Research Of Snow Surface Relfectance In Pixel Scale

Land surface reflectance is a key parameter in the entironment monitoring. It isimportant for the study of surface energy budget and land-atmosphere interactions. Forits angle distribution, surface reflectance needs to be measured in different directions.This kind of experiment is expensive and temporal limited. Remote sensor can providelong term observations over worldwide range. It is the only method for monitoring andresearch of surface reflectance in reginal and global scale.Since snow reflects almost80%radiation, it significantly influences the energybudget of the earth. However, the retrieval of snow BRDF/albedo is always a difficultissue in the application of remotely sensed information. Therefore, in this dissertation,we comprehensively evaluate the major satellite-estimate albedo products and surfacereflectance models used in retreival algorithms by in-situ measurements, establish thesurfave reflectance database from multi remote sensors, and extract the characteristicsof reflection for pure snow and vegetation-snow mixed land surface. Belows are themajor work conducted in this dissertation:Firstly, surface albedo products of MODIS, MISR and SEVIRI are evaluated usingin-situ measurements from44flux sites. Snow surface albedo products are validated byground measured data over Greenland. Influence of seasonal snow on quanlity ofvegetation albedo products are assessed over North America, Europe and Asia. Theresults show that albedo products match ground measurements well except some caseswith snow cover. Then the decrease of albedo products' quality caused by snow isquantified.Secondly,evaluation of five surface reflectance models, including Ross-Li model,Roujean model, Walthall model, modified Rahman model and Staylor model, incapturing snow reflectance signatures using ground measurements in Antarctica. Wecompare the outputs from models and the ground measurements. Throuth the bias,RMSE and the R2performances of these five models, we can get that Ross-Li modeland Roujean model have high capability of capturing the magnitude and angledistribution of snow surface reflectance. Compared with in-situ experiments, remote sensors can hardly provide samplings with such a high viewing resolution and abundantangular samplings. Combination of observations from multi sensors can provide moresurface reflectance samplings and may improve the retrieval quality.Thirdly, combination of MODIS and MISR snow surface reflectance products withtemporal and spatial variation is obtained. Extract snow surface reflection charictoristicsfrom this combination using surface reflectance models. Because these two sensors cancomplement the angular samplings of surface reflectance in almost perpendiculardirection for each other, the extracted surface reflectance has more angular informationthan data from single sensor. It can be used as empirical data for snow reflectance andbackground for model use.Fourthly, a surface reflectance database is established using MISR reflectance data.This databse also includes albedo, view geometry, LAI and snow mask from othersatellite products and covers six kinds of land surface types. For the inherent correlationbetween surface reflectance and albedo, the reflectance data are classified by albedo fordifferent vegetation type and snow cover degree.At last, previous models are applied in simulating surface reflectance data of fiveland cover types in the database. The outputs describe and quantify the vegetation-snowmixed surface reflectance property for multi land cover types.