The Quantitative Inversion and Space-time Variety Analysis o

合集下载
  1. 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
  2. 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
  3. 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。

The Quantitative Inversion and Space-time Variety Analysis of VWC Based
on the Remote Sensing Spectrum
YU-XIA LI
Institute of Geo-Spatial Information Technology University of Electronic Science and Technology of China,UESTC
Chengdu, China.
Email: liyuxia@
WU-NIAN YANG
Institute of RS&GIS,
Chengdu University of
Technology,CDUT
Chengdu, China
Email: ywn@
LING TONG
Institute of Geo-Spatial Information Technology
University of Electronic Science and
Technology of China,UESTC
Chengdu,China.
Email: tongling@
Abstract:The paper primary makes certain some bands or band combination as the RS spectral indices of vegetation water content based on analyzing the vegetation spectral characteristics and the relativity between spectrum and vegetation water content. With the ground survey data, the best function models are respectively established between the vegetation water content and spectral indices. Based on analyzing the different functions of the spectral index and its relative error RE, The paper confirms the spectral index1600820
SR=R/R as the characteristic parameters of the vegetation water content model in this study area. The monitoring model is established between VWC and SR spectral index, and the correlative physics parameters are educed by the means of RS spectrum and math statistics. According to the RS monitoring model of vegetation water content, ETM and ASTER remote sensing data, the quantitative inversion of vegetation water content were achieved by the program based on IDL7.0 platform. The measured data and background data surveyed in the study area is used to evaluate and analyze synthetically the inversion results. The study results show that the SR spectral index can eliminate the outside impact of the background environment, the canopy structure, and other factors. The precision of remote sensing inversion of vegetation water content is superior. It can truly reflect the time and space variety characteristics of the vegetation water content in the study area.
Keywords:Spectral Index, Quantitative Remote Sensing, Vegetation Water Content, Spectral Reflectance
1. Introduction
The quantitative monitoring and estimation of Vegetation Water Content can use a variety of means in the field, but the obtained information only reflects the situation of surrounding circumstances in the relatively small sampling points and short period [1]. If people want to obtain dynamical monitoring status of the vegetation water by using the field survey means in larger area, it need to dispose a large number of sample points and then have to last for a long time, which are not feasible on economy and technology. For large-scale vegetation water dynamical quantitative monitoring, Remote Sensing is a good way to meet the requirements of the space and time continuity [2].
As a result, the paper selects the method of quantificational Remote Sensing to monitor and estimate the Vegetation Water Content in Mao-Er-Gai district in the upper river of the Minjiang. Using large numbers of data based on the different equipments, example for ASD Field Spec ® Pro FR (350nm-2500nm), the paper analyzes a variety of data received in the experimental field, studies the relation between the models of vegetation water content and vegetation spectral reflectance.
2. Construct VWC Model Based on Remote Sensing Spectrometry
The reflectance spectrum curve of vegetation has apparent variation in ups and downs, and has the characteristics of more peaks and valleys[3]. in Fig.1, near the visible light 550nm (green)there is a wave crest whose reflection rate is 10% -20%, in the visible light 450nm (Blue) and 670nm (red) there are two absorption bands caused by chlorophyll absorption. In the near-infrared 800-1000nm, there is a steep slope formed by reflection on cell structure. In the near-infrared 1300-2500nm due to vegetation water content, the rate of absorption increase, reflection rate decrease and several water absorption bands were formed. The shape and characteristics of Vegetation, in different spectral paragraph, in different spectral reflection of the curve, form the physical basis where we can identify or distinguish other features in remote sensing images.
Wavelength(λ/μm)
r
e
f
l
e
c
t
i
v
i
t
y
(
α
/
%
)
Fig. 1 Effective spectral response characteristics of green plants(Mei an-xin,2001)2.1 Constructing Spectral Index
According to the vegetation spectral characteristics, reflection valley can be seen in the vicinity of 980nm, 1200nm, 1450nm and 1940nm, which are some sensitive bands of moisture[4]. Through a comprehensive analysis, the paper chooses the following bands to study correlation between vegetation water and spectral index with LOPEX'93 database.
By regarding 900nm as a reference band, the paper constructs the water index:
1970900
WI=R/R,
2950900
WI=R/R
(1)
Near 860nm and 1240nm the ratio index of reflectivity is treated by non-linear normalized index, and then normalized difference index of water can be obtained as follows:
8601240
8601240
NDW I=(R-R)/(R+R)(2) To 820nm and 1600nm ,that the ratio index defined by reflectivity:
2008 International Workshop on Education Technology and Training & 2008 International Workshop on Geoscience and Remote Sensing 978-0-7695-3563-0/08 $25.00 © 2008 IEEE
DOI 10.1109/ETTandGRS.2008.122
119
1600820SR=R /R (3)
Soil moisture adjustable index :
82016008201600SWAI=(R -R )*(1+L)/(R +R +L) (4) Normalized infrared index:
82016008201600II=(R -R )/(R +R ) (5)
The ratio index whose center wavelength located in 975nm and 1200nm can construct water index:
)975960-990920-94010901110Ratio =2*R /(R R −+ (6) )12001180-12201090-111012651285Ratio =2*R /(R R −+ (7)
The paper introduces the percentage FMC(Fuel Moisture Content) to express the vegetation water content[5], its expression is defined as:
100%FW DW FMC FW
−=× (8) Where FW represents fresh weight, which is the total
vegetation weight including the water quality; Where DW represents dry weight, which subtracts the water quality from the total vegetation weight.
With the database of LOPEX'93, the paper founds the optimal regress relations by using 52 samples of the FMC and the spectral index value. Through analysis, WI 1,R 2=0.901, has the highest correlation with FMC; then SR, R 2= 0.769; the worst relation is 1200Ratio , R 2 = 0.239. Selecting Another 20 samples to verify the statistics relationship among the different spectral index, and illustrate the accuracy of its statistics by using relative error RE, The results show that the different spectral indices have the different impact to FMC, and the relative error changes from 10.21% to 16.42%. SR has the best forecast accuracy for FMC, and the relative error is 10.21%; then II, the relative error is 10.29 %; the next are 1WI ,NDWI ,
SWAI ,975Ratio and 1200Ratio , as follows 10.38%, 10.86%, 11.56%, 11.97% and 12.01%. The worst accuracy is 2WI , and the relative error is 16.42 %.
The experiments show that the 1600nm is a strong water absorption band, and the 820nm is a weak water absorption band, whose ratio can enhance the water content information. When FMC is used to respond vegetation water content, the selection of near-infrared band combination, or combination of near-infrared and short-wave infrared band has a relatively better result, however, the SR spectral index is the most effective. In order to construct optimal model of FMC in the study area, and obtain the better inversion precision, the spectral index-SR is made as the model feature parameters.
2.2 FMC Modeling
According to the spectral and FMC data measured, and other reflective data, the paper establishes the optimal model between FMC (Y) and the spectral Index SR (X), and draws up the logarithm equation with least squares, which is expressed as a formula (9), and then gives the residual analysis (Figure 2).
0.1233()0.2735Y Ln x =−+ 20.9275R = (9)
Fig.2 The standard residual error distribution of the FMC model
There is a logarithm function relation between FMC and reflection spectrum index SR according to the formula (9). From the analysis of the residual error and standard residual error, the result shows that the residual error value is small and its distribution is even. Actual results and inversion results shows that the results are coincident between the actual measurement values and repeated testing variables (n = 20, the data showed
normal distribution), and have higher accuracy. In general, SR spectral index is able to better forecast vegetation water content. When the model relations are applied to test the data, the results shows that the RE value is small and has a equality distribution, and the model is very effective.
3. Inversion and Evaluation of VWC 3.1 Remote Sensing Inversion of VWC
According to the VWC inversion model, ETM image selects short-wave infrared 5-band (1550-1750nm) and near-infrared (775-900nm) 4-band after the atmosphere correction, and ASTER images chooses 4-band (1600-1700nm) and 3N-band (760-860nm); Through programming on IDL7.0 platform[6] ,the paper carries out the remote sensing inversion results of vegetation water content in the study area, FMC distribution plot is Figure 3 and 4.
By synthetically analyzing the diverse survey data and inversion outcomes in the study area, the paper educes that the VWC are closely related with the cover types, vegetation growth conditions, Vegetation coverage, soil types, soil moisture, elevation, gradient and other conditions of the investigation area. Based on the ETM data in the upper Minjiang Mao-er-gai region, statistic analysis shows that the minimum FMC is 0.593 %, the maximum is 80.869%, with an average of 39.867%. in Figure 3, Mao-er-gai river and other small rivers,its water content is also very high .in the inversion results chart, the performance is near-black green; in the lower right corner of the map there is a long strip of white region. According classification maps surveyed by remote sensing in the study area, it can be found that this is a harvesting regional tracks and the water content is relatively lower, so the performance of light white in the map is correct . However, in Figure4, for river and other small water system, the color is light white, this phenomena appears because the acquisition time of ASTER image was November, when the river water has been the ice of different thickness.
In general, the inversion results show that spectral index SR can better remove the influence of background and the canopy structure, and extract vegetation water content truly.
3.2 Results Evaluation
The paper utilizes two methods to evaluate the model inversion results. one uses the data measured from the 16 samples to evaluate the accuracy of the model inversion results in study region.The Figure 5 is the discrete plot which belongs the data measured and the model inversion results (ETM data) in the corresponding point, which denotes that it has the equality distribution and smaller error between the value measured and inversion value. Meanwhile, the individual data points measured with relatively large error may be due to the impact by the apparatus or man, when measured data was acquire and process. Figure 5 indicates that the results of VWC are basically the same between the inversion value and survey data, and the inversion model has higher precision.
Another method is taking the vegetation classification data surveyed in the study area as one of the basis to evaluate the VWC model. From the anterior analysis, it shows that vegetation water content has closely relation with vegetation types, soil types and water system distribution. By analyzing the VWC distribution and vegetation types plot, the RS model of
VWC is right from another side.
R1600/R820 Residual Plot
-0.2
0.20
Fig.3 FMC superposition of vegetation distribution and ETM inversion result
Fig.4 FMC superposition of vegetation distribution and ASTER inversion result
In Figures 3, the results of the superposition about the FMC distribution image of ETM data inversion and the distribution maps surveyed by remote sensing about ecological environment in the experimental area (the meaning of vegetation classification code can be seen in table 1) the results of the stack. Figure 4, the image pointed to the ASTER data, the others are like figure 3. The yellow line in the map is the type line of vegetation distribution; the red letter is the code for the type of vegetation. As can be seen from the chart, the inversion of vegetation water content and the distribution maps surveyed by remote sensing about ecological environment can be fit on in space consequence. For example, evergreen forest (classification code 312) distribution have more water content than natural grass (classification code 333) distribution; bush (classification code 331) has more water content than the shrub-grass transition zone (classification code 332) Regional distribution; the region covered by clouds and snow, Uncovered Rock (classification code 623) ,logging slash (classification code 342), and other non-vegetation have little water content. In Figure 3 and figure 4, the inversion results of vegetation water content can better reflect the actual situation of vegetation water content in the experimental area, and the inversion results are
superior.
Fig.5 comparison the VMC inversion data to the survey data
Chart 1 vegetation distribution chart surveyed by remote sensing about ecological environment in the study area Classification code 213 214 241 244 311 312 313 323 3231
Factual meaning Irrigated
land
Dry
fields
Inter-planting
orchard crops
agro-fores
try area
Deciduou
s forest
Evergree
n forest
mixed
forest
Shelter
forest
Young shelter
forest
Classification code 325 331 332 333 342 511 513 621 623
Factual meaning Other
forest
shrubber
y
Shrub-grass
transition zone
Natural
grassland
Cutover land river lake Wasteland Uncovered rock
4. Conclusion and Discussion
This paper analyzes the remote sensing data, ground observation data and moisture sensitive band of vegetation spectral characteristics. The results shows that the spectral index SR has a better forecast for FMC, and there are highly negative related functions between SR and FMC, its relevance R2=0.769. When the model is verified by the test data, the relative error is smaller, and RE is 10.21 %, and the error distribution is even. The spectral index SR can better remove the impacts of the background environment, canopy structure and other factors. According to the inversion model, ETM and ASTER remote sensing data, the paper achieves the quantitative inversion of VWC in the study area by programming on IDL7.0 platform, and the inversion accuracy is 88%. A large number of on-site verification and analysis shows that the inversion results can truly reflect the changes about the characteristics and pattern on time and space. In this paper, the test sample collection covers more than 10 vegetation types and a wider geographical distribution, so it can better represent the actual situation in the study area. If the study considers more biochemical parameters, extend the samples space, and add other factors, it benefits to further perfect the RS inversion model. Acknowledgment
This paper is supported by the 11th Five-Year Plan Item (MYHT0601), Youth Science and Technology Fund of UESTC.
References
[1].Zarco-Tejada, P. J.,Rueda, C. A., & Ustin, S. L. Water content estimation in vegetation with MODIS reflectance data and model inversion methods[J].Remote Sensing of Environment. Vol 85, Aug. 2003, pp.109-124.
[2].Tucker, C. J.. Remote sensing of leaf water content in the near infrared [J]. Remote Sensing of Environment,Vol 123, Oct. 2005, pp.129-132.
[3]. Sims, D. A., & Gamon, J. A. Estimation of vegetation water content and photosynthetic tissue area from spectral reflectance: A comparison of indices based on liquid water and chlorophyll absorption features [J]. Remote Sensing of Environment, Vol 84, Oct. 2002, pp. 526– 537.
[4]. Serrano, L., Ustin, S. L., Roberts, D. A., Gamon, J. A.. & Penuelas. Deriving water content of chaparral vegetation from AVIRIS data [J]. Remote Sensing of Environment, Vol 74, Aug. 2000. pp. 570– 681. [5].JIngjing-Dong,Zheng-Niu,Yan-Shen,etc.the comparison of the methods to extract water content of leaves by using the Reflectance spectrum information [J]. Journal of Jiangxi Agricultural University, Vol 28, Apr. 2006, pp. 588-593.
[6]. Hui-Lin,daisheng-Tang,Guanghua-Ye.The characteristic study of the main categories of spectrum in Zhuzhou city [J]. Journal of Central South Forestry college, Vol 23, Jan. 2003, pp. 94-98.。

相关文档
最新文档