|
|
|
Poster Session
|
Estimation of Areal Evapotranspiration Using Landsat TM Data Alone
Yongfen Wei and Kimiteru Sado
Department of Civil Engineering, kitami Institute of Technology
Koen-cho 165, Kitami, 090 Hokkaido, Japan
Abstract
The objective of this study is to investigate the possibility of using landsat TM 7 bands' data alone to estimate evapotranspiration (ET) over large areas. To achieve this purpose, various regression analyses were conducted, and the relationship between these data from satellites and the values of ET calculated from Morton or Penman equation was established. In addition, to examine the normalization of residual terms, the significance of regression analyses and the model';s linear quality, regression diagnosis was also carried out. As a result, the reliability of using only Landsat TM 7 bands' data to estimate ET over large areas was elearly demonstrated. Moreover, this study also made elear that of the totao 7 bands examined, the band 6 and 3 were most important for the estimation of ET.
Introduction
To utilize water resources more effectively, quicker and more accurate estimation of ET is considered to be very important. Methods based on micro-meteorological observations and heat balance models, or methods with the combined use of remote sensing data have been generally used for ET estimation. However, as these methods require a large number of observations and parameters, the estimation of ET over large areas is considered difficult. On the contrary, as remote sensing data can cover large areas, using these data to estimate the ET of such areas can be considered possible. In this study, the estimation of the ET over large areas by application of landsat TM data alone was tried, based on the considerastion that certain correlations exist between remote sensing data and the areal ET, and from the models established in regression analyzing, the polssibility of using Landsat TM 7 bands' data alone to estimate the areal ET was demonstrated.
Studied Areas and Estimation Methodology
1) Studied Areas
In this study, Tokoro river basin and abashiri river basin , both of which are located in the north - easterm of Hokkaido, Japan , with a length of 120 and 115 km , an area of 1930 and 1380 km2 , respectively , as well as a part of Kitmai city with an area of 114km2 located in the area of Tokoro river basin , all of which are shwon in fig.1, were used as the studied areas.
Fig 1 Studied areas
Landsat TM data of these areas collected on may 19,1989 were utilized and the corresponding land cover classifications by the supervised land classification method were conducted . The total number of pixels for Tokoro and Abashiri river basin were 34,003and 24,540 respectively , with each pixel possessing an area of 250m* 250m . For each of these river basins , eight land cover categories were obtained , with forest, grassland , plowed field , and the others including urbon area , residential area , bar ground , paddy field and water occuping 56.97 ,22.98,6.2 and 13.85% of the total area ,respectively , for Tokoro river basin and 42. 80 38.85 5.11 and 13.24% respectively, for Abashiri river basin. For Kitami city the total number o mash was 1277 (1 mesh=10x10 pixels) with each mesh having an area of 300m x 300m, and of which the ocupation of forest wasx 28.36% residential area 30.5% bare ground 15.47% the others including urban area, paddy field gressland and plowed field 26.12%.
2) Calculation of NDVI and Ground Surface Temperature
a) Calculation of NDVI : NDVI , used as an index for the evaluation of the distribution condition of the distribution condition of vegetation, was calculated by the following equation.
NDVI = (NIR-VIS) / (NIR +VIS)
In which NIR and VIS represent the CCT values in the near infrared range (band 4) and in the visible or visible red range (band 3) respectively.
b) Calculation of the ground surfacr temperature: As band 6 can detect the tempreratures, it is called as the thermal band. The temperatures in this study was calculated from the following equation proposed by K. Sado, et al. (1993) which was established according to the ground surface temperatures measured through ground truth.
H = 22 50 + 0.2643 V ……………..(1)
In which , H means the ground surfacr temperature (°C) and V the CCT value of band 6
3) Estimation of ET U sing TM data
The real ET or the potential ET calculated from Morton or Penman equation for each pixel (mesh) was used as a dependent variable and the data collected from Landsat TM as an Explanatory variable, The simple and multiple regression analyses were conducted, based on the assumption that a certain linear correlation existed between them Four models, as represented in the following are presented:
|
|
|
|
|
|
|
