for the amount of ozone of the stratosphere. While introducing into an integrated model the hybrid perpendicular coordinates developed by the last fiscal year, a new radiation code, and non-orographic gravity wave resistance parameterization and performing fine tuning, improvement in the speed of a calculation code is attained towards long-term integration. The wave motion of the stratosphere and the research of general circulation by a high resolution air model required for improvement of gravity wave resistance parameterization are done succeedingly.

e. Fruits of work in the Heisei 17 fiscal year

e.1 The result about upper bed extension of an integrated model

The extension to the middle atmosphere of an integrated model and a physical process component (a σ-p hybrid coordinate system, a new radiation code, and Hines parameterization) required in the stratosphere were introduced. The experimental findings at the time of already using an AGCM simple substance for the report of a degree about improvement in the reproducibility of physical climate the year before last by a σ-p hybrid coordinate system and new radiation code introduction were described in detail among these. Below, it describes about the experimental findings using the integrated model which actually introduced Hines non-orographic gravity wave resistance parameterization.

First, the derivation procedure of the non-orographic gravity wave sauce inputted into Hines gravity wave parameterization (Hines 1997) is explained in full detail. One are all ball unity and it is isotropic and the means of assuming the gravity wave sauce which does not carry out time diversification, and launching it from the troposhere lower part (e. g. and Manzini and McFarlane 1998). The big demerit of this method is that the propagation direction and the seasonal variation of the geographic distribution and the gravity wave of sauce are disregarded. Another technique is a method of asking for gravity wave sauce somewhat experientially inside a model using non-insulating heating which can be found from cumulus convection parameterization of a model, and information, such as front activity. The law learned by experience used here is a relation with the phase velocity spectrum of gravity wave momentum flux to the heat source distribution of a cumulus convection and the distribution of the wind of a background place which were simulated in the experiment using the cloud resolving model (the many are two dimensions) performed separately which appeared over the cumulus convection (e. g., Beres et al.2004). However, there is a problem whether the convection of several level km scale and the relation of a gravity wave are applicable to non-insulating heating (average value of about 300 km four quarters) called for from cumulus convection parameterization of AGCM. Moreover, an argument is still going to be just only by the relationship between non-insulating heating and a background place also about the ability to ask for the phase velocity spectrum of a gravity wave in approximation primarily (Chun et al.2004). Instead of these, into our group, the climate value of the gravity wave of the lower stratosphere is calculated, and it is inputted from the simulation result by high resolution AGCM called T213L256. The merit of this method is the point which can include that the wave of a required level scale can be extracted alternatively, and realistic geographic distribution, realistic propagation direction, and seasonal variation of a gravity wave in sauce.

It was total and integrated with T213L256 AGCM for two years. From the result, with the horizontal resolution of an integrated model, only the wave motion ingredient of under about 950 level wavelength km that cannot be expressed is taken out with the number filter of high waves, and it asks for gravity wave sauce. Although about 60km and the minimum level wavelength are about 190km and its level grid interval which can be expressed with the horizontal resolution of T213 is inadequate for resolving the gravity wave which comes out of each convection, squall line, etc., the turbulence of the level scale of the envelope of the cold front or a thunderhead is resolvable. It is actually thought that the structure of the general circulation of the middle atmosphere of this model is quite realistic, and the slowdown action of the average style by the breaking wave in a mesosphere can be expressing it as the momentum flux by a gravity wave correctly somewhat quantitatively (Kawamiya et al.2005). As sauce being advanced, in consideration of the influence of the horizontal transmission of the inertio-gravitational wave which excels in the lower stratosphere, it asked in respect of 70 hPa, and we are the relation from which the zonal wind of Equator QBO gets down below 70 hPa side, and decided to ask in respect of somewhat low 100 hPa by low latitude in inside and high latitude. After taking out an ingredient with a level wavelength of less than 950km, the semi- regular ingredient which is equivalent to a ground formation gravity wave by deducting the time average value of 48 hours is removed from the data of the perpendicular style as the level style outputted on the average per hour. In the above-mentioned constant pressure surface top, it asks for the perpendicular flux of the east and west and the north-south momentum in each grid, and it is assumed that the gravity wave which had the momentum flux of the size in eight nearest directions from the orientation of the vector has spread. In fact, although the wave of various character is flying in the various directions, as the wave has spread in the one direction to one moment value as the set, it is dealt with here. The data for January is repeated, this is calculated and averaged, and it asks for the momentum flux of as another gravity wave as the methods of eight of a monthly average, and distribution of level wind velocity. From the data of AGCM, these both are inputted into parameterization, and although distribution of level wind velocity and a typical level wave number are inputted and it asks for momentum flux in original Hines parameterization using the dispersion relation of a gravity wave, since distribution of momentum flux and level wind velocity can be found easily, the program is changed in part so that it may not be contradictory in dispersion relation and a level wave number can be found (refer to the report in the last fiscal year).

Fig. 41 is an example of the sauce for which it asked in this way. It has limited and drawn in eight days containing the summer solstice of the Northern Hemisphere. Although not stated in detail here, the gravity wave accompanying the storm track of the Southern Hemisphere, the gravity wave by the convection which stands in connection with the mountains slope near the South Andes and the gravity wave accompanying the tropical cyclone near the Bay of Bengal, the gravity wave accompanying the convective activity near the seasonal rain front of East Asia, etc. serve as large sauce especially. On the other hand, since what has a comparatively large level wavelength is main as for the gravity wave accompanying the strong convective activity near ITCZ of Nishi Pacific Ocean, it has not appeared in this sauce distribution not much strongly.

Fig. 41: Distribution of gravity wave momentum flux in 70 hPa side on 21-June 28. An arrow is the square root of the size of the gravity wave momentum flux of eight directions, and a color is the square root of an average of the size of all the direction ingredients of the gravity wave momentum flux of each grid. (The square root was shown the decipherment nature of a figure sake)

The experiment using an integrated model was conducted inputting such gravity wave sauce into Hines parameterization. In Hines parameterization, a gravity wave spreads a part high up in the sky in response to the wind of an integrated model, or the reaction of thermal stratification, and a part dissipates and influences the wind of an integrated model, and distribution of temperature. Here, since emphasis was put on the upswing in the reproducibility of physical climate, the atmospheric chemistry model CHASER conducted the experiment for four years after a adequate initial condition, and compared three years of the last of a result with the observation climate value without joining together. Fig. 42 shows latitude-advanced distribution of a beltlike average zonal wind of the winter (December to February) of the Northern Hemisphere, and a summer (June to August) average. Left-hand side is the climate value of CIRA86, and right-hand side is the result of an integrated model. A stratosphere polar night jet and the structure near a mesosphere jet are well reproducible qualitatively in Northern Hemisphere winter (upper row of a figure). This means that the gravity wave sauce which carried out same actual distribution produces loss in the same actual place, and is weakening the west wind on that occasion. The structure of such a jet is unreproducible in the conventional Rayleigh friction and the model which used uniform gravity wave sauce. On the other hand, while the structure where the axis of a polar vortex inclines to the equator side in high altitude is qualitatively reproduced in the mesosphere, the bias whose polar night jet of the stratosphere is too strong compared with observation is seen in Southern Hemisphere winter (lower berth of a figure). The main causes of this bias have important pole orientation propagation of the latitude gravity wave in the Southern Hemisphere in the ability not to express by Hines parameterization in the actual atmosphere (refer to the report in the last fiscal year).

Fig. 42: Latitude-advanced section of seasonal average beltlike average zonal wind. left line: CIRA86 climate value. right line: Integrated model. Isoline interval: 10 m s^-1.

Fig. 43 is the bias from the British Weather Bureau assimilation data climate value (1994 -2001 annual average) of the beltlike mean temperature of a seasonal average. If near the tropopause of the altitude more than 1 hPa with inadequate observation, and inside and high latitude is removed, it turns out that the integrated model using Hines parameterization (a new radiation code is also important) can be expressing atmospheric temperature structure correctly about. It can be said that it is a big result that there is no low-temperature bias which was especially conspicuous in the South Pole lower stratosphere (near 50 hPa of SON) of Southern Hemisphere spring important for reappearance of an ozone hole.

Fig. 43: Latitude-advanced section of bias from British Weather Bureau assimilation data (1994 -2001 average) of integrated model seasonal average beltlike mean temperature. Isoline interval: 2K.

Finally, although it is about QBO of the equatorial sky, or the reproducibility of SAO, at present, tuning of the physical climate of the troposhere is not settled completely, but it is connected also with the wave motion ingredient which an integrated model can resolve explicitly not being expressed correctly enough, either, and is in the tendency for the cycle of QBO to be shorter than the observation fact (about 26 months) (about 16 months). After tuning up troposhere climate succeedingly also next year, it is necessary to examine this point.

Compared with the integrated model of the troposhere version which did not include the stratosphere and a chemical process, calculation cost of the integrated model extended to the stratosphere increases sharply by the increase in a perpendicular total (20 layers → 80 layers), reduction of the time step by a mesosphere jet being contained, the increase in the number of tracers, calculation of a chemical process, etc. For this reason, if the same parallelization technique as the former is used, in case a warming prediction experiment is conducted, execution high-speed enough cannot be performed. Then, in addition to the conventional "domain division of the direction of latitude by a MPI process", improvement in the speed by "hybrid parallelization" which newly uses "the micro task in a node" was performed.

On the program structure where the parallelization by the conventional MPI performs domain division in the direction of latitude, in the case of the model of T42, 64 division is maximums, and if vector length is further taken into consideration, it can parallelize only to 32 division actually. Since horizontal division is a limit, it is necessary to divide by the kind of a vertical direction or tracer etc. but, and the domain division by MPI affects the whole source code. Dividing into a vertical direction in a MPI process further in addition to the direction of latitude has large work cost. On the other hand, when parallelizing by a micro task, since an automatic parallelization function can be used, there are few amounts of work and they end. Moreover, generally the performance of the communication accompanying parallelization is better than between processes between threads (between micro tasks). For this reason, it can be said that the inside of a node is a method with natural parallelizing a vertical direction by a micro task.

By dividing at new dimensions (a vertical direction, tracer kind, etc.), PE number (parallel number) which can be used can be increased conventionally. It is because each MPI process can be further divided by a micro task after division by MPI. The usual maximum value of the number of micro tasks is the number of processors in a node, and, in the case of an Earth Simulator, it is 8. Therefore, it is more possible than the case of only an MPI process to use 8 times as many PE as this. However, if performance per 1PE is not maintained even if it increases PE number, it must be careful of a meaningless point.

Since hybrid-ization does not change a deed and the ocean side does not change calculation cost only as for the atmosphere side, parallelization only using a MPI process as usual is performed. Since it is the form where the structure where the atmosphere side differs from a hybrid and the ocean side differs only from MPI is intermingled, a program serves as MPMD inevitably. The thing of only a MPI process is called flat MPI to calling hybrid MPI the style which combined the MPI process and the micro task. Although the concrete work about hybrid parallelization is not explained in full detail, after performing the optimization which was possible in addition to removal of a SAVE failure of a program, an initialization failure, etc., and parallelization, it is necessary to add a directions line required for the parallelization in a node to the part which the automatic parallelization function of a compiler does not commit. Furthermore, the parallelization directions line by Open MP is added to some source files which cannot use an automatic parallelization function.

This target has been attained although it was that important one does not drop the performance per 1PE about performance. In order to compare the performance per 1PE, flat MPI and hybrid MPI were performed and tested by the same PE number. As for both, a use PE number is atmosphere side 32 and ocean side 16 (in flat MPI, in 32 and hybrid MPI, the number of the MPI processes by the side of the atmosphere is 4.). The ocean side does not change both but is 16. When Real time was compared, hybrid MPI was as quick as 2690 seconds to 2807 seconds of flat MPI. It can be said that the parallelization by the micro task in a node has attained by this the rate only of parallelization which is not inferior as compared with flat MPI.

It was checked how far PE number could be increased. 256PE(s) at the time of making the number of processes by the side of the atmosphere the same as 32 which is the realistic maximum of flat MPI are thought as a maximum. However, this is a maximum, and considering parallelization efficiency, it can be predicted using this number that efficient parallelization is impossible. Although the increase in a perpendicular layer is working in favor of hybrid parallel, it is because 3-dimensional calculation which all calculation includes for a perpendicular layer is not performed. Moreover, even if it is 3-dimensional calculation, dependability is in a vertical direction and it cannot parallelize in many cases. In such a case, a horizontal grid will be divided by both the MPI process and a micro task, and sufficient particle size or vector length cannot secure.

Change of the performance when increasing PE number was summarized in Table 8. If it asks for the rate of parallelization, those with 99.3% or more and the use application of 144PE are possible. However, in 144PE(s), the performance has fallen considerably. After introducing stratosphere chemistry from now on, performance must be evaluated again, but if efficiency is thought as important, it will be considered that it is better to perform at least by 80 PE.

In the integrated model, the gravity wave climate value of high resolution AGCM is used as sauce of Hines non-orographic gravity wave resistance parameterization. In case a warming prediction experiment is conducted using this, change of the gravity wave momentum flux accompanying global warming may do big influence. For example, if the regional distribution, and the intensity and the frequency of convective activity of the troposhere differ from present condition climate greatly, change of the character of the gravity wave which comes out from there may be unable to be disregarded, either. Such a problem is considered to influence the cycle of Equator QBO etc., and is beginning (for example, Giorgetta and Doege 2005) to attract attention in recent years. However, there is no implement at present besides investigating in the warming experiment using AGCM as a response of the convective activity like all balls to global warming, and, moreover, status quo is that is not so high as the horizontal resolution of the usual AGCM can resolve a gravity wave. the warming assumption experiment by the high resolution air sea joint model which the Sumi group of a symbiosis first division title performed under such circumstances be conduct by combine T106L56 AGCM with the ocean, and in that the gravity wave to a minimum level wavelength 380 km grade be resolvable, when investigate gravity wave change of all the ball scales at the time of warming, it can be say that it be the only material.

It asked for the perpendicular flux of the east-and-west quantity of motion respectively accompanying the severe storm ingredient (from the level wavelength of 380km to 930km) of the lower 70hPa side of stratospheres as a par condition for 20 years of the date equivalent to the carbon dioxide redoubling climate of a pre-industrial par experiment (CTL), and 1% / year carbon dioxide gradual increase experiment (GHG), and the difference was defined as diversification of the gravity wave momentum flux accompanying warming. It summarizes only by the text below about the main change at the time of warming, and the factor which brings it about. First, although change of the convective activity of the troposhere was setting distribution of strengthening in the storm track of inside and high latitude to tropical ITCZ (especially on eastern part Pacific Ocean), corresponding to this, the increase in the momentum flux of the gravity wave in the sky was seen. And the momentum flux by a west gravity wave increased by strengthening of a subtropical jet in the sky. These can be qualitatively called expected change. On the other hand, the momentum flux by a gravity wave of the observed 70hPa side increased mostly about 20 to 40% all over the districts. This change is not based on change of the intensity of the sauce of the troposhere, or the wind of a background, and in order that the upper levels of the troposhere which are the generation region of the main gravity waves may also shift to high altitude with regards to the tropopause carrying out a 1-2 km grade rise in all balls in connection with warming, it thinks for the gravity wave which arrives at the lower stratosphere to increase.

Although it must investigate about the reliability of these experimental findings from now on also, supposing it happens that it is actually such, it may be able to be said that it is not appropriate to continue using for the warming experiment of an integrated model the gravity wave sauce of 70 obtained from the present condition climate experiment of T213L250 GCM or a 100 hPa side. The gravity wave research using high resolution AGCM must be continued, and new gravity wave sauce must be derived from now on also if needed, or a still newer measure must be taken.

f. Consideration

Introduction of a required physical process was able to be completed in the extension and the stratosphere to the middle atmosphere of an integrated model. About seasonal advance of stratosphere general circulation, there is also worth of Hines parameterization introduction which gives original sauce, and the best reproducibility in the newest model of the present other organizations was able to be acquired. From now on, the check of tuning of troposhere climate and the climate sensitivity at the time of warming etc. must do the work towards a warming experiment. Furthermore, after improving a point inadequate about the present condition about reappearance of the stratosphere equator QBO, the experiment using the integrated model which combined stratosphere chemistry is started. Moreover, in order to make the gravity wave sauce inputted into Hines parameterization into what was more suitable for the warming experiment, it is thought that it is necessary to conduct continuously the experiment which used high resolution AGCM. When using a present Earth Simulator and a present integrated model about improvement in the speed of a calculation code, as a result of accelerating the limitation considered, an integrated model without stratosphere chemistry can be performed now by about 60 percent of the conventional computation time. It is expected by the induction of a stratosphere chemical process being completed at the beginning of next year that it can decide upon the plot of a concrete warming experiment using an integrated model.

g. Bibliography

Beres, J. H., M. J. Alexander, and J. R. Holton, A method of specifying the gravity wave spectrum above convection based on latent heating properties and background wind, J. Atmos. Sci., vol.61, 324-337, 2004.

Chun, H.Y., I. S. Song, and T. Horinouchi, Momentum flux spectrum of convectively forced gravity waves: Can diabatic forcing be a proxy for convective forcing?, J. Atmos. Sci., vol.62, 4113--4120, 2005.

Giorgetta, M., and M. C. Doege, Sensitivity of the quasi-biennial oscillation to CO2 doubling, Geophys. Res. Lett., vol.32, L08701, doi:10.1029/2004GL021971, 2005.

Hines, C. O., Doppler-spread parameterization of gravity-wave momentum deposition in the middle atmosphere. Part 2: Broad and quasi monochromatic spectra, and implementation, J. Atmos. Solar Terr. Phys., vol.59, no.4, pp.387--400, 1997.

Kawamiya, M., C. Yoshikawa, H. Sato, K. Sudo, S. Watanabe and T. Matsuno, Development of an Integrated Earth System Model on the Earth Simulator, J. Earth Simulator, vol.1, 2005.

Manzini, E., and N. A. McFarlane, The effect of varying the source spectrum of a gravity wave parameterization in a middle atmosphere general circulation model, J. Geophys. Res., vol.103, no.D24, 31523--31539, 1998.

Watanabe, S., T. Nagashima, and S. Emori, Impact of global warming on gravity wave momentum flux in the lower stratosphere, SOLA, vol.1, 189-192, 2005.

h. The announcement of a result

Watanabe, S., K. Sato, and M. Takahashi, Orographic gravity waves over Antarctica excited by Katabatic winds; a GCM study, J. Geophys. Res., 2006 (submitted).

Watanabe, S., T. Nagashima, and S. Emori, Impact of global warming on gravity wave momentum flux in the lower stratosphere, SOLA, vol.1, 189-192, 2005.

S. Watanabe, M. Takahashi, and K. Sato, Orographic gravity waves over Antarctica excited by Katabatic winds; a GCM study, IAGA2005 Scientific Assembly, July 21, Toulouse, France.

S. Watanabe, Development of Chemistry Coupled Models at CCSR/NIES/FRCGC, IAGA2005 Scientific Assembly, July 21, Toulouse, France.

S. Watanabe, M. Takahashi, and K. Sato, GCM Studies on Atmospheric Gravity Waves: Gravity Waves over Antarctica, CAWSES workshop, September 13, Nagoya University.

S. Watanabe, On source spectra for the Hines gravity wave drag parameterization in KISSME, The 8th International Workshop on Next Generation Climate models for Advanced High Performance Computing Facilities, February 24, Albuquerque, USA.