MODIS 地表产品数据的相关算法及处理过程

以MODIS积雪覆盖产品和地表温度产品为例,详细介绍了MODIS 地表产品数据的相关算法、处理过程及产品级别,对MODIS产品的数据结构及数据项的具体含义做了较为详尽的阐述。最后介绍了处理MODIS数据的常用软件,为MODIS地表产品在各领域的应用打下了良好的基础。     

基于MODIS数据的我国天山典型区积雪特征研究

准确监测天山地区积雪面积和积雪日数对合理利用水资源及分析区域气候变化有重要意义。MODIS每日积雪产品可以为大面积快速积雪制图与监测提供依据,但因云量较高成为其应用的瓶颈。利用结合MODIS产品的时间与空间信息有效地减少了云对MODIS积雪产品的影响,并利用改进的MODIS积雪数据和DEM分析2002~2009年天山地区积雪面积和积雪日数的变化特征。结果表明:积雪频率总体上随着海拔升高而增大;不同坡向积雪面积差异明显,西北坡积雪覆盖率最高,北坡、西坡和东北坡次之,南坡和东南坡的积雪覆盖率最低;2006~2008年研究区积雪面积出现低值,年内最大积雪面积呈逐年减少的趋势;随着海拔下降,积雪日数逐渐变小,天山南部地区积雪日数仅为40 d以下;积雪日数大的区域年际积雪日数变化相对稳定,积雪日数少于40 d的区域积雪日数的变异系数最大,年际积雪日数变化不稳定。     

青藏高原MODIS积雪面积比例产品的精度验证与去云研究

MODIS积雪产品的精度验证和去云处理是积雪监测研究的基础。首先利用青藏高原典型地区的ETM+数据作为“真值”影像,对MODIS积雪面积比例(FSC)产品在无云条件下的精度进行验证,发展了一个基于三次样条函数插值的去云算法,并采用基于“云假设”的检验和地面站积雪覆盖日数(SCD)检验两种方法对去云算法的精度进行了分析评价。结果表明:MODIS FSC产品在青藏高原地区具有较高的精度,与FSC“真值”相比,其平均绝对误差、均方根误差以及相关系数分别为0.098、0.156和0.916;去云算法能够有效地获取云遮蔽像元的FSC信息,平均绝对误差为0.092,用新生成的无云MODIS FSC产品计算得到的SCD与地面观测值具有较高的一致性(87.03%),平均绝对误差为3.82 d。     

基于ART模型的MODIS积雪反照率反演研究

积雪反照率是研究局地或全球的能量收支平衡和气候变化中的重要参数,遥感反演为积雪反照率的获取提供了便利的手段。积雪反照率大小主要取决于积雪的自身物理属性(雪粒径、形状和污染物等因子)以及天气状况,遥感反演反照率大多基于双向反射模型(BRDF),积雪BRDF模型常使用积雪辐射传输模型获得。采用考虑了雪粒径、粒子形状以及污染物影响的渐进辐射传输理论(ART)模型,建立了MODIS积雪反照率反演算法,得到了MODIS 8d合成积雪反照率产品。将此算法应用于具有均一积雪地表的格陵兰岛地区,并使用GC-Net实测数据进行了验证,反演的总均方根误差(RMSE)为0.018,相关系数(r)为0.83,结果表明考虑了积雪特性的ART模型能够较好地反演积雪反照率,而且反演需要的参数较少。     

基于遥感与地埋信息系统的额敏河流域积雪变化分析

2005年3月新疆北部部分地区发生融雪洪水灾害,给当地人民的生活和国民生产都带来了严重的影响。分析积雪的分布及其变化可为防洪抗灾提供决策依据,同时精确的流域积雪制图和雪盖消融曲线可为融雪径流的模拟提供参数。本文介绍了MODIS数据积雪监测的方法及流域雪盖的分带提取,利用MODIS影像,结合地理信息系统技术分析了额敏河流域3月4日—3月12日每天、每个海拔高度带的积雪变化情况,并利用逐步回归法对积雪变化与气象因子作了回归分析,结果表明400~900m海拔高度带积雪变化与气温降水相关性很好,相关系数R=0.9。     

东北-内蒙古地区基于MODIS单、双卫星积雪数据及常规积雪观测结合的积雪日数研究

结合Terra和Aqua卫星的积雪产品,获取2001~2008年东北-内蒙古地区逐年积雪日数分布,并利用此数据对比Terra卫星积雪数据获取的逐年积雪日数。结果表明随海拔的升高,双星与单颗卫星积雪日数差异呈现明显增加的趋势。整个东北-内蒙古地区双星积雪日数平均高出单颗卫星积雪日15 d,但与台站积雪日数对比发现,双星积雪日数平均仍然偏低27 d。这说明,利用Terra和Aqua双卫星积雪监测数据能明显改善山区云对遥感监测的影响,同时也可以减少降雪初期和消融期由于积雪消融较快带来的积雪漏测,但不足以消除云等因素的影响。考虑到获取的2001~2006年台站年积雪日数与MODIS年积雪日数与有良好的统计关系,利用两者建立的线性统计关系修正整个东北-内蒙古地区的MODIS积雪日数,能够很好地消除云等因素带来的MODIS双卫星积雪日数偏小的问题,修正后台站与双星积雪日数之间的绝对误差由原来的27 d减小到18 d。     

面向亚像元级积雪信息提取的MODIS影像波段融合方法研究

MODIS影像因其共享性和时间序列的完整性而成为大区域积雪监测研究广泛使用的数据源,进行MODIS影像波段间融合,能够为积雪研究提供较高分辨率的影像数据源。为了充分利用MODIS影像250 m分辨率波段的空间和光谱信息,提取亚像元级的积雪面积,使用两种具有高光谱保真度的影像融合方法：基于SFIM变换和基于小波变换的融合方法,采取不同的波段组合策略,对MODIS影像bands 1~2和bands 3~7进行融合,并以Landsat TM影像的积雪分类图作为“真值”,对融合后影像进行混合像元分解得到的积雪丰度图的精度进行评价。结果表明：利用基于SFIM变换和小波变换方法融合后影像提取的积雪分类图精度较高,数量精度为75%,比未融合影像积雪分类图的精度提高了6%,表明MODIS影像波段融合是一种提取高精度积雪信息的有效方法。     

基于FY-MWRI的中国西部被动微波积雪判识算法北大核心CSCD

积雪是冰冻圈中分布最广泛的要素,在气候变化以及水文循环中扮演着重要角色。微波遥感因其全天时全天候工作、具有一定穿透性等优势,成为积雪监测的重要手段。利用FY-3C卫星同步观测获取的微波成像仪(MWRI)被动微波亮度温度数据、融合可见光红外扫描仪(VIRR)与中等分辨率成像光谱仪(MERSI)数据得到的积雪产品,结合MODIS地表分类数据、地表温度数据,发展了基于国产卫星数据的被动微波积雪判识算法。首先提取无云覆盖的不同地表类型被动微波数据像元样本,然后对各地表类型的微波特征进行分析,利用空间聚类的方法,得到TB19V-TB19H、TB19V-TB37V、TB22V、TB22V-TB89V、(TB22V-TB89V)—(TB19V-TB37V)这五类可以较好地区分积雪和其他类似积雪地表的指标。最后应用MODIS积雪产品为参考对该积雪判识算法进行精度评价,该算法在中国西部积雪判识总体精度为87.1%,漏判率为4.6%,误判率为23.3%;Grody算法判识总体精度为78.6%,漏判率为9.8%,误判率为30.7%,该算法判识精度高于Grody算法;通过Kappa系数分析比较,该算法积雪判识结果的Kappa系数值为47.3%,高于Grody算法判识结果的Kappa系数值39.9%,表明该算法积雪判识结果与MODIS积雪产品判识结果一致性更好。     

C波段SAR山区积雪面积提取研究

合成孔径雷达(SAR)不仅具有穿云透雾,全天候观测地表的能力,而且可穿透地表覆盖一定深度获取地表覆盖物内部特征信息。利用2011年10景ENVISAT\|ASAR可变极化模式精细图像(ASA_APP_1P)数据,分析比较了黑河上游祁连山冰沟流域不同时段积雪SAR后向散射特性,应用同期的MODIS积雪面积产品确定研究区积雪的累积和消融背景信息。研究表明:由于融雪期积雪含水量上升,SAR图像后向散射系数相比干雪或无雪图像明显降低,经过分析认为广泛应用的-3 dB阈值会明显低估湿雪覆盖范围,-2 dB阈值更适合该地区湿雪面积参数提取。山区积雪融化过程中低海拔区域积雪融化而高海拔山区积雪仍可能为干雪,在提取湿雪像元的基础上,根据Sigmoid函数阈值获取的像元湿雪百分比及DEM信息来提取干雪像元,最终获取整个流域积雪面积信息。通过与Landsat ETM+图像积雪面积分类结果进行比较,总体精度达到78%。积雪累积和消融背景信息的分析表明:误差主要源于流域东北部与西北部低海拔区域积雪快速消融。     

欧亚大陆中高纬度区逐日无云积雪产品研发及验证

利用多源遥感数据,结合光学遥感数据高空间分辨率及被动微波数据不受云干扰的优势,利用MODIS逐日积雪标准产品和AMSR-E雪水当量产品,生成了欧亚大陆中高纬度区500m分辨率的逐日无云积雪产品,并利用更高分辨率的Landsat-TM数据生成的积雪产品作为"真值"影像,对研发的逐日无云积雪覆盖产品的精度进行了验证。结果表明:MOD10A1和MYD10A1受云影响均较为严重,无法直接用于地表积雪面积的监测。而本研究合成的逐日无云产品具有较好的精度,与TM积雪图具有较高的一致性。但不同的土地覆盖类型对积雪分类精度有一定的影响。其中,裸地和草原覆盖区精度最好,Kappa系数分别为0.655和0.644,均为高度一致性;其次精度较好的是灌丛和耕地覆盖区,Kappa系数分别为0.584和0.572,均为中等的一致性;而森林覆盖区由于受到高大植被的影响,Kappa系数仅为0.389,合成产品相对TM积雪产品明显高估了森林区积雪面积。整体Kappa均值达到0.569,接近高度一致,研究结果对实时监测欧亚大陆积雪面积具有一定的应用价值。     

Development and evaluation of a cloud-gap-filled MODIS daily snow-cover product

The utility of the Moderate Resolution Imaging Spectroradiometer (MODIS) snow-cover products is limited by cloud cover which causes gaps in the daily snow-cover map products. We describe a cloud-gap-filled (CGF) daily snow-cover map using a simple algorithm to track cloud persistence, to account for the uncertainty created by the age of the snow observation. Developed from the 0.05° resolution climate-modeling grid daily snow-cover product, MOD10C1, each grid cell of the CGF map provides a cloud-persistence count (CPC) that tells whether the current or a prior day was used to make the snow decision. Percentage of grid cells “observable” is shown to increase dramatically when prior days are considered. The effectiveness of the CGF product is evaluated by conducting a suite of data assimilation experiments using the community Noah land surface model in the NASA Land Information System (LIS) framework. The Noah model forecasts of snow conditions, such as snow-water equivalent (SWE), are updated based on the observations of snow cover which are obtained either from the MOD10C1 standard product or the new CGF product. The assimilation integrations using the CGF maps provide domain-averaged bias improvement of ~11%, whereas such improvement using the standard MOD10C1 maps is ~3%. These improvements suggest that the Noah model underestimates SWE and snow depth fields, and that the assimilation integrations contribute to correcting this systematic error. We conclude that the gap-filling strategy is an effective approach for increasing cloud-free observations of snow cover.     

Snow-cover mapping in forests by constrained linear spectral unmixing of MODIS data

A snow-cover mapping method accounting for forests (SnowFrac) is presented. SnowFrac uses spectral unmixing and endmember constraints to estimate the snow-cover fraction of a pixel. The unmixing is based on a linear spectral mixture model, which includes endmembers for snow, conifer, branches of leafless deciduous trees and snow-free ground. Model input consists of a land-cover fraction map and endmember spectra. The land-cover fraction map is applied in the unmixing procedure to identify the number and types of endmembers for every pixel, but also to set constraints on the area fractions of the forest endmembers. SnowFrac was applied on two Terra Moderate Resolution Imaging Spectroradiometer (MODIS) images with different snow conditions covering a forested area in southern Norway. Six experiments were carried out, each with different endmember constraints. Estimated snow-cover fractions were compared with snow-cover fraction reference maps derived from two Landsat Enhanced Thematic Mapper Plus (ETM+) images acquired the same days as the MODIS images. Results are presented for non-forested areas, deciduous forests, coniferous forests and mixed deciduous/coniferous forests. The snow-cover fraction estimates are enhanced by increasing constraints introduced to the unmixing procedure. The classification accuracy shows that 96% of the pixels are classified with less than 20% error (absolute units) on 7 May 2001 when all forested and non-forested areas are included. The corresponding figure for 4 May 2000 is 88%.     

Statistical evaluation of remotely sensed snow-cover products with constraints from streamflow and SNOTEL measurements

Using streamflow and Snowpack Telemetry (SNOTEL) measurements as constraints, the evaluation of the Moderate Resolution Imaging Spectroradiometer (MODIS) daily and 8-day snow-cover products is carried out using the Upper Rio Grande River Basin as a test site. A time series of the snow areal extent (SAE) of the Upper Rio Grande Basin is retrieved from the MODIS tile h09v05 covering the time period from February 2000 to June 2004 using an automatic Geographic Information System (GIS)-based algorithm developed for this study. Statistical analysis between the streamflow at Otowi (NM) station and the SAE retrieved from the two MODIS snow-cover products shows that there is a statistically significant correlation between the streamflow and SAE for both products. This relationship can be disturbed by heavy rainstorms in the later springtime, especially in May. Correlation analyses show that the MODIS 8-day product has a better correlation (r=−0.404) with streamflow and has less percentage of spurious snowmelt events in wintertime than the MODIS daily product (r=−0.300). Intercomparison of these two products, with the SNOTEL data sets as the ground truth, shows that (1) the MODIS 8-day product has higher classification accuracy for both snow and land; (2) the omission error of misclassifying snow as land is similar for both products, both are low; (3) the MODIS 8-day product has a slightly higher commission error of misclassifying land as snow than the MODIS daily product; and (4) the MODIS daily product has higher omission errors of misclassifying both snow and land as clouds. Clouds are the major cause for reduction of the overall accuracy of the MODIS daily product. Improvement in suppressing clouds in the 8-day product is obvious from this comparison study. The sacrifice is the temporal resolution that is reduced from 1 to 8 days. The significance of the results is that the 8-day product will be more useful in evaluating the streamflow response to the snow-cover extent changes, especially from the long-term point of view considering its lower temporal resolution than the daily product. For clear days, the MODIS daily algorithm works quite well or even better than the MODIS 8-day algorithm.     

A blended global snow product using visible,passive microwave and scatterometer satellite data

A joint US Air Force/National Aeronautics and Space Administration (NASA) blended global snow product that uses Earth Observation System Moderate Resolution Imaging Spectroradiometer (MODIS), Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) and Quick Scatterometer (QuikSCAT or QSCAT) data has been developed. Existing snow products derived from these sensors have been blended into a single, global, daily, user-friendly product by using a newly developed Air Force Weather Agency (AFWA)/NASA Snow Algorithm (ANSA). This initial blended snow product uses minimal modelling to expeditiously yield improved snow products, which include, or will include, snow-cover extent, fractional snow cover, snow water equivalent (SWE), onset of snowmelt and identification of actively melting snow cover. The blended snow products are currently 25-km resolution. These products are validated with data from the lower Great Lakes region of the USA, from Colorado obtained during the Cold Land Processes Experiment (CLPX), and from Finland. The AMSR-E product is especially useful in detecting snow through clouds; however, passive microwave data miss snow in those regions where the snow cover is thin, along the margins of the continental snowline, and on the lee side of the Rocky Mountains, for instance. In these regions, the MODIS product can map shallow snow cover under cloud-free conditions. The confidence for mapping snow-cover extent is greater with the MODIS product than with the microwave product when cloud-free MODIS observations are available. Therefore, the MODIS product is used as the default for detecting snow cover. The passive microwave product is used as the default only in those areas where MODIS data are not applicable due to the presence of clouds and darkness. The AMSR-E snow product is used in association with the difference between ascending and descending satellite passes or diurnal-amplitude variations (DAV) to detect the onset of melt, and a QSCAT product will be used to map areas of snow that are actively melting.     

Satellite-derived snow coverage related to hydropower production in Norway: Present and future

Hydropower derived from snow-melt runoff is a major source of electricity in Norway. Therefore, amount of snow-melt runoff is key to the prediction of available water. The prediction of water quantity may be accomplished through the use of hydrological models. These models, which may be run for individual basins, use satellite-derived snow-covered area in combination with snow-cover depletion curves. While it is known that snow albedo information would increase the accuracy of the models, large-scale albedo measurements have not yet been obtained from satellites on a regular basis. This paper presents Landsat-5 Thematic Mapper (TM) reflectances recorded in May 1989 from a mountainous catchment at Kvikne, Norway. Satellite-derived albedo values are analysed, and compared with simultaneously measured in situ albedo. The satellite-derived shortwave snow albedo is comparable with bare ground albedo and values as low as 0.19 were found in areas where the snow was highly metamorphosed and heavily blackened by organic material. To map snow-covered areas, the contrast between snow and snow-free areas can be improved by using a normalized TM Band 2-5 difference image. While TM Band 2 alone shows varying degrees of snow surface contamination within the study area, the normalized difference snow index (NDSI) is not affected by impurities. This paper also discusses the use of NASA's EOS (Earth Observing System) Moderate Resolution Imaging Spectroradiometer (MODIS) sensor, which is planned to be launched in the summer of 1999 for mapping of large-scale geophysical parameters including snow-cover. MODIS will enable snow cover and albedo to be mapped in Norway on a daily basis, and should enhance our ability to estimate snow coverage and thus manage hydropower production.     

被动微波遥感在青藏高原积雪业务监测中的初步应用

积雪范围、积雪深度和雪水当量等参数的遥感监测与反演对气候模式的建立以及积雪灾害的评估具有重要意义。被动微波遥感在这些参数的反演方面具有明显优势,但目前尚未应用到青藏高原地区的积雪遥感业务监测上来。2001年10月至2002年4月,利用SSM/I数据对青藏高原地区的积雪范围和积雪深度进行了实时监测,为西藏、青海遥感应用部门提供逐日的雪深分布图。对这次监测的总效果进行了分析和评价,并对发生在青海省内一次较大的降雪过程进行了遥感分析,结果表明:SSM/I反演的积雪范围变化趋势与MODIS结果总体上较为一致;SSM/I的雪深监测结果为当地遥感部门对大于10 cm的雪深做出正确判断提供了重要信息,是对雪灾定位的重要信息源。     

Retrieval of snow grain size over Greenland from MODIS

This paper presents a new automatic algorithm to derive optical snow grain size at 1 km resolution using Moderate Resolution Imaging Spectroradiometer (MODIS) measurements. The retrieval is conceptually based on an analytical asymptotic radiative transfer model which predicts spectral bidirectional snow reflectance as a function of the grain size and ice absorption. The snow grains are modeled as fractal rather than spherical particles in order to account for their irregular shape. The analytical form of solution leads to an explicit and fast retrieval algorithm. The time series analysis of derived grain size shows a good sensitivity to snow melting and snow precipitation events. Pre-processing is performed by a Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm, which includes gridding MODIS data to 1 km resolution, water vapor retrieval, cloud masking and an atmospheric correction. MAIAC cloud mask is a new algorithm based on a time series of gridded MODIS measurements and an image-based rather than pixel-based processing. Extensive processing of MODIS TERRA data over Greenland shows a robust discrimination of clouds over bright snow and ice. Because in-situ grain size measurements over Greenland were not available at the time of this work, the validation was performed using data of Aoki et al. (Aoki, T., Hori, M., Motoyoshi, H., Tanikawa, T., Hachikubo, A., Sugiura, K., et al. (2007). ADEOS-II/GLI snow/ice products — Part II: Validation results using GLI and MODIS data. Remote Sensing of Environment, 111, 274-290) collected at Barrow (Alaska, USA), and Saroma, Abashiri and Nakashibetsu (Japan) in 2001-2005. The retrievals correlate well with measurements in the range of radii ~ 0.1-1 mm, although retrieved optical diameter may be about a factor of 1.5 lower than the physical measured diameter. As part of validation analysis for Greenland, the derived grain size from MODIS over selected sites in 2004 was compared to the microwave brightness temperature measurements of SSM/I radiometer which is sensitive to the amount of liquid water in the snowpack. The comparison showed a good qualitative agreement, with both datasets detecting two main periods of snowmelt. Additionally, MODIS grain size was compared with predictions of the snow model CROCUS driven by measurements of the automatic weather stations of the Greenland Climate Network. We found that the MODIS value is on average a factor of two smaller than CROCUS grain size. This result agrees with the direct validation analysis indicating that the snow reflectance model may need a “calibration” factor of ~ 1.5 for the retrieved grain size to match the physical snow grain size. Overall, the agreement between CROCUS and MODIS results was satisfactory, in particular before and during the first melting period in mid-June. Following detailed time series analysis of snow grain size for four permanent sites, the paper presents maps of this important parameter over the Greenland ice sheet for the March-September period of 2004.     

MODIS和VEGETATION雪盖产品在北疆的验证及比较

雪盖产品的准确性评估对于水文模型中的遥感应用具有重要的意义,利用北疆47个气象站实测雪深资料,并将气象站根据海拔和下垫面进行分类,对我国可使用的3种光学遥感雪盖产品MOD10A1、MOD10A2和VGT-S10雪盖产品进行验证。研究表明,MOD10A1、MOD10A2和VGT-S10雪盖产品识别总体精度分别为91.3%、90.6%和87.9%,3种产品在农田、草地、城镇和建筑用地总体精度更高 |在稀疏灌木林、裸地与稀疏植被识别总体精度较低,特别是在山区,3种产品识别精度均较低,分别为66.3%、75.7%和61.9%。进一步统计3种雪盖产品的错分误差、漏分误差,发现3种产品错分误差都比较小,但在山区站的漏分误差比较严重,分别为32.4%、21.7%和36.3%,3种产品在山区都低估了雪盖面积。3种不同时间分辨率的雪盖产品云影响率分别为61.8%、7.6%和1.8%。最后将MODIS合成与VGT-S10时间分辨率相同的雪盖产品,并对两种产品在积雪积累期和消融期进行相互比较,比较发现MODIS识别精度要优于VGT-S10雪盖产品,3种产品中VGT-S10由于合成天数最多,所以雪盖产品受云的影响最小。     

Using crowdsourced and weather station data to fill cloud gaps in MODIS snow cover datasets

We present the design, development, and testing of a new software package for generating snow cover maps. Using a custom inverse distance weighting method, we combine volunteer snow reports, cross-country ski track reports and station measurements to fill cloud gaps in the Moderate Resolution Imaging Spectroradiometer (MODIS) snow cover product. The method is demonstrated by producing a continuous daily time step snow probability map dataset for the Czech Republic region. For validation, we checked the ability of our method to reconstruct MODIS snow cover under cloud by simulating cloud cover datasets and comparing estimated snow cover to actual MODIS snow cover. The percent correctly classified indicator showed accuracy between 80 and 90% using this method. The software is available as an R package. The output data sets are published on the HydroShare website for download and through a web map service for re-use in third-party applications.     

A prototype MODIS-SSM/I snow-mapping algorithm

Data in the wavelength range 0.545-1.652 w m from the Moderate Resolution Imaging Spectroradiometer (MODIS), launched aboard the Earth Observing System (EOS) Terra in December 1999, are used to map daily global snow cover at 500 m resolution. However, during darkness, or when the satellite's view of the surface is obscured by cloud, snow cover cannot be mapped using MODIS data. We show that during these conditions, it is possible to supplement the MODIS product by mapping the snow cover using passive microwave data from the Special Sensor Microwave Imager (SSM/I), albeit with much poorer resolution. For a 7-day time period in March 1999, a prototype MODIS snow-cover product was compared with a prototype MODIS-SSM/I product for the same area in the mid-western USA. The combined MODIS-SSM/I product mapped 9% more snow cover than the MODIS-only product.