Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6408329 | Geoderma | 2016 | 11 Pages |
Abstract
Soil salinity is a global threat to agricultural production and necessitates the monitoring thereof. Saline conditions in South African irrigation schemes generally occur in small patches (some only a few metres in diameter) and this study, which forms part of a Water Research Commission (WRC) project, evaluates the use of very high resolution (VHR) satellite imagery, in particular those produced by the WorldView 2 (WV2) sensor, for the detection of salt accumulation in irrigated areas. A range of features derived from the WV2 image were evaluated, namely 8 WV2 bands, 10 vegetation indices (VIs), 25 texture measures and 2 principle component analysis (PCA) components. These features were generated at six spatial resolutions (0.5Â m, 2Â m, 6Â m, 10Â m, 15Â m and 20Â m) to investigate the value of high spatial resolution for detecting affected areas. The relationships between the image features and electric conductivity measurements of 30 soil samples were studied by means of regression analysis and classification and regression tree (CART) modelling. The regression analysis results showed that a spatial resolution of 6Â m or higher is ideal when VIs are used as input. When texture measures are used, higher spatial resolution (0.5Â m) produced better models. The regression results also showed that the relatively high spectral resolution of the WV2 sensor (compared to other VHR sensors) did not provide a significant improvement in accuracy. The CART results provided a categorical quantification of salt accumulation and showed that VIs generated at 0.5Â m resolution were the best features to use. Because the use of WV2 images is not financially viable for operational use in very large irrigation schemes, the study concludes with some guidelines on the spectral and spatial requirements of images when monitoring salt accumulation in irrigation schemes with similar conditions.
Keywords
EVISAVIROCsPLSNIRSoil-adjusted vegetation indexWorldView 2BREVHRCARTAUCPCAVIsvilImage textureDecision tree analysisPrincipal component analysisRandom forestsPartial least squaresDecision treeRegressionRemote sensingnormalized difference vegetation indexEnhanced Vegetation IndexNDVIVegetation indicesSalinityClassification and regression treeSVMSupport vector machinesarea under the curveNear-infraredNearest neighbourelectric conductivityVery high resolutionreceiver operating characteristics
Related Topics
Physical Sciences and Engineering
Earth and Planetary Sciences
Earth-Surface Processes
Authors
Sybrand Jacobus Muller, Adriaan van Niekerk,