|Chesson, Joseph - RETIRED USDA-ARS|
|Ojala, John - RETIRED USDA-ARS|
Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: March 16, 2008
Publication Date: May 2, 2008
Citation: Detar, W.R., Chesson, J.H., Penner, J.V., and Ojala, J.C. 2008. Detection of soil properties with airborne hyperspectral measurements of bare fields. Transactions of the ASABE 51(2):463-470. Interpretive Summary: Typically soil survey maps show very large field areas as having the same soil type. For precision farming, however, there is a need for soil information as it changes from yard to yard. Cameras capable of measuring the amount and quality of light and radiation reflected from a field surface were mounted in an airplane and flown at an elevation of 7200 ft over recently-planted and nearly bare cotton fields. Data from the cameras and from soil samples were analyzed with sophisticated image analysis and statistical software and a mathematical formula was discovered capable of determining the percent sand (and several other properties) at 4 ft intervals over two farm-size fields, with a total area of 128 ha. This information can eventually be used for predicting differences in irrigation management, growth regulators, salinity treatment, and fertilizer applications throughout the field.
Technical Abstract: Airborne remote sensing data, using a hyperspectral (HSI) camera, were collected for a flight over two fields with a total of 128 ha. of recently seeded and nearly bare soil. The within-field spatial distribution of several soil properties was found by using multiple linear regression to select the best combination of wave bands, taken from among a full set of 60 narrow bands in the wavelength range of 429 to 1010 nm. Two-, three-, and four-term predictive equations made it possible to calculate the value of the soil property at every pixel, with a spatial resolution of 1.2 m. Both surface and subsurface soil samples were taken from the center of 321 equal-sized grids and tested in a laboratory for 15 properties. The percent sand in the surface samples was the easiest to detect, with a coefficient of determination of 0.806 for the four-term model; the best combination of wavelengths was 627, 647, 724, and 840 nm. The correlation for silt, clay, chlorides, electrical conductivity, and phosphorus, was a little lower. The poorest fit was found with carbon, organic matter and saturation percentage. Intermediate results were obtained with pH, Ca, Mg, Na, K, and bicarbonates. An image map is presented showing the percent sand at every pixel location in one field. A common spectral index was found for 5 of the properties, EC, Ca, Mg, Na, and Cl, possibly related to the salinity effect on surface color and roughness. The main finding was that an HSI flight over nearly bare soil can lead to a fine-resolution soil map for some soil properties; this map should be very useful to site-specific farm management.