associated to moisture and organic

associated to moisture and organic matter (Baret et al., 1993), and
departures from the soil line are in turn strongly related to biophysical
parameters such as the Fraction of Green Vegetation, FGV, or the
Fraction of Absorbed Photosynthetically Active Radiation, FAPAR
(Pinty and Verstraete, 1992). The soil line is therefore a constraint in
the R/NIR spectral space that greatly contributes to the design of new
vegetation indices that are insensitive to the soil background while
remaining responsive to vegetation (Pinty et al., 2008). Examples of
improved alternatives to the traditional Normalized Difference
Vegetation Index, NDVI (Rouse et al., 1973) are the Perpendicular
Vegetation Index, PVI (Richardson and Wiegand, 1977), the SoilAdjusted Vegetation Index, SAVI (Huete, 1988) and the Global
Environment Monitoring Index, GEMI (Pinty and Verstraete, 1992).
However, to the best of our knowledge, no similar constraint has
been found in the MIR/NIR space, a circumstance that may have
impaired the design of optimal vegetation indices, which have been
heuristically derived from indices already developed in the R/NIR
domain. This is the case of VI3 (Kaufman and Remer, 1994), a
modification of NDVI, as well as of GEMI3 (Pereira, 1999) that directly
resulted from GEMI. As pointed out by the developers of VI3 and
GEMI3, the derivation of the indices was primarily based on the fact
that MIR and R reflectance are strongly correlated. On the other hand,
as also stressed by the authors, the processes that govern reflectance
in R and MIR are not expected to lead to similar results and the
existence of other processes that may change reflectance in the two
channels cannot be ignored.
The aim of the present paper is to investigate the possibility of
defining a transformation in the MIR/NIR space that leads to an
enhancement of the spectral information about vegetation. For this
purpose, and taking into account the methodology suggested by
Verstraete and Pinty (1996) to design optimal indices, a new space is
proposed and an appropriate coordinate system is then defined that is
suitable to discriminate vegetation and is sensitive to its water
content. The rationale adopted may be viewed as comparable to that
followed to derive the tasseled cap transformation (Crist and Cicone,
1984; Kauth and Thomas, 1976; Cohen et al., 1995), where a new
coordinate system is introduced in order to optimize data for
vegetation studies. Using satellite imagery, it will be then shown
that the proposed coordinate system is particularly appropriate to
operationally monitor vegetation and to detect vegetation changes, in
particular those caused by droughts and fire events.
Accordingly, the three specific goals of the present study may be
stated as follows:
1. To study the possibility of defining a transformation in the MIR/NIR
space leading to an enhancement of the spectral information about
vegetation;
2. To define a new coordinate system representing an improved
combination of the MIR and NIR channels when the two spectral
bands are used to detect vegetation changes, in particular those
caused by droughts and fire events;
3. To assess the added value brought by the proposed coordinate
system when applied to real satellite data.
2. Data
The present study relies on data from remotely-sensed observations, as well as from laboratory measurements. Remotely-sensed
observations were gathered over two main Brazilian biomes, namely
the Amazon Forest and the Cerrado region (see Fig. 1 and Table 1) as
covered by 16 Landsat ETM+images. Data consist of top of the
atmosphere (TOA) values of MIR radiance, NIR reflectance and
thermal infrared (TIR) brightness temperature, acquired by the
Moderate Resolution Imaging Spectrometer (MODIS) instrument
on-board Terra satellite during the year of 2002, together with the
respective solar zenith angles. Data were obtained from the Terra/
MODIS Level 1B 1 km V5 product, MOD021 (MCST, 2006) and
correspond to channels 2 (centered at 0.858 μm), 20 (centered at
3.785 μm), and 31 (centered at 11.017 μm). Surface values of MIR
reflectance were then retrieved by applying the methodology
developed by Kaufman and Remer (1994), paying special attention
to the possible drawbacks previously pointed out by Libonati et al.
(2010).
Validation of results from the analysis performed on MODIS images
was mainly carried out based on ETM+imagery. Direct validation
of results in the MIR domain is, however, a difficult task because of
the lack of “in-situ” (direct) measurements of MIR reflectance. This
limitation may be partially circumvented by laboratory measurements

Results (English) 3:

水分和有机质结合（巴雷特et al.，1993），和从土壤中的偏离是在密切相关的生物物理如绿色植被，分数和参数，或分数的吸收光合有效辐射，FAPAR（pinty和是1992）。因此，土壤线是一个约束近红外光谱空间，大大有助于设计新的植被指数对土壤背景不敏感其余响应的植被（pinty et al.，2008）。实例改进的替代传统的标准差植被指数NDVI（Rouse et al.，1973）是垂直的植被指数PVI（理查德森和韦根，1977），该土壤可调节植被指数SAVI（韦特，1988）和全球环境监测指标，就（pinty和是1992）。然而，到我们所知，没有类似的约束在红外/近红外空间的情况下，可能会发现，受损的最佳植被指数的设计，已启发来自R /近红外发展指标域。这是VI3案例（考夫曼和雷纳，1994），一个修改的NDVI，以及gemi3（佩雷拉，1999）直接由于工作。指出通过VI3开发商和gemi3，指数的推导是基于事实那我和R反射密切相关。另一方面，正如作者所强调的，治理反射的过程在R和我预计不会导致类似的结果，存在的其他过程，可能改变反射率在2渠道不能被忽视。本文件的目的是调查的可能性在红外/近红外空间，导致一个定义的变换植被光谱信息的增强。这目的，并考虑到建议的方法（1996）pinty Verstraete和设计最佳的指标，一个新的空间提出并定义了一个适当的坐标系适用于区分植被，对水敏感内容。所采用的理由可以被看作是可比的然后导出缨帽变换（克里斯特和cicone，1984；考思和托马斯，1976；科恩等人，1995），在一个新的为了优化数据，引入了坐标系植被研究。使用卫星图像，它会显示建议的坐标系统是特别合适的在植被监测和检测植被的变化，在特别是由干旱和火灾事件引起的。因此，本研究的三个具体目标可能是陈述如下：1。在研究红外/近红外定义转换的可能性空间导致的光谱信息的增强植被；2。定义一个新的坐标系，表示一种改进的组合的红外通道时，两个谱乐队是用来检测植被的变化，特别是那些由干旱和火灾事件引起的；3。要评估拟议的协调所带来的附加价值应用于真实卫星数据的系统。2。数据本研究的数据来自遥感观测，以及从实验室测量。遥感观察两主要聚集在巴西的生物群落，即亚马逊森林和稀树草原区（见图1和表1）作为覆盖16 Landsat ETM+影像。数据包括顶部大气（TOA）红外辐射值，近红外反射率热红外（TIR）亮度温度，得到的中分辨率成像光谱仪（MODIS）仪器星载地上卫星，在2002年，连同各自的太阳高度角。数据从土地/MODIS 1B 1公里V5的产品，mod021（MCST，2006）和对应渠道2（以0.858μm为中心），203.785μm），31（以11.017μm为中心）。米尔的表面价值反射率，然后检索应用的方法由考夫曼和里默发达（1994），特别注意对可能存在的弊端曾指出的利博纳蒂等人。（2010）。从分析MODIS图像进行结果验证主要是进行基于ETM+影像。直接验证在我的领域的结果，然而，一个困难的任务，因为现场“缺乏”（直接）红外反射率的测量。这限制可能部分通过实验室测量