Soil Mapping Unit Discrimination by Using Remote Sensing in Varamin Area of Iran.
Fereydoon Sarmadian and Kamran Moravej. Soil Science Dept Faculty of Soil and Water, Univ of Tehran, Daneshkadeh Street, Karaj, Iran
In this research, we used digital image processing techniques produced from T.M sensor in order to make soil map units. This region is part of Varamin plain and watershed catchments of Jajroud River. Based on the information obtained from the nearest weather station, the average yearly rainfall for the region is between 145 – 150 mm and the average of temperature is 18 oc, too. Soils of region are classified in two Aridisols and Entisols orders (U.S.D.A, 2003) and based on the F.A.O method are often Fluvisols and Cambisols and for some units are Calcisols, Gypsisols and Solonchaks. Soil map produced using the maximum likelihood classification method. Agreement ratio between soil map derived from this manner and ground truth map derived from traditional methods and its Kappa index were 82% and 75% respectively. Analyzing the matrix error of this research shows that unit 2and 9 have not well delineation from unit 6 and 7 respectively(deliniation of Haplocalcids and Calcaric Fluvisol from Haplocambids and Calcaric Cambisols). In other hand, the presence of plentiful silt in topsails that has high spectral reflectance causes that there are often problem in digital image processing, especially image classification (because spectral reflectance of plentiful silt is similar to saline soils).This research suggest in soil science studies beside selection suitable spectral band, data were used that have not vegetation. Also, we can use images that have plant cover, simultaneously. It is better digital images were generated by other sensors for example SPOT or I.R.S and also other classification methods for instance classification based on coordination of condition were studied more further than classification based on difference of spectral reflectance.Key words: digital processing, T.M sensor, soil map units, agreement ratio, matrix error, Kappa index, maximum likelihood classification, spectral reflectance.