重現跨日關係和日夜循環的空間–時間降尺度方法

A novel spatiotemporal statistical downscaling method preserving inter-day correlation and diurnal cycle

張明瑞(Ming-Jui Chang)

指導教授 ： 林國峰

國立臺灣大學/工學院/土木工程學研究所/博士(2017年)

https://doi.org/10.6342/NTU201703683

摘要

資料的空間及時間解析度是模擬氣候變遷對中、小流域水文衝擊評估的關鍵。然而，目前的研究沒有考慮到水文循環中的跨日關係(Inter-daily connection)及日夜循環(Diurnal cycle)。因此，本研究提出一個空間–時間統計降尺度方法，可以重現跨日關係、日夜循環，及日尺度和時尺度的統計特性。
本研究使用提出的方法建立大尺度的NCEP/NCAR(National centers for environmental prediction/National center for atmospheric research ) 再分析資料和GCM(General circulation model )輸出值與區域尺度的雨量和溫度關係。方法分成兩階段，分別為空間降尺度和時間降尺度階段。空間降尺度階段使用K最近鄰居法(k-nearest neighbor method, KNN)，建立大尺度大氣因子及雨量關係，獲得測站尺度日雨量和日溫度序列。接著，時間降尺度階段使用KNN結合基因演算法(Genetic algorithm, GA)的GAKNN模式，並同時考慮跨日關係及日夜循環，由測站尺度日雨量和溫度序列，降尺度獲得測站尺度時雨量和時溫度序列。接著，直接使用日和時尺度的雨量和溫度序列當作水文模式的輸入項，本研究使用HBV模式 (Hydrologiska byråns vattenbalansavdelning) 建立水文模式。最後使用CGCM3.1 (Third generation coupled global climate model) and BCCR-BCM2.0 (Bergen climate model version 2)模式下的A2、A1B及B1三種情境之中期(2046-2065)和長期(2081-2100)資料，映射未來雨量、溫度和流量的改變。
本研究提出的方法應用於台灣的石門水庫集水區，結果證明此方法可以準確地重現日和小時尺度的雨量、溫度和流量，並保有觀測資料的統計特性。總而言之，本研究提出的空間–時間統計降尺度方法可有效的得到小時雨量和溫度，並可直接使用於水文建模中。瞭解未來雨量、溫度和流量的改變，可協助計畫與管理水資源系統。

關鍵字

氣候變遷 ； 降尺度 ； 雨量 ； 溫度 ； 流量 ； K最近鄰居法

並列摘要

Finer spatiotemporal resolution information is essential for assessing hydrological impacts of climate change on medium and small basins. Previous researches have shown the potential of the downscaling methods for assessing hydrological impacts. However, these methods pay less attention to the inter-day correlation and diurnal cycle, which can strongly influence the hydrological cycle. To preserve this correlation, this study presents a spatiotemporal downscaling method that is capable of reproducing the inter-day correlation, the diurnal cycle, and the statistics on daily and hourly scales, as well as directly use the downscaled rainfall and temperature to simulate daily and hourly streamflow.
The large-scale datasets, which are obtained from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis dataset (NNR) and general circulation model (GCM) outputs, and local rainfall and temperature are analyzed. The proposed method consists of three steps: spatial downscaling, temporal downscaling, and hydrological modelling. The spatial downscaling is first applied to obtain the relationship between large-scale datasets and daily rainfall and temperature at a site scale using a k-nearest neighbor method (KNN). Then, hourly downscalings of daily rainfall and temperature are conducted in the second step using a genetic algorithm-based KNN (GAKNN) with the inter-day correlation and the diurnal cycle. For hydrological modelling, the rainfall and temperature series on daily and hourly scales from the proposed method are directly used as input to hydrological models, which is Hydrologiska Byråns Vattenbalansavdelning model (HBV). Furthermore, changes in statistics of rainfall, temperature and streamflow for the periods 2046–2065 and 2081–2100 under the A2, A1B, and B1 scenarios of the third version of the Canadian Centre for Climate Modelling and Analysis Coupled Global Climate Model (CCCma-CGCM3.1) and Bjerknes Centre for Climate Research Bergen Climate Model version 2 (BCCR-BCM2.0) are analyzed.
An application of the proposed method to the Shihmen Reservoir basin (Taiwan) has shown that it can accurately reproduce local rainfall, temperature, streamflow and their statistics on daily and hourly scales. Overall, the results demonstrated that the proposed spatiotemporal method is a powerful tool for downscaling hourly rainfall and temperature data from a large-scale dataset. The downscaled rainfall and temperature series can be directly used for hydrological modelling. The understanding of future changes of rainfall, temperature and streamflow characteristics through this thesis is also expected to assist the planning and management of water resources systems.

並列關鍵字

Climate change ； Rainfall ； Temperature ； Discharge ； k-nearest neighbor method ； Genetic algorithm

參考文獻

Bergström S (1995) The HBV model. In: Singh, V.P. (Ed.). Computer Models of Watershed Hydrology. Water Res Publications 443–476.

Anandhi A, Srinivas VV, Kumar DN, Nanjundiah RS (2009) Role of predictors in downscaling surface temperature to river basin in India for IPCC SRES scenarios using support vector machine. Int J Climatol 29:583–603

Burger G, Murdock TQ, Werner AT, Sobie SR, Cannon AJ (2012) Downscaling extremes-an intercomparison of multiple statistical methods for present climate. J Clim 25:4366–4388

Castro CL, Pielke RA, Leoncini G (2005) Dynamical downscaling: Assessment of value retained and added using the Regional Atmospheric Modeling System (RAMS). J Geophys Res 110(D5):D05108

Chen H, Xiang T, Zhou X, Xu CY (2012) Impacts of climate change on the Qingjiang Watershed’s runoff change trend in China. Stoch Environ Res Risk Assess 26:847–858