Remote sensing parameterization of the processes of energy and water cycle over desertification area

Vol. 45 Supp.SCIENCE IN CHINA (Series D)December 2002Remote sensing parameterization of the processes of energyand water cycle over desertification areasMA Yaoming (马耀明)', Tsukamoto Osamu? & Ishikawa Hirohiko3P9 A1. Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lan-zhou 730000 China;2. Faculty Sciences, Okayama University, Okayama 700, Japan;3. Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto 611, JapanReceived September 15, 2002Abstract In order to understand the processes of land surface-atmosphere interaction over de-serification area, it is indispensable to utilize of satellite remote sensing. Two scenes of LandsatTM were used to produce a set of maps of surtace reflectance, MSAVI, vegetation coverage, sur-face temperature, net radiation, soil heat flux, sensible heat fux and latent heat flux. Statisticalanalysis based on these maps revealed some quantitative significant land surface characteristics.Future developments of the method are also discussed.Keywords: energy and water cycle, desertiflcation area, Landsat TM, field observation.The study on the energy exchanges between the land surface and atmosphere is of paramountimportance for desertification area, e.g. HEIhe River Field Experiment (HEIFE) and Arid Envi-ronmental Comprehensive Monitoring Plan, 95 (AECMP'95) areas in northwestem China. Someinteresting detailed studies on the land surface heat flux densities over two areas have been re-ported"- 51. These researches were, however, on point- level or local-patch-level. Since pointwiseinformation of land-surface atmosphere interaction is required, it is necessary to integrate the in-dividual results into a regional scale. Remote sensing from satellites offers the possibility to deriveregional distribution of land surface heat flux densities.The purpose of this study is to up-scale the point or patch scale field observations of land sur-face variables and land surface heat fluxes to regional distribution by using Landsat TM data andfield observations.1 Landsat TM data and field observation dataLandsat Thematic Mapper (TM) provides a spectral radiance in seven narrow bands, with aspatial resolution of about 28.5X 28.5 m' for three visible bands (Bands 1, 2, 3) and three nearinfrared bands (Bands 4, 5, 7), and 120X 120 m2 for the thermal infrared band 6. Two TM imagesused in this paper were at 10:00 (local time) July 7, 1991 and 10:00 (local time) August 21, 1995over the areas of HEIFE and AECMP'95 respectively.Most relevant data, collected at HEIFE and AECMP'95 surface stations to be used in theparameterization of land surface heat flux densities and analysis of TM images, include surface中国煤化工MYHCNMHGSCIENCE IN CHINA (Series D)Vol. 45radiation budget components, surface radiation temperature, surface reflectance, vertical profilesof air temperature, humidity, wind speed and direction measured at the PBL towers, Sodar, ra-diosonde, tethersonde and turbulent fluxes measured by eddy. correlation technique, soil heat flux,soil temperature profiles, soil moisture profiles, and the vegetation state.2 Theory and schemeBy combining satellite remote sensing (e.g. Landsat TM data) with field observations, theland surface heat flux densities over inhomogeneous land surface can be derived.2.1 Net radiationThe regional net radiation flux density can be derived fromR(xy)=(1-n(x.y)). K(x.y)+ L(xy)- E(x,)oTfe(x,y),1)where rn0(x, y), is surface reflectance, and Tsf (x, y) is surface temperature. They can be derivedfrom landsat TM datal'. The incoming short-wave radiation flux density and incoming long-waveradiation flux density KI (x, y and L↓(x, y) in eq. (1) can be derived from radiative transfer modelMODTRANI%. Surface emissivity e(x,y) will be determined by Valor and Caslles's methodl7IE(x,y)=&,(x, y)P.(x.y)+eE(x,y)(1- R(x,))+4<8>(1- R(x,y),(x,y)，(2)and vegetation coverageP(x,y)=|「 NDVI(x,y) - NDVIm 7”NDVImx - NDVIme」2.2 Soil heat flux densityThe regional soil heat flux density Go(x, y) is determined by using a parameterization basedon MSAVI (modified soil adjusted vegetation index!8)Go(x,y)= R.(x.y) (Tgc(x)/o(x.y)). (a+b丽+c%2)-[1+ dMSAVI(x,y)"], (4)where the constants a, b, c, d and e are determined by using the field data observed at HEIFE andAECMP'95 observation stations;百is a daily mean reflectance value, i.e. for HEIFE caseG(x,y)= R,(x.y)Tse(x.y)二-((0.00025 +0.00436 +0.00845元2 )[1-0.979MSAVI(x,y91. (5)For AECMP'95 caseGo(x,y)= R(.y)Tye(x.y)(0.00028 + 0.00424万+ 0.00875元)[1- 0.982MSAVI(x,y)"]. (6)6(x,y) .whereMSAVIx. y)= 24(.)+1- J2+4(.y)+1I 84(.1)-5(.明(7)22.3 Sensible heat flux densityThe regional distribution of sensible heat flux density can be estimated from中国煤化工MYHCNMHGSupp.REMOTE SENSING PARAMETERIZATION OF PROCESSES OF ENERGY & WATER CYCLE4[e(xy)-T.(x,y)]H(x,y)= pCpk"u (x.y)(8In二-d(x.y),+ kB*(xy)-y(x,)| Ini三- d6(xy),-(x.y)||n"Zml(t,y)ZEm(x.y)To simulate sensible heat flux density on a large scale, a straightforward method is to scale-up oraggregate the regional sensible flux by a weighted average of the contributions from different sur-face elements based on the principle of flux conservation. Two approaches will be introduced andaddressed here.The approach of "blending height" (for HEIFE area), i.e.[Tye(x,y)-T(x,y)]H(x,y)= pCpkiug(9"Z0m(x,y)Im ag -d(x.y)Vm(x.y)where Zg is blending height, 4B is wind speed at the blending height; Zg and up can be deter-mined by using field measurements or numerical models. In this study, they will be determinedwith the aid of field measurements of radiosonde, tethersonde and sodar. Ta(x, y) in eq. (9) is theregional distribution of air temperature at the reference height. An improved interpolation methodis proposed here to derive the regional distribution of air temperature over the oasis desert system.In other words, the regional distribution of air temperature Ta (x, y) can be derived by using thisimproved numerical interpolation method based on a number of field observations of air tempera-ture and regional surface temperature asT(x,y)=Tge(x,y)- DT.(x,y).(10)The efctive roughness lengths in eq. (9) over HEIFE area including the effect of topography, lowvegetation (e.g. gras), taller plants (e.g. wheat canopy, trees and shrubs) can be determined by theTaylor's mode19! Raupach's methodol will be used to derive the zero plane displacement d(x,y)in eq. (9) over HEIFE area,1. d(x,y)_ 1-exp(-VCqLAI(x,y)h(x,y)VcgLAI(x,y)kB'(x, y) is determined by using the relationship between kB '(x, y and Tye(x, y)Approach of“lassification" for the AECMP'95 case.The study area can be classified into n categories according to land surface characteristics.The land surface variables can be measured or derived at the reference height over each kind ofsurface as: u), *..** Tal, ...... doi, .....* ZOoml, e....o.m kB"I, kB .......kB~'n; Ym1 ...... Ymm and WhI, ..... Yhm. The sensible heat flux is then calculated over dif-ferent land surfaces[T:fe(x,y)-Ta]H(x.y)= pCpk2un二“a +kB*'-Vnn=doa-yn|LZomi中国煤化工MYHCNMHGSCIENCE IN CHINA (Series D)Vol.45[Tfe(x,y)-Tg2](12)H2(xy)= pCpk"nn Co2 +kB~'2-Vn2ln=°02-Vm2l"Zom2Zom2[Tye(x.y)-Tgn]H。(x,y)= pCpk2unInζ°on+kB-'π←ψmZomThe sensible heat flux over the whole area can be derived using the aggregation method (conser-vation of energy) asH(x,y)=之w()H(xy),(13)=where w(i) is the weight of each kind of land surface. The regional sensible heat flux overAECMP'95 area can also be derived by using the approach of "blending height", i.e. eq. (9) canalso be used for this area. The variables in eq. (9) are determined by the same procedure as shownpreviously.2.4 Latent heat flux densityThe regional latent heat flux density E(x,y) will be derived as the residual of the energybudget theorem for land surface, i.e.AE(x,y)= R,(x,y)- H(x,y)-G(x,y).(14)3 Cases study and validationFig.1 shows the distibution maps of land surface heat flux densities over the HEIFE andAECMP'95 areas. The frequency distributions of land surface flux densities over two areas areshown in fig. 2. The derived land surface heat flux densities are validated by field measurements.In fig. 3 are plotted the derived results against the measured values in the field of HEIFE andAECMP'95 for four terms of the energy balance. The 1 : 1 line is also plotted in the graphs. Themean absolute percent difference (MAPD) is computed as a quantitative measure of the differencebetween the derived results (Heived () and measured values (Hmesured() here, and100 t| Hderived(i) - H measured(i)|(15)H mesured()The results on HEIFE area show that: () the derived heat flux densities over the HEIFE area are ingood accordance with the land surface status. These parameters show a wide range due to thestriking contrast of surface features and there are two peaks in the figures of all frequency distri-butions in this area; (i) the derived net radiation is very close to that of the field measurementwith MAPD less than 5%. It is better than the previous rsut1l; (i) the parameterization中国煤化工MYHCNMHGSupp.REMOTE SENSING PARAMETERIZATION OF PROCESSES OF ENERGY & WATER CYCLE51加40000600.00。1000100.0 .300015004.020002500- 500Fig 1. Mops of land srfoce heat 0ux desitie. (a) HEIFE: (b)AECMP"95.20N10Mean-186SMD-155SD-47”2|200400600800%2040608010010200000002004000600800Net radiation R/W .m"Soilheat flux G/W.m Sensible het tux H/W.m2 Lacent heat Aux iE/W.m'(a10p-30一3020|2020)0|25035450、550600920304050607080 00. 200 30'o 100 200 300 400 500R/W.m-'G/w.m~H/w.m2E/W.mFig.2. Frequency distribution of land surface heat flux densitis for HEIFE and AECMP 95 areas. (a) HEIFE: (b) AECMP'95.method based on MSAVI for soil heat flux density is suitable for heterogeneous land surface. Al-though the derived regional soil heat flux is slightly higher than the measured value, the MAPD issmaller than the value derived based on the NDV12); (iv) the derived regional sensible heat flux中国煤化工MYHCNMHG52SCIENCE IN CHINA (Series D)Vol. 45“。 750p.110E 250.8。Desert1“自o●calL2 Linze7soi zhngy35nr 5 9o0cal2 Colythoed400 sCal.1; s570|主200Gobif;400? Zhangyel10Limpe45050LingZhangye是100|; Zhangye2000350d4505506507505930507090、710550100150200250300/W.m; 600Rmnawww. m'Hesurns/W .m*600p-。800p。Bleding heighf。Bleding beight ip目 70{60|●clasifcation' E 350{ . lasficatioitinξ 00ξ6主250{400504030g500.600“506070、80408012016020050150250350450RomatowedW . m*Gomesuer/W.m*Hmesund/W.m'hEmewrerW " m'Fig. 3. Validation of the derived results vs. the field measurements for the surface reflectance, surface temperature and landsurface heat f0ux densities over the HEIFE area, logether with l:I line. Cal.l: formal rsulss2; Cal.2: derived results in thispaper.densities with MAPD around 5% at four validation sites are in good accordance with those of thefield measurements. This is due to the improved calculating scheme of regional air temperature Ts(x,y), Zom(x, y), d om(x, y) and kB~(x, y). The previous rsul.. derived from SEBAL could befit for the Gobi and sand desert surface (MAPD = 9.37% and MAPD = 6.21%), but there is a largedifference between the derived results and the field-measured values over oasis (MAPD = 36.76%in Linze and MAPD = 34.40% in Zhangye); and (V) the derived regional latent heat flux density,which is based on the energy balance equation, is acceptable for the whole HEIFE area. The valuecalculated from MAPD is 10% less than that of the four sites.The results of AECMP'95 area indicate that: () the derived land surface variables, vegetationvariables and heat flux densities over the AECMP'95 area are in good accordance with the landsurface status. These parameters keep wide ranges due to the strong contrast of surface featuresand there are two peaks in all the distribution histograms in the study area; (i) the derived net ra-diation flux density, soil heat flux density and sensible heat flux density are close to those of fieldmeasurements. The difference between the derived results and the field observation MAPD is lessthan 10%; (ii) the distribution of net radiation flux over the Gobi desert is almost the same as thatderived over the Gobi desert in HEIFE, but it is lower over the oasis than that derived over theoasis in HEIFE. The reason is that surface temperature of Gobi desert area is almost the same inthe two study areas, but it is higher in the oasis of AECMP'95 area (corn canopy) than in the oasisof HEIFE area(wheat canopy); and (iv) the estimated latent heat flux density over the 01 site is ingood accordance with the measured values (MAPD bending heigh= 7.32% and MAPIEeasifcation =6.76%). However the derived values are higher than the measured ones for latent heat flux overthe rest observation sites. The reason is that the derived latent heat flux density is based on theequation of surface energy balance. But the energy is imbalanced over the whole study area, espe-cially in the fringe region between the oasis and the Gobi desert due to their strong interactions.Another reason may be the accuracy of the turbulence measurement sensors.中国煤化工MYHCNMHGSupp.REMOTE SENSING PARAMETERIZATION OF PROCESSES OF ENERGY & WATER CYCLE54 Concluding remarksIn this paper, the regional distributions of land surface heat flux densities (net radiation, soilheat flux, sensible and latent heat flux over heterogeneous areas of HEIFE and AECMP'95 arederived with the aid of Landsat TM data and the field observations. The results are in goodagreement with filed observations. Compared with existing research results, the new method hasproved to be a better approach to getting related air-land parameters over heterogeneous land-scape.The apprach of deriving regional latent heat flux density as the residual of the energy budgetmay not be a good method for the fringe region between oasis and Gobi desert (e.g. AECMP'95area) due to the imbalance of energy and strong advection over the fringe region between the oasisand the Gobi desert. Further improvements are to be made to derive more accurate regional latentheat flux density of such areas. The vegetation variables cannot be validated in this research forlack of such measurements during this experiment. We should pay more attention to the measure-ments of vegetation variable, such as NDVI, LAI and vegetation coverage, in the future experi-ments.Acknowledgements The authors wish to thank Profs. M. Maiani, E. Ohtaki and K.Sahashi, Dr. Zzh.Su and Dr. M. Me-nenti for their very kind helps and useful discussions. This work was supported by the Chinese National Key Project (Grant No.G1998040900) and the Innovation Project of Cold and Arid Regions, Envionmental and Engineering Research Instue, theChinse Academy of Sciences (Grant Nos. CACX210072 and CACX210039). Some parts of this study were done as coperaiveresearch work in Disaster Prevention Research Institute, Kyoto University, Japan and the Alteera Green World Research,Wageningen UR, the Netherlands.References1. TTsukamoto. 0.. Wang. J. Mitsuta, A significant evening peak of vapour pressure at an oasis in the semi-arid region, Jour-nal of the Meteorological Society of Japan. 1992, 70<6): I15 -1159.2. Tsukamoto. 0.. Sahashi, K. Wang. J. Heat budget and evaporanspiration at an oasis surface surounded by deser, Jour-nal of the Meteorological Society of Japan, 1995, 73(5): 925--935.3. Mitsuta, Y. Tamagawa. 1. Sahashi, K. et al, Estimation of annual evaporation from the Linze Desert during HEIFE, Jour.nal of the Meleorological Society of Japan, 1995. 73(5): 967- _974.4. Hu, Y. Gao, Y. Wang, J. et al, Some achievements in scientific research during HEIFE, Platcau Meteorology, 1994, 13(3):225- -236.5. Maitani, T. Sahashi, K.. Otaki, E. et al., Measurements of turbulent fluxes and model simulation of micrometeorology ina wheat field at Zhangye Oasis, Journal of the Meteorological Society of Japan, 1995, 73(5): 959- 965.. Ma, Y. Parameterization of land surface heat fux densities over inhomogencous landscape by combining salie remotesensing with field observations, Doctoral Disertation, Okayama Universit, 200 Japan, 195.7. Valor, E. Caselles, V., Mapping land surface emissivity from NDVI: Application to European, African, and South Ameri-can areas, Remote Sensing of Environment, 1996, S7: 167-184.8. Qi.J. A.. Chehbouni, A. R. Huete. Y. H. et al, A modifed soil adjusted vegetation index. Remole Sensing of Envioment,1994, 8.48: 119-126.9. Taylor, P A.. Sykes. R. L., Mason, P. J.. On the pramerization of drag over small scale lopograply in nurally sratfedboundary flow, Boundary Layer Meteorology, 1999, 48: 409- 422.10. Raupach, M. R.. Simplied expesios for vetation roughness lengh and zero-plane displacements as functions ofcanopy height and area index. Boundary Layer Mcteorology, 1994, 71:11-216.11. Ma, Y. Wang. J. Meneni. M. et al. Estimation of flux densities over the heterogeneous land surface with the aid ofsatelite remote sensing and field observation, ACTA Meteorological Sinica, 199 S7(2): 180--189.12. Wang, J. Ma, Y. Meneni, M. The sealing-up of processes in the heterogeneous landscape of HEIFE with the aid of satel-lite remote sensing. Journal of the Meteorological Society of Japan, 1995. 73(6): 1235- 1244.中国煤化工MYHCNMHG