CAJ | 학술논문

一致性假设一直是洪水频率分析的基本假设条件，但在气候变化与人类活动综合影响下，水文极值序列的一致性假设受到极大挑战。基于此，以东江流域为例，运用 GAMLSS模型（广义可加模型），将时间、气候指标（北极涛动AO、北太平洋涛动NPO、太平洋年代际振荡PDO和南方涛动SOI）以及水库指标（Reservoir Index）统一纳入洪水频率分析中，并对东江流域1954―2009年年最大流量序列（AMS）进行频率分析，结果表明：1）龙川、河源和岭下站年最大流量序列均值与时间呈线性关系，方差为常量，而博罗站均值和方差与时间均呈非线性关系；2）水库对各水文站点AMS均值有显著线性影响；3）NPO对各站点AMS均值有显著线性影响，NPO值较高时，东江流域可能面临着较低的洪水风险，在一致性假设前提下，可能高估洪水设计值，反之亦然；而PDO对各站点方差有显著线性（岭下、博罗）/非线性（龙川、河源）影响；4）以时间为协变量构建非一致性模型，研究得出：龙川、河源和岭下3站T年一遇洪水设计值均呈单调下降趋势，博罗站1954―1995年左右洪水设计值呈下降趋势，而在1996―2009年呈上升趋势；以气候与水库为协变量构建非一致性模型，研究表明：龙川、河源和岭下3站T年一遇洪水设计值均因水库影响呈向下跃变；5）以气候和水库为协变量的非一致性模型对洪水频率具有良好的预测能力，为非一致条件下设计洪水的预测提供了新的预测方法。
일치성가설일직시홍수빈솔분석적기본가설조건，단재기후변화여인류활동종합영향하，수문겁치서렬적일치성가설수도겁대도전。기우차，이동강류역위례，운용 GAMLSS모형（엄의가가모형），장시간、기후지표（북겁도동AO、북태평양도동NPO、태평양년대제진탕PDO화남방도동SOI）이급수고지표（Reservoir Index）통일납입홍수빈솔분석중，병대동강류역1954―2009년년최대류량서렬（AMS）진행빈솔분석，결과표명：1）룡천、하원화령하참년최대류량서렬균치여시간정선성관계，방차위상량，이박라참균치화방차여시간균정비선성관계；2）수고대각수문참점AMS균치유현저선성영향；3）NPO대각참점AMS균치유현저선성영향，NPO치교고시，동강류역가능면림착교저적홍수풍험，재일치성가설전제하，가능고고홍수설계치，반지역연；이PDO대각참점방차유현저선성（령하、박라）/비선성（룡천、하원）영향；4）이시간위협변량구건비일치성모형，연구득출：룡천、하원화령하3참T년일우홍수설계치균정단조하강추세，박라참1954―1995년좌우홍수설계치정하강추세，이재1996―2009년정상승추세；이기후여수고위협변량구건비일치성모형，연구표명：룡천、하원화령하3참T년일우홍수설계치균인수고영향정향하약변；5）이기후화수고위협변량적비일치성모형대홍수빈솔구유량호적예측능력，위비일치조건하설계홍수적예측제공료신적예측방법。
Under the influences of human activities and climate change, the validity of the stationarity assumption in flood frequency analysis is violated. In this case, the GAMLSS (Generalized Additive Models in Location, Scale and Shape) is used to model the annual maximum streamflow (AMS) in terms of flood frequency analysis by taking time, climate indices (AO, NPO, PDO and SOI) and reservoir index (RI) as covariates. The results indicate that:1) linear dependence is identified between AMS and time at 3 stations, i.e. Longchuan, Heyuan and Lingxia, while nonlinear dependence at Boluo station;2) the reservoirs have a significant linear impact on AMS at all the stations;3) NPO has evident linear impacts on AMS at all the stations, and higher NPO is related to lower flood risk, and vice versa. PDO, however, has evident linear impacts on variance at Lingxia and Boluo stations, and nonlinear impacts on variance at Longchuan and Heyuan stations;4) With time as covariate, the flood with return periods of T is decreasing monotonically;However, with climate indices and reservoir index as covariates, the influences of water reservoirs on AMS changes are clearly identified, showing powerful performance of nonlinear models with considerations of combined influences of climate change and water reservoirs in modeling AMS processes. This study would provide a novel thinking way for flood modeling and flood mitigation.