Abstract
摘要:總計畫:
本整合型計畫是「都市積淹水即時預報」的先導研究。根據預報系統的需求,本整合型計劃分為七項子題,邀請這方面的專家共同參予完成系統雛型。主要工作包括:
1. 精密降水估計雷達觀測策略研擬:利用中央大學的雙偏極化雷達,研擬掃描策略,以期獲得高時空解析度和精確度的降雨觀測。
2. 都市地區及其外圍集水區降雨逕流模式之建立:根據土地利用和下水道管線狀況,選擇監測地點,研發適用的逕流模式,和估計逕流參數方法。
3. 應用微機電感測器與Zigbee無線監測網路於即時水位監測之研發:選點進行下水道水位密集監測,建立通訊與資料系統,傳輸即時監測資料,資料儲存與提供給模式應用的服務介面。
4. 渠道管網演算與資料同化模式研發:研發可跨明渠/滿管流演算、整合下水道和市街流渠管網演算的模式;並研發可同化水位觀測資料,改善預報初始條件的技術。
5. 都會區溢淹演算即時預報模式之研發:除了二維淹水模式外,亦採用類神經網路研發推估淹水潛勢;另配合SWMM模式進行案例模擬分析。
6. 都市地貌與積淹水地理資訊4D展示模組研發:採用地理資訊系統化的三維虛擬實境展示,提供直觀的淹水範圍演算成果展示,並以開放原始碼開發圖形展示程式。
7. 分散式都市淹水模擬與支援決策系統框架之研發:研發能夠讓各子模式分散演算的系統,包括資料交換標準介面,平行化或分散化以提升模式即時演算效率,並將模式模組化、執行程序腳本化的演算流程設計。
本研究以台北市玉成抽水站集水區作為「自然實驗室」,利用中央大學的雙偏極化都卜勒雷達,選擇東側山區集水區和易淹水子集水區,設置水位與流量觀測儀器,透過實地密集觀測,了解都市降雨-逕流歷線轉換關係,以及可能發生積淹水的過程。同時,研究發展數學模式,逐步達成「都市積淹水即時預報」必須具備的各項科學技術、模擬程式、演算系統和展示工具等。
子計畫:
都市區下水道管線多,地文條件複雜。暴雨時期地表水與下水道內之水流交換頻繁,地表沒有清楚的集水區邊界(淹水時可能跨區漫淹),因此在模擬都市積淹水時,必須同時考量垂直方向(地表水與下水道),及水平方向(都市區塊間)的交換機制,才能得到合理的計算結果。
下水道幹管內流況複雜,可能存在超臨界流與亞臨界流混合之跨臨界流況,且同一幹管內可能同時存在明渠流與滿管流,因此需要一套穩定性高之數值方法,才能確保「都市積淹水即時預報」系統在作業時期持續運作。本研究擬採用NewC法建立市街流與下水道水理數值模式,並結合Preissmann slot法克服滿管流計算問題,除了可有效解決超臨界流與亞臨界流混合流況外,也能夠完成渠管內多變流況的模擬。
由於降雨觀測和降雨-逕流歷線演算均存在不確性,其影響都將反映在下水道水理模式模擬結果之中。加上都市集水區面積小、渠道管網容量小的特性,模擬計算結果與觀測資料的誤差往往不可忽略。本計畫擬採用資料同化技術,結合伴隨狀態法(adjoint state method,簡稱ASM)與「擬牛頓法」,利用下水道水位觀測資料修正模式狀態,避免因為不確定因素的影響使誤差持續累積。
本研究將以台北市玉成抽水站之集水範圍進行模式測試與結果分析,並評估模式平行化計算之可行性。最後對於建置「都市積淹水即時預報系統」提供詳實之建議。
Abstract: 總計畫:
This is an integrated pilot study for the development of the Urban Inundation-Potential Nowcasting System. It is subdivided into 7 topics. They are:
1. Development of radar scanning strategies for accurate estimate of quantitative precipitation. Utilize the dual-polarization radar of National Central University to test and to enhance the performance of the strategies.
2. Runoff model development: Choose small urban catchment to do special observation and hence develop the rainfall-runoff transfer function of the catchments.
3. Develop pressure sensor, velocemetry and transducer using MEMS:Install sensors, communication system and data server for the special observation of stage hydrograph in storm sewer.
4. Development of pipe/open channel simulation model that is able to do trans-critical flow as well as trans-pressured/free surface flow. A scheme to assimilate stage data will be implemented using the Adjoint State Method.
5. Development of 2-D and ANN overland flow model. The SWMM will be utilized for case studies.
6. Construct the urban morphology and 4D GIS display using Open Source code.
7. Design a distributed computation system.
Using the catchment of Yu-Cheng pumping station as the "natural laboratory", this study will select a few sites and conduct rainfall-runoff special observations. Stage hydrograph data are to be collected for rainfall/runoff relationship calibration or estimation. The mathematical model, computation environment and graphical display needed to establish an Urban Inundation-Potential Nowcasting System prototype are developed.
子計畫:
In urban area, 2-way vertical flow between surface runoff and storm sewer is frequently. Horizontal exchange among urban catchments may also occur during inundation. A storm sewer numerical model must have the capability to simulate them to be able to correctly reproduce inundation events.
In storm sewer system, flow regimes include supercritical flow, subcritical flow, transcritical flow, free surface flow and pressurized flow. A robust and stable numerical model, able to simulate all the combinations, is mandatory for an operational `urban inundation forecast system`. In this study, we intend to develop the 1-D street flow and sewer pipe flow model by NewC scheme. The scheme is proven stable in trans-critical flow computation without the need of changing governing equation, which is a major improvement over the currently available schemes in engineering applications. The `Preissmann slot` technique will be implemented to facilitate cross open channel and pressure pipe flow computation.
For the catchment areas and the sewer storage capacities are small in urban area, rainfall observation and runoff model uncertainty may make simulated water stage differs from the measured. In order to avoid error accumulation, a data-assimilation scheme using the adjoint state method and quasi-Newton solution technique are applied. Lateral discharge is considered uncertain and is adjusted by a ratio such that the differences between the observed and simulated stage are minimized.
The catchment area of Yu-Cheng pumping station in Taipei City is chosen be the case study area. Feasibility of parallel computation using PC cluster will be studied. The results will be used to provide sensible suggestions for establishing a `real-time urban inundation forecast system`.
Keyword(s)
NewC法
市街流與下水道水理數值模式
Preissmann slot
資料同化
伴隨狀態法
都市積淹水即時預報系統
NewC scheme
urban street flow and storm sewer pipe flow model
Preissmann slot
data assimilation
adjoint state method
urban inundation forecast system