Mapping olive varieties and within-field spatial variability using high resolution QuickBird imagery

Abstract

[Abstract]: The growth of the Australian olive (Olea europaea L.) industry requires support from research to ensure its profitability and sustainability. To contribute to this goal, our project tested the ability of remote sensing imagery to map olive groves and their attributes. Specifically, this study aimed to: (a) discriminate olives varieties; and to (b) detect and interpret within-field spatial variability. Using high spatial resolution (2.8m) QuickBird multispectral imagery acquired over Yallamundi (southeast Queensland) on 24 December 2003, both visual interpretation and statistical (divergence) measures were employed to discriminate olive varieties. Similarly, the detection and interpretation of within-field spatial variability was conducted on enhanced false colour composite imagery, and confirmed by the use of statistical methods. Results showed that the two olive varieties (i.e. Kalamata and Frantoio) can be visually differentiated and mapped on the enhanced image based on texture. The spectral signature plots showed little difference in the mean spectral reflectance values, indicating that the two varieties have a very low spectral separability. In terms of within-field spatial variability, the presence or absence of Rhodes grass (Chloris gayana) was detected using visual interpretation, corroborated by the results of quantitative statistical measures. Spatial variability in soil properties, caused by the presence of a patch of sandy soil, was also detected visually. Finally, the “imprint” of former cover-type or land-use prior to olive plantation establishment in 1998 was identified. More work is being done to develop image classification techniques for mapping within-field spatial variability in olive varieties, biomass and condition using hyperspectral image data, as well as interpreting the cause of observed variability.