Soil Property Analysis using Principal Components Analysis, Soil Line, and Regression Models

where SOC is the surface organic content; R, G, and B are image intensity values in the red, blue and green Past research has attempted to relate surface characteristics of soils bands, respectively; and a, b, c and d are curve-fit paramto reflectance from remotely sensed images to provide a means for quantifying spatial heterogeneity. Existing procedures have proven eters. The Chen et al. (2000) regression technique has valuable, but no research has been performed to compare these techa drawback in that this technique requires field specific niques. The objective of this research was to compare existing methodregression parameters that change significantly deologies, that is, principal components analysis (PCA), Chen et al.’s pending on parent material. Chen et al. (2000) reported regression model, and the soil line Euclidean distance (SLED) techthat the relationship between soil OM and reflectance nique, for quantifying spatial heterogeneity in soil surface organic was poor when measurements were taken across large matter (OM) and cation exchange capacity (CEC). The three existing geographical areas, suggesting that the difficulties may techniques were compared using five bare soil images from three be the result of different types of parent materials (Ferdifferent silt loam to loam fields in the Midwest USA. At the same nandez et al., 1988; Henderson et al., 1992; Schulze et time as image acquisition, surface (upper 2.54 cm [1 in]) soil properties al., 1993). were measured in situ. Organic matter and CEC were highly correlated (R 2 0.70) to the first principal component (PC1) for three bare soil Another technique, called SLED, was proposed and images, moderately correlated (R 2 0.40) for one image, and only evaluated by Fox and Sabbagh (2002). This technique slightly correlated (R 2 0.25) for the final image. The lower correlaestimated soil OM from R and NIR image intensity tions were hypothesized to be because of the range in the soil OM values by using the soil line concept. The soil line is a and CEC and image exposure. Principal Component 1 accounted for widely researched linear relationship between reflecapproximately 95% of the total variance in all the fields; therefore, tance or image intensity in the R and NIR wavelengths no correlation was observed between the upper 2.54 cm (1-in) surface (Campbell, 1996; Baret et al., 1993; Richardson and soil properties and the second, third, or fourth principal components Wiegand, 1977): (PC2, PC3, and PC4, respectively). All three techniques equivalently predicted OM and CEC. However, PCA does not require field-specific NIR R [2] regression or soil lines parameters. It is also suggested that PC1 can where NIR and R are near-infrared and red reflectance replace the soil line in a technique for identifying soil-sampling locations. values, is the soil line slope, and is the soil line intercept. The soil line extends from a lower region representing darker soils to an upper region having high R and NIR values representing the brighter soils within T relationship between soil OM and remotely the field. Pixels with reflectance values to the left of the sensed measurements has been the subject of consoil line correspond to the vegetation (Curran, 1983). siderable research (Baumgardner et al., 1970; Al-Abbas The SLED technique requires the identification of et al., 1972; Vinogradov, 1982; Shonk et al., 1991; Chen the minimum point along the calculated soil line. The et al., 2000; Fox and Sabbagh, 2002; Fox et al., 2004; minimum point refers to the pixel with the lowest R Hong et al., 2004). Numerous researchers have atand NIR intensity values, corresponding to the lefttempted to relate surface characteristics of soils to remost extreme point on the soil line and representing flectance from remotely sensed images thus providing the darkest soils within the field. The routine then calcua means for quantifying spatial heterogeneity without lates the distance (D) of each pixel’s intensity values the collection of a large number of in situ soil samples. away from the soil line’s minimum point. The SLED Baumgardner et al. (1970) and Al-Abbas et al. (1972) technique relates back to the properties that influence performed the first airborne experiments to study the the soil line (i.e., soil texture, soil moisture, soil roughness, relationship between OM and reflectance in the visible etc.), and therefore, overcomes the difficulties associand near-infrared (NIR) wavelengths. More recently, ated with the Chen et al. (2000) regression technique. Chen et al. (2000) proposed a technique which relates The SLED technique also lends itself to a methodology surface organic C (SOC) content in the upper 15 cm of for less intense soil sampling. the soil profile to image intensities in the red (R), green Evaluation of the SLED technique included in situ (G), and blue (B) bands of the visible spectrum: soil samples of OM from the upper 2.54-cm (inch) of SOC exp(a bR cG dB) [1] the soil profile and digital, aerial, bare-soil images of two fields in the Midwest, USA. Fox and Sabbagh (2002) G.A. Fox and R. Metla, Dep. of Civil Engineering, Univ. of Missisused an exponential function to characterize the relasippi, University, MS 38677-1848. Received 22 Nov. 2004. *Corresponding author ([email protected]). Abbreviations: B, blue; CEC, cation exchange capacity; DCL, data collection location; G, green; NIR, near infrared; OM, organic matter; Published in Soil Sci. Soc. Am. J. 69:1782–1788 (2005). Pedology PC1, first principal component; PC2, second principal component; PC3, third principal component; PC4, fourth principal component; doi:10.2136/sssaj2004.0362 © Soil Science Society of America PCA, principal component analysis; R, red; SLED, soil line Euclidean distance; SOC, soil organic carbon. 677 S. Segoe Rd., Madison, WI 53711 USA 1782 Published online September 29, 2005