趙君 劉雨 徐進(jìn)超 王國(guó)慶 邵月紅 楊林
摘要:目前的降水產(chǎn)品依然存在較大的不確定性,采用多源降水?dāng)?shù)據(jù)融合可以更準(zhǔn)確地估計(jì)降水量和空間分布情況。為實(shí)現(xiàn)無資料地區(qū)的數(shù)據(jù)融合,本文在不使用任何先驗(yàn)信息的前提下,通過整合站點(diǎn)插值、衛(wèi)星遙感和再分析的降水產(chǎn)品,基于貝葉斯三角帽(Bayesian-Three Cornered Hat,BTCH)法,融合多源降水?dāng)?shù)據(jù),探究不同輸入數(shù)量的降水產(chǎn)品對(duì)于融合數(shù)據(jù)精度的影響以及每個(gè)降水產(chǎn)品對(duì)于融合數(shù)據(jù)精度的貢獻(xiàn)率,并在黃河源區(qū)進(jìn)行應(yīng)用。結(jié)果表明:在月尺度上,融合數(shù)據(jù)性能優(yōu)于原始降水產(chǎn)品;在日尺度上,融合數(shù)據(jù)性能明顯高于衛(wèi)星遙感和再分析降水產(chǎn)品,但低于基于站點(diǎn)的降水產(chǎn)品CHM_PRE;2套基于站點(diǎn)的降水產(chǎn)品CN05.1和CHM_PRE對(duì)于融合數(shù)據(jù)有最大的貢獻(xiàn)率。在黃河源區(qū)的應(yīng)用表明,該數(shù)據(jù)融合方法確實(shí)能夠更準(zhǔn)確地估計(jì)降水量,可應(yīng)用于無實(shí)測(cè)降水資料地區(qū),為數(shù)據(jù)融合分析及應(yīng)用提供參考。
關(guān)鍵詞:多源降水;數(shù)據(jù)融合;不確定性分析;貝葉斯三角帽
中圖分類號(hào):P426.6
文獻(xiàn)標(biāo)志碼:A
文章編號(hào):1001-6791(2023)05-0685-12
降水是陸地水文循環(huán)的主要驅(qū)動(dòng)因素。研究表明,徑流預(yù)報(bào)的誤差主要由降水?dāng)?shù)據(jù)的偏差主導(dǎo)[1],因此量化評(píng)估和減少這種誤差,可以提高對(duì)水文系統(tǒng)和模型模擬的理解[2]。目前,地面站點(diǎn)、衛(wèi)星遙感和氣象雷達(dá)的觀測(cè)數(shù)據(jù)都存在一定的不確定性[3-5]。為了解決降水觀測(cè)數(shù)據(jù)存在的不確定性,通常采用多源數(shù)據(jù)融合來提高對(duì)降水時(shí)空分布的估計(jì)[6]?,F(xiàn)在已有大量關(guān)于數(shù)據(jù)融合的算法,如最優(yōu)插值、卡爾曼濾波、概率密度函數(shù)最優(yōu)插值等[7-9]。然而,這些方法大多需要地面實(shí)測(cè)數(shù)據(jù),對(duì)于數(shù)據(jù)匱乏或無資料地區(qū),基于機(jī)器學(xué)習(xí)的遷移學(xué)習(xí)[10]以及基于TC(Triple Collocation)算法[11]的多源降水?dāng)?shù)據(jù)融合都以得到應(yīng)用,但關(guān)于無資料地區(qū)的數(shù)據(jù)融合研究依然相對(duì)較少。
針對(duì)缺乏實(shí)測(cè)數(shù)據(jù)的區(qū)域,三角帽(Three Cornered Hat,TCH)法已被用于量化降水、GRACE、土壤濕度和蒸散發(fā)等[12-15]在區(qū)域或全球尺度上的不確定性。這為無資料地區(qū)的數(shù)據(jù)融合提供了思路,即利用三角帽法計(jì)算數(shù)據(jù)集的不確定性,并根據(jù)不確定性的大小為每個(gè)數(shù)據(jù)集分配權(quán)重,從而實(shí)現(xiàn)無實(shí)測(cè)數(shù)據(jù)地區(qū)的數(shù)據(jù)融合[16-17]。Xu等[18]利用廣義三角帽分析了13套月降水?dāng)?shù)據(jù)集和11套日降水?dāng)?shù)據(jù)集在全球尺度上的相對(duì)不確定性,并根據(jù)各降水產(chǎn)品的不確定性通過加權(quán)進(jìn)行多源降水?dāng)?shù)據(jù)融合,結(jié)果表明基于廣義三角帽法的多源數(shù)據(jù)融合要優(yōu)于其他方法;He等[19]基于一種貝葉斯三角帽(Bayesian-Three Cornered Hat,BTCH)方法,通過整合多源地表蒸散發(fā)產(chǎn)品來提高地表蒸散發(fā)(ET)的估計(jì),結(jié)果表明BTCH方法能夠有效地減少ET產(chǎn)品之間的差異,并提高ET估計(jì)的精度和穩(wěn)定性。這些研究主要集中在融合算法的發(fā)展方面,但融合數(shù)據(jù)的精度不僅被算法影響,同時(shí)也受到輸入源的影響[20]。全球范圍內(nèi)降水產(chǎn)品種類繁多,如何選擇降水產(chǎn)品進(jìn)行數(shù)據(jù)融合是值得考慮的問題。因此,在融合數(shù)據(jù)之前,分析并選擇不同數(shù)量和類型的降水產(chǎn)品對(duì)融合數(shù)據(jù)精度也是非常重要的[21]。如何更加合理地構(gòu)建多源降水融合框架有待深入研究。
本文基于貝葉斯三角帽方法融合多源降水?dāng)?shù)據(jù),采用2001—2020年中國(guó)大陸地區(qū)8套基于站點(diǎn)、衛(wèi)星遙感和再分析原始降水產(chǎn)品,以實(shí)測(cè)站點(diǎn)數(shù)據(jù)作為參照,定量分析不同輸入數(shù)量下融合數(shù)據(jù)的精度差異以及每個(gè)降水產(chǎn)品對(duì)于融合數(shù)據(jù)精度的貢獻(xiàn)率。
1 研究區(qū)域與數(shù)據(jù)
1.1 研究區(qū)概況
第5期趙君,等:基于貝葉斯三角帽法的多源降水?dāng)?shù)據(jù)融合分析及應(yīng)用
水科學(xué)進(jìn)展第34卷
本次研究選取中國(guó)大陸作為研究區(qū)域,選取了834個(gè)地面氣象站點(diǎn),站點(diǎn)分布如圖1(a)所示。研究區(qū)包括熱帶、亞熱帶、溫帶、亞寒帶等多種氣候類型。
為驗(yàn)證融合降水?dāng)?shù)據(jù)的效果,選取黃河源區(qū)作為驗(yàn)證區(qū)。黃河源區(qū)地處青藏高原東部邊緣,目前區(qū)域內(nèi)國(guó)家設(shè)立的氣象站點(diǎn)僅有12個(gè),數(shù)量嚴(yán)重不足且已有站點(diǎn)大部分分布在河谷地帶,在空間分布上不具有代表性,屬于典型的資料匱乏地區(qū),數(shù)據(jù)的缺失嚴(yán)重制約了黃河源區(qū)的水文預(yù)報(bào)精度[22]。黃河源區(qū)的地理位置分布如圖1(b)所示。
1.2 研究數(shù)據(jù)
本文中的實(shí)測(cè)降水資料來源于國(guó)家氣象信息中心(CMA),所選用的降水產(chǎn)品包括CMORPH(Climate Prediction Center MORPHing technique)數(shù)據(jù)集[23]、中國(guó)科學(xué)院氣候變化研究中心CN05格點(diǎn)化觀測(cè)數(shù)據(jù)集[24]、PERSIANN(The Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks)數(shù)據(jù)集[25]、國(guó)家青藏高原科學(xué)數(shù)據(jù)中心CHM_PRE數(shù)據(jù)集[26]、歐洲中期天氣預(yù)報(bào)中心ERA5-Land降水?dāng)?shù)據(jù)集[27]、日本宇宙航空研究開發(fā)機(jī)構(gòu)(JAXA)GSMaP(Global Satellite Mapping of Precipitation)數(shù)據(jù)集[28]、IMERG(Integrated Multi-satellitE Retrievals for GPM)數(shù)據(jù)集[29]以及CHIRPS(Rainfall Estimates from Rain Gauge and Satellite Observations)數(shù)據(jù)集[30]。數(shù)據(jù)詳細(xì)信息如表1所示。選取2001—2020年作為研究時(shí)段,采用雙線性插值法(Bilinear)將所有降水產(chǎn)品的空間分辨率統(tǒng)一為0.25°,時(shí)間分辨率統(tǒng)一為1 d。
2 研究方法
2.1 基于貝葉斯理論的三角帽方法
2.2 評(píng)估方法與統(tǒng)計(jì)指標(biāo)
3 結(jié)果與分析
3.1 降水產(chǎn)品不確定性分析
圖2為2001—2020年使用TCH法計(jì)算的單個(gè)降水產(chǎn)品不確定性的箱線圖。其中,CMORPH、ERA5-Land和GSMaP相對(duì)不確定性的中位數(shù)分別為20.6、18.3和26.1 mm,表明它們的不確定性比其他降水產(chǎn)品高得多;CHIRPS和PERSIANN不確定性的中位數(shù)分別為14.3和13.3 mm,表明這兩者的不確定性相對(duì)較低;而CN05.1、CHM_PRE和IMERG不確定性的中位數(shù)分別為9.3、10.0和10.1 mm,表明這3套降水產(chǎn)品擁有最小的相對(duì)不確定性。
圖3顯示了2001—2020年使用TCH法計(jì)算的不確定性空間分布圖??梢钥闯觯?套降水產(chǎn)品的不確定性呈現(xiàn)從東南沿海向西北內(nèi)陸遞減的趨勢(shì),這與中國(guó)的降水分布類似,因此降水多的地區(qū)通常存在較大的不確定性。
特別地,IMERG、CHM_PRE和CN05.1在全國(guó)范圍內(nèi)具有較小的不確定性;此外,ERA5-Land、CHIRPS和GSMaP在南方地區(qū)和青藏高原的部分地區(qū)存在較大的不確定性。
3.2 月尺度融合數(shù)據(jù)精度對(duì)比
為了探究輸入降水產(chǎn)品數(shù)量對(duì)融合數(shù)據(jù)精度的影響,按照?qǐng)D2中各降水產(chǎn)品不確定性從小到大的順序,將不確定性最小的前3個(gè)降水產(chǎn)品組成融合數(shù)據(jù)BTCH3(融合了CN05.1、CHM_PRE、IMERG),以此類推,不確定性最小的前4個(gè)降水產(chǎn)品組成BTCH4,直至前8個(gè)產(chǎn)品組成BTCH8。基于貝葉斯三角帽方法分別計(jì)算BTCH3—BTCH8相應(yīng)的融合數(shù)據(jù),每個(gè)分組中,不同降水產(chǎn)品在每個(gè)格點(diǎn)上的平均權(quán)重如圖4所示??梢钥闯?,不確定性最小的CHM_PRE和CN05.1在每個(gè)分組中都擁有最大的權(quán)重,隨著降水產(chǎn)品數(shù)量的增加,每種降水產(chǎn)品的權(quán)重都不同程度的下降。這會(huì)“稀釋”精度高的降水產(chǎn)品在融合數(shù)據(jù)中的分量,或是隨著精度不高的降水產(chǎn)品的增多“拉低”融合數(shù)據(jù)的整體精度。使用EKG、CC、ERMS和RB定量評(píng)估不同分組融合數(shù)據(jù)的精度也證實(shí)了這一點(diǎn)。圖5展示了融合數(shù)據(jù)隨輸入數(shù)據(jù)集數(shù)量增加各評(píng)價(jià)指標(biāo)中位數(shù)的變化情況。BTCH3在各項(xiàng)指標(biāo)的綜合表現(xiàn)優(yōu)于其他組合,其EKG(0.859)、CC(0.955)最高,ERMS(21.718 mm)最小。因此,對(duì)于多源數(shù)據(jù)融合單純?cè)黾訑?shù)據(jù)集的數(shù)量可能并不會(huì)提升融合數(shù)據(jù)的精度。
為了更直觀地展示評(píng)價(jià)指標(biāo)的空間分布情況,圖7展示了融合數(shù)據(jù)BTCH3和參與融合的原始降水產(chǎn)品在月尺度上的精度評(píng)價(jià)指標(biāo)空間分布圖。由圖7(a)、7(d)、7(g)、7(j)可知,融合數(shù)據(jù)和各原始降水產(chǎn)品反映的中國(guó)大陸地區(qū)相關(guān)系數(shù)分布格局總體上相似,即東部季風(fēng)區(qū)相關(guān)系數(shù)較高(0.91~0.99),西北地區(qū)和青藏高原地區(qū)相關(guān)系數(shù)偏低(0.43~0.72),融合數(shù)據(jù)和CHM_PRE在總體精度和空間分布類似,CN05.1表現(xiàn)弱于兩者,IMERG在總體精度上表現(xiàn)最差。圖7(b)、7(e)、7(h)、7(k)顯示ERMS沿西北地區(qū)向東南沿海遞增。圖7(c)、7(f)、7(i)、7(l)表明融合數(shù)據(jù)和各降水產(chǎn)品在東部季風(fēng)區(qū)的部分區(qū)域存在低估降水的情況,但大部分地區(qū)都不同程度地高估了降水。
3.3 日尺度融合數(shù)據(jù)的精度對(duì)比
采用與融合月尺度降水相同的方法,將8套不同降水產(chǎn)品按照不同組合融合為一個(gè)新的日降水?dāng)?shù)據(jù)。圖8為不同分組下各評(píng)價(jià)指標(biāo)的中位數(shù),結(jié)果顯示融合數(shù)據(jù)在日尺度上的表現(xiàn)和月尺度上相似。其中,最佳融合數(shù)據(jù)由不確定性最小的3套降水產(chǎn)品(BTCH3)組成。一個(gè)有趣的現(xiàn)象是,日尺度融合數(shù)據(jù)受降水產(chǎn)品數(shù)量的影響更顯著。在日尺度上,融合數(shù)據(jù)的修正Kling-Gupta效率系數(shù)隨降水產(chǎn)品數(shù)量的增加下降了25.6%(波動(dòng)范圍為0.497~0.668),而在月尺度上下降了5.1%(波動(dòng)范圍為0.815~0.859)。換句話說,在日尺度上,融合數(shù)據(jù)對(duì)于輸入的降水產(chǎn)品數(shù)量更加敏感。
表2為精度最高的融合數(shù)據(jù)(BTCH3)和原始降水產(chǎn)品在日尺度上各種評(píng)價(jià)指標(biāo)的中位數(shù)。在參與數(shù)據(jù)融合的原始降水產(chǎn)品中,CHM_PRE在各項(xiàng)評(píng)價(jià)指標(biāo)上均明顯優(yōu)于其他原始降水產(chǎn)品和融合數(shù)據(jù)。IMERG的精度較低,在所有降水產(chǎn)品和融合數(shù)據(jù)中表現(xiàn)最差。盡管融合數(shù)據(jù)的精度明顯優(yōu)于IMERG和CN05.1,但總體精度不及CHM_PRE。造成融合數(shù)據(jù)在日尺度上的融合效果不如月尺度的原因,一方面是因?yàn)镃HM_PRE是基于中國(guó)境內(nèi)及周邊共2 839個(gè)雨量站點(diǎn),利用月值降水約束和地形特征矯正得到的數(shù)據(jù)集[26],其數(shù)據(jù)精度足夠高;另一方面,隨著時(shí)間尺度減小,降水序列中周期性的成分不斷降低,隨機(jī)性成分和背景噪聲的增加導(dǎo)致融合算法難以獲取真正的降水信息,這也導(dǎo)致在日尺度上融合數(shù)據(jù)的精度對(duì)于不同輸入數(shù)量的降水產(chǎn)品更加敏感。
3.4 降水產(chǎn)品貢獻(xiàn)率的定量評(píng)估
為了定量評(píng)估不同降水產(chǎn)品對(duì)于融合數(shù)據(jù)精度的影響,采用式(12)來計(jì)算從BTCH4至BTCH3和BTCH5至BTCH4等變化情況下每個(gè)降水產(chǎn)品的相對(duì)貢獻(xiàn)率。表3總結(jié)了不同輸入下各降水產(chǎn)品的相對(duì)貢獻(xiàn)率。例如,去除BTCH4中的PERSIANN(即由BTCH4變?yōu)锽TCH3)導(dǎo)致融合數(shù)據(jù)的精度(EKG)由0.581變?yōu)?.668,因此PERSIANN對(duì)融合數(shù)據(jù)的相對(duì)貢獻(xiàn)率為-14.896%;同樣,去除BTCH4中的CHM_PRE使得融合數(shù)據(jù)的精度由0.581變?yōu)?.478,因此CHM_PRE對(duì)于BTCH4精度的相對(duì)貢獻(xiàn)率為17.73%,其他數(shù)據(jù)依次類推。
通過表3可以看出,2套基于站點(diǎn)的降水產(chǎn)品(CHM_PRE和CN05.1)對(duì)于融合數(shù)據(jù)的貢獻(xiàn)率最大,而衛(wèi)星遙感和再分析降水產(chǎn)品對(duì)于融合數(shù)據(jù)精度的相對(duì)貢獻(xiàn)率基本為負(fù)。隨著降水產(chǎn)品數(shù)量的增加,CHM_PRE和CN05.1的相對(duì)貢獻(xiàn)率不斷降低,這與3.2節(jié)中的增加過多的降水產(chǎn)品會(huì)“稀釋”融合數(shù)據(jù)精度的結(jié)論相符。Wei等[21]關(guān)于數(shù)據(jù)融合的研究也表明基于站點(diǎn)的CPC(Climate Prediction Center)數(shù)據(jù)集對(duì)于融合數(shù)據(jù)精度的影響最大,在貝葉斯模型平均(BMA)中增加CPC數(shù)據(jù)集顯著提高了融合數(shù)據(jù)的精度。因此,數(shù)據(jù)融合過程中的數(shù)據(jù)集選擇至關(guān)重要。
3.5 融合數(shù)據(jù)的適用性分析
為了驗(yàn)證基于貝葉斯三角帽法的融合數(shù)據(jù)在資料匱乏地區(qū)的適用性,選取黃河源區(qū)內(nèi)的瑪多、興海、河南、達(dá)日和若爾蓋5個(gè)典型氣象站點(diǎn)進(jìn)行驗(yàn)證。由于該地區(qū)的地面降水觀測(cè)數(shù)據(jù)匱乏,嚴(yán)重制約了該區(qū)域的水文預(yù)報(bào)精度。本文使用ENS來驗(yàn)證融合數(shù)據(jù)BTCH3在相應(yīng)站點(diǎn)處的模擬精度,并與精度最高的降水產(chǎn)品CHM_PRE進(jìn)行對(duì)比。結(jié)果如圖9所示,基于貝葉斯三角帽法的融合數(shù)據(jù)在各個(gè)站點(diǎn)上的ENS均優(yōu)于CHM_PRE。這表明基于貝葉斯三角帽法的數(shù)據(jù)融合方法確實(shí)可以更準(zhǔn)確地估計(jì)降水量,適用于資料匱乏的地區(qū)。
4 結(jié)論
本文基于貝葉斯三角帽法,使用了8套不同的降水產(chǎn)品,包括CMORPH、CN05.1、PERSIANN、CHM_PRE、ERA5-Land、GSMaP、IMERG和CHIRPS,探究了不同輸入數(shù)量的降水產(chǎn)品對(duì)于融合數(shù)據(jù)精度的影響和各降水產(chǎn)品的相對(duì)貢獻(xiàn)率,并驗(yàn)證了融合數(shù)據(jù)在資料匱乏的黃河源區(qū)的適用性。主要結(jié)論如下:
(1) 在8套不同的降水產(chǎn)品中基于站點(diǎn)的CN05.1和CHM_PRE以及基于衛(wèi)星遙感的IMERG相較于其他降水產(chǎn)品擁有較小的不確定性。各降水產(chǎn)品的不確定性存在明顯的空間分布差異,基本呈現(xiàn)出從東南沿海向西北內(nèi)陸遞減的趨勢(shì)。
(2) 使用過多的降水產(chǎn)品會(huì)降低融合數(shù)據(jù)的精度,本文中融合數(shù)據(jù)的修正Kling-Gupta效率系數(shù)隨著降水產(chǎn)品數(shù)量的增加逐漸降低,在月尺度上降低了5.1%,在日尺度上降低了25.6%。精度最高的融合數(shù)據(jù)是由CN05.1、CHM_PRE和IMERG這3套不確定性最小的降水產(chǎn)品組成的。
(3) 各降水產(chǎn)品中CHM_PRE和CN05.1對(duì)于融合數(shù)據(jù)有最大的相對(duì)貢獻(xiàn)率。黃河源區(qū)的適用性分析表明,基于貝葉斯三角帽法的數(shù)據(jù)融合方法可以更準(zhǔn)確地估計(jì)降水量。
雖然基于貝葉斯三角帽法的數(shù)據(jù)融合方法在無資料或數(shù)據(jù)匱乏的地區(qū)得到成功應(yīng)用,但本研究依然存在一些不確定因素和限制。例如,在本文中,貝葉斯三角帽法在日尺度上的融合效果不如月尺度融合效果好,該數(shù)據(jù)融合模型還需進(jìn)一步優(yōu)化。此外,有很多因素可能會(huì)限制三角帽法的準(zhǔn)確性,如數(shù)據(jù)集中樣本的數(shù)量和異常值、數(shù)據(jù)集的真實(shí)偏差以及未知誤差的相關(guān)性,相關(guān)問題還需要進(jìn)一步研究。
參考文獻(xiàn):
[1]YATHEENDRADAS S,WAGENER T,GUPTA H,et al.Understanding uncertainty in distributed flash flood forecasting for semiarid regions[J].Water Resources Research,2008,44(5):w05S19.
[2]REICHERT P,MIELEITNER J.Analyzing input and structural uncertainty of nonlinear dynamic models with stochastic,time-dependent parameters[J].Water Resources Research,2009,45(10):w10402.
[3]POLLOCK M D,O′DONNELL G,QUINN P,et al.Quantifying and mitigating wind-induced undercatch in rainfall measurements[J].Water Resources Research,2018,54(6):3863-3875.
[4]MCMILLAN H,KRUEGER T,F(xiàn)REER J.Benchmarking observational uncertainties for hydrology:rainfall,river discharge and water quality[J].Hydrological Processes,2012,26(26):4078-4111.
[5]TAN X H,YONG B,REN L L.Error features of the hourly GSMaP multi-satellite precipitation estimates over nine major basins of China[J].Hydrology Research,2018,49(3):761-779.
[6]BECK H E,WOOD E F,PAN M,et al.MSWEP V2 global 3-hourly 0.1° precipitation:methodology and quantitative assessment[J].Bulletin of the American Meteorological Society,2019,100(3):473-500.
[7]SAPIANO M R P,SMITH T M,ARKIN P A.A new merged analysis of precipitation utilizing satellite and reanalysis data[J].Journal of Geophysical Research,2008,113(D22):D22103.
[8]JOYCE R J,XIE P P.Kalman filter-based CMORPH[J].Journal of Hydrometeorology,2011,12(6):1547-1563.
[9]SHEN Y,ZHAO P,PAN Y,et al.A high spatiotemporal gauge-satellite merged precipitation analysis over China[J].Journal of Geophysical Research:Atmospheres,2014,119(6):3063-3075.
[10]LIU Z Y,YANG Q L,SHAO J M,et al.Improving daily precipitation estimation in the data scarce area by merging rain gauge and TRMM data with a transfer learning framework[J].Journal of Hydrology,2022,613:128455.
[11]CHEN C,HE M N,CHEN Q W,et al.Triple collocation-based error estimation and data fusion of global gridded precipitation products over the Yangtze River basin[J].Journal of Hydrology,2022,605:127307.
[12]LIU Y,ZHENG Y J,LI W P,et al.Evaluating the performance of satellite-based precipitation products using gauge measurement and hydrological modeling:a case study in a dry basin of Northwest China[J].Journal of Hydrometeorology,2022,23(4):541-559.
[13]YAN X,ZHANG B,YAO Y B,et al.GRACE and land surface models reveal severe drought in Eastern China in 2019[J].Journal of Hydrology,2021,601:126640.
[14]LIU J,CHAI L N,DONG J Z,et al.Uncertainty analysis of eleven multisource soil moisture products in the third pole environment based on the three-corned hat method[J].Remote Sensing of Environment,2021,255:112225.
[15]XU T R,GUO Z X,XIA Y L,et al.Evaluation of twelve evapotranspiration products from machine learning,remote sensing and land surface models over conterminous United States[J].Journal of Hydrology,2019,578:124105.
[16]SHANGGUAN Y L,MIN X X,SHI Z.Inter-comparison and integration of different soil moisture downscaling methods over the Qinghai-Tibet Plateau[J].Journal of Hydrology,2023,617:129014.
[17]SHAO X M,ZHANG Y Q,LIU C M,et al.Can indirect evaluation methods and their fusion products reduce uncertainty in actual evapotranspiration estimates?[J].Water Resources Research,2022,58(6):e2021WR031069.
[18]XU L,CHEN N C,MORADKHANI H,et al.Improving global monthly and daily precipitation estimation by fusing gauge observations,remote sensing,and reanalysis data sets[J].Water Resources Research,2020,56(3):e2019WR026444.
[19]HE X L,XU T R,XIA Y L,et al.A Bayesian three-cornered hat (BTCH) method:improving the terrestrial evapotranspiration estimation[J].Remote Sensing,2020,12(5):878.
[20]TANG G Q,CLARK M P,PAPALEXIOU S M,et al.Have satellite precipitation products improved over last two decades? A comprehensive comparison of GPM IMERG with nine satellite and reanalysis datasets[J].Remote Sensing of Environment,2020,240:111697.
[21]WEI L Y,JIANG S H,DONG J Z,et al.Fusion of gauge-based,reanalysis,and satellite precipitation products using Bayesian model averaging approach:determination of the influence of different input sources[J].Journal of Hydrology,2023,618:129234.
[22]雍斌,張建云,王國(guó)慶.黃河源區(qū)水文預(yù)報(bào)的關(guān)鍵科學(xué)問題[J].水科學(xué)進(jìn)展,2023,34(2):159-171.(YONG B,ZHANG J Y,WANG G Q.Key scientific issues of hydrological forecast in the headwater area of Yellow River[J].Advances in Water Science,2023,34(2):159-171.(in Chinese))
[23]JOYCE R J,JANOWIAK J E,ARKIN P A,et al.CMORPH:a method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution[J].Journal of Hydrometeorology,2004,5(3):487-503.
[24]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 J,GAO X J.A gridded daily observation dataset over China region and comparison with the other datasets[J].Chinese Journal of Geophysics,2013,56(4):1102-1111.(in Chinese))
[25]ASHOURI H,HSU K L,SOROOSHIAN S,et al.PERSIANN-CDR:daily precipitation climate data record from multisatellite observations for hydrological and climate studies[J].Bulletin of the American Meteorological Society,2015,96(1):69-83.
[26]HAN J Y,MIAO C Y,GOU J J,et al.A new daily gridded precipitation dataset for the Chinese mainland based on gauge observations[J].Earth System Science Data,2023,15(7):3147-3161.
[27]MUOZ-SABATER J,DUTRA E,AGUST-PANAREDA A,et al.ERA5-Land:a state-of-the-art global reanalysis dataset for land applications[J].Earth System Science Data,2021,13(9):4349-4383.
[28]KUBOTA T,SHIGE S,HASHIZUME H,et al.Global precipitation map using satellite-borne microwave radiometers by the GSMaP project:production and validation[J].IEEE Transactions on Geoscience and Remote Sensing,2007,45(7):2259-2275.
[29]HOU A Y,KAKAR R K,NEECK S,et al.The global precipitation measurement mission[J].Bulletin of the American Meteorological Society,2014,95(5):701-722.
[30]FUNK C,PETERSON P,LANDSFELD M,et al.The climate hazards infrared precipitation with stations:a new environmental record for monitoring extremes[J].Scientific Data,2015,2:150066.
[31]TAVELLA P,PREMOLI A.Estimating the instabilities of N Clocks by measuring differences of their readings[J].Metrologia,1994,30(5):479-486.
[32]GALINDO F J,PALACIO J.Estimating the instabilities of N correlated clocks[C]∥Proceedings of the 31th Annual Precise Time and Time Interval Systems and Applications Meeting.Dana Point,California:Institute of Navigation,1999:285-296.
Multi-source precipitation data fusion analysis and application based on
Bayesian-Three Cornered Hat method
The study is financially supported by the National Key R&D Program of China (No.2021YFC3201101) and Key R&D Project of Jiangsu Province,China (No.BE2020633).
ZHAO Jun1,2,LIU Yu1,XU Jinchao1,2,WANG Guoqing2,SHAO Yuehong1,YANG Lin1
(1. School of Hydrology and Water Resources,Nanjing University of Information Science & Technology,Nanjing 210044,China;
2. The National Key Laboratory of Water Disaster Prevention,Nanjing Hydraulic Research Institute,Nanjing 210029,China)
Abstract:At present,precipitation products still have great uncertainty.Precipitation and its spatial distribution can be estimated more accurately by using multi-source precipitation data fusion.To achieve data fusion in no-gauged areas,Bayesian-Three Cornered Hat method is adopted to integrate precipitation products based on gauged data,satellite remote sensing and reanalysis data without any prior information,to explore the influence of precipitation products with different input quantities on the accuracy of fusion data,and to study the contribution rates of each precipitation product to the accuracy of fusion data.It is applied in the source region of the Yellow River.The results show that the performance of the fusion data is better than that of the original precipitation products on the monthly scale.On the daily scale,the performance of the fusion data is obviously better than that of satellite remote sensing and reanalysis precipitation products,but lower than that of the gauge-based precipitation product CHM_PRE.Two gauge-based precipitation products,CN05.1 and CHM_PRE,have the largest contribution rates to the fusion data.The application in the source region of the Yellow River shows that the Bayesian-Three Cornered Hat method can estimate precipitation more accurately.It is suitable for no-gauged areas,and can provide the reference basis for data fusion analysis and its application.
Key words:multi-source precipitation;data fusion;uncertainty analysis;Bayesian-Three Cornered Hat