Open Collections

Using spectral reflection and multi-dye layer pixel values to quantify soil patterns for assessing field fertility conditions

University of British Columbia

The variability of soil chemical properties exerts a great influence on the practice of fertilization and other soil management. Quantitative reliable measurements of the soil variability are vital to the accuracy of fertilizer recommendations and the effective uses of fertilizer. The aim of the thesis is to determine whether soil classifications and variability assessments can be facilitated by the use of quantitative remote sensing techniques. An agricultural field with very contrasting soils was selected for this study and field variability in total and organic C, exchangeable cations, CEC, major fertility elements N, P, and K, soil water content and coarse fragments was examined using three different sampling techniques and laboratory analysis. The remote sensing techniques evaluated in this study were: 1) laboratory spectral reflection measurements of soil samples in the green, red and two near IR bands using a multi-channel radiometer, and 2) multi-dye layer pixel value analysis of digitized color aerial photos taken at the time of sampling. Conventional, selective and stratified random sampling techniques were used to quantify the soils in the field and although the variability in K, Ca, and P was high no significant differences were obtained in the mean values among the three techniques. Three distinct soil types were identified in the field, which included type I - very dark soils, type II - gravelly, very light colored soils, and type III - average brown or dominant soils. All three categories could be separated by Munsell value and chroma data. Significant differences in C, N, K, CEC, moisture content and coarse fragment content were obtained among the three soil types. Once the chemical data were translated into fertilizer requirements it became evident that soil type II (gravelly light colored soils) needed a higher K fertilizer rate than either type I or type III, thus suggesting that a differential fertilizer rate application within the field should be beneficial to crop performance. Correlation and regression studies of soil parameters with spectral reflection and dye-layer pixel values revealed the nature of the relationships between soil spectral properties and physical and chemical conditions. Significant correlations were found between reflectance values and most of the chemical parameters, and between pixel values, soil chemistry and moisture content. In both cases, % organic C showed the highest correlation. The results from stepwise regression and discriminant analysis revealed that organic C, water content and color value were the most-dominant soil parameters to influence spectral or pixel value variations. The relationship between water content and pixel value was significant suggesting that the variation in water content mignt be quantified by an analysis of dye-layer pixel values. Soil organic matter and soil color proved to be best predicted by laboratory reflectance measurements. Multi-variate cluster and discriminant analysis revealed that the soil types could be quantified with both spectral and multi-dye layer pixel value analysis and that the remote sensing data were best related to organic matter, soil color and soil moisture content in the field. The pattern of soil types in the field was determined visually by planimetry and by quantitative dye-layer pixel value analysis. The two results were found to be in close agreement and provided quantitative values for the spatial extent of the three soil types. These values were used to determine the total amount of fertilizers required for the field and the quantified spatial pattern is an excellent medium to facilitate soil sampling for fertilizer assessment for future cultivation.

Text

2010-07-08

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/26212

Soil Science

University of British Columbia

Graduate