Hyperspectral Surface Reflectance Data Detect Low Moisture Status of Pecan Orchards during Flood Irrigation

摘要

For large fields, remote sensing might permit plant low moisture status to be detected early, and this may improve drought detection and monitoring. The objective of this study was to determine whether canopy and soil surface reflectance data derived from a handheld spectroradiometer can detect moisture status assessed using midday stem water potential (psi(smd)) in pecan (Carya illinoinensis) during cyclic flood irrigations. We conducted the study simultaneously on two mature pecan orchards, one in a sandy loam (La Mancha) and the other in a clay loam (Leyendecker) soil. We were particularly interested in detecting moisture status in the -0.90 to -1.5 MPa psi(smd) range because our previous studies indicated this was the critical range for irrigating pecan. Midday stem water potential, photosynthesis (A) and canopy and soil surface reflectance measurements were taken over the course of irrigation dry-down cycles at psi(smd) levels of -0.40 to -0.85 MPa (well watered) and -0.9 to -1.5 MPa (water deficit). The decline in A averaged 34% in La Mancha and 25% in Leyendecker orchard when psi(smd) ranged from -0.9 to -1.5 MPa. Average canopy surface reflectance of well-watered trees (psi(smd) -0.4 to -0.85 MPa) was significantly higher than the same trees experiencing water deficits (psi(smd) -0.9 to -1.5 MPa) within the 350- to 2500-nm bands range. Conversely, soil surface reflectance of well-watered trees was lower than water deficit trees over all bands. At both orchards, coefficient of determinations between psi(smd) and all soil and canopy bands and surface reflectance indices were less than 0.62. But discriminant analysis models derived from combining soil and canopy reflectance data of well-watered and water-deficit trees had high classification accuracy (overall and cross-validation classification accuracy >80%). A discriminant model that included triangular vegetation index (TVI), photochemical reflectance index (PRI), and normalized soil moisture index (NSMI) had 85% overall accuracy and 82% cross-validation accuracy at La Mancha orchard. At Leyendecker, either a discriminant model weighted with two soil bands (690 and 2430 nm) or a discriminant model that used PRI and soil band 2430 nm had an overall classification and cross-validation accuracy of 99%. In summary, the results presented here suggest that canopy and soil hyperspectral data derived from a handheld spectroradiometer hold promise for discerning the psi(smd) of pecan orchards subjected to flood irrigation.