Cloud amounts significantly control the earth's radiation budgets through the change in albedo for short-wave radiation and the out-going long wave radiation. Also, cloud systems cause large radiative and latent heat evaporation cooling/heating to the atmosphere. Inhomogeneous diabatic heating associated with clouds produces available potential energy and drives the atmospheric motion with various scales. In numerical simulation models, shortcomings of cloud parameterization considerably limit their predictability for short and medium-range predictions and bring large uncertainties to climate projection. Therefore there is a need of higher-resolution advanced regional models alongwith accurate inclusion of cloud microphysical processes in numerical models for reliable weather forecasting and climate prediction, which requires a reliable specification of the initial atmospheric conditions.

Therefore to have accurate estimate of the spatial distributions and temporal variations of the cloud microphysics parameters, we need to develop a variational data assimilation system for cloud microphysics based on satellite observations. This system has potential to assimilate passive microwave remote sensing observations into the cloud microphysics scheme and whose output can be used to improve initial conditions of the atmospheric model. By the downscaling manner, initial conditions for local grid scale model can be obtained for having reliable precipitation forecast over the ungauged basins. Furthermore the application and validation of such a cloud microphysics data assimilation system over the land can be further performed by coupling it with a land data assimilation system （LDAS）. Adequate estimates of land surface variables can be obtained with this system by providing detailed spatial patterns of precipitation and quantitative precipitation data to the soil moisture data assimilation.

The selection of variational data assimilations scheme in the present research is also due to its characteristics of offering the possibility of achieving the optimal performance of Kalman filters with the computational efficiency of sub optimal methods. Also variational methods do not explicitly evaluate the large error covariance matrices which are propagated by Kalman filters. Instead, variational algorithms simultaneously process all of the data within a given time period （or "assimilation window"） and the adjustment process is simultaneous. The adjusted states at all times are influenced by all of the observations over the assimilation window.

It is believe that the initialization and forecast of the state variables of the variational cloud microphysics data assimilation system for warm and cold cloud processes might be improved by the assimilation of remotely sensed data. Therefore having such intention in mind, an efficient algorithm of one dimensional variational （1DVAR） Cloud Microphysics Data Assimilation System （CMDAS） for warm cloud processes and 1DVAR Ice Cloud Microphysics Data Assimilation System （IMDAS） for cold cloud rain processes have been developed. Both assimilation systems have been used to solve the initialization of the state variables using available in-situ and satellite observations over the ocean by considering the integrated cloud liquid water （ICLWC） and integrated water vapor （IWV） content as assimilation parameters.

The general framework of CMDAS and IMDAS includes the Kessler warm-rain cloud microphysics scheme and Lin ice microphysics scheme as the model operator respectively. The common framework of the two data assimilation systems consists of a 4-stream fast microwave radiative transfer model in the atmosphere and a heuristic minimization approach called Shuffled Complex Evolution （SCE）, which is capable of minimizing the cost function without using an adjoint model （gradient vector）. SCE can result in a robust global searching scheme that conducts an efficient search on the feasible space. Due to high computational cost problem, the environmental forcing effect is neglected at the moment in both systems.

Both assimilation systems are applied to the microwave radiometer data set obtained by the international cooperative observation experiment, "Wakasa Bay Experiment 2003", in Japan. The potential of the CMDAS and IMDAS are investigated to modify the cloud properties by considering the assimilation parameters of ICLWC and IWV and to introduce the heterogeneity into the initial state of the atmosphere. From these two assimilation systems, it is also possible to perform the sensitivity analysis under different weather events to see which has more potential to produce the better initial conditions and how much these improved initial conditions will have positive impact on the reliable estimation of precipitation by the Advanced Regional Prediction System （ARPS） model.

The simulation results of CMDAS and IMDAS with the observed AMSR-E Tb 89.0H GHz and 23.0H GHz values identify clearly their effects on the cloud distribution mapping and show the comparable structure of cloud system with Moderate-resolution Imaging Spectroradiometer （MODIS） image for cloud top. Both systems have improved the performance of cloud microphysics schemes significantly by the intrusion of heterogeneity into the external Global Reanalysis （GANAL） data, which resultantly improved atmospheric initial conditions. It also concluded that IMDAS comparatively work well compared to CMDAS due to capability of considering the all types of precipitation hydrometeors.

Furthermore, total precipitation rate derived by ARPS with improved initial conditions provided by CMDAS and IMDAS reveal a good agreement of the spatial distribution of the precipitation rate with precipitation rate derived by 3-D Doppler radar reflectivity data. But at a few places, it has been observed that ARPS has over-predicted the precipitation along with some spatial displacement which needs to explore in details.

短期および中期の数値気象予報精度の向上に必要な水蒸気，雲の初期値推定精度向上を目的として，本研究は衛星観測に用いられる放射伝達モデルと数値気象予測に用いられる雲微物理モデルを組み合わせたデータ同化手法の開発に取り組んだ．

本研究では，暖かな雲のための効率的な1次元変分法による雲微物理同化システム（CMDAS）と冷たい雲のための1次元変分法による氷晶プロセスを含んだ雲微物理過程の同化システム（IMDAS）を開発している．これらのシステムでは，積算雲水量と積算水蒸気量を同化パラメータとして考慮し，初期値とCMDASおよびIMDASの両者に関する状態変数の問題を解決している．CMDASおよびIMDASではそれぞれ，ケスラーによる暖かなスキームとリンによる雲氷微物理過程モデルとをモデル操作子（Model Operator）としている．

これらふたつのデータ同化システムは共に大気中での4ストリームマイクロ波放射伝達モデルとShuffled Complex Evolution（SCE）という発見的な誤差最小化手法を組み込んでいる．計算量が大きくなることによる負荷を下げる目的で，雲微物理過程以外の外部の力学的な環境についての同化は現段階では含まれていない．さらに本研究では，これらのデータ同化システムと衛星搭載受動型マイクロ波放射計データを組み合わせて，全球大循環モデルの出力を領域モデルあるいはメソスケールモデルにダウンスケーリングに適用し，高い精度の水蒸気や雲水量の初期値を得ることを可能にしている．

CMDASおよびIMDASの二つのシステムは，冬の日本海沿岸域に適用され，人工衛星Aqua搭載のマイクロ波放射計AMSR-Eによって観測された89GHz水平偏波と23.0GHz水平偏波の輝度温度を用いて，雲の空間分布の算定可能性を示した．これは同じ衛星に搭載されている可視赤外センサであるMODISによる雲頂画像による雲の構造とも良い対応がみられた．ふたつのシステムは，全体の場を記述している全球解析データに不均一性を入れ込むことによって，雲微物理の性能を大いに向上させ，結果として大気に関する初期値を改善した．またIMDASは全てのタイプの大気水象を考慮することが可能なため，CMDASと比較して良い結果を得るということも結論付けられた．

以上，本研究は，雲水量と水蒸気量の初期値算定精度の向上を通して数値気象予報精度の著しい向上をもたらし，豪雨災害の軽減に資する．これらの成果は科学的側面だけでなく，社会に貢献するところが大きく，社会的有用性に富む独創的な研究成果と評価できる．よって本論文は博士（工学）の学位請求論文として合格と認められる．