THE TEMPORAL CHANGING INFORMATION EXTRACTION
ON SPECTRAL REFLECTANCE AND GROWTH PARAMETERS OF TYPICAL CROPS
Xiao Yueting, Song Jinling, Wang Jindi
School of Geography and Remote Sensing Science, State Key Laboratory of Remote Sensing Science, Beijing Key Laboratory of
Environmental Remote Sensing and Digital City, Beijing Normal University, 100875, Beijing, China
xiaoyueting2099@163.com, wangjd@bnu.edu.cn, songjl@bnu.edu.cn
Commission VII, WG VII/3
KEY WORDS: A Priori Knowledge, Spectral Reflectance, Leaf Area Index, Vegetation Index, Model Inversion
ABSTRACT:
When we estimate the land surface variables by remote sensing model inversion, a priori knowledge is needed for given the initial
values of the model parameters to be retrieved. In this paper, based on the measured datasets accumulated in the Spectrum Database
of Typical Land Objects, we analyzed the measured data of typical crops and presented the parameters prior knowledge as the
statistics data on canopy spectral reflectance and leaf spectral reflectance. We calculated the mean values and variances for
constructing the initial values and uncertainties of model parameters. By taking the winter wheat as the typical crop, with the ground
measured crop canopy spectral reflectance and leaf spectral reflectance data in the spectrum database, integrating the spectral
response function of Beijing-1 multi-spectral sensor, we got the needed a priori spectral knowledge in three spectral band (Green:
523nm—605nm, Red: 630nm—690nm and Near Infrared: 774nm—900nm) and achieved the transformation from narrow band to
broad band in order to get the corresponding vegetation index, such as RVI,NDVI,SAVI and so on. With this method, we obtained
the statistical relationship between vegetation index and LAI for estimate LAI using Beijing-1 data. The Beijing-1 multi-spectral data
with DN value is transformed into different vegetation index. Then we could get LAI data from Beijing-1 multi-spectral image with
the statistical relationship expressed above. The data and methods presented in this paper could be considered as valuable reference
while estimating model parameter’s initial value and its associated uncertainty in physical model inversion.
1. INTRODUCTION
New views are provided to understand the earth system more
comprehensively for people via remote sensing ways. Beijing-1
small satellite was launched in Oct. 10, 2005 in China
successfully. Two type of sensors were carried on the satellite,
the 3-band multi-spectral sensor with 32m spatial resolution and
the the panchromatic sensor with 4m spatial resolution.
Preliminary processing system has been developed for Beijing-1
data receiving, preprocessing, and distribution (Han Dong,
2007). Some relative research works are needed for retrieving
more information on land surface.
Taking the remotely sensed data application in Beijing region as
sample, MU Fengyun concluded the land use characteristic of
Beijing between 1984-2005 urbanization based on three
Landsat TM images of 1984, 1996, 2001 and one “Beijing-1”
data obtained in 2005, which shows that “Beijing-1” data could
service as the data base for the studies on the city planning, the
ecological environment monitoring, etc (Mu Fengyun, 2007).
Yang Shengtian integrated “Beijing-1”, SPOT5 and QuickBird
images to detect and analyze the vegetation fractional coverage
in riparian buffer zones of Guanting Reservoir. For reed
marshes, compared with the result of using vegetation fractional
coverage calculation equation, the mean absolute error of
vegetation fractional coverage measured by “Beijing- T’image
was reduced by 22.7 % after applying statistic models (Yang
Shengtian, 2006). Wangqian studied “Beijing-1” satellite image
application potential in land use field, showed that “Beijing-1”
image quality is fine and land use classification precision is high
(Wang Qian, 2007).
For applying broadband and high spatial resolution remotely
sensed data in agricultural applications at the field measurement
scale, an improve method was developed to evaluate commonly
used broadband vegetation indices (Vis) for the estimation of
LAI with VI-LAI relationships. For the agricultural applications,
MSAVI-LAI relationships are easy-to-apply and reasonably
accurate for estimating LAI. (Wu Jindong, 2007). Researchers
also have found that the broadband SAVI2 index is least
affected by background reflectance for both LAI and CCD
estimation, and is also the best predictor of LAI. RVI is
marginally better than SAVI2 for CCD estimation in terms of
canopy effects, but it becomes increasingly sensitive to
atmospheric effects and solar zenith angle with increasing
vegetation densities. Further, SAVI2 proved to be least affected
by illumination geometry changes (N.H. Brogea, 2000).
In this paper, based on the spectrum database of typical land
surface objects, we first statistically analyze the collected
data(including five typical crops, such as winter wheat, summer
maize, cotton, rice and rape) and construct the prior knowledge
of the canopy and leaf reflectance(obtaining the mean values
and variances of measured canopy reflectance). Then we
applied the prior knowledge to estimate LAI by using VI-LAI
relationship from “Beijing-1” images obtained in 2006.
2. MATERIALS
2.1 Spectral Reflectance Datasets
The five typical crops included in this study are winter wheat,
summer maize, cotton, rice and rape. These datasets come from
the Spectral Database System of Typical Objects in China