Novel Early flood Warning and Risk Assessment System 中小尺度集合洪水预报系统 Novel Early flood Warning and Risk Assessment System – a next generation expert system software for flood warning and risk analysis 中小尺度集合洪水预报系统现由河海大学和伦敦国王学院在淮河上游共同试验开发, 英国合乐集团独家受聘进行市场调研以及应用技术推广.

Outline Introduction of NEWS in Chinese TIGGE Global weather forecast centres Flood forecasting procedure Test application 1: Upper Severn catchment (with results of probabilistic forecasts) Test application 2: Upper Huai catchment (ongoing) NEWS potential market value and endusers NEWS in the public domain (website) 2

Summary 　　NEWS中小尺度集合洪水预报系统 背景概述 　　集合预报系统从其实质上讲又可称之为概率预报系统，其最终目的是提供大气变量的完全概率预报。近几年集合预报技术经历了不断的发展完善，从以前仅考虑初值的不确定性发展为同时考虑模式的不确定性，进而发展到多模式和多分析集合预报技术。TIGGE集合预报是世界气象组织的“观测系统研究和预报实验”项目的重要组成部分，在全球范围组织各气象业务中心的集合预报开发与合作，并计划发展成为未来的“全球交互式预报系统”。该技术在世界范围被认同，并逐渐成为天气预报的主流发展趋势。越来越多的国家投入准业务运行和业务运行，有代表意义的是欧盟联合研究中心(EU JRC)正在欧洲范围内大力推广应用的EFAS ‘欧盟洪水预警系统’。EFAS正是建立在这样一个集合预报系统基础上结合其现有的分布式水文模型开展其准业务运行，为欧盟各国提供3-10天内的中尺度洪水预报。 　　在EFAS技术雏形的基础之上，NEWS‘中小尺度集合洪水预报系统’对其功能进行扩充和完善：(1) 引入尺度转化方法，提高精度，加强了该项技术用于到小尺度流域的可行性；(2) 完整的气象，水文和水力模型耦合，一套系统可以连续运行提供流域范围内降雨量，洪水过程线和洪水淹没区域及水深预报；(3) 对统计预报的后续处理和预报矫正；(4) 预报洪水统计风险区域图(flood risk map)。NEWS 已经在英国中部塞文河 (Severn river) 流域上游子流域 (面积大约4062Km2)进行了试验洪水预报(水文预报为1公里网格精度，水力预报为50米网格精度)。与传统预报技术相比，其预报准确性和命中率显著提高，虚警率降低。从时间上讲，洪水过程预报可以提前至3到10天，从而为疏散居民和转移物资提供了宝贵的时间。 　　NEWS中小尺度集合洪水预报系统现由河海大学和伦敦国王学院在淮河上游共同试验开发，英国合乐集团独家受聘进行市场调研以及应用技术推广。NEWS项目同时得到了安徽省水文局和欧洲中期天气预报中心(ECMWF) 的大力支持。ECMWF的Florian Pappenberger博士同时参与该项目，为NEWS的气象数据采集和处理提供技术支持。 　　尽管集合概率预报结果与传统方法比较有显著提高，但其发布还需要做大量的普及推广和培训工作。目前伦敦国王学院在塞文河流域的试点项目是世界范围内第一次将TIGGE多模式集合气象预报与分布式水文水力模型耦合应用，其推广前景还需要在不同流域范围内，不同气候条件的区域进行试验，以便确定其应用技术的普遍性。中国在面向TIGGE的集合预报关键应用技术已经做了大量工作，为水文水力领域中应用该技术做好了准备工作。这可以促使NEWS产品在实际洪水预报中尽早发挥作用。 Will you like to have an English summary here? 3

Global weather forecast centres 全球集合气象预报中心 Centre Code Country/Region Forecast Center Centre Abbr. Ensemble Members ammc Australia Bureau of Meteorology BOM 32+1 babj Canada Meteorological Service of Canada MSC 14+1 cwao China China Meteorological Administration CMC 20+1 ecmf Europe European Centre for Medium-Range Weather Forecasts ECMWF 50+1 egrr UK UK Met Office UKMO 23+1 kwbc USA National Centre for Environmental Predictions NCEP rjtd Japan Japan Meteorological Agency JMA sbsj Brazil Brazilian Centro de Previsao de Tempo e Estudos Climatico CPTEC All these ensemble prediction systems are based on n time integrations of a numerical weather prediction model, with one (the control forecast) starting from a ‘central’ analysis, usually the unperturbed analysis generated by a data-assimilation procedure, and the others (the perturbed forecasts) starting from perturbed initial conditions defined to simulate the effect of initial uncertainties. 全球集合气象预报中心: 每个气象中心在每天12点和0点(GMT时间,比北京时间晚8个小时)发布10天的集成气象预报(15到51套,数目不等). 我们从欧盟气象局提取所有预报中心10天内每6小时的中尺度集成气象预报, 之后转换为小尺度, 输入到水文模型. 左图: 气象预报中心开始输出集合气象预报 并加入TIGGE数据库的时间 n 代表所有可能的集合气象预报组合(供用户单位在试验阶段针对不同流域筛选最可行的气象预报中心) 4

Flood forecasting procedure in testing mode Numerical Weather Forecast (NWP) models 气象预报数学模型 Hydrological/Hydraulics model calibration 水文和水力模型率定 Initial status, forecast starting time 水文模型用观测气象值模拟初始状态 Ensemble forecasts (replace conventional deterministic single forecast) 集成预报替代单一预报 Model simulation using calibrated parameter set 水文模型用预报气象值 模拟未来10天内河道流量 Downscale forecasts to the catchment model spatial unit 中尺度集成预报重新构造成小尺度(100公里1公里或更小尺度) Schematic flowchart of the test procedure (Slide 5) The forecast procedure is tested in an iterative way such that the disaggregated precipitation input maps could display satisfactory spatial patterns and produce comparable discharge level as that driven by rain-gauge observation. Once the ensemble of forecasted discharges is obtained, a hydraulic model was coupled to the end of the cascade to forecast flood water depth and inundation extent. The time used in this paper is the UTC time. NEWS key model components: Ensemble weather forecasts (obtained from a medium-range global ensemble system – TIGGE - to provide probabilistic forecasts for 6 hours up to 10 days and for the whole globe) Rainfall-runoff model (to produce hydrograph using the ensemble weather forecasts) Hydraulics model (to simulate flood inundation using predicted ensemble hydrograph as input) NEWS immediate output: Downscaled ensemble weather forecasts (in the second project phase, a long-term ensemble weather prediction within a timeframe of 30 to 50 years) Predicted ensemble hydrograph Predicted ensemble flood inundation map [a flood risk map with probabilistic diagnosis - (1) inundation area; (2) likelihood of being inundated; (3) inundation depth] Models selected TIGGE forecast data are disaggregated from approximately 100×100 km to 1×1 km and spatially corrected before being fed into a rainfall-runoff model. The LISFLOOD-RR model (LFRR) has been used within European Flood Alert System (EFAS) and thus selected for the Upper Severn. LFRR is a fully distributed and semi-physically based rainfall-runoff model. The LISFLOOD-FP model (LFFP) has been selected for simulating flood inundation both in the channel and on the flood plain. LFFP uses the Saint-Venant equations of open channel flow within the channel, and a 2D raster model based on the Manning equation for the floodplain. For testing/quality control purpose, not applicable for operational purposes 上图: 集合预报过程示意(试验阶段) 5

Severn Basin - DEM Already tested area: Severn river basin located in the Midlands region of England The River Severn collects water from the Welsh mountainous area and its tributaries after passing through many steep sided valleys in Wales. It then flows eastwards and contributes significantly high levels of discharge to Montford (see figure on Slide 6). Between Montford and Buildwas, the river channel meanders through a low-lying flood plain. Elevation on both sides of Buildwas rises above the flood plain forcing the river to cut through and form a gorge behind Buildwas. The river diverts its flow southwards just before passing through Bridgnorth. The drainage area up until Bridgnorth is referred to as the Upper Severn catchment. It is approximately 4062 km2 with urban or sealed areas, forest and agricultural land accounting for 3%, 7.1% and 48.5% of the area respectively. Loosely packed peat soil dominates the Upper Severn catchment. Flow levels are generally high in autumn and winter and low in summer. The Shropshire town of Shrewsbury is situated within the incised meander loop of this reach and vulnerable to floods. The simulation of flood inundation is therefore carried out on the river section between Montford Bridge (UK Ordnance Survey grid reference SJ412144) and Buildwas (SJ644044). 6

Severn Basin – meteo domain Metrological input domain is overlaid with the Severn catchment boundary to give you an impression of the spatial scale of the TIGGE ensemble weather forecast data as well as that of the Regional Climate Model (RCM) [RCM is considered for the second stage of the project when we will like to have NEWS produce both short/medium-term forecast and long-term prediction, the latter is of special interest for insurance]. As you can see from the figure (Slide 7), the meteorological domain is in rather coarse resolution. Approximately 6 points cover the Upper Severn catchment (4062km2). Downscaling or disaggregation is therefore required to bridge the spatial gap between meteorological input and driving inputs for hydrological and hydraulics models. 7

Severn Basin – ensemble event hydrograph Ensemble discharge forecasts (Qforecast) in comparison with observed discharge (top); ensemble precipitation forecasts issued on 18/01/08 by ECMWF (bottom) Lines with diamond, squares & stars: Q5, Q50 and Q95 Horisontal dashed line: EA warning level The light grey lines: the ensemble forecasts Jan 2008 8

Severn Basin – ensemble area of inundation Ensemble area of inundation by using ensemble forecasts from the ECWMF centre 9

Severn Basin – probabilistic inundation estimate 2008年1月塞文河上游子流域 洪水淹没区域概率预报 由于此次洪水强度不高, 洪泛区域较小 100% inundated  not inundated 完全淹没  安全区域 Probabilistic area of inundation by using ensemble forecasts from the ECWMF centre (a blow-up of area of figure (c) on Slide 6 between Montford and Buildwas) The inundated area isn't extensive in this case due to the low magnitude of the flood event took place in Jan. 2008. 10

Severn Basin – a flood case study Percentage of ensemble Q above flooding threshold at Montford The horisontal bars represent the 7 centres in the same order as in table (Slide 4, not the sbsj centre) plus the Grand Ensemble (all 7 centres together). The use of more than one forecast centre gives the forecaster more information at an earlier stage and potentially increases the ability to detect a flood. The flooding signal is visible in some centres as early as 10 days ahead, although only on the 25%-level. The signal becomes persistent approximately 4 days ahead, which ascertains floods are going to take place. Jan 2008 11

NEWS project selected area Upper Huai Catchment (30672 km2) Upstream of Wangjia Ba reservoir (王家坝水库上游) Huai River Basin (淮河流域) Upper Huai catchment has been selected to replace the originally proposed Luohe catchment due to the abundant data of the Upper Huai catchment. 12

Huai Basin 13

Huai Basin 14

Huai Basin Upper Huai Catchment overlaid with the TIGGE ensemble forecast grids The differences between Upper Severn (US) and Upper Huai (UH) are (1) hydrological model – LISFLOOD-RR for US, Xinanjiang for UH – use existing models to facilitate coupling with meteorological inputs; (2) hydraulics model – LISFLOOD-FP for US, 1D flood routing model for UH – due to the availability of river cross sections and DEM data. Hohai is now coupling Xinanjiang model with meteorological inputs. Results will be then used to drive the 1D hydraulics model. Preliminary results are anticipated in early April. 15

NEWS market value Market value in both business and public sectors Insurance companies: to derive flood risk maps to estimate loss values and calculate premium to be allocated to specific regions; Engineering consulting firms: to provide sound climate-proof design for flood defenses and other hydraulic structures. The system can also be used to provide strategic advice on flood insurance including institutional and cost-recovery mechanisms to assist in the development of resilience to climate change through mitigation; Research institutions: use tailor made modules from the NEWS system for simulation exercises for research and development purposes; Educational institutions: teaching purposes especially targeted at local government authorities and graduate students at advanced learning institutions including universities; Public authorities: to perform flood risk analysis to inform planning and new investment in real estate especially in low lying areas. The forecasting component can be used to issue flood warning, improve public awareness of flood risk and assist regional level policy formulation and decision making. Significant market value through impact in both business and public sectors. 16

NEWS in the public domain - website Elaborated on the next slide 17