MODIS大气校正精度评价及其对表层雪密度提取影响

引用本文

刘艳, 汪宏, 张璞, 李杨. MODIS大气校正精度评价及其对表层雪密度提取影响[J]. 国土资源遥感, 2011,23(1): 128-132
LIU Yan, WANG Hong, ZHANG Pu, LI Yang. Accuracy Evaluation of MODIS Atmospheric Correction and Its Effects on Surface-Snow Density Extraction[J]. REMOTE SENSING FOR LAND & RESOURCES,2011,23(1): 128-132 复制到剪切板

Permissions

《国土资源遥感》编辑部

MODIS大气校正精度评价及其对表层雪密度提取影响

刘艳¹, 汪宏², 张璞¹, 李杨¹

1.中国气象局乌鲁木齐沙漠气象研究所,乌鲁木齐 830002

2.四川省农业科学院遥感应用研究所,成都 610066

第一作者简介: 刘艳(1978-),女,副研究员,主要从事遥感在积雪、植被变化监测等方面的研究。

收稿日期: 2010-04-06

基金: 中央级公益性科研院所基本科研业务项目(编号: IDM200808、IDM200704)和新疆气象局基金项目(编号: C2006003)共同资助

摘要

以古尔班通古特沙漠为研究区,以中分辨率成像光谱仪(MODIS)为遥感数据源,结合ASD FieldSpec准同步实测积雪反射光谱数据对FLAASH大气校正能力进行了评价。研究表明: ①校正后的MODIS各波段积雪反射率与准同步实测积雪反射率波形相似, 在第1~7波段整体相关系数达0.82,表明FLAASH大气校正能极大地提高MODIS地物识别能力; ②校正后的MODIS 第6波段反射率和归一化差值积雪指数(NDSI)与实测雪密度呈线性相关,可用回归拟合构建MODIS雪密度遥感计算模式。

关键词: 大气校正; 雪密度; 反演; 同步地物反射光谱; MODIS

中图分类号:TP751.1 文献标志码:A 文章编号:1001-070X(2011)01-0128-05

Accuracy Evaluation of MODIS Atmospheric Correction and Its Effects on Surface-Snow Density Extraction

LIU Yan¹, WANG Hong², ZHANG Pu¹, LI Yang¹

1.Institute of Desert Meteorology, CMA, Urumqi 830002, China

2.Institute of Remote Sensing Applications, Sichuan Academy of Agricultural Sciences, Chengdu 610066, China

Abstract

With Gurbantunggut desert as the study area and by using moderate resolution imaging spectrometer (MODIS) data, which is a kind of optical remote sensing data,in combination with snow reflectance spectra simultaneously measured by ASD Field Spec, the authors evaluated the correction capability of FLAASH model. Some results have been attained: ① Snow reflectance from 1st to 7th band of MODIS is similar to the simultaneously measured snow reflectivity,and the correlation coefficient of all bands is 0.82 on the whole, indicating that FLAASH can greatly enhance the capability for identifing surface features of M0DIS . ② On the basis of linear relationship between snow reflectance after correction at MODIS channel 6 and NDSI and measured snow density, a snow density remote computation model can be built by way of regression and fitting.

Keyword: Atmospheric correction; Snow density; Retrieval; Synchronous ground reflectance spectra; MODIS

Show Figures

0 引言

目前, 获取积雪密度的方法主要有: ①直接应用气象站积雪密度观测记录插值计算平均积雪密度^{[1, 2]}; ②利用雪密度函数计算积雪密度; ③利用星载扫描式多通道微波辐射仪 (SMMR)和特别微波辐射/成像计(SMM/I)等遥感数据获取积雪密度^{[3, 4, 5, 6]}。微波传感器虽可全天候获取雪层信息并穿透积雪层获取地表信息, 但空间分辨率较低, 难以准确地从空间上获取积雪密度。目前, 利用可见光-短波红外甚高分辨率辐射计(AVHRR)和多时相中等分辨率成像光谱仪(MODIS)能在晴空条件下比较准确地获取较高分辨率的积雪覆盖信息^{[7, 8, 9, 10, 11, 12, 13, 14, 15]}, 但对利用光学遥感数据定量获取雪密度和雪水当量的研究甚少。找寻适宜的大气校正模型对光学遥感数据进行精确的大气校正, 乃是积雪密度参数光学遥感反演的一个关键步骤和前提条件。

本文以古尔班通古特沙漠地区多时相MODIS为例, 采用FLAASH(Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes)大气校正模型进行大气校正; 结合该地区ASD FieldSpec准同步实测积雪反射光谱数据, 对MODIS 1B FLAASH大气校正效果进行评价, 分析校正后的MODIS各波段积雪反射率与实测雪密度的相关性, 寻找对雪密度具有良好响应的波段及波段反射比, 为雪密度光学遥感定量研究奠定基础。

1 研究区概况

古尔班通古特沙漠地处新疆北部, 位于准噶尔盆地中央, 地理范围为44° 11'~46° 20'N、84° 31'~90° 00'E, 是我国最大的固定、半固定沙漠。区内年降雨量为100~150 mm, 冬春有降雪, 积雪深度可达10~35 cm, 积雪时间可达4个月。冬季积雪和冻土对沙漠地表形成有利的保护层, 植被覆盖和生物结皮使得沙面整体处于稳定的状态。

本实验在沙漠边缘建立了积雪参数野外采集试验场并布设数据采集样方。试验场地势较平坦, 积雪覆盖率达90%以上, 积雪平均深度达6 cm。样方大小为360 m× 360 m, 被分割为90 m× 90 m大小的16个样方, 用于与MODIS同步观测。

2 数据源与计算方法

2.1 遥感数据源

遥感数据来自2008年12月6、8和20日获取的MODIS 1B数据, 覆盖了古尔班通古特沙漠地区, 已经过辐射校正和几何纠正等预处理。

2.2 积雪准同步观测

2008年12月6、8和20日在野外采集场样方内(图1)进行了积雪表面反射光谱、雪密度和雪深准同步观测, 观测时间为MODIS卫星过顶前后5 min内。其中, 使用FieldSpec Pro FR(Full Range)ASD光谱仪(350~2 500 nm)完成了雪表面反射光谱采集; 雪密度用Snow Fork雪密度分析仪采集(将距地表2 cm处的雪密度定义为表层雪密度)。

	Figure Option View Download New Window
	图1 MODIS准同步试验场示意图Fig.1 Sketch map of quasi-synchronism test ground for MODIS

2.3 实测积雪反射光谱归一化计算

归一化反射率可部分消除积雪表面光滑度和反射光谱随时间和空间变化的干扰。天气与周围环境的细微变化、光谱仪位置移动和太阳角变化等都会影响积雪反射光谱测量。为了对不同样点采集的积雪反射光谱进行对比分析, 对所测反射光谱进行归一化处理, 即以积雪在0.3~2.5 μ m范围内的平均反射率为归一化点, 利用ENVI软件Spectral Math计算归一化积雪反射率, 即

S=(S₁- $\begin{matrix} \bar{S} \end{matrix}$ )/(S₁+ $\begin{matrix} \bar{S} \end{matrix}$ )(1)

式中, S为归一化积雪反射率; S₁ 为实测积雪反射率; $\begin{matrix} \bar{S} \end{matrix}$ 为积雪平均反射率。

2.4 MODIS FLAASH大气校正参数输入

FLAASH大气校正基于传感器处单个像元接受到的太阳波谱范围内(不包括热辐射)标准平面朗伯体(或近似平面朗伯体)反射光谱辐射亮度^[15]。光谱辐射亮度计算公式为

L^*=Aρ /(1-ρ _eS)+Bρ _e/(1-ρ _eS)+ $\begin{matrix} L_{a}^{*} \end{matrix}$ (2)

式中, L^*为传感器接收到的单个像元辐射亮度; ρ 为该像元地表反射率; ρ _e为该像元和其周边像元的混合平均地表反射率; S为大气球面反照率; $\begin{matrix} L_{a}^{*} \end{matrix}$ 为大气层辐射进入传感器的辐射亮度; A、B分别为大气条件和下垫面几何条件决定的系数。A、B、S和 $\begin{matrix} L_{a}^{*} \end{matrix}$ 值分别由特定大气模式、气溶胶类型、能见度、卫星观测角度、大阳角度和地表平均海拔等决定, 这些参数均可通过其他方式或利用标准大气模式和气溶胶模式获取, 并与大气中的水蒸气含量、气溶胶光学厚度有密切关系。

FLAASH利用暗目标法反演气溶胶光学厚度, 利用MODTRAN 4模拟大气吸收并形成一个查找表, 然后利用查找表逐像元地进行水蒸气含量估算。ρ 和ρ _e的区别主要来自与大气散射引起的“ 邻近像元效应” , 若ρ =ρ _e, 则校正过程会忽略邻域效应, 但在有薄雾或地表存在强烈对比的条件下会使短波波段范围内的大气校正出现明显误差。FLAASH利用大气点扩散函数对邻域效应进行纠正。ρ _e的估算公式为

L_e≈ ρ _e(A+B)/(1-ρ _eS)+ $\begin{matrix} L_{a}^{*} \end{matrix}$ (3)

式中, L_e 为某像素和周边像元的空间平均值, 可以通过原始影像计算得到。

当取得所有参数后, 利用式(2)、(3)即可逐像元地求出整幅图像的真实地表反射率。用于本实验的大气校正实验数据的输入参数见表1^[16]。大气模式和气溶胶类型因缺乏实测数据而选择标准模式; 能见度参数选择沙漠地区天气较好情况下的能见度^[16]。

表1 FLAASH大气校正输入参数 Tab.1 Input parameters of FLAASH atmospheric correction

3 结果分析

3.1 校正前后视觉对比

对比大气校正前后的MODIS 6、2、3波段假彩色合成图像(图2)可以看出, 大气校正前图像整体偏暗, 对比度不高(图2(a)); 大气校正后图像对比度明显提高, 地物显得更加清晰可辨(图2(b))。大气校正前后图像视觉效果和反射辐射亮度发生了明显的变化, 校正后的图像更加清楚、色彩明亮, 说明FLAASH 校正有效地去除了大气气溶胶的影响。

	Figure Option View Download New Window
	图2 大气校正前后MODIS图像视觉对比Fig.2 Visual comparison between MODIS images before and after atmospheric correction

3.2 校正前后积雪反射光谱比较

在MODIS图像上选取积雪像元, 对MODIS第1~7波段计算反射率、FLAASH大气校正反射率和实测积雪重采样反射率进行了比较(图3)。

	Figure Option View Download New Window
	图3 积雪反射率比较Fig.3 Comparison of snow reflectance

实测反射率数据用ASD光谱测量仪测得, 光谱分辨率在350~1 000 nm谱段为3 nm, 在1 000~2 500 nm 谱段为10 nm; 采样间隔在350~1 000 nm谱段为1.4 nm, 在1 000~2 500 nm谱段为2 nm; 数据间隔为1 nm。本实验利用ENVI软件光谱重采样模块对ASD所测数据进行重采样, 得到MODIS图像第1~7波段反射率数据。校正后图像的积雪反射率与实测反射率波形相似, 在第1~7波段范围整体相关系数达0.82; 校正后图像的高反射波段较实测反射率低, 吸收波段与实测光谱接近。

3.3 校正前后NDSI比较

MODIS全球积雪面积产品包括多种类型, 并针对不同下垫面提供了一系列不同的算法, 但基本方法是归一化差值积雪指数(NDSI)法, 其计算公式为

NDSI=(B4-B6)/(B4+B6)(4)

式中, B4和B6分别为MODIS近红外波段和红波段像元的反射率值。

为进一步评价和验证FLAASH大气校正效果, 本试验对校正前后的NDSI进行了比较。按式(4)分别计算实测积雪光谱数据、MODIS图像DN值、MODIS计算反射率和MODIS校正反射率及其NDSI值, 并提取上述各参数在试验场观测点的3像元× 3像元范围的亮度平均值(表2)。

表2 积雪在MODIS 第4和6波段的DN值/计算反射率/校正反射率/NDSI Tab.2 Snow DN values and reflectance before and after correction at the B4 and B6 of MODIS image and NDSI values computed from these images

经分析认为: ①理论上, DN值未经过任何校正, 只是进入传感器中的辐射能的一种数字转换形式, 不能从本质上地反映地物的辐射特性(实际上, 用DN值计算的NDSI明显低于用实测数据计算的NDSI); ②大气顶层MODIS计算反射率虽然经过了辐射定标校正, 但它是地面反射率和大气反射率的总和, 仍不能真实地反映地物的辐射特性(因而用MODIS计算反射率计算的NDSI明显低于用实测数据计算的NDSI); ③用校正后的地表反射率计算的NDSI与实测数据计算的NDSI比较接近(相关系数达0.6), 这表明经大气校正后的图像能基本消除大气的影响, 能比较真实地反映地物的辐射特性。校正前后NDSI变化趋势基本上相同(图4), NDSI峰值和谷值的出现位置整体上十分吻合, 说明大气校正前后图像总体特征相同。

	Figure Option View Download New Window
	图4 MODIS影像大气校正前后某一剖面的NDSI值Fig.4 NDSI of MODIS images before and after atmospheric correction

图像均值代表图像的整体亮度水平; 标准差反映图像中灰度值的离散程度, 标准差越大, 灰度级分布越分散, 图像目视效果越好, 图像所含信息量也越丰富。统计校正前、后的NDSI数据, 得到校正前NDSI平均值和标准差分别为0.283和0.236、校正后NDSI平均值和标准差分别为0.311和0.241, 这说明经大气校正后NDSI的信息量更丰富。同时, 大气校正后计算的NDSI值更接近于用地面同步实测积雪光谱数据计算的NDSI值。因此, FLAASH大气校正在一定程度上有效地降低了大气对遥感图像的影响, 获得了地表真实反射率, 达到了降低大气影响、增强图像信息的目的。

3.4 与实测雪密度比较

以2008年12月6日和20日在试验场样方内同步采集的23组数据(反射率和雪密度)为例, 进行实测积雪密度与MODIS图像校正前后波段反射率相关分析。首先, 计算MODIS数据在可见光和近红外波段的中心波长, 提取对应各中心波长处的ASD实测归一化积雪反射率, 得到基于MODIS图像波段的归一化ASD反射率数据; 其次, 在ERDAS软件的定位器中输入实测雪密度时记录的地理坐标, ERDAS自动在经过几何纠正的MODIS影像上进行定位。为了减小GPS记录实测点地理坐标存在的误差和图像几何纠正时存在的误差, 本实验在图像匹配点周围选取3像元× 3像元的窗口, 以所选像元的平均值作为最终像元值; 最后, 进行雪密度和ASD反射率、真实反射率和计算反射率相关分析, 分别就单波段和波段组合模式进行了分析(表3)。

表3 不同积雪反射率与同步实测雪密度的相关系数 Tab.3 Correlation coefficients between different snow reflectance and density

结果表明, FLAASH大气校正后近红外波段(B6、B7)反射率和NDSI与实测雪密度间相关系数得到显著提高。雪密度与B6和B7具有很高的相关性, B7常被用来识别陆地和云, B6被用来区别云雪差异。因此, 本试验以校正后的MODIS 第6波段和NDSI与实测雪密度数据进行回归拟合, 得到两种回归模式(表4)。

表4 雪密度的遥感回归模式及相关分析 Tab.4 Remote sensing arithmetic model of snow density and regression analysis

同时, 选取试验区8个样本点的实测值对雪密度的遥感回归模式进行了验证(表5)。

表5 雪密度验证结果 Tab.5 Results of accuracy evaluation of snow density

4 结论

(1)FLAASH校正前后MODIS影像的视觉效果和反射辐射亮度发生了明显变化, 校正后的图像更加清楚。大气校正大大恢复了积雪光谱反射特征。校正后的MODIS图像积雪波段反射率与ASD实测反射率波形相似, 高反射波段较实测反射率低, 吸收波段与实测光谱接近。

(2)FLAASH校正后的MODIS NDSI更加接近根据ASD实测积雪光谱数据计算的NDSI。校正后MODIS的第6波段和NDSI同SnowFork实测雪密度间具有很好的线性相关性, 适合用于雪密度的光学遥感反演; 对回归拟合建立的MODIS雪密度计算模式进行的精度评价, 验证了利用积雪反射率反演雪密度的可行性。经FLAASH大气校正的MODIS数据能为雪密度的光学定量遥感反演提供良好的数据源。

The authors have declared that no competing interests exist.

参考文献

View Option

[1]	黄慰军, 黄镇, 崔彩霞, 等. 新疆雪密度分布研究[J]. 中国农业气象, 2007, 28(4): 383-385. [本文引用:1]
[2]	黄慰军, 黄镇, 崔彩霞, 等. 新疆雪密度时空分布及其影响特征研究[J]. 冰川冻土, 2007, 29(1): 66-72. [本文引用:1]
[3]	Chang A T C, Foster J L, Hall D K. Nimbus-7 SMMR Derived Global Snow Cover Parameters[J]. Annals of Glaciology, 1987(9): 39-44. [本文引用:1] [JCR: 2.524]
[4]	车涛, 李新, 高峰. 青藏高原雪深和雪水当量的被动微波遥感反演[J]. 冰川冻土, 2004, 26(3): 363-368. [本文引用:1]
[5]	SUN Zhi-wen, SHI Jian-cheng, JIANG Ling-mei. Development of Snow Depth and Snow Water Equivalent Algorithm in Western China Using Passive Microwave Remote Sensing Data[J]. Advances in Earth Science, 2006, 21(12): 1363-1369. [本文引用:1] [CJCR: 1.16]
[6]	李新, 车涛. 积雪被动微波遥感研究进展[J]. 冰川冻土, 2007, 29(3): 487-469. [本文引用:1]
[7]	史培军, 陈晋. RS与GIS支持下的草地雪灾监测试验研究[J]. 地理学报, 1996, 51(4): 296-305. [本文引用:1]
[8]	王世杰. 利用NOAA/AVHRR影像资料估算积雪量的方法探讨[J]. 冰川冻土, 1998, 20(1): 68-71. [本文引用:1]
[9]	周咏梅, 贾生海, 刘萍. 利用NOAA-AVHRR资料估算积雪参量[J]. 气象科学, 2001, 21(1): 117-121. [本文引用:1]
[10]	延昊. NOAA 16卫星积雪识别和参数提取[J]. 冰川冻土, 2004, 6(3): 369-371. [本文引用:1]
[11]	梁天刚, 吴彩霞, 陈全功, 等. 北疆牧区积雪图像分类与雪深反演模型的研究[J]. 冰川冻土, 2004, 26(2): 160-165. [本文引用:1]
[12]	Yoram J K, Richard G K, Dorothy K H, et al. Remote Sensing of Sub-pixel Snow Cover Using 0. 66 and 2. 1μm Channels[J]. Geophysical Research Letters, 2002, 29: 16-20. [本文引用:1] [JCR: 4.456]
[13]	王健, 陈子丹, 李文君, 等. 中分辨率成像光谱仪图像积雪反射特性的初步分析研究[J]. 冰川冻土, 2000, 22(2): 165-169. [本文引用:1]
[14]	Dorothy K H, George A R, Vincent V S. Development of Methods for Mapping Global Snow Cover Using Moderate Resolution Imaging Spectrometer Data[J]. Remote Sensing of Environment, 1995, 54: 127-140. [本文引用:1] [JCR: 4.769]
[15]	宋晓宇, 王纪华, 刘良云, 等. 基于高光谱遥感影像的大气纠正: 用AVIRIS数据评价大气纠正模块FLAASH[J]. 遥感技术与应用, 2005, 20(4): 393-398. [本文引用:2]
[16]	李国砚, 张仲元, 郑艳芬, 等. MODIS影像的大气校正及在太湖蓝藻监测中的应用[J]. 湖泊科学, 2008, 20(2): 160-166. [本文引用:2]

2007

0.0

黄慰军, 黄镇, 崔彩霞, 等. 新疆雪密度分布研究[J]. 中国农业气象, 2007, 28(4): 383-385.

新疆9个代表站点实测资料显示,新疆雪密度变化有明显稳定期和不稳定期之分,稳定期雪密度随时间呈斜率为0.0059的直线变化;进一步通过对新疆29个有雪密度观测资料的气象站40a(1964-2003年)气象资料进行聚类和回归分析,结果表明降水、雪深、大风、吹雪日数等因子与稳定期雪密度有显著或极显著正相关关系,从而建立了雪密度估算的气候模型.利用雪密度气候模型,以79个站雪密度为基础,采用mapgis65,结合地形、高程和卫星遥感数据,完成了新疆稳定期雪密度的空间分布图,该图是可用于计算的电子图集,可随时提取新疆及其任意区域的雪密度.

... 0 引言目前, 获取积雪密度的方法主要有: ①直接应用气象站积雪密度观测记录插值计算平均积雪密度^[1,2] ...

2007

0.0

黄慰军, 黄镇, 崔彩霞, 等. 新疆雪密度时空分布及其影响特征研究[J]. 冰川冻土, 2007, 29(1): 66-72.

对29个有雪密度观测的气象站40a气象资料进行聚类和回归分析。结果表明：降水、雪深、大风、吹雪等因子与雪密度有密切正相关关系，由此建立雪密度与气候因子关系模型。另选无雪密度观测的50个站40a气象资料，用关系模型计算出各站雪密度，从而使有雪密度值的站点增加到79个，为深入细致研究新疆雪密度时空分布打下了基础。研究表明，新疆雪密度有明显稳定期和不稳定期之分，稳定期雪密度是时间的函数。采用mapgis65，结合卫星遥感资料，研制了新疆雪密度(稳定期)空间分布图。新疆雪密度的分布呈现从盆地及其周边到山地及其周边最后到海拔3800-4000m以上高山带，随高度的升高雪密度依次升高。依据雪密度分布图和时间函数计算得出：稳定期新疆雪密度最大平均值为0.191g.cm-3。

... 0 引言目前, 获取积雪密度的方法主要有: ①直接应用气象站积雪密度观测记录插值计算平均积雪密度^[1,2] ...

1987

2.524

0.0

... ③利用星载扫描式多通道微波辐射仪 (SMMR)和特别微波辐射/成像计(SMM/I)等遥感数据获取积雪密度^[3,4,5,6] ...

2004

0.0

车涛, 李新, 高峰. 青藏高原雪深和雪水当量的被动微波遥感反演[J]. 冰川冻土, 2004, 26(3): 363-368.

Snow depth and snow water equivalent are the most import factors in the hydrologic model and climate model. So far, there is not an operational algorithm to estimate the snow water equivalent from passive microwave remote sensing data (SSM/I) in the Tibetan Plateau. In this study the SSM/I brightness temperature data in January 1993 are used to estimate the snow water equivalent SWE in the plateau. The frequencies of SSM/I data used to retrieve snow depth are 19 and 37 GHz in horizontal polarization. The results show that all available algorithms overestimate the snow depth in the Tibetan Plateau. In this paper the reasons of overestimation of snow depth from several aspects are analyzed, such as the water content of snow pack, large water bodies (e.g. lakes), and the abnormal field snow depth data. After eliminating some futile data (including the passive microwave brightness temperature values and snow depth data in the weather stations), an improved algorithm has been established to retrieve the snow depth from the difference of 19 and 37 GHz brightness temperatures in horizontal polarization. Here, snow density is obtained by a time function of fresh snow density. The snow depth and density were converted to the snow water equivalent, and are regarded as the ground truth. In finally, the TB vertically polarized differences of 19 and 37 GHz are regressed with the SWE. Using the statistical method, a simple and practical algorithm is developed to estimate the snow water equivalent from the differences of 19 and 37 GHz in vertical polarization.

Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou Gansu 730000, China

利用1993年1月份的SSM/I亮度温度数据反演了青藏高原的雪水当量,首先使用被动微波SSM/I数据19和37GHz的水平极化数据来反演雪深,根据积雪时间的函数来计算实时的雪密度,由雪的深度和密度计算出雪水当量.最后,利用SSM/I数据的19和37GHz的垂直极化亮度温度梯度对计算出的雪水当量进行回归分析,得到了利用SSM/I数据直接反演雪水当量的算法.

... ③利用星载扫描式多通道微波辐射仪 (SMMR)和特别微波辐射/成像计(SMM/I)等遥感数据获取积雪密度^[3,4,5,6] ...

2006

0.0

1.16

... ③利用星载扫描式多通道微波辐射仪 (SMMR)和特别微波辐射/成像计(SMM/I)等遥感数据获取积雪密度^[3,4,5,6] ...

2007

0.0

李新, 车涛. 积雪被动微波遥感研究进展[J]. 冰川冻土, 2007, 29(3): 487-469.

Research progress of the passive microwave remote sensing for snow cover at home and abroad in recent years was overviewed, and common passive microwave sensors SMMR（Scanning Multichannel Microwave Radiometer）, SSM/I（Special Sensor Microwave/Imager）, AMSRE (Advanced Microwave Scanning RadiometerEOS) and MWRI（Microwave Radiation Imager）were compared for each other. Some problems were found as follows: 1) The discriminant threshold of dry snow and wet snow is influenced by many factors that do not determined still. 2) The measured value of snow depth can not by reflected by meteorological situation. 3) The resolution is not high enough to used in monitoring the shallow snow zone information. 4) It seems to be overestimated for range and depth of snow and so on. In addition, the accuracy of the inversion results was affected by snow density, particle size, precipitation, cold desert, vegetation, permafrost and other factors. Aim to further study intensively, passive microwave remote sensing of snow cover was prospected as follows: 1) To operational monitoring, visible light and passive microwave data should be integrated; 2) Combination of remote sensing and geographic information systems would improve the accuracy and application breadth; 3) Improvements of the algorithm, higher spatial resolution data of passive microwave AMSRE, and passive microwave radiation imaginer MWRI of our FY3 satellite with the snow model would enhance importance of snow monitoring; 4) Range of snow monitoring is more consistent with passive microwave remote sensing and MODIS visible light.The passive microwave remote sensing can play an alternative role when visible light images can not be obtained satisfactorly due to more clouds; and 5) Passive microwave remote sensing still have more issues in operational snow monitoring, continuous monitoring of ground operations and Algorithm improvement. Its validation is an important guarantee to improve the technology, which need to be verified in snow storms.

被动微波遥感可用于进行全天候的积雪监测。概述了国内外近年来积雪被动微波遥感研究进展，比较了常用的被动微波传感器SMMR（Scanning Multichannel Microwave Radiometer）、SSM/I（Special Sensor Microwave/Imager）、AMSR-E(Advanced Microwave Scanning Radiometer-Earth Observing System)和MWRI（Microwave Radiation Imager）的参数特征，指出目前积雪被动微波遥感研究中存在的问题：1)干雪和湿雪的判别阈值受多种因素的影响，仍难以确定;2)各气象台站的实测雪深值缺乏代表性;3)分辨率不高，无法监测浅雪区信息等问题;4)容易造成积雪范围估计过大、雪深过深等问题。另外，雪的密度、粒径、降水、寒漠、植被、冻土等因素都在一定程度上影响了反演结果的精度。并对积雪被动微波遥感的应用前景提出了展望：1)在积雪遥感业务监测中，对可见光和被动微波数据进行融合，可充分发挥二者的优势；2)遥感和地理信息系统的结合，可以提高积雪遥感监测的精度和应用范围；3)随着积雪模型算法的进一步完善以及更高空间分辨率的AMSR-E和我国FY3卫星的被动微波数据MWRI的应用, 被动微波遥感必将成为积雪监测的重要手段。4)被动微波遥感与可见光MODIS的积雪监测范围较为一致, 当由于云的影响而无法得到满意的可见光影像时, 则被动微波遥感就可发挥替代作用。5)被动微波遥感在积雪业务监测中还存在较多的问题, 持续不断的地面监测和算法改进及验证是完善这项技术的重要保证。

... ③利用星载扫描式多通道微波辐射仪 (SMMR)和特别微波辐射/成像计(SMM/I)等遥感数据获取积雪密度^[3,4,5,6] ...

1996

0.0

史培军, 陈晋. RS与GIS支持下的草地雪灾监测试验研究[J]. 地理学报, 1996, 51(4): 296-305.

Snow cover is an improtant factor influencing animal husbandry in winter in the north part of China. The relationship among grass height,snow cover area and snow depth is a key to determine the harmfulness of snow disaster.So.using NOAA/AVHRR image to get the snow cover area and using snowfall and topographic indices from DEM to measure snow depth will provide a practical way to monitor large area snow disasters. In this paper,Xilingol League in Inner Mongolia is selected as a study area,and following problems are studies:1.Clouds and snow surfaces have similar characteristics in both visible and infrared spectra.How to identify them in determining snow cover area is necessary but difficult. Multi-temporal composite technique in processing NOAA/AVHRR image is presented and testified as an effective method in this paper.2. Two topographic indices of wind sheltering fetch and relief from DEM are put forward,which are main factors influencing snow depth.Then the model of the relationship between snow cover depth and snowfall and above indices is built,and the distribution pattern of snow cover depth is revealed.3. By overlaying snow depth map and grass height map. the harmfulness of snow disaster is evaluated in the study area.

冬季降雪是影响我国北方草原牧区畜牧业发展的重要因子。在草地雪灾监测和危险性评价过程中，确定积雪区域和积雪深度与牧草高度之间的关系是关键步骤。通过在内蒙古锡林郭勒草原的试验研究，可以认为：应用NOAA/AVHRR气象卫星资料确定积雪分布范围、应用地面气象站观测的降雪量资料及由DEM生成的地形参数确定积雪深度是在较大范围内进行雪灾监测的有效方法。在确定积雪分布范围时，提出并尝试应用多时相最小亮度合成法，来剔除在可见光和红外波段与积雪具有相似特征的云的干扰。文中还提出了两个主要影响积雪深度的地形参数F和R,用来刻画风对积雪量的再分配作用。通过建立地形参数、积雪深度和降雪量的统计模型即可估算积雪深度的分布。通过叠加分析研究区积雪深度图和牧草高度图进行了雪灾危险性的评价。

... 目前,利用可见光-短波红外甚高分辨率辐射计(AVHRR)和多时相中等分辨率成像光谱仪(MODIS)能在晴空条件下比较准确地获取较高分辨率的积雪覆盖信息^{[7,8,9,10,11,12,13,14,15]},但对利用光学遥感数据定量获取雪密度和雪水当量的研究甚少 ...

1998

0.0

王世杰. 利用NOAA/AVHRR影像资料估算积雪量的方法探讨[J]. 冰川冻土, 1998, 20(1): 68-71.

Snowpack volume in the Urumqi District and the mid-section of the Tianshan Mountains etc. is estimated from the NOAA/AVHRR data. Rectification of orientation, regional automatic windowing, zonal altitute superposition and measurement of snow depth and volume are suggested in this paper, along with their verification, supplementation and modification. The followings are mainly discussed: 1) Every step of comprehensive analysis on snowpack volume;2) Conversion formulas among snow area, depth and volume; 3) Error treatment; 4) Interannual difference in snowpack volume in the same season. For error treatment, some correction coefficients are presented.

结合实际考察的各种资料数据,在我们以往研究的理论技术、估算用数学模型和应用程序的基础上,提出和论述了估算航卫遥感影像信息中积雪量的整套计算机分析流程和处理新方法,并在技术措施和适用性上,进一步进行了验证、增补和修改。在时间上对不同年份而在同一季节时期局部地区的遥感积雪资料进行相对量的结果分析和比较,给出了一些实际数据图表。

2001

0.0

2004

0.0

延昊. NOAA 16卫星积雪识别和参数提取[J]. 冰川冻土, 2004, 6(3): 369-371.

Based on a spectral analysis of snow, soil, vegetation and cloud, in this paper it is indicates that the first two bands of traditional NOAA-AVHRR has trouble to distinguish snow from cloud, and the low reflectance of snow in the middle infrared spectral region can be used to distinguish snow from cloud. In this paper winter snow in northern China can be successfully detected by using newly 1.6 μm middle infrared band of NOAA16-AVHRR. Further, a method to estimate the snowpack parameters is presented, including snow time, snow area and snow depth. The difference of Band 1 and Band 3a for NOAA16-AVHRR is correlated with snow depth data from agri-meteorologic observation stations, and then an exponential fit equation of snow depth is built using curve regression. It is verified that the accuracy of the snow depth is good.

通过对积雪、地物和云进行光谱分析,指出传统的NOAA-AVHRR可见光和近红外波段进行云雪识别存在困难,而雪在红外波段的低反射性特点是区分云雪的一个可行途径.利用NOAA16气象卫星新增的1.6μm红外波段,对中国北方冬季的卫星积雪图象进行识别,结果显示,云雪可以准确区分.同时,提出了利用AVHRR资料估算积雪面积、积雪深度和积雪时间的方法,并对积雪深度进行了精度检验.

2004

0.0

梁天刚, 吴彩霞, 陈全功, 等. 北疆牧区积雪图像分类与雪深反演模型的研究[J]. 冰川冻土, 2004, 26(2): 160-165.

Establishing snow depth monitoring models using NOAA/AVHRR data received in the sunshiny weather condition, and dynamically monitoring the spatial distribution of snow on large scale during disaster taking place, have an important significance for comprehensively estimating and preventing the disaster caused by snow in pastoral areas. In this paper, four NOAA satellite digital images under sunshiny condition during two snow disasters from 1996 to 1997 and corresponding ground observation in 20 climate stations in the North of Xinjiang are used to study the approaches of identifying snow and clouds, and cutting off the image blocks with cloud. Meanwhile, a new method, coupled with geographic information system technology, is put forward to solve the problem of absent data for the image blocks with cloud. Using linear mixture spectrum disassembling method, the snow coverage rate and snow spatial classification are studied, and the remotely sensed monitoring models are simulated in the North of Xinjiang. Results indicated that: 1)The image cells covered by cloud may be identified and cut off from NOAA images of channels 1 and 2, and the spectral data of the cells covered by cloud can be restored using the image difference between channels 3 and 4 and GIS technology. Accordingly, the snow and cloud in the visible and near-infrared images can be identified. 2)The snow coverage rate based on image cells can be calculated by linear mixture spectrum disassembling method to improve the precision in snow classification. 3)Snow depth can be simulated by use of NOAA satellite data of channels 1 and 2. The study results will provide scientific basis for the construction of dynamically monitoring and analyzing, as well as integrated evaluating the condition of snow disaster in the pastoral areas.

利用NOAA/AVHRR晴空气象条件下的资料,建立积雪监测反演模型,动态监测雪灾期间大范围的积雪空间分布状况,对牧区雪灾综合评价及防灾救灾具有重要的意义.使用北疆地区1996-97年2次雪灾期间的4个时相的晴空NOAA卫星数据及20个地面气象台站观测资料,研究了云与雪的判识及图像去云处理方法,提出了云层覆盖下图像缺值插补处理的一种新算法;采用线性混合光谱分解方法,研究了基于像元的积雪覆盖率及积雪空间分类算法,模拟出北疆地区积雪深度遥感地学反演模型.研究结果可为牧区雪情动态监测、分析与综合评价系统的建设提供科学依据.

2002

4.456

0.0

2000

0.0

王健, 陈子丹, 李文君, 等. 中分辨率成像光谱仪图像积雪反射特性的初步分析研究[J]. 冰川冻土, 2000, 22(2): 165-169.

Among the information resources of remote sensing application, Moderate Resolution Imaging Spectroradiometer (MODIS) is more and more important because of its continence, wide wavelength and narrow bands. According to the spectral reflectance properties of different research objects, the optimum bands can be chosen for image processing and comparative analyzing. It is the first time that the data of MODIS were applied in the west of China. The origin data can be transformed to TIFF format using ArcView. It includes 26 bands, and every band has an image file, so all data are in 26 image files. The front 25 bands are in visible and near-infrared, with a resolution of 250 m. The last one is far-infrared, with a resolution of 500 m. In this paper, the snow cover in the Qilian Mountains has been studied by using the MODIS data. According to the principle of radiation and the theory of projection, the data were rectified firstly. Then a group processing file was built to obtain the values of the same points in order to plot a completely spectral reflectance curve from visible bands to near-infrared bands. And then based on the reflectance properties of snow in different bands, different kinds of snow training areas were built and certified. Using the same method, 572samples were chosen and they were classified into three kinds named old snow, coarse snow and firn. Their reflectance property curves wereplotted. After that, comparative analysis of spectral reflectance properties of snow between samples from the images and observation in situ have been operated and researched. The result shows: 1) In the range of 0.42～0.70 μm where snow reflectance is sensitive, the sample values radically conform to the observations, but in the range of 0.70～0.90 μm, the image sample values are all greater than the observations;2) The fluctuations of coarse snow and firn are smaller than that of old snow. So the spectral reflectance curves of same snow type are analogous for these trends, and the image classification of snow can be achieved by building signature file for each snow type and by supervised classification filter. MODIS data have a good relationship with the observations, especially in the periods of snow aging and refreezing, and training sample pixel files can be built based on the spectral reflectance properties. Therefore more and more calculating models for snow classification could be established aiming at image processing of different bands.

通过对中分辨率成像光谱仪图像26个连续波段的同点采样,获得从可见到红外光谱一个完整的积雪反射特性系列.在背景气候、气象数据信息的支持下,进行图像积雪反射特性与同类积雪实测反射特性的对比分析.结果表明:本研究所使用的中分辨率成像光谱仪图像积雪反射特性反映了积雪变化的过程,同时与地面光谱仪实测数据有一定的对应关系.同类积雪在可见光波谱区的光谱反射基本吻合.据此,反演积雪类型的图像分类是可行的

1995

4.769

0.0

2005

0.0

宋晓宇, 王纪华, 刘良云, 等. 基于高光谱遥感影像的大气纠正: 用AVIRIS数据评价大气纠正模块FLAASH[J]. 遥感技术与应用, 2005, 20(4): 393-398.

With its combination of good spatial and spectral resolution, visible to near infrared spectral maging from aircraft or spacecraft is a highly valuable technology for remote sensing of the earth' s surface. In practice, it is desirable to eliminate atmospheric effects on the imagery, a process known as atmospheric correction or atmospheric compensation. At present, there are many atmospheric correction software packages for imagery atmospheric correction, such as ATREM(Atmospheric REMoval program), ACORN(Atmospheric CORrection Now) and FLAASH(Fast Line of Sight Atmospheric Analysis of Spectral ypercubes). In this paper, the latest version of FLAASHatmosphere correction code derives its physicsbased algorithm from the MODTRAN4 radiative transfer code was introduced.and some comparisons of ith FLAASH-processed AVIRIS data, including results obtained using different processing options wereshowed. A preliminary evaluation to FLAASHalgorithm was done in this paper and the result showed thatthe new automated spectral recalibration algorithm, which has been incorporated into FLAASH, is an extremely valuable addition. It improve the quality of the output reflectance cube for measured radiance datacontaining wavelength calibration errors and it also can be used as a tool to measure varying wavelengthshifts in corss-track spatial dimension of an image, such are found in data from the AVIRIS sensor.

高光谱遥感影像由于集中了高光谱分辨率和高空间分辨率的优点,在对地观测中具有不可替代的优势。实际应用当中,往往需要从遥感影像获取地物的地表反射率信息,这就要求首先从影像中去除大气的影响,即进行大气纠正及补偿。目前,对遥感影像进行大气纠正的算法有很多,详细介绍了基于遥感影像自身信息的大气纠正模块FLAASH(Fast Line of Sight Atmospheric Analysis f Spectral Hypercubes)所涉及的算法,并利用该模块对AVIRIS航空遥感影像进行了大气纠正, 对不同的结果进行了分析对比,从而对该算法进行了初步的评价。

... 4 MODIS FLAASH大气校正参数输入FLAASH大气校正基于传感器处单个像元接受到的太阳波谱范围内(不包括热辐射)标准平面朗伯体(或近似平面朗伯体)反射光谱辐射亮度^[15] ...

2008

0.0

李国砚, 张仲元, 郑艳芬, 等. MODIS影像的大气校正及在太湖蓝藻监测中的应用[J]. 湖泊科学, 2008, 20(2): 160-166.

MODIS 数据有免费、波段丰富、时间分辨率高等优点,是进行太湖蓝藻监测的重要数据源,由于MODIS传感器接收的是地物反射太阳辐射的信号,太阳辐射与地球大气的相互作用会引起传感器接收到的信号失真,为了提高利用MODIS数据监测太湖蓝藻的精度,必须对其进行大气校正.本文介绍了FIAASH大气校正模型的基本原理,并对2007年4月25日MODIS数据的前七个波段进行试验,对比分析了影像大气校正前后的NDVI值以检测大气校正的效果;分析表明,大气校正前后NDVI的变化趋势基本上相同,但大气校正后的NDVI动态范围更大,校正后NDVI的平均值和标准差增大,大气校正在一定程度上有效地降低了大气对遥感影像的影响,达到了增强信息的目的;最后,利用大气校正获取的地表真实反射率数据的第二波段与第一波段的比值,运用阈值法提取蓝藻信息,经试验当阈值为1.9时提取出来的蓝藻分布图基本上与实际相符.利用MODIS影像可以快速、及时地监测蓝藻爆发的位置及爆发程度.

... 用于本实验的大气校正实验数据的输入参数见表1^[16] ...

... 能见度参数选择沙漠地区天气较好情况下的能见度^[16] ...