吳海江,粟曉玲,祁繼霞,張 特,朱興宇,武連洲
·農(nóng)業(yè)水土工程·
Vine Copula與貝葉斯模型平均結(jié)合的月徑流預(yù)測(cè)及應(yīng)用
吳海江1,2,粟曉玲1,2※,祁繼霞2,張 特2,朱興宇2,武連洲2
(1. 西北農(nóng)林科技大學(xué)旱區(qū)農(nóng)業(yè)水土工程教育部重點(diǎn)實(shí)驗(yàn)室,楊凌 712100;2. 西北農(nóng)林科技大學(xué)水利與建筑工程學(xué)院,楊凌 712100)
準(zhǔn)確可靠且預(yù)見(jiàn)期較長(zhǎng)的月徑流預(yù)測(cè)對(duì)水資源配置、防汛抗旱以及生態(tài)環(huán)境保護(hù)等具有重要意義。徑流變化與降水、氣溫、潛在蒸散發(fā)以及前期徑流等存在密切聯(lián)系。鑒于Vine Copula可以靈活地將多個(gè)隨機(jī)變量的邊緣分布函數(shù)通過(guò)Copula對(duì)的形式聯(lián)結(jié)起來(lái)構(gòu)造多維聯(lián)合分布函數(shù)以及貝葉斯模型平均(Bayesian Model Averaging,BMA)在處理多模型集合預(yù)報(bào)方面的優(yōu)勢(shì),該研究基于BMA集合多個(gè)Vine Copula模型提出了一種BVC徑流預(yù)測(cè)模型(簡(jiǎn)稱(chēng)BVC模型),應(yīng)用于黃河流域上游4個(gè)水文站(唐乃亥站、民和站、紅旗站和折橋站)的月徑流預(yù)測(cè),采用確定性系數(shù)(2)、納什效率系數(shù)(Nash-Sutcliffe Efficiency coefficient,NSE)和均方根誤差(Root Mean Squared Error,RMSE)評(píng)價(jià)模型的預(yù)測(cè)性能。結(jié)果表明,驗(yàn)證期內(nèi)預(yù)見(jiàn)期為1~3個(gè)月時(shí),BVC模型在各水文站的2均大于等于0.83、NSE均大于等于0.78且RMSE均維持在較低水平;與隨機(jī)森林(Random Forest,RF)模型和長(zhǎng)短期記憶神經(jīng)網(wǎng)絡(luò)(Long Short-Term Memory Neural Network,LSTM)模型相比,BVC模型能夠很好地預(yù)測(cè)各水文站月徑流的變化過(guò)程,特別是月徑流極值的變化。研究表明BVC模型在預(yù)見(jiàn)期為1~3個(gè)月時(shí)的月徑流預(yù)測(cè)性能明顯優(yōu)于RF模型和LSTM模型。該研究構(gòu)建的BVC模型為流域的水資源管理和風(fēng)險(xiǎn)評(píng)估等提供參考。
水資源;機(jī)器學(xué)習(xí);徑流;預(yù)測(cè);貝葉斯模型平均;Vine Copula;黃河流域
河川徑流的變化直接關(guān)系到水資源安全、防汛抗旱、農(nóng)業(yè)生產(chǎn)活動(dòng)以及生態(tài)環(huán)境健康等[1-2]。近年來(lái),受全球氣候變暖和人類(lèi)活動(dòng)的影響,與徑流有關(guān)的極端事件的頻率和強(qiáng)度呈增加趨勢(shì)[3-6],如由徑流減少誘發(fā)的干旱事件以及徑流快速增加引發(fā)的洪水事件等[7-9]。這些極端事件對(duì)工農(nóng)業(yè)生產(chǎn)、人民生命財(cái)產(chǎn)和物種多樣性等造成了災(zāi)難性的影響。因此,可靠的徑流預(yù)測(cè)對(duì)于規(guī)避這些潛在風(fēng)險(xiǎn)具有重要意義,而預(yù)見(jiàn)期越長(zhǎng),越有利于盡早采取措施緩解和規(guī)避相關(guān)潛在風(fēng)險(xiǎn)。
徑流預(yù)測(cè)模型大致分為基于物理機(jī)制的預(yù)測(cè)模型和基于統(tǒng)計(jì)關(guān)系的預(yù)測(cè)模型?;谖锢頇C(jī)制的預(yù)測(cè)模型成因機(jī)制較強(qiáng),但其結(jié)構(gòu)較為復(fù)雜、需要輸入的實(shí)測(cè)水文氣象變量較多、模型構(gòu)建過(guò)程中需對(duì)參數(shù)和輸出結(jié)果進(jìn)行多次率定和誤差訂正[4-5,10-12],普遍存在模型參數(shù)難以確定及模型通用性較差等缺陷。在水文實(shí)踐中,由于高寒地區(qū)準(zhǔn)確可靠的水文氣象實(shí)測(cè)資料難以獲取,極大地限制了基于物理機(jī)制的預(yù)測(cè)模型的推廣和準(zhǔn)確預(yù)測(cè)[13]?;诮y(tǒng)計(jì)關(guān)系的預(yù)測(cè)模型可以根據(jù)歷史水文氣象變量之間的統(tǒng)計(jì)關(guān)系實(shí)現(xiàn)預(yù)測(cè),能夠較好地揭示各要素之間的聯(lián)系且需要輸入的水文氣象變量較少[14]。常用的統(tǒng)計(jì)預(yù)測(cè)模型主要有時(shí)間序列統(tǒng)計(jì)模型、機(jī)器學(xué)習(xí)模型和多元線性回歸模型等[15-17]。其中,時(shí)間序列統(tǒng)計(jì)模型如自回歸滑動(dòng)平均模型(Autoregressive Moving Average,ARMA)和自回歸差分滑動(dòng)平均模型(Autoregressive Integrated Moving Average,ARIMA模型)只能捕捉自身變量間的線性關(guān)系,不能較好地反映變量間的非線性特征,預(yù)測(cè)效果普遍較差;機(jī)器學(xué)習(xí)模型包括支持向量機(jī)(Support Vector Machine,SVM)[16]、長(zhǎng)短期記憶神經(jīng)網(wǎng)絡(luò)(Long Short-Term Memory Neural Network,LSTM)[18]、自適應(yīng)模糊推理系統(tǒng)(Adaptive Neuro-Fuzzy Inference System,ANFIS)[13,19]及隨機(jī)森林(Random Forest,RF)[15,17,20]等,多為灰箱或黑箱模型,容易出現(xiàn)過(guò)擬合問(wèn)題且多不能給出明確的解析表達(dá)式;多元線性回歸模型基于解釋變量與預(yù)測(cè)變量之間的線性關(guān)系來(lái)構(gòu)建預(yù)測(cè)模型[16,21],不能很好地反映變量之間存在的復(fù)雜非線性關(guān)系。徑流作為水文循環(huán)的重要組成部分之一,受降水、氣溫、潛在蒸散發(fā)以及前期來(lái)水狀況等多種因素影響[16]。現(xiàn)有的模型在進(jìn)行徑流預(yù)測(cè)時(shí),考慮的影響因素不夠全面[21-22],從而導(dǎo)致模型的預(yù)測(cè)精度在預(yù)見(jiàn)期較長(zhǎng)時(shí)表現(xiàn)較差,不能滿足水文預(yù)報(bào)、水資源管理以及政策制定的需求。
Copula函數(shù)可以有效處理多變量問(wèn)題,被廣泛用于水-能源-糧食共生安全風(fēng)險(xiǎn)概率評(píng)估[23]和水資源優(yōu)化配置[24-25]。Copula函數(shù)的提出和應(yīng)用可以較好地解決上述統(tǒng)計(jì)預(yù)測(cè)模型存在的缺陷。其中,Vine Copula函數(shù)具有明確的解析表達(dá)式,可以將多個(gè)隨機(jī)變量的邊緣分布函數(shù)通過(guò)Copula對(duì)的形式聯(lián)結(jié)起來(lái)構(gòu)造多維聯(lián)合分布函數(shù),能夠刻畫(huà)不同變量之間存在的正負(fù)依賴(lài)性以及尾部相依性,在干旱風(fēng)險(xiǎn)評(píng)估、干旱預(yù)測(cè)和洪水風(fēng)險(xiǎn)分析等研究中得到廣泛應(yīng)用[8,11,26-28]。在高維情形下,由于Vine Copula函數(shù)的形式與變量的順序密切相關(guān),即不同的變量順序(或分解形式)對(duì)應(yīng)的樹(shù)型結(jié)構(gòu)有差異,可能影響預(yù)測(cè)的可靠性[13]。貝葉斯模型平均(Bayesian Model Averaging,BMA)作為一種集合預(yù)報(bào)模型,通過(guò)對(duì)不同的模型(集合成員)分配不同的權(quán)重,可有效綜合不同集合成員的優(yōu)勢(shì),從而提高預(yù)測(cè)性能[29-32]。因此,將BMA與Vine Copula函數(shù)結(jié)合起來(lái)有望提高徑流的預(yù)測(cè)精度,并拓寬徑流預(yù)測(cè)的理論框架?;诖?,本文基于BMA結(jié)合Vine Copula函數(shù)提出了BVC徑流預(yù)測(cè)模型(簡(jiǎn)稱(chēng)BVC模型),并將其應(yīng)用于黃河流域上游唐乃亥、民和、紅旗和折橋4個(gè)水文站的月徑流預(yù)測(cè),以期為流域的防汛抗旱以及水資源配置等提供參考。
黃河流域上游4個(gè)水文站1963—2016年的月徑流資料來(lái)源于黃河流域水文年鑒。同時(shí)段逐月的降水、最高氣溫、最低氣溫、日照時(shí)數(shù)、2 m風(fēng)速等氣象數(shù)據(jù)資料來(lái)自中國(guó)科學(xué)院氣候變化研究中心提供的CN05.1數(shù)據(jù)集[33-34](http://ccrc.iap.ac.cn/resource/detail?id=228),其空間分辨率為0.25°×0.25°。此外,使用CN05.1氣象數(shù)據(jù)資料基于Penman-Monteith公式[35]計(jì)算月潛在蒸散發(fā)。根據(jù)各水文站集水控制面積(圖1)范圍內(nèi)逐月的降水、氣溫和潛在蒸散發(fā)的格網(wǎng)均值得到各水文站逐月的降水、氣溫和潛在蒸散發(fā)序列。
圖1 黃河流域上游4個(gè)水文站的位置及其相應(yīng)的集水面積
從水文循環(huán)的角度分析,徑流變化主要受降水、氣溫、潛在蒸散發(fā)以及前期來(lái)水狀況的影響。因此,后期的徑流變化與前期的降水、氣溫、潛在蒸散發(fā)和徑流存在一定的成因聯(lián)系。設(shè)前期的降水(P-l;為時(shí)滯)、氣溫(T-l)、潛在蒸散發(fā)(E-l)和徑流(Q-l)等解釋變量分別為1、2、3和4,目標(biāo)月份的徑流(即預(yù)測(cè)變量)為5。根據(jù)卡方檢驗(yàn)最小原則[13,36],基于極大似然估計(jì)方法分別為變量1~5選擇最優(yōu)的邊緣分布函數(shù)(正態(tài)分布、伽馬分布、韋伯分布或?qū)?shù)正態(tài)分布)。根據(jù)Skalar準(zhǔn)則[37],五維連續(xù)隨機(jī)變量=[1,,5]的分布函數(shù)(1,,5)可以表示為
式中表示五維Copula函數(shù);u=F(x)為變量x(=1~5)的累積概率。
Vine Copula函數(shù)包含2種子類(lèi)型:C-vine Copula函數(shù)和D-vine Copula函數(shù)[38]。與D-vine Copula函數(shù)相比,C-vine Copula函數(shù)聯(lián)合變量時(shí)可以共用一個(gè)節(jié)點(diǎn)且樹(shù)型結(jié)構(gòu)相對(duì)簡(jiǎn)單,因此,本文選用C-vine Copula函數(shù)來(lái)聯(lián)合變量1~5,則(1,,5)的概率密度函數(shù)(1,,5)可表示為[8]
式中f和c分別表示變量x和雙變量Copula函數(shù)的概率密度函數(shù);和分別表示樹(shù)型和邊(圖2);擬選用的雙變量Copula函數(shù)有Elliptical族Copula、Archimedean族Copula以及Tawn型Copula等(共計(jì)39種Copula函數(shù)[39])。對(duì)于式(2)中的每個(gè)條件概率密度函數(shù)(|)可以寫(xiě)成[40]
式中-k表示從向量中去除第個(gè)變量后的向量。引入函數(shù),(|)的表達(dá)式為[8,27,38]
C-vine Copula(VC)模型結(jié)構(gòu)與變量順序(或分解形式)密切相關(guān)[13,38]。例如,當(dāng)固定預(yù)測(cè)變量后,對(duì)于一個(gè)維變量,與VC模型有關(guān)的變量順序共有(-1)!種[9]。在VC模型中與每個(gè)特定的變量順序有關(guān)的樹(shù)型有–1個(gè)(圖2)。以變量順序1、2、3、4和5為例,簡(jiǎn)記為“12345”(圖2),則根據(jù)式(4)可以推導(dǎo)出在給定1、2、3和4的條件下5的條件分布函數(shù)(5|1,2,3,4)滿足
式中θ為第個(gè)樹(shù)型第條邊對(duì)應(yīng)的Copula對(duì)參數(shù);為函數(shù)(圖2)。利用分位數(shù)曲線[41](∈[0, 1])通過(guò)遞歸調(diào)用對(duì)式(5)關(guān)于5求逆[8,13]后得到
式中-1和-1分別表示函數(shù)和變量5服從某一分布(正態(tài)分布、伽馬分布、韋伯分布或?qū)?shù)正態(tài)分布)的逆函數(shù)。
注:變量x1~x5順序記為“12345”,帶箭頭的線表示邊。模型包含4棵樹(shù)和10條邊。C15表示C(u1, u5);C24|1表示C(u4|u1,u2|u1);θij為第i個(gè)樹(shù)型第j條邊對(duì)應(yīng)的Copula對(duì)參數(shù)。
類(lèi)似地,根據(jù)以上步驟可以推導(dǎo)出五維的VC模型中其他23種變量順序?qū)?yīng)的函數(shù)表達(dá)式(共(5-1)!即24種VC模型)。以預(yù)見(jiàn)期為1個(gè)月時(shí)2006年8月的徑流預(yù)測(cè)為例(假設(shè)已知1963年1月—2006年7月的降水、氣溫、潛在蒸散發(fā)和徑流信息),首先借助于蒙特卡洛模擬在區(qū)間[0, 1]上產(chǎn)生500個(gè)均勻分布的隨機(jī)數(shù),然后由式(6)(或其他23種變量順序下對(duì)應(yīng)的VC模型)得到500個(gè)徑流預(yù)測(cè)值并取其均值,以此作為2006年8月徑流預(yù)測(cè)值。注意,不同變量順序下的VC模型得到的徑流預(yù)測(cè)值可能不同。
由于不同的變量順序?qū)?yīng)的VC模型做出的預(yù)測(cè)結(jié)果可能有差異,因此,基于某些統(tǒng)計(jì)指標(biāo)(如AIC)僅選擇最優(yōu)的VC模型可能會(huì)導(dǎo)致模型的預(yù)測(cè)結(jié)果與實(shí)際情況存在較大偏差[13]。BMA作為一種被廣泛使用的多模型集合預(yù)報(bào)方法,可以有效地耦合多個(gè)模型的預(yù)測(cè)結(jié)果,從而降低單一模型預(yù)測(cè)的不確定性[13,30-31]。設(shè)五維情形下Vine Copula函數(shù)包含的24種預(yù)測(cè)模型分別用VC1~VC24表示,即VC = [VC1, …, VC24],給定訓(xùn)練集,則基于BMA結(jié)合Vine Copula函數(shù)(即BVC模型)關(guān)于預(yù)測(cè)變量5的表達(dá)式為
各水文站點(diǎn)的率定期和驗(yàn)證期分別為1963—2006年和2007—2016年,即留20%的數(shù)據(jù)來(lái)驗(yàn)證模型的預(yù)測(cè)效果和有效性。采用確定性系數(shù)(2)、納什效率系數(shù)(Nash-Sutcliffe Efficiency coefficient,NSE)和均方根誤差(Root Mean Squared Error,RMSE)評(píng)價(jià)模型在率定期和驗(yàn)證期的預(yù)測(cè)性能,其表達(dá)式分別為
分析前期的降水、氣溫、潛在蒸散發(fā)和徑流與后期徑流之間相關(guān)性的強(qiáng)弱可以表征前期的氣象水文因子對(duì)后期徑流變化的影響程度。如表1所示,各水文站的預(yù)測(cè)因子與預(yù)測(cè)變量之間呈正相關(guān),且相關(guān)性隨著時(shí)滯的增加逐漸降低。除唐乃亥站、民和站和紅旗站的前期徑流與當(dāng)前時(shí)段的徑流在3個(gè)月時(shí)滯下的相關(guān)性(徑流自相關(guān)性)沒(méi)有通過(guò)顯著性檢驗(yàn)外(> 0.05),其余站點(diǎn)的預(yù)測(cè)因子與預(yù)測(cè)變量之間在1~3個(gè)月時(shí)滯下均表現(xiàn)為顯著正相關(guān)(< 0.05)。這表明前期的降水、氣溫、潛在蒸散發(fā)和徑流均顯著影響后期的徑流變化,即這些水文氣象變量之間存在密切的成因聯(lián)系,選取的這些預(yù)測(cè)因子可以為后期的徑流預(yù)測(cè)提供有效的信息。
由于本文選取的4個(gè)水文站點(diǎn)的集水控制面積位于黃河流域上游(圖1),屬于高海拔低溫地區(qū),其徑流變化過(guò)程對(duì)氣候非常敏感。在該區(qū)域,氣溫除了影響潛在蒸散發(fā)外,還可能影響凍土的消融,進(jìn)而影響徑流的變化過(guò)程(注意,本文是按整個(gè)時(shí)間序列來(lái)計(jì)算潛在蒸散發(fā)(氣溫)與徑流之間的相關(guān)性)。依據(jù)潛在蒸散發(fā)(氣溫)和徑流的年內(nèi)變化規(guī)律(即春冬兩季的潛在蒸散發(fā)(氣溫)和徑流相對(duì)較小,而夏秋兩季則相反),徑流量較大的月份,其潛在蒸散發(fā)量也較大(氣溫也較高)。因此,徑流與潛在蒸散發(fā)(氣溫)之間呈顯著正相關(guān)[16]。對(duì)于某一特定的月份,徑流與潛在蒸散發(fā)(氣溫)多呈負(fù)相關(guān)[26]。
表1 黃河流域上游4個(gè)水文站1~3個(gè)月時(shí)滯下預(yù)測(cè)變量與預(yù)測(cè)因子之間的相關(guān)性
注:*表示在0.05水平顯著。
Note: * denotes significance at 0.05 level.
根據(jù)卡方檢驗(yàn)最小原則[13,36],黃河流域上游4個(gè)水文站的降水、氣溫、潛在蒸散發(fā)、徑流分別服從伽馬分布、正態(tài)分布、韋伯分布、對(duì)數(shù)正態(tài)分布。在率定期內(nèi),BVC模型在各水文站1~3個(gè)月預(yù)見(jiàn)期下的徑流預(yù)測(cè)值與其對(duì)應(yīng)的觀測(cè)值吻合較好(圖3),且較好地捕捉到了各站的徑流極值,表明BVC模型能夠有效地集合不同VC模型的優(yōu)點(diǎn),并利用前期的降水、氣溫、潛在蒸散發(fā)及徑流所包含的預(yù)測(cè)信息提高模型的月徑流預(yù)測(cè)能力。如表2所示,BVC模型在各水文站1~3個(gè)月預(yù)見(jiàn)期下的性能評(píng)價(jià)指標(biāo)NSE和2均在0.73及以上且RMSE均較低,表明BVC模型在1~3個(gè)月預(yù)見(jiàn)期下能夠作出可靠的月徑流預(yù)測(cè)。鑒于隨機(jī)森林(Random Forest,RF)模型在徑流預(yù)測(cè)方面的優(yōu)勢(shì)[15,17],本文選用RF模型作為參考模型以進(jìn)一步評(píng)價(jià)BVC模型的徑流預(yù)測(cè)性能。RF模型中的解釋變量和預(yù)測(cè)變量與BVC模型保持一致。如圖3所示,RF模型在1~3個(gè)月預(yù)見(jiàn)期下也較好地預(yù)測(cè)了各站點(diǎn)的徑流變化過(guò)程,但對(duì)徑流峰值的預(yù)測(cè)效果較差,存在明顯低估現(xiàn)象。由于汛期洪水造成的災(zāi)難性影響較大,而B(niǎo)VC模型相較于RF模型對(duì)某些汛期徑流存在一定的高估現(xiàn)象(但所占比例很小),BVC模型在汛期偏安全的預(yù)測(cè)結(jié)果有利于規(guī)避汛期洪水風(fēng)險(xiǎn)和有效管理水資源。此外,RF模型對(duì)枯水期徑流的預(yù)測(cè)能力也較差(圖3b和圖3d)。從各水文站的性能評(píng)價(jià)指標(biāo)來(lái)看,BVC模型在1~3個(gè)月預(yù)見(jiàn)期下的2和NSE均大于RF模型而RMSE均小于RF模型,表明BVC模型在率定期的徑流預(yù)測(cè)效果明顯優(yōu)于RF模型(表2)。以唐乃亥站為例(表2),1~3個(gè)月預(yù)見(jiàn)期下BVC模型的2分別為0.92、0.95、0.88,NSE分別為0.91、0.94、0.87,而RF模型的2分別為0.73、0.57、0.44,NSE分別為0.73、0.57、0.44。可見(jiàn)BVC模型隨著預(yù)見(jiàn)期的延長(zhǎng)仍能保持良好的預(yù)測(cè)性能,而RF模型的預(yù)測(cè)性能則隨著預(yù)見(jiàn)期的延長(zhǎng)衰減較快。
表2 BVC模型與RF模型在率定期的徑流預(yù)測(cè)性能評(píng)價(jià)
為定量比較評(píng)價(jià)BVC模型和RF模型在枯水期和汛期對(duì)各水文站月徑流的預(yù)測(cè)能力,本文基于連續(xù)3個(gè)月累積徑流量最大(最?。┑玫礁魉恼狙雌冢菟冢?duì)應(yīng)的時(shí)段。其中,唐乃亥站和紅旗站的汛期為7—9月,民和站和折橋站的汛期為8—10月,而各水文站的枯水期均為1 —3月。由表3可知,在率定期內(nèi)1~3個(gè)月預(yù)見(jiàn)期下,BVC模型在各水文站枯水期和汛期的月徑流預(yù)測(cè)能力均明顯優(yōu)于RF模型,如BVC模型在枯水期和汛期的2均大于等于0.57、NSE多大于等于0.61且RMSE均維持在較低水平,且在3個(gè)月預(yù)見(jiàn)期下仍能保持良好的預(yù)測(cè)性能。
圖3 貝葉斯模型平均集合Vine Copula模型和隨機(jī)森林模型在率定期(1963—2006年)1~3個(gè)月預(yù)見(jiàn)期下的徑流預(yù)測(cè)結(jié)果對(duì)比
為進(jìn)一步說(shuō)明BVC模型優(yōu)越的徑流預(yù)測(cè)性能,選擇長(zhǎng)短期記憶神經(jīng)網(wǎng)絡(luò)(Long Short-Term Memory Neural Network,LSTM)模型作為驗(yàn)證期(2007—2016年)的另一個(gè)參考模型。驗(yàn)證期BVC模型、RF模型和LSTM模型在各水文站不同預(yù)見(jiàn)期下徑流預(yù)測(cè)值和觀測(cè)值的散點(diǎn)圖以及線性擬合曲線如圖4所示。RF模型和LSTM模型的線性擬合曲線隨著預(yù)見(jiàn)期的延長(zhǎng)明顯偏離1∶1線且向右傾斜,尤其是3個(gè)月預(yù)見(jiàn)期,表明RF模型和LSTM模型對(duì)徑流峰值(枯期徑流)存在低估(高估)。此外,根據(jù)性能評(píng)價(jià)指標(biāo)2、NSE和RMSE(表4),可以發(fā)現(xiàn)RF模型的預(yù)測(cè)性能隨著預(yù)見(jiàn)期的延長(zhǎng)衰減較快,這是因?yàn)镽F模型在預(yù)測(cè)過(guò)程中損失掉了許多重要的預(yù)測(cè)信息。同樣條件下,LSTM模型的徑流預(yù)測(cè)性能總體上優(yōu)于RF模型(表4)。以紅旗站為例,RF模型由1個(gè)月預(yù)見(jiàn)期NSE0.45降為3個(gè)月預(yù)見(jiàn)期NSE0.08,而LSTM模型由1個(gè)月預(yù)見(jiàn)期NSE0.62降為3個(gè)月預(yù)見(jiàn)期NSE0.40。BVC模型在1~3個(gè)月預(yù)見(jiàn)期下的徑流預(yù)測(cè)值與其對(duì)應(yīng)的觀測(cè)值之間的散點(diǎn)圖均勻地分布在1∶1線附近,說(shuō)明BVC模型在各水文站1~3個(gè)月預(yù)見(jiàn)期下的徑流預(yù)測(cè)能力均較強(qiáng)。BVC模型在4個(gè)水文站的徑流預(yù)測(cè)性能均明顯優(yōu)于RF模型和LSTM模型,且BVC模型在各水文站的2均大于等于0.83、NSE均大于等于0.78且RMSE均維持在較低水平(表4)。例如,在3個(gè)月預(yù)見(jiàn)期下,BVC模型在折橋站的RMSE僅為6.48 m3/s,而RF模型和LSTM模型的RMSE卻分別增為15.53 m3/s和10.94 m3/s。這些結(jié)果進(jìn)一步表明BVC模型在1~3個(gè)月預(yù)見(jiàn)期下良好的月徑流預(yù)測(cè)性能,有利于提前進(jìn)行水資源協(xié)調(diào)配置和規(guī)避風(fēng)險(xiǎn)等,可為站點(diǎn)或流域尺度可靠的徑流預(yù)測(cè)提供理論依據(jù)。
表3 率定期BVC模型與RF模型在枯水期和汛期徑流的預(yù)測(cè)性能評(píng)價(jià)
注:唐乃亥站和紅旗站的汛期為7—9月,民和站和折橋站的汛期為8—10月。
Note: The wettest season was in the July-September period at the Tangnaihai and Hongqi stations, while that appeared in the August-October period at the Minhe and Zheqiao stations.
圖4 BVC模型、隨機(jī)森林(RF)模型和長(zhǎng)短期記憶神經(jīng)網(wǎng)絡(luò)模型在驗(yàn)證期(2007—2016年)1~3個(gè)月預(yù)見(jiàn)期下的徑流預(yù)測(cè)對(duì)比
表4 BVC模型、RF模型和LSTM模型在驗(yàn)證期的徑流預(yù)測(cè)性能評(píng)價(jià)
黃河上游各水文站點(diǎn)位于不同的干支流,汛期時(shí)間不一致,唐乃亥站和紅旗站的汛期為7—9月,民和站和折橋站的汛期為8—10月,這種水文站點(diǎn)集水控制范圍內(nèi)的水文情勢(shì)變化特點(diǎn)不同,以及地形等差異可能導(dǎo)致前期的水文氣象要素對(duì)后期的徑流變化影響具有一定的滯后性[44],從而可能影響模型的預(yù)測(cè)結(jié)果。RF模型和LSTM模型是兩類(lèi)不同類(lèi)型的算法且均屬于黑箱模型。其中,RF模型是由多個(gè)相互獨(dú)立的決策樹(shù)組成的集合,在每次抽樣訓(xùn)練時(shí),只考慮了部分特征分量,且極值樣本可能未被充分抽樣,導(dǎo)致其對(duì)極值的模擬預(yù)測(cè)能力較差。LSTM模型則是由相互連接(非獨(dú)立)的神經(jīng)元組成的網(wǎng)絡(luò),模型中超參數(shù)的設(shè)置非常關(guān)鍵,若設(shè)置不合理則可能導(dǎo)致預(yù)測(cè)結(jié)果不理想。相較于RF模型和LSTM模型,本文構(gòu)建的BVC模型可以給出明確的解析表達(dá)式,一方面可以通過(guò)Vine Copula模型將變量間存在的依賴(lài)關(guān)系充分地挖掘出來(lái),另一方面BMA基于貝葉斯理論為不同的VC模型集合成員分配合理的權(quán)重,可以最大限度地利用各個(gè)VC模型的預(yù)測(cè)信息,從而作出準(zhǔn)確可靠的預(yù)測(cè)結(jié)果。本文僅基于BVC模型對(duì)黃河流域上游的月徑流進(jìn)行了預(yù)測(cè)并取得了良好的預(yù)測(cè)效果,而B(niǎo)VC模型是否適用于其他流域還有待進(jìn)一步研究。此外,如何改善和提高短時(shí)間尺度上(如日、周、旬尺度)的徑流預(yù)測(cè)精度也值得重點(diǎn)關(guān)注。然而,在實(shí)際場(chǎng)景中,當(dāng)月、當(dāng)旬、當(dāng)周、當(dāng)日的水文氣象數(shù)據(jù)很難實(shí)現(xiàn)近實(shí)時(shí)獲取,可以考慮使用國(guó)家氣象科學(xué)數(shù)據(jù)中心提供的陸面數(shù)據(jù)同化系統(tǒng)(CLDAS-V2.0)實(shí)時(shí)產(chǎn)品數(shù)據(jù)集和歐洲中期天氣預(yù)報(bào)中心(European Center for Medium-Range Weather Forecast,ECMWF)提供的數(shù)值天氣預(yù)報(bào)產(chǎn)品作為水文氣象代理數(shù)據(jù)實(shí)現(xiàn)短預(yù)見(jiàn)期近實(shí)時(shí)訂正的徑流預(yù)測(cè),但需注意評(píng)估由此引入的預(yù)測(cè)誤差。
本文利用前期的降水、氣溫、潛在蒸散發(fā)以及徑流作為后期徑流的預(yù)測(cè)因子,基于貝葉斯模型平均(Bayesian Model Averaging,BMA)結(jié)合Vine Copula函數(shù)提出了一種BVC(Bayesian model averaging ensemble Vine Copula)徑流預(yù)測(cè)模型,并將其應(yīng)用于黃河流域上游4個(gè)水文站(唐乃亥站、民和站、紅旗站和折橋站)1~3個(gè)月預(yù)見(jiàn)期下的月徑流預(yù)測(cè),取得了良好的預(yù)測(cè)效果,得到的主要結(jié)論如下:
1)BVC模型在1~3個(gè)月預(yù)見(jiàn)期下很好地預(yù)測(cè)了不同水文站的月徑流變化過(guò)程,可以較為準(zhǔn)確地捕捉到各水文站的月徑流極值,BVC模型在枯水期和汛期的2均大于等于0.57、NSE多大于等于0.61且RMSE均維持在較低水平,且在3個(gè)月預(yù)見(jiàn)期下仍能保持良好的預(yù)測(cè)性能;
2)與隨機(jī)森林(Random Forest,RF)模型和長(zhǎng)短期記憶神經(jīng)網(wǎng)絡(luò)(Long Short-Term Memory Neural Network,LSTM)模型相比,BVC模型在各個(gè)水文站1~3個(gè)月預(yù)見(jiàn)期下的預(yù)測(cè)性能評(píng)價(jià)指標(biāo)2、NSE和RMSE均明顯優(yōu)于RF模型和LSTM模型,驗(yàn)證期內(nèi)BVC模型的2均大于等于0.83、NSE均大于等于0.78且RMSE均維持在較低水平。
[1] 劉章君,郭生練,李天元,等. 貝葉斯概率洪水預(yù)報(bào)模型及其比較應(yīng)用研究[J]. 水利學(xué)報(bào),2014,45(9):1019-1028.
Liu Zhangjun, Guo Shenglian, Li Tianyuan, et al. Comparative study of Bayesian probabilistic flood forecasting models[J]. Journal of Hydraulic Engineering, 2014, 45(9): 1019-1028. (in Chinese with English abstract)
[2] 郝麗娜,粟曉玲,黃巧玲. 基于小波廣義回歸神經(jīng)網(wǎng)絡(luò)耦合模型的月徑流預(yù)測(cè)[J]. 水力發(fā)電學(xué)報(bào),2016,35(5):47-54.
Hao Lina, Su Xiaoling, Huang Qiaoling. Monthly runoff prediction using wavelet transform and generalized regression neural network model[J]. Journal of Hydroelectric Engineering, 2016, 35(5): 47-54. (in Chinese with English abstract)
[3] 郭生練,郭家力,侯雨坤,等. 基于Budyko假設(shè)預(yù)測(cè)長(zhǎng)江流域未來(lái)徑流量變化[J]. 水科學(xué)進(jìn)展,2015,26(2):151-160.
Guo Shenglian, Guo Jiali, Hou Yukun, et al. Prediction of future runoff change based on Budyko hypothesis in Yangtze River basin[J]. Advances in Water Science, 2015, 26(2): 151-160. (in Chinese with English abstract)
[4] 王子龍,何馨,姜秋香,等. 氣候變化下東北中等流域冬季徑流模擬和預(yù)測(cè)[J]. 水科學(xué)進(jìn)展,2020,31(4):575-582.
Wang Zilong, He Xin, Jiang Qiuxiang, et al. Simulation and forecast of winter runoff in medium basin of Northeast China under climate change[J]. Advances in Water Science, 2020, 31(4): 575-582. (in Chinese with English abstract)
[5] 田晶,郭生練,劉德地,等. 氣候與土地利用變化對(duì)漢江流域徑流的影響[J]. 地理學(xué)報(bào),2020,75(11):2307-2318.
Tian Jing, Guo Shenglian, Liu Dedi, et al. Impacts of climate and land use/cover changes on runoff in the Hanjiang River basin[J]. Acta Geographica Sinica, 2020, 75(11): 2307-2318. (in Chinese with English abstract)
[6] 賈路,任宗萍,李占斌,等. 基于耦合協(xié)調(diào)度的大理河流域徑流和輸沙關(guān)系分析[J]. 農(nóng)業(yè)工程學(xué)報(bào),2020,36(11):86-94.
Jia Lu, Ren Zongping, Li Zhanbin, et al. Relationship between runoff and sediment load in Dali River Basin based on coupling coordination degree[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2020, 36(11): 86-94. (in Chinese with English abstract)
[7] Wu J F, Chen X H, Yao H X, et al. Hydrological drought instantaneous propagation speed based on the variable motion relationship of speed‐time process[J]. Water Resources Research, 2018, 54(11): 9549-9565.
[8] Liu Z Y, Cheng L Y, Hao Z C, et al. A framework for exploring joint effects of conditional factors on compound floods[J]. Water Resources Research, 2018, 54(4): 2681-2696.
[9] Bevacqua E, Maraun D, Hob?k H I, et al. Multivariate statistical modelling of compound events via pair-copula constructions: analysis of floods in Ravenna (Italy)[J]. Hydrology and Earth System Sciences, 2017, 21(6): 2701-2723.
[10] 劉江濤,徐宗學(xué),趙煥,等. 基于改進(jìn)降水輸入模塊的融雪徑流模擬:以拉薩河為例[J]. 水利學(xué)報(bào),2018,49(11):1396-1408.
Liu Jiangtao, Xu Zongxue, Zhao Huan, et al. Simulation of snowmelt runoff processes based on enhanced precipitation input module: Case studies in the Lhasa River basin[J]. Journal of Hydraulic Engineering, 2018, 49(11): 1396-1408. (in Chinese with English abstract)
[11] Zhang B E, Wang S, Wang Y M. Probabilistic projections of multidimensional flood risks at a convection-permitting scale[J]. Water Resources Research, 2021, 57(1): e2020WR028582.
[12] 楊肖麗,李文婷,任立良,等. 渭河流域藍(lán)綠水對(duì)土地利用變化的響應(yīng)模擬[J]. 農(nóng)業(yè)工程學(xué)報(bào),2021,37(11):268-276.
Yang Xiaoli, Li Wenting, Ren Liliang, et al. Simulation of the response of blue and green water to land use change in the Weihe River Basin[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2021, 37(11): 268-276. (in Chinese with English abstract)
[13] Liu Z Y, Cheng L Y, Lin K Y, et al. A hybrid bayesian vine model for water level prediction[J]. Environmental Modelling & Software, 2021, 142: 105075.
[14] 郝增超,侯?lèi)?ài)中,張璇,等. 干旱監(jiān)測(cè)與預(yù)報(bào)研究進(jìn)展與展望[J]. 水利水電技術(shù),2020,51(11):30-40.
Hao Zengchao, Hou Aizhong, Zhang Xuan, et al. Research progresses and prospects of drought monitoring and prediction[J]. Water Resources and Hydropower Engineering, 2020, 51( 11): 30-40. (in Chinese with English abstract)
[15] 胡義明,陳騰,羅序義,等. 基于機(jī)器學(xué)習(xí)模型的淮河流域中長(zhǎng)期徑流預(yù)報(bào)研究[J]. 地學(xué)前緣,2022,29(3):284-291.
Hu Yiming, Chen Teng, Luo Xuyi, et al. Medium to long term runoff forecast for the Huai River Basin based on machine learning algorithm[J]. Earth Science Frontiers, 2022, 29(3): 284-291. (in Chinese with English abstract)
[16] 梁浩,黃生志,孟二浩,等. 基于多種混合模型的徑流預(yù)測(cè)研究[J]. 水利學(xué)報(bào),2020,51(1):112-125.
Liang Hao, Huang Shengzhi, Meng Erhao, et al. Runoff prediction based on multiple hybrid models[J]. Journal of Hydraulic Engineering, 2020, 51(1): 112-125. (in Chinese with English abstract)
[17] 張軒,張行南,王高旭,等. 長(zhǎng)江上游水庫(kù)入庫(kù)流量的中長(zhǎng)期預(yù)報(bào)[J]. 水資源保護(hù),2022,38(4):131-136,165.
Zhang Xuan, Zhang Xingnan, Wang Gaoxu, et al. Medium and long term forecast of reservoir inflow in upper reaches of the Yangtze River[J]. Water Resources Protection, 2022, 38(4): 131-136, 165. (in Chinese with English abstract)
[18] 范宏翔,何菡丹,徐力剛,等. 基于長(zhǎng)短記憶模型的鄱陽(yáng)湖流域徑流模擬及其演變的歸因分析[J]. 湖泊科學(xué),2021,33(3):866-878.
Fan Hongxiang, He Handan, Xu Ligang, et al. Simulation and attribution analysis based on the long-short-term-memory network for detecting the dominant cause of runoff variation in the Lake Poyang Basin[J]. Journal of Lake Sciences, 2021, 33(3): 866-878. (in Chinese with English abstract)
[19] Yaseen Z M, Ebtehaj I, Bonakdari H, et al. Novel approach for streamflow forecasting using a hybrid ANFIS-FFA model[J]. Journal of Hydrology, 2017, 554: 263-276.
[20] 王婕,張建云,鮑振鑫,等. 糧食產(chǎn)量對(duì)氣候變化驅(qū)動(dòng)水資源變化的響應(yīng)[J]. 水科學(xué)進(jìn)展,2021,32(6):855-866.
Wang Jie, Zhang Jianyun, Bao Zhenxin, et al. Response of grain yield to climate change driving water resources change[J]. Advances in Water Science, 2021, 32(6): 855-866. (in Chinese with English abstract)
[21] Wang L L, Li X, Ma C F, et al. Improving the prediction accuracy of monthly streamflow using a data-driven model based on a double-processing strategy[J]. Journal of Hydrology, 2019, 573: 733-745.
[22] 袁喆,楊志勇,史曉亮,等. 灰色微分動(dòng)態(tài)自記憶模型在徑流模擬及預(yù)測(cè)中的應(yīng)用[J]. 水利學(xué)報(bào),2013,44(7):791-799.
Yuan Zhe, Yang Zhiyong, Shi Xiaoliang, et al. Differential hydrological grey self-memory model for runoff simulation and prediction[J]. Journal of Hydraulic Engineering, 2013, 44(7): 791-799. (in Chinese with English abstract)
[23] 李紅芳,王會(huì)肖,趙茹欣,等. 基于Copula函數(shù)的水-能源-糧食共生安全風(fēng)險(xiǎn)概率[J]. 農(nóng)業(yè)工程學(xué)報(bào),2021,37(8):332-340.
Li Hongfang, Wang Huixiao, Zhao Ruxin, et al. Estimating the symbiosis risk probability of water-energy-food using Copula function[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2021, 37(8): 332-340. (in Chinese with English abstract)
[24] 李茉,曹凱華,付強(qiáng),等. 不確定條件下考慮水循環(huán)過(guò)程的灌區(qū)多水源高效配置[J]. 農(nóng)業(yè)工程學(xué)報(bào),2021,37(18):62-73.
Li Mo, Cao Kaihua, Fu Qiang, et al. Efficient allocation of multiple water sources in irrigation areas considering water cycle process under uncertainty[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2021, 37(18): 62-73. (in Chinese with English abstract)
[25] 郭愛(ài)軍,暢建霞,王義民,等. 考慮水庫(kù)來(lái)用水過(guò)程關(guān)聯(lián)性的多維隨機(jī)動(dòng)態(tài)規(guī)劃算法[J]. 農(nóng)業(yè)工程學(xué)報(bào),2022,38(2):139-148.
Guo Aijun, Chang Jianxia, Wang Yimin, et al. Multi-dimensional stochastic dynamic programming algorithms considering the relevancy between streamflow and water demand[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2022, 38(2): 139-148. (in Chinese with English abstract)
[26] Wang W Z, Dong Z C, Lall U, et al. Monthly streamflow simulation for the headwater catchment of the yellow river basin with a hybrid statistical‐dynamical model[J]. Water Resources Research, 2019, 55(9): 7606-7621.
[27] Wu H J, Su X L, Singh V P, et al. Agricultural drought prediction based on conditional distributions of Vine Copulas[J]. Water Resources Research, 2021, 57(8): e2021WR029562.
[28] Sarhadi A, Burn D H, Concepción A M, et al. Time-varying nonstationary multivariate risk analysis using a dynamic Bayesian copula[J]. Water Resources Research, 2016, 52(3): 2327-2349.
[29] Long D, Pan Y, Zhou J, et al. Global analysis of spatiotemporal variability in merged total water storage changes using multiple GRACE products and global hydrological models[J]. Remote Sensing of Environment, 2017, 192: 198-216.
[30] Duan Q Y, Pappenberger F, Wood A, et al. Handbook of Hydrometeorological Ensemble Forecasting[M]. Berlin: Springer, 2019.
[31] 董磊華,熊立華,萬(wàn)民. 基于貝葉斯模型加權(quán)平均方法的水文模型不確定性分析[J]. 水利學(xué)報(bào),2011,42(9):1065-1074.
Dong Leihua, Xiong Lihua, Wan Min. Uncertainty analysis of hydrological modeling using the Bayesian Model Averaging Method[J]. Journal of Hydraulic Engineering, 2011, 42(9): 1065-1074. (in Chinese with English abstract)
[32] 張更喜,粟曉玲,劉文斐. 考慮CO2濃度影響的中國(guó)未來(lái)干旱趨勢(shì)變化[J]. 農(nóng)業(yè)工程學(xué)報(bào),2021,37(1):84-91.
Zhang Gengxi, Su Xiaoling, Liu Wenfei. Future drought trend in China considering CO2concentration[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2021, 37(1): 84-91. (in Chinese with English abstract)
[33] Xu Y, Gao X J, Shen Y, et al. A daily temperature dataset over China and its application in validating a RCM simulation[J]. Advances in Atmospheric Sciences, 2009, 26(4): 763-772.
[34] 吳佳,高學(xué)杰. 一套格點(diǎn)化的中國(guó)區(qū)域逐日觀測(cè)資料及與其它資料的對(duì)比[J]. 地球物理學(xué)報(bào),2013,56(4):1102-1111.
Wu Jia, Gao Xuejie. A gridded daily observation dataset over China region and comparison with the other datasets[J]. Chinese Journal Geophysics, 2013, 56(4): 1102-1111. (in Chinese with English abstract)
[35] Monteith J L. Evaporation and environment[J]. Symposia of the Society for Experimental Biology, 1965, 19: 205-234.
[36] Pourghasemi H R, Yousefi S, Kornejady A, et al. Performance assessment of individual and ensemble data-mining techniques for gully erosion modeling[J]. Science of the Total Environment, 2017, 609: 764-775.
[37] Sklar A. Fonctions de répartition àn dimensions et leurs marges[J]. Publications de l'Institut de Statistique de l'Université de Paris, 1959, 8: 229-231.
[38] Aas K, Czado C, Frigessi A, et al. Pair-copula constructions of multiple dependence[J]. Insurance: Mathematics and Economics, 2009, 44(2): 182-198.
[39] Bevacqua E. CDVineCopulaConditional: Sampling from conditional C- and D-vine copulas, R package: Version 0. 1. 1[R/OL]. CRAN. (2017-07-28) [2021-10-10]. https: //CRAN. R-project. org/package=CDVineCopulaConditional.
[40] Xiong L H, Yu K X, Gottschalk L. Estimation of the distribution of annual runoff from climatic variables using copulas[J]. Water Resources Research, 2014, 50(9): 7134-7152.
[41] Chen X H, Koenker R, Xiao Z J. Copula-based nonlinear quantile autoregression[J]. Econometrics Journal, 2009, 12: 50-67.
[42] 金君良,舒章康,陳敏,等. 基于數(shù)值天氣預(yù)報(bào)產(chǎn)品的氣象水文耦合徑流預(yù)報(bào)[J]. 水科學(xué)進(jìn)展,2019,30(3):316-325.
Jin Junliang, Shu Zhangkang, Chen Min, et al. Meteo-hydrological coupled runoff forecasting based on numerical weather prediction products[J]. Advances in Water Science, 2019, 30(3): 316-325. (in Chinese with English abstract)
[43] Fraley C, Raftery A E, Sloughter J M, et al. EnsembleBMA: Probabilistic Forecasting using Ensembles and Bayesian Model Averaging, R package: Version 5. 1. 8[R/OL]. CRAN, (2022-09-22) [2021-10-10]. https: // CRAN. R-project. org/package=ensembleBMA
[44] 祁繼霞,粟曉玲,張更喜,等. VMD-LSTM模型對(duì)不同預(yù)見(jiàn)期月徑流的預(yù)測(cè)研究[J]. 干旱地區(qū)農(nóng)業(yè)研究,2022,40(6):258-267.
Qi Jixia, Su Xiaoling, Zhang Gengxi, et al. Research on monthly runoff prediction of VMD-LSTM model in different forecast periods[J]. Agricultural Research in the Arid Areas, 2022, 40(6): 258-267. (in Chinese with English abstract)
Prediction and application of monthly streamflow based on Vine Copula coupled Bayesian model averaging
Wu Haijiang1,2, Su Xiaoling1,2※, Qi Jixia2, Zhang Te2, Zhu Xingyu2, Wu Lianzhou2
(1.,,,712100,;2.,,712100,)
Streamflow (channel runoff) is one of the paramount components in the hydrological cycle from the land to waterbodies. Reliable prediction of monthly streamflow in the long lead time is of great significance for the water resource allocation, flood defense, drought mitigation, and ecological environment. The streamflow over time is closely related to precipitation, temperature, potential evapotranspiration, and antecedent streamflow. Fortunately, vine copulas can easily establish the multivariate distribution function by decomposing multidimensional variables into pair copula constructions. And, the Bayesian Model Averaging (BMA) provides outstanding advantages in multi-model ensemble prediction. In this study, a novel streamflow prediction model was proposed to integrate the multiple vine copula models with BMA, (i.e., Bayesian model averaging ensemble Vine Copula (BVC) model). The monthly streamflow predictions of Tangnaihai, Minhe, Hongqi, and Zheqiao hydrological stations in the upstream of Yellow River basin were selected as four cases. The spatial average of precipitation, temperature, and potential evapotranspiration data were calculated across the watershed controlled by each hydrological station. The precipitation, temperature, potential evapotranspiration, and streamflow in each month were firstly fitted with the best marginal distribution functions from the pool of Normal, Gamma, Weibull, and Log-Normal functions. The vine copulas model was leveraged to couple these variables (incorporated four explainable variables and a predicted variable) under five-dimensional situations. The BMA was then employed to combine the streamflow predictions of these candidate vine copula models to reduce the uncertainties caused by distinct variable ordering of individual vine copula model. Finally, the Random Forest (RF) model and the Long Short-Term Memory neural network (LSTM) model were adopted as two reference models. The results show that the best-fitted marginal distributions for precipitation, temperature, potential evapotranspiration, and streamflow were Gamma, Normal, Weibull, and Log-Normal based on the chi-square test, respectively. The minimum coefficient of the determination (2) (Nash-Sutcliffe Efficiency coefficient (NSE)) was all above 0.83 (0.78) and the Root Mean Squared Error (RMSE) was all sustained at a lower level for the 1-3-month lead streamflow predictions using the BVC model during the validation period (1963-2006). Compared with the RF model, the BVC model greatly was captured the variations in the monthly streamflow at these hydrological stations, especially for the extreme streamflow. The prediction performances of BVC and RF models were further evaluated by leveraging the precipitation, temperature, potential evapotranspiration, and streamflow time series over the driest and wettest seasons (corresponding to the average lowest and highest streamflow of three consecutive months during 1963-2006, respectively). Among them, the driest season was found in the January-March period at four hydrological stations; the wettest season was in the July-September period at the Tangnaihai and Hongqi hydrological stations, whereas the Minhe and Zheqiao hydrological stations were found in the August-October period. Similarly, in comparison with the RF model, the BVC model yielded a better performance for streamflow predictions with 1-3-month lead times during the driest and wettest seasons, and the minimum2(NSE) values all exceeded 0.57 (0.61). Moreover, the BVC model also outperformed the RF and LSTM models for the 1-3-month lead times during the validation period (2007-2016), in terms of2, NSE, and RMSE. The findings can provide a theoretical framework for streamflow prediction, and can serve as a guidance for water resources management and risk assessment.
water resource; machine learning; streamflow; prediction; Bayesian model averaging; Vine Copula; the Yellow River basin
10.11975/j.issn.1002-6819.2022.24.008
TV11;P338
A
1002-6819(2022)-24-0073-10
吳海江,粟曉玲,祁繼霞,等. Vine Copula與貝葉斯模型平均結(jié)合的月徑流預(yù)測(cè)及應(yīng)用[J]. 農(nóng)業(yè)工程學(xué)報(bào),2022,38(24):73-82.doi:10.11975/j.issn.1002-6819.2022.24.008 http://www.tcsae.org
Wu Haijiang, Su Xiaoling, Qi Jixia, et al. Prediction and application of monthly streamflow based on Vine Copula coupled Bayesian model averaging[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2022, 38(24): 73-82. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2022.24.008 http://www.tcsae.org
2022-07-14
2022-10-10
國(guó)家自然科學(xué)基金項(xiàng)目(51879222,52079111)
吳海江,博士生,研究方向?yàn)楦珊殿A(yù)測(cè)及高溫干旱復(fù)合事件風(fēng)險(xiǎn)評(píng)估。Email:haijiangwu@nwafu.edu.cn
粟曉玲,教授,研究方向?yàn)樗哪M。Email:xiaolingsu@nwafu.edu.cn