Extreme weather events have been the leading cause of disasters and damage all over the world. Recent events have led to mass displacement, loss of income, and hampered access to clean water and health to many. The primary ingredient to these disasters especially floods is rainfall which over the years, despite advances in modeling, computing power and use of new data and technologies, has proven to be difficult to predict. Numerical weather prediction (NWP) and climate models using global forecasts and emission scenarios as initial and boundary conditions have provided short to midterm forecasts and climate projections in many parts of the world. The Lake Victoria Basin (LVB) in East Africa supports over three million livelihoods and hosts the world record for elaboration of vertebrate species diversity, species extinctions, and exotic species invasions. The second largest freshwater lake in the world is a valuable resource to five East African countries as a source of water and means of transport. However, with a Mesoscale climate regime driven by land and lake dynamics, extreme Mesoscale events have been prevalent and the region has been on the receiving end during anomalously wet years in the region. This has resulted in loss of lives, displacements, and food insecurity. As an essential aspect of early warning there is a need to strengthen adaptation through improved prediction of rainfall and floods.

It is now more evident than ever that climate change will have adverse impacts on the global population in coming years. Since the release of Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report: Climate Change 2007 (AR4) in 2007, it has been shown that impacts of climate change are already being felt, with increases in sea level rise, retreating glaciers and more frequent weather extremes. In the LVB, the effects of climate change are increasingly being recognized as a significant contributor to poverty, by its linkage to agriculture, food security and water resources. One aspect that is investigated widely with respect to climate change is the impacts it will have on extreme weather events. There is a general consensus that changes in frequency and intensity of extreme weather and climate events will have adverse effects on both humanity and nature.

Adaptation to climate change needs an understanding of climate change impacts at local scales. There is a need to connect the global scale projections with impacts it may have on people through downscaling and other means. The LVB is a regional basin with multiple land uses and water resource needs. In addition, it is one of the basins that have been heavily affected by extreme weather, especially storm-induced floods. Of particular importance are the likely impacts of climate change in frequency and intensity of extreme events. To tackle this aspect, this study adopted an integrated regional, mesoscale and basin scale approach to climate change assessment. The study investigated the projected changes in mean climate over East Africa, diagnosed the signals of climate change in the atmosphere, and transferred this understanding to mesoscale and basin scale. Changes in rainfall were analyzed and similar to the IPCC AR4 report; the selected three General Circulation Models (GCMs) project a wetter East Africa with intermittent dry periods in June-August.

The highest projected changes are in the October-December season whose variability is closely linked to inter-annual variability of sea surface temperature anomalies in the Indian Ocean. Weakening of the Walker circulation, anomalous monsoons, and moisture advection into East Africa in the two seasons are some of the phenomena associated with increased rainfall. Investigation on the Nino3 sea surface temperature (SST) anomaly index reveals an increasing trend with the selected target period of study projected to experience increased frequency of El-Nino/Nina events. GCMs projections for the past (1981-2000) and future (2045-2065) were bias corrected and projected changes at a river basin scale investigated. In the Nyando river basin, a perennial flood basin, the selected three GCMs project a wetter future with extreme events occurring almost two times the magnitude of the past climate. The trend by the three GCMs is the same; increased flood probability in the future and about 10-20% increase in mean discharge.

In addition to climate change assessment; the study also focused on short-term weather forecasting as a step towards adapting to a changing climate. This involved dynamic downscaling of global weather forecasts to high resolution with a special focus on extreme events. The skill of operational weather forecasts has increased over the last five decades. This improvement has taken place gradually and relatively steadily; driven by advances in scientific understanding of physical processes and rapidly increasing computational resource developments. The last decade has seen remarkable progress in exploring satellite observations especially microwave measurements and the launching of new platforms. Numerical weather prediction and assimilation satellite data in the microwave band has been shown to improve predictability. In this tropical inland lake basin with a sizeable water body and locally controlled Mesoscale weather system; an improved prediction would be greatly useful for flood early warning and water resources management.

By utilizing complex model dynamics, the system was able to reproduce the Mesoscale dynamics well, simulated the land/lake breeze and diurnal pattern but was inadequate in some aspects. The quantitative prediction of rainfall was inaccurate with overestimation and misplacement but with reasonable occurrence. This was a remarkable improvement compared to the coarse resolution of 100km which had not factored in local scale dynamics induced by the land-lake interaction and low-high altitude contrast in this region. To address these shortcomings this thesis investigated the value added by assimilating Advanced Microwave Scanning Radiometer (AMSR-E) brightness temperature during the event. By assimilating 23GHz (sensitive to water) and 89GHz (sensitive to cloud) frequency brightness temperature; the predictability of an extreme rain weather event was investigated. The assimilation of AMSR-E brightness temperature through a Cloud Microphysics Data Assimilation (CMDAS) into the weather prediction model considerably improved the spatial distribution of this event.

Data assimilation improved the simulation of the rain event by more than 50% with the spatial location of maxima at 00hrs and 03hrs after assimilation matching Tropical Rainfall Measuring Mission (TRMM). In addition, overestimation of rainfall predicted by the NWP (without data assimilation) was reduced considerably. The assimilation of brightness temperature is implemented through a Radiative transfer model (RTM). The RTM involves the calculation of the surface emissions, scattering and absorption of passive microwave energy on soil, land surface and the atmosphere. A key component of this determination is the computation of surface emissivity which is derived from surface temperature. Results in this region showed that the NWP had a tendency to overestimate surface temperature which in turn affected the assimilated brightness temperature. This had the negative impact of reduced water vapor in the system and the model could not keep the induced model state for long. Sensitivity experiments conducted by subtracting 2-10K from the surface temperature at the assimilation time showed that changes in the amount of water vapor corresponded to the magnitude of temperature change. This highlighted the uncertainty in water vapor assimilation due to model errors in estimation of surface temperature. This study identified this challenge and proposes a major validation / improvement of the RTM.

The downscaled/predicted quantitative rainfall was then assessed for its suitability as an input to a decision making process with respect to floods. Hydrological modeling using the QPF from the NWP showed promising flood simulation results even though there was a tendency to overestimate flow. Nonetheless, it shows a potential for future application as an input to an early warning system.

ナイル川最上流に位置するビクトリア湖流域は、洪水や気象災害に対して脆弱であり、人的被害に加え、農業、漁業や湖上交通に大きな被害が生じている。加えて気候の変化に伴う気候の変動性の増加と、異常気象の頻度の増加によって、同地域の災害リスクの増大が懸念されている。そこで、気候の変化に伴う地域スケールの気候レジームの変化を明らかにして、ビクトリア湖周辺の異常気象のハザードの変化と、ビクトリア湖に注ぐ流域スケールでの洪水の変化を定量的に把握する必要が高まっている。また、気候の変化に対する適応策の提示も重要な課題である。

本研究は、ビクトリア湖に注ぐ一つの河川流域のスケール、ビクトリア湖流域スケール、地域スケールで、気候の変化の影響を明らかにするとともに、気候変動への適応策の一つとして、ビクトリア湖流域スケールの豪雨予測の精度向上のために、衛星データ同化手法の適用可能性を検討している。

河川流域スケールでは、第3次結合モデル比較実験(CMIP3)に参加した24の気候変動予測モデル(GCM)から、現在気候の降水時空間分布特性、気圧場、気温場、風場、海面水温分布の再現性の良い3つのGCMを選択し、地上降水量データを用いて、日降水量のバイアスを補正し、過去と現在の日降水量時系列を作成した。また、ビクトリア湖北東岸に位置するニャンド川流域を対象に、エネルギー・水循環分布型モデル(WEB-DHM)を構築し、バイアス補正した降水量と各GCMの他の気象要素を入力としてWEB-DHMで河川流量を計算した。その結果、流域各地点で、選択された3つのGCMの入力を用いた場合、いずれも気候変化に伴う洪水流量の顕著な増加が示された。

ビクトリア湖流域スケールについては、ビクトリア湖流域周辺で、豪雨傾向を示す指標として、日降水量40mm以上の日数(ECAR40MM)が20-50%、単純降雨強度指数(ECASDII)が4-10%増加し、豪雨の増加傾向は当該スケールで共通の現象であることが分かった。

そこで、インド洋、アフリカ大陸を含む地域スケールで解析したところ、気候の変化に伴い、赤道付近でコンゴを上昇域、アフリカ大陸東岸を下降域とするウォーカー循環が、10-12月の雨季において弱化する傾向が明らかとなり、アフリカ大陸東岸での下降の弱化に伴い、降水強度が増加している様子が明らかとなった。

一方、気候変動に伴う豪雨災害の適応策として、観測が十分ではないビクトリア湖流域の豪雨予測精度の向上を目的として、衛星搭載マイクロ波放射計を用いた大気-陸面結合データ同化手法(CALDAS)の適用を試みている。適用の結果、初期には衛星観測で得られる降水量分布と適合する予測雨量を得ることができたが、予測計算3時間で豪雨を形成する擾乱が消滅することが分かった。

その原因を明らかにするために、CALDASで用いられる同化プロセス、収束発散条件、陸域からの射出等、各条件に関する感度実験を行った。その結果、現システムでは、降水粒子の同化が強く効きすぎて潜熱放出が妨げられることが、擾乱を弱化する要因になっていることを明らかとなった。また、陸面データ同化で、地表面温度が高めに推定されていることが、大気水蒸気の過小評価につながり、擾乱を維持するための対象領域の水蒸気量の著しい低下が、降水システムの早期の消滅につながっていることを明らかにした。そのうえで、この2点を改善したシステムで予測計算を行ったところ、長時間にわたって、実際の雨域で降雨を予測する割合が増加し、無降雨域で降雨を予測する割合が低下する結果を得て、豪雨の予測精度向上に貢献する結果を得た。

上記のように、本研究では各研究項目において優れた科学的知見、予測精度向上の優れた実績を挙げるとともに、流域スケールから地域スケール、解析研究から予測研究まで、気候変化に伴う豪雨災害の増加のメカニズムに関する科学的知見を得ている。また、適応策に関する具体的な貢献も果たしており、その包括的な研究の枠組みそのものを確立したところに意義がある。

以上、本研究は、気候の変化に伴うビクトリア湖周辺の気象ハザードの変化の理解とその定量的な評価に優れた科学的知見を提供しているとともに、適応策としての豪雨予測の精度向上に大きく貢献している。この研究成果は、現在の洪水管理、および気候変動を考慮した将来の洪水管理計画に資するところが大きく、科学的、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。