پديد آورندگان :
ابراهيمي خوسفي ، زهره نويسنده دانشجوي دكتراي بيابانزدايي، دانشكدهي منابع طبيعي و علوم زمين، دانشگاه كاشان , , ولي ، عباسعلي نويسنده استاديار دانشكدهي منابع طبيعي و علوم زمين، دانشگاه كاشان , , قضاوي ، رضا نويسنده استاديار دانشكدهي منابع طبيعي و علوم زمين، دانشگاه كاشان , , حق پرست، حامد نويسنده دكتراي محيط زيست، دانشگاه پونا، هند ,
كليدواژه :
ميانگين هندسي قطر ذرهها , بافت خاك , پاسخ طيفي , رگرسيون , مدل
چكيده لاتين :
Extended Abstract
Introduction
Identification of soil texture with special respect to its variability in specific location and stability in time periods affects many physical and chemical characteristics of soil and has been widely highlighted in geomorphology, agricultural and environmental sciences. Introduction of remote sensing techniques which has emerged approach towards science development and its high special resolution abilities, having multi spectral bands and processing vast range of data has brought, more facile and cost-efficient soil study techniques. (Casa et al., 2013) studied hyper spectral images and highlighted the importance of short waves infrared bands (SWIR) for soil texture identification; whereas (Feng et al., 2012) used MODIS images for preparation of soil texture maps. It is safe to say that there has been no study on geometric mean of particle diameter difference by means of ASTER multi spectral imaging in Iran. So, this paper describes the relationship between geometric mean of particle diameter and soil texture components (sand, silt, clay) versus ASTER multi spectral images. For this purpose Khatam plain in Yazd province, having 713.5 km2 was selected.
Methodology
The study was undertaken using Terra images taken in august 2007, ENVI4.2 software, SPSS 16, and Geographic information system. In the present study, Measurement of soil texture components was done in 2007/08/23, parallel to the date of imaging. In order to measure values of soil texture, 76 profiles in the khatam plain were selected randomly and drilled. The location of each profile was recorded using Extra Vista and Cx GPS set and transformed in to the map by means of Geography Information System. The soil samples were analyzed in the laboratory in reference to hydrometer method (Soil Survey Staff, 1996).then the values of silt,clay and sand variables were determined and these values were used in forming correlation models. Eventually, geometric mean of particles diameter were determined for each sample point. FLAASH algorithm was used to convert radiance values to ground reflectance and also to remove atmospheric hazes and after geometric correction, 3*3 mean filter was applied on the images .After that, some processing operations were done such as: Normalized Difference Vegetation Index, Principal Component Analysis and Soil Line Euclidean Distance Then the sampling points were crossed with all original and artificial bands and the pixel value of these points were extracted. To study the effect of soil texture and geometric mean particle on the ground reflection ,Multivariate linear regression was used within the estimated values of each layer and laboratory values in the training (56 samples). Finally, the accuracy of models have been evaluated by fitting the regression line between observations and estimated values in the validation data (21 samples).
Results and Discussion
The results indicated that the sand and geometric mean of particles diameter have the highest level of statistical significance associated with band3 (Pvalue < 0.05), but this relationship is reverse. Also, silt and clay have the highest level of statistical significance associated with band3 but this relationship is direct. Also, the results of this study showed that the particles of sand, clay, silt, and the geometric mean has a significant difference at 5% level with the soil line Euclidean distance. However, based on multivariate regression equations, there is no significant relationship between spectral reflectance values of other major bands and experimental values of the soil line Euclidean distance, silt, sand and dg, but there is significant correlation between clay with two bands 7 and 9 in the mid-infrared range with a relatively high coefficient and lower error. But it is important that the interaction coefficient in these bands has been enhanced determined coefficient. Therefore, evaluating the accuracy of models based on control points can have a significant role in the selection of the optimal model. Therefore we can say that only the near-infrared band (band 3) has an important role in the soil texture spectral in the khatam plain.
Conclusion
Generally, the results showed that Near Infra Red band of ASTER images Can be effective in determining the percentages of sand (with R2=0.5, SE=11. 9), silt(with R2=0.4, SE=10.09), clay(with R2=0.57, SE=3.46) and geometric mean of soil particles, (with R2=0.4, SE=0.09). Also, the results showed that the significant relation isn,t between spectral reflectance of the rest of original bands and SLED with values of silt, clay, sand and geometric mean of particles diameter on the soil At the end, was determined that these parameters have a remarkable effect on the spectral behavior of soil and we can use ASTER data to evaluate the physical properties of soil in the study region.
