積雪數(shù)據(jù)同化方法

1、積雪數(shù)據(jù)同化方法及其研究進(jìn)展黃春林寒區(qū)旱區(qū)環(huán)境與工程遙感與地理信息科學(xué)研究室2014年11月21日 蘭州中國(guó) 提綱1. 數(shù)據(jù)同化2. 積雪數(shù)據(jù)同化方法3. 積雪數(shù)據(jù)同化研究進(jìn)展üü同化微波數(shù)據(jù)同化積雪面積數(shù)據(jù)4. 結(jié)論21.數(shù)據(jù)同化定義數(shù)據(jù)同化是在考慮數(shù)據(jù)時(shí)空分布以及觀測(cè)場(chǎng)和背景場(chǎng)誤差的基礎(chǔ)上，在數(shù)值模型的動(dòng)態(tài)運(yùn)行過程中融合新的觀測(cè)數(shù)據(jù)的方法（根據(jù) Talagrand, 1997）。陸面數(shù)據(jù)同化的思想是在陸面過程模型的動(dòng)力框架內(nèi)，融合不同來源和不同分辨率的直接與間接觀測(cè)，將陸面過程模型和各種觀測(cè)算子（如輻射傳輸模型）集成為不斷地依靠觀測(cè)而自動(dòng)調(diào)整模型軌跡，并且減小誤差的預(yù)報(bào)

2、系統(tǒng)（李新等，2007）。模型觀測(cè)31.數(shù)據(jù)同化數(shù)學(xué)表達(dá)4最優(yōu)估計(jì) Xa (t )k觀測(cè)方程 Yo (t ) = H (X f (t ) +e(t )Y ÎÂm´1e ÎÂm´1kkkk觀測(cè)算子 觀測(cè)誤差狀態(tài)方程 X f (t ) = M (Xa (t) +h(t)XÎÂn´1 h ÎÂn´1kkk -1k -1模型算子模型誤差1.數(shù)據(jù)同化數(shù)學(xué)表達(dá)變分算法連續(xù)方法濾波算法順序方法5Xa (t ) = X f (t ) + K Yo (t ) - H X f (t )kkkkk

3、kPa (t ) = (I - K H )P f (t )kkkkK= P f (t )HT (H P f (t )HT + R(t )-1kkkkkkk模型誤差J (b )T-1(x - x b ) 2000觀測(cè)算子n 觀測(cè)+ 1 å(M (x ) -)T-1(H (x ) - yO )20ii0iii=0模型算子觀測(cè)誤差RiyO iHi MB1.數(shù)據(jù)同化算法數(shù)據(jù)同化方法順序方法經(jīng)驗(yàn)方法優(yōu)化插值連續(xù)方法Monte Carlo全局優(yōu)化方法Kalman濾波變分方法多項(xiàng)式插值等價(jià)擴(kuò)展Kalman模擬退火法濾波3DVar4DVar集合Kalman 濾波連續(xù)修遺傳算法粒子濾波洗牌復(fù)形算法Ba

4、yes濾波等價(jià)松弛法李新, 擺玉龍等, 2010. 順序數(shù)據(jù)同化的Bayes框架. 地球科學(xué)進(jìn)展, 25(5): 515-523.62.積雪數(shù)據(jù)同化方法2.1遙感數(shù)據(jù)源7微波：雪深和SWE光學(xué)遙感：積雪面積2.積雪數(shù)據(jù)同化方法2.2積雪面積同化1. 直接遙感反演的積雪面積（Rl & Houser，2004，JH）(1) 增加雪層：模型沒有積雪，MODIS積雪面積大于40%，增加5mm雪水當(dāng)量；(2) 刪除雪層：模型有積雪， MODIS積雪面積小于10%，刪除雪層；2. 利用 “積雪面積-雪深”衰減曲線作為觀測(cè)算子來約束積雪模型模擬的雪深/雪水當(dāng)量hsno(W)SCA = tanh(A=

5、 A2.5z (r/ r）afdcmin(W, SI )0snonewmaxSu., 2008, JGRAndreadis & Lettenmaier, 2006，82.積雪數(shù)據(jù)同化方法2.3微波亮度溫度數(shù)據(jù)同化積雪表面微波輻射亮度溫度的正向模擬Durand & Margulis, 2008, JGR; Durand, 2009, GRL92.積雪數(shù)據(jù)同化方法同化微波亮溫Durand., 2009, GRLChe., 2014, RSE103.積雪數(shù)據(jù)同化研究進(jìn)展3.1同化微波亮溫?cái)?shù)據(jù)積雪粒徑模型和分層策略對(duì)積雪輻射數(shù)據(jù)同化的影響： 正向模擬評(píng)價(jià)同化主微波遙感數(shù)據(jù)改進(jìn)積雪和土壤

6、凍融模擬的可行性分析3.2同化積雪面積實(shí)驗(yàn)（北疆）113.1同化微波亮溫?cái)?shù)據(jù)積雪粒徑模型和分層策略對(duì)積雪輻射數(shù)據(jù)同化的影響：正向模擬評(píng)價(jià)Huang, C., S. Margulis, K. Musselman, M. Durand, 2012. Assessment of grain-size mand stratigraphyrepresentation impacts on snow radiance assimilation: Forward Ming Evaluation, IEEE Transactions onGeoscience & Remote Sensing, 50(

7、11), 4551-4564復(fù)雜程度不同的積雪粒徑增長(zhǎng)模型如何影響亮度溫度？不同模型的積雪分層策略如何影響亮度溫度？13Jordan m（J91）dDg = g1 D(1000)(T)6 C¶T dtDDe0sP273.15KT ¶zooaMelt metamorphismìg (q + 0.05)ï2v0 < q< 0.09dDïDv g = ígdtïg2 ´ 0.14q³ 0.09ïDvîgg1UvSNOWPACK m（SP）Equilibrium growth me

8、tamorphismdDAA ( 1 - 1 )3g = 4s( A +2 )eTR Tdt1DgKinetic growth metamorphism2a3(t)dD4(a (t)JL (t) +DJL2 L (t) g =Dzdtfr D (0)D (t)ggiggMelt metamorphismdD2(1.27 ´ 10-17 + 4.28 ´ 10-13q3 )g =mdtp D2gParameterization of Doì 0.1d + (1 - d )(0.3s + 0.4(1 - s)d > 0ïDo = íD

9、9;sD + max(0.4,g )(1 - s)d = 0îg2積雪粒徑增長(zhǎng)模型Flanner-Zender m（FZ06）p- p ambsdD =4peqdtDrLLRvTice (- 1)+RvTKTpeq Dv 1dDdDétùko =o ´ êúdtdt2( D - D) + t0ëoo0ûMelt metamorphismdD2(4.28 ´ 10-13q3 ) o =m dtp D2o試驗(yàn)區(qū)和積雪粒徑模擬CLPX LSOS2002年10月-2003年3月IOP-3 (1826 Februa

10、ry 2003)14積雪粒徑和相關(guān)長(zhǎng)度的關(guān)系SP-DgSP-DoFZ06-DoJ91-Dg00:00 February 23, 200315p= a Dexip= 2 b D (1 - v)ex3o粒徑增長(zhǎng)模型對(duì)亮溫的影響SNOWPACKFZ06J9116積雪分層對(duì)亮溫的影響N-Layer5-Layer3-Layer1-Layer17同化主微波遙感數(shù)據(jù)改進(jìn)積雪和土壤凍融模 擬的可行性分析在驅(qū)動(dòng)數(shù)據(jù)不確定的條件下，能否利用數(shù)據(jù)同化方法恢復(fù)土壤凍融狀態(tài)和SWE？結(jié)合主動(dòng)和被微波遙感數(shù)據(jù)，能否提高土壤凍融狀態(tài)和SWE的估計(jì)？哪個(gè)頻率（極化）包含的土壤凍融狀態(tài)和SWE信息？Bateni S. M.,

11、Huang C., Margulis S. A., Podest E., McDonald K., 2013, Feasibility of Characterizing Snowpack and the Freeze-Thaw State of Underlying Soil Using Multi-frequency Active/Passive Microwave Data, IEEE Transactions on Geoscience & Remote Sensing, 51(7), 4085-4102模型算子： CoLM + Jordan m觀 Passive Microw

12、ave Radiative Transfer MTB = M1(Tsoil ,qi,soil ,qw,soil ,Tsnow , rsnow , Dsnow ,qsnow , hsnow )測(cè)算 Radar Backscatter M子 s = M2 (qi,soil ,qw,soil ,qsnow , rsnow , hsnow , Dsnow , rsnow )觀測(cè)微波：1.4 GHz, 18.7GHz, 36.5GHz數(shù)據(jù) 主動(dòng)微波：L-band (1.4GHz), C-band (5.4GHz), Ku-band (12GHz)實(shí)驗(yàn)設(shè)計(jì)19（1） 同化多頻亮溫?cái)?shù)據(jù)（2） 同化多頻主動(dòng)后

13、向散射系數(shù)（3） 同化“主動(dòng)-”數(shù)據(jù)Open-loopNLDAS驅(qū)動(dòng)數(shù)據(jù)驅(qū)動(dòng)CoLM模型的結(jié)果synthetic truth 站點(diǎn)驅(qū)動(dòng)數(shù)據(jù)驅(qū)動(dòng)CoLM模型的結(jié)果實(shí)驗(yàn)結(jié)果：同化多頻亮度溫度20實(shí)驗(yàn)結(jié)果：同化多頻后向散射系數(shù)L-band (1.4GHz), C-band (5.4GHz), Ku-band (12GHz)21實(shí)驗(yàn)結(jié)果：同化“主動(dòng)+”數(shù)據(jù)土壤溫度土壤含冰量雪水當(dāng)量土壤含水量22實(shí)驗(yàn)結(jié)果：同化“主動(dòng)+”數(shù)據(jù)233.2積雪面積同化實(shí)驗(yàn)同化MODIS積雪面積改進(jìn)雪深估計(jì)（DirectInsert）同化MODIS積雪面積改進(jìn)雪深估計(jì)Filter）（EnsembleKalman243.2積雪面

14、積同化實(shí)驗(yàn)1. 直接遙感反演的積雪面積（Rl & Houser，2004，JH）(1) 增加雪層：模型沒有積雪，MODIS積雪面積大于40%，增加5mm雪水當(dāng)量；(2) 刪除雪層：模型有積雪， MODIS積雪面積小于10%，刪除雪層；2. 利用 “積雪面積-雪深”衰減曲線作為觀測(cè)算子來約束積雪模型模擬的雪深/雪水當(dāng)量hsno(W)SCA = tanh(A= A2.5z (r/ r）afdcmin(W, SI )0snonewmaxAndreadis & Lettenmaier, 2006，Su., 2008, JGR25（1）直接遙感反演的積雪面積26同化策略 增加雪層：模型沒

15、有積雪，MODIS積雪面積大于40%，增加5mm雪水當(dāng)量； 刪除雪層：模型有積雪， MODIS積雪面積小于10%，刪除雪層；研究區(qū)：空間分辨率：0.05deg 陸面過程模型：CoLM 觀測(cè)數(shù)據(jù)：MOD10C1 數(shù)據(jù)同化算法：EnKF 驅(qū)動(dòng)數(shù)據(jù)：GLDAS 觀測(cè)數(shù)據(jù)：MODIS積雪面積數(shù)據(jù)（MOD10C1) MOD Cloud FractionMOD Snow Fraction27同化結(jié)果：積雪面積MOD Cloud FractionSIMMOD Snow FractionASS28同化結(jié)果：雪深和雪水當(dāng)量SIMSIM雪水雪當(dāng)量深A(yù)SSASS29同化結(jié)果：站點(diǎn)驗(yàn)證30同化結(jié)果：誤差統(tǒng)計(jì)31（2）

16、利用“積雪面積-雪深”衰減曲線同化實(shí)驗(yàn)CoLM模型積雪過程模塊觀測(cè)算子 “SCF-雪深”衰減曲線模型算子（CoLM）模型誤差協(xié)方差矩陣EnKF數(shù)據(jù)同化算法同化算法（EnKF）32改進(jìn)雪深/雪水當(dāng)量預(yù)報(bào)更新的雪深集合觀測(cè)誤差協(xié)方差矩陣模擬的SCF集合觀測(cè)的SCF集合去云后的MOIDSSCF雪深預(yù)報(bào)集合地表參數(shù)集大氣驅(qū)動(dòng)數(shù)據(jù)MODIS積雪面積去云云的影響云覆蓋日數(shù)Terra：60.95%55.72%56.08%Aqua:58.64%55.04%57.06%云覆蓋時(shí)間序列變化33MODIS積雪面積去云總積雪面積變化34“積雪面積-雪深”衰減曲線（積累期）35“積雪面積-雪深”衰減曲線（消融期）36“

17、積雪面積-雪深”衰減曲線消融期積累期= 1- e-0.1565xf1= 1- e-0.134xf1xf2 = 0.9548x + 3.649 xf2 = 1 068x + 2 005f = 2 arctan(0.2905x)p32f3 = p arctan(0 3438x)= 1- e-0.1432xf1= 1- e-0.0884xf1xf =x20.9119x + 4.747f =20 8702x + 7 294f = 2 arctan(0.2387x)2f = arctan(0.1714x)p33p= 1- e-0.1224xx0.6575x + 7.904f1= 1- e-0 0758x

18、x-0 0492x + 23 47f1f =2=2f3 = p arctan(0.1958x) 2f3 = p arctan(0 1172x)= 1- e-0.8824xf= 1- e-0.0355x1f1x0.7376x + 9.685f =xf2 = 0 9913x + 20 782f = 2 arctan(0 0511x) 2>1500f3 = p arctan(0.1377x)f4 = 0.0513x3pf4 = 0 0263x-11f5 = 0 1066x 2-f = 0.1811x 2537驗(yàn)證站點(diǎn)海拔<=500m：“阿拉山口”和“烏蘇”500m<海拔<=1

19、000m：“裕民”和“哈巴河”1000m<海拔<=1500m：“木壘”和“托里” 海拔>1500 m：“昭蘇”和“天池”38雪深和積雪面積同化結(jié)果39同化結(jié)果誤差統(tǒng)計(jì)積雪面積雪深MBE(m)RMSE(m)MBE(%)RMSE(%)站點(diǎn)名稱海拔(m)積雪季NER站點(diǎn)名稱海拔(m)積雪季NER模擬同化模擬同化模擬同化模擬同化2004200520052006200620072004200520052006200620072004200520052006200620072004200520052006200620072004200520052006200620072004200520

20、05200620062007200420052005200620062007200420052005200620062007-20.288.88-32.8-30.55-9.82-39.17-24.57-57-22.73-43.1-42.51-35.85-37.57-41.58-43.77-19.07-61.52-35.97-48.68-58.14-44.12-15.39-12.1-19.24.117.768.759.5110.178.260.49-1.461.09-20.18-17.64-15.35-4.33-6.64-5.85-1.74-6.74-4.3-37.09-44.37-33.7-1

21、0.241.48-11.7131.0540.1949.4440.3618.6452.134.1966.3444.375456.1851.1950.6851.8255.0129.7467.9947.7260.2668.1855.2531.6927.9734.6110.4919.8319.1816.216.5816.146.735.159.1927.2825.9222.428.59.6810.476.259.167.1346.2452.5942.7421.7119.1524.480.660.510.610.60.110.690.80.920.790.490.540.560.830.810.810.

22、790.870.850.230.230.230.320.320.292004200520052006200620072004200520052006200620072004200520052006-0.020.01-0.02-0.06-0.01-0.04-0.04-0.13-0.01000.010.01-0.010-0.030.040.040.030.090.030.060.060.150.030.020.020.030.030.030.040.070.250.500.330.670.000.500.330.53阿拉山口286阿拉山口286烏蘇478烏蘇478哈巴河534哈巴河53420062

23、007-0 020 020 040 05-0 25200420052005200620062007200420052005200620062007-0.09-0.06-0.06-0.06-0.06-0.07-0.05-0.02-0.010.020.030.020.130.080.10.10.080.090.090.050.070.060.060.060.310.380.300.400.250.33裕民716裕民716托里1078托里107820042005-0 020 020 050 06-0 20200520062006200720042005200520062006200720042005

24、2005200620062007-0.14-0.06-0.12-0.11-0.05-0.04-0.12-0.1400.02-0.08-0.06-0.02-0.03-0.07-0.120.160.090.180.150.070.060.150.170.060.050.130.10.040.060.140.160.630.440.280.330.430.000.070.06木壘1271木壘1271昭蘇昭蘇18551855天池1935天池193540 4.結(jié)論數(shù)據(jù)同化方法可以用于提高雪深/雪深當(dāng)量的估計(jì)；多源遙感（光學(xué)、主動(dòng)、）結(jié)合的數(shù)據(jù)同化策略是當(dāng)前積雪數(shù)據(jù)同化研究的主要方向；積雪面積數(shù)據(jù)同化的關(guān)

25、鍵問題：觀測(cè)算子：“積雪面積-雪深”衰減曲線觀測(cè)數(shù)據(jù)：去云、積雪面積比例的計(jì)算微波數(shù)據(jù)同化的關(guān)鍵問題： 模型算子：雪粒徑的參數(shù)化 觀測(cè)/觀測(cè)算子：多極化和多頻率數(shù)據(jù)同化41 參考文獻(xiàn)Rl M, Houser P R. Updating a land surface mwith MODIS-derived snow coverJ. Journalof Hydrometeorology, 2004, 5(6)Sun C, Walker J P, Houser P R. A methodology for snow data assimilation in a land surfacemJ. Jou

26、rnal of Geophysical Research: Atmospheres (19842012), 2004, 109(D8).Clark M P, Slater A G, Barrett A P,hydrologic and land-surface m. Assimilation of snow covered area information intosJ. Advances in water resources, 2006, 29(8): 1209-1221.Andreadis K M, Lettenmaier D P. Assimilating remotely sensed

27、 snow observations into amacroscale hydrology mJ. Advances in Water Resources, 2006, 29(6): 872-886.Durand M, Margulis S A. Feasibility test of multifrequency radiometric data assimilation to estimate snow water equivalentJ. Journal of Hydrometeorology, 2006, 7(3).Su H, Yang Z L, Niu G Y,. Enhancing

28、 the estimation of continentalscale snow waterequivalent by assimilating MODIS snow cover with the ensemble Kalman filterJ. Journal of Geophysical Research: Atmospheres (19842012), 2008, 113(D8).Durand M, Kim E J, Margulis S A. Radiance assimilation shows promise for snowpack characteriz ationJ. Geo

29、physical Research Letters, 2009, 36(2).Su H, Yang Z L,inson R E,. Multisensor snow data assimilation at the continental scale:The value of Gravity Recovery and Climate Experiment terrestrial water storage informationJ. Journal of Geophysical Research: Atmospheres (19842012), 2010, 115(D10)Takala M,

30、Luojus K, Pulliainen J,. Estimating northern hemisphere snow water equivalent forclimate research through assimilation of space-borne radiometer data and ground-basedmeasurementsJ. Remote Sensing of Environment, 2011, 115(12): 3517-3529.42Su H, Yang Z L, Niu G Y,. Parameter estimation in ensemble ba

31、sed snow data assimilation: Asynthetic studyJ. Advances in Water Resources, 2011, 34(3): 407-416.Yatheendradas S, Lidard C D P, Koren V,. Distributed assimilation of satellitebased snow extentfor improving simulated streamflow in mountainous, dense forests: An example over the DMIP2 western basinsJ.

32、 Water Resources Research, 2012, 48(9):1-54.Thirel G, Salamon P, Burek P, Kalas M. Assimilation of MODIS Snow Cover Area Data in a DistributedHydrological MUsing the Particle FilterJ. Remote Sensing, 2013, 5(11): 5825-5850.Fletcher S J, Liston G E, Hiemstra C A,. Assimilating MODIS and AMSR-E Snow Observations ina Snow Evolution MJ. Journal of Hydrometeorology, 2012, 13(5).Huang, C. L., Margulis S. , Mus