The data‐driven approach as an operational real‐time flood forecasting model

Accurate water level forecasts are essential for flood warning. This study adopts a data‐driven approach based on the adaptive network–based fuzzy inference system (ANFIS) to forecast the daily water levels of the Lower Mekong River at Pakse, Lao People's Democratic Republic. ANFIS is a hybrid system combining fuzzy inference system and artificial neural networks. Five ANFIS models were developed to provide water level forecasts from 1 to 5 days ahead, respectively. The results show that although ANFIS forecasts of water levels up to three lead days satisfied the benchmark, four‐ and five‐lead‐day forecasts were only slightly better in performance compared with the currently adopted operational model. This limitation is imposed by the auto‐ and cross‐correlations of the water level time series. Output updating procedures based on the autoregressive (AR) and recursive AR (RAR) models were used to enhance ANFIS model outputs. The RAR model performed better than the AR model. In addition, a partial recursive procedure that reduced the number of recursive steps when applying the AR or the RAR model for multi‐step‐ahead error prediction was superior to the fully recursive procedure. The RAR‐based partial recursive updating procedure significantly improved three‐, four‐ and five‐lead‐day forecasts. Our study further shows that for long lead times, ANFIS model errors are dominated by lag time errors. Although the ANFIS model with the RAR‐based partial recursive updating procedure provided the best results, this method was able to reduce the lag time errors significantly for the falling limbs only. Improvements for the rising limbs were modest. Copyright © 2011 John Wiley & Sons, Ltd.     

Moving boulders in flash floods and estimating flow conditions using boulders in ancient deposits

Boulders moving in flash floods cause considerable damage and casualties. More and bigger boulders move in flash floods than predicted from published theory. The interpretation of flow conditions from the size of large particles within flash flood deposits has, until now, generally assumed that the velocity (or discharge) is unchanging in time (i.e. flow is steady), or changes instantaneously between periods of constant conditions. Standard practice is to apply theories developed for steady flow conditions to flash floods, which are however inherently very unsteady flows. This is likely to lead to overestimates of peak flow velocity (or discharge). Flash floods are characterised by extremely rapid variations in flow that generate significant transient forces in addition to the mean‐flow drag. These transient forces, generated by rapid velocity changes, are generally ignored in published theories, but they are briefly so large that they could initiate the motion of boulders. This paper develops a theory for the initiation of boulder movement due to the additional impulsive force generated by unsteady flow, and discusses the implications.     

The Early Cretaceous Jacuí Group,a newly discovered volcaniclastic–epiclastic accumulation at the top of the Paraná Basin,southern Brazil

The presence of volcaniclastic rocks related to the silicic magmatism within the Serra Geral Formation has been a matter of long-standing debate. In this paper, we present extensive documentation that supports the presence and abundance of these rocks in the Jacuí Group, a newly discovered volcaniclastic and epiclastic accumulation in southern Brazil. The Jacuí Group is composed of two interfingered stratigraphic units, the Volta Alegre and Tupanciretã formations, and it represents the uppermost stratigraphic unit of the Paraná Basin. The Volta Alegre Formation is primarily composed of resedimented volcaniclastic tuffites, the pyroclasts which were sourced from the Santa Maria subgroup of the Palmas-type of the Serra Geral Formation. The Tupanciretã Formation is composed of fluvial and aeolian deposits transported towards the north–northwest. Deposition of the Jacuí Group began in the Early Cretaceous (∼132 Ma) and was coeval with the acidic volcanism of the Santa Maria subgroup. This group was deposited in a probable interior sag basin that represents either the beginning of the extension in the inner part of the continent that subsequently migrated to the east or the far-field impact of extensional processes that preceded the break-up of Gondwana and the opening of the South Atlantic Ocean.     

湖北省旱涝灾害致灾规律的初步研究

利用1960—2005年湖北省76个地区气象灾害的灾情普查数据和逐日降水量观测资料,对湖北省旱涝灾害的时空分布特征及其致灾规律进行分析。结果表明:干旱灾害的频发区呈东西走向的带状分布,而洪涝灾害的发生频次和频发区面积均明显少于干旱;干旱和洪涝灾害年平均发生站次在1996年以后出现相反的变化趋势,干旱发生站次增加,而洪涝发生站次减少,且两种灾害均主要集中发生在夏季;1996—2001年湖北省部分地区连续出现严重干旱灾害,干旱的累积增强效应导致农业经济损失出现跳跃性增长并在2001年达到最大值;洪涝的致灾强度呈准周期的起伏振荡,农作物受洪涝影响面积最大、损失最多的年份集中在20世纪90年代,农作物受害面积与农业经济损失的决定系数为0.8;受害人口与直接经济损失具有较好的相关特征,且直接经济损失随受害人口增多而增加的速度加快,但近年来人口对洪涝灾害的抵御能力也显著提高;急转干旱和急转洪涝主要发生在鄂西北和鄂东南的夏季,农作物的脆弱度增加,农业经济损失随受害面积增大而增加的速度加快,但所造成的农业经济损失远小于仅发生干旱和洪涝时的数值。     

新疆昌吉市主汛期降水日变化特征

基于昌吉市2008—2015年逐时自动降水资料,分析了主汛期(5—8月)降水日变化特征。结果表明,降水主要集中在夜间21:00至翌日03:00,最大值出现在02:00,最小值出现在14:00;逐时降水频次为明显的单峰型,降水易发生在21:00至翌日08:00,降水频次的高峰值出现在01:00,降水最不易产生于午后15:00至18:00;降水强度变化的波动性较大,大值区出现在21:00至翌日02:00和午后15:00至19:00,最高值出现在18:00,最低值出现在04:00至08:00;在≥0.1 mm、≥1 mm和≥3 mm的逐时降水频次中,夜间降水频次较白天高,≥0.1 mm的降水出现次数较多;降水主要以夜雨,且以短时间(1—4h)的降水为主,贡献率最大的是持续7h的降水,最小的为12h;总云量和低云量的变化与降水量成显著正相关关系。     

1991年5-7月江淮洪涝:水汽和冷空气的作用比较

对1991年5-7月江淮流域持续性暴雨的环流异常进行分析,比较造成此次洪涝的水汽输送和冷空气活动重要性。结果表明：（1）东亚大槽维持在中国东北上空发展明显,同时鄂霍次克海上空没有建立强大阻高,这种形势有利于冷空气持续向南侵袭,盛行经向环流。此外,低纬度地区,西太平洋副热带高压主体位置于常年同期相比,明显偏西、偏强,有利于暖湿气流沿副热带高压北上到达江淮流域,与北方冷空气辐合形成强降水。（2）通过定义I_T和I_q分别表征温度与水汽对降水的贡献,发现此次江淮流域地区强降水是由对流层低层水汽异常增多和气温异常偏低共同造成的,作用基本相当。     

运用社交网络分析淹水空间信息之关联性

随着社交网络的普遍发展,大量的讯息透过智能手机发布在个人的微博或其他社交网站。台湾地区的社交网站以脸书(Facebook)的使用量最大,平均每天有近千万笔的讯息量,大多数的讯息多以食衣住行或个人讯息为主,但从本研究所撷取自2010年至2015年的数据中显示,公众在社交网站所分享的信息中具有降雨、淹水或相关灾情的讯息,而这些讯息具有极高比例的正确性。由于社交网站无法提供私人讯息,故本研究将从社交信息中,以地点为单位撷取大量的数据信息再辅以语意关键词萃取出有关可作为淹水预判的讯息数据。为检核资料的可性度,本研究透过历史台风数据FLO-2D仿真重建淹水之空间信息进行检核。从研究比对分析中发现,经萃取后的公众信息其与灾害的关联性及正确性相当显着,故透过社交网站中大量的非结构讯息,透过语意及空间的转换,可萃取转化为防灾信息,对广域的都市治理而言,此一讯息将可作为预判区域淹水或防救灾情报之有效参考。     

Geochemistry of Paraná-Etendeka basalts from Misiones,Argentina: Some new insights into the petrogenesis of high-Ti continental flood basalts

The Early Cretaceous (∼135–131 Ma) Paraná-Etendeka continental flood basalts, preserved in bulk in the Paraná basin of southern Brazil and vicinity, have been divided into low-Ti and high-Ti types that govern the southern and northern halves of the basin, respectively. We have examined a new sample set from the southern margin of the northern high-Ti segment of Paraná basalts in Misiones, northeastern Argentina. These basalts are strongly to moderately enriched in TiO₂ (2–4 wt.%), have relatively high Ti/Y (300–500), low MgO (3.5–6.5 wt.%), and high Fe (FeO(tot) 12–14 wt.%) and belong to the Pitanga and Paranapanema magma types of Peate et al. (1992). Nd and Sr isotope compositions are quite unvarying with ε_Nd (at 133 Ma) values of −4.6 to −3.6 and initial ⁸⁷Sr/⁸⁶Sr of 0.7054–0.7059 and show no variation with fractionation. Compared to high-Ti lavas in the central and northern parts of the Paraná high-Ti basalt segment, the lavas from Misiones are similar to those in the northeastern magin of the basin but less radiogenic in initial Nd isotope composition than those in the central part. This variation probably reflects mixed EM1-EM2 source components in the sublithospheric mantle. A polybaric melt model of a sublithospheric mantle source at the garnet lherzolite-spinel lherzolite transition is compatible with the observed Ti budget of the Pitanga and Paranapanema lavas, regardless of the Nd isotope composition of their purported source.     

Rainfall flood hazard at nuclear power plants in India

Flood hazard evaluation is an important input for Nuclear Power Plants external events safety studies. In the present study, flood hazard at various nuclear sites in India due to rainfall has been evaluated. Hazard estimation is a statistical procedure by which rainfall intensity versus occurrence frequency is estimated from historical records of rainfall data and extrapolated with asymptotic extreme value distribution. Rainfall data needed for flood hazard assessment are daily annual maximum rainfall (24?h data). The observed data points have been fitted using Gumbel, power law and exponential distribution, and return period has been estimated. To study the stationarity of rainfall data, a moving window estimate of the parameters has been performed. The rainfall pattern is stationary in both coastal and inland regions over the period of observation. The coastal regions show intense rainfall and higher variability than inland regions. Based on the plant layout, catchment area and drainage capacity, the prototype fast breeder reactor (PFBR) site is unlikely to be flooded.     

融雪洪水跳跃变异对其参数估计不确定性影响分析

利用Pettitt非参数检验法和Mann-Kendall非参数趋势检验法,分析年最大洪峰流量序列的非一致性,确定序列的变异形式,采用“分解-合成”理论对其进行一致性修正,得到过去、现状两种条件下年最大洪峰流量序列,根据贝叶斯理论对序列一致性修正前后参数不确定性进行估计,并对其预报区间优良性进行评价。研究结果表明：年最大洪峰流量序列变异点发生在1993年,序列整体上升趋势不显著,在1957-1993年子序列呈显著下降趋势,而1994-2006年子序列变化趋势不显著,跳跃变异为序列主要变异形式;给出了实测、还原及还现序列参数后验分布估计值及95%置信区间,将其结合优化适线法进行P-Ⅲ型频率分析,得到修正前后设计频率年最大洪峰流量预报区间估计值;还原、还现序列与实测序列相比,预报区间覆盖率均提高24%,平均带宽分别减少39.59%、23.17%,平均偏移幅度分别减少28.45%、11.39%。通过对非一致性年最大洪峰流量序列还原/还现计算,可减小参数估计不确定性对其计算产生的影响,从而提高预报区间的可靠性。