Publications - Faculty - Recently Submitted Faculty Publications 2008
<- Back
Parsons - Soil Science Society of America Journal
Wije Bandaranayake and L. R. Parsons
Abstract. The range of available water is narrow in central Florida sandy ridge soils (>95% sand). Therefore, the volumetric water content (θv) estimation needs to be precise for good irrigation control. Soil water estimates utilizing soil water sensors require an accurate calibration equation to obtain reliable θv estimates. In this study, a laboratory calibration was performed using Decagon Devices EC-5 soil water sensors in a Florida ridge sandy soil. The data were then regressed using an MS Excel Spreadsheet and SAS Statistical software. The accuracy of estimated θv (using the two regression equations and a simplified equation derived from the two equations) was then compared to that of θv values measured gravimetrically in the laboratory. Results indicated that the SAS generated regression equation is more accurate than the Excel generated equation. Reducing the number of decimal places from a linear regression equation can negatively affect the accuracy of estimates. A linear regression equation is more accurate between air-dry and field capacity θv , and when wetter than field capacity, sensor output rate will start decreasing. The estimated θv using the SAS generated regression equation within that θv range deviated from that of measured between +0.013 and -0.011 m3 m-3. In terms of plant available water for this soil, this range was 16.3% on the positive side and 13.8% on the negative side. The deviation of estimated θv from measured (i.e. an index of accuracy) varied for different sensors at the same θv value and for the same sensor at different θv values. The deviation of the estimated θv is considerable both when the soil is very dry and when the soil is very wet (close to field capacity θv). The negative effect is more critical if one waits to apply an irrigation when the soil is too dry (below 50% of available water).
