Mapping canegrub damage from high spatial resolution satellite imagery

Abstract

CANEGRUBS FEED on the roots of sugarcane plants, reducing plant vigour and yield, and if left untreated within a growing season they have the potential to rapidly increase the impacted area in the following year. For the targeted control of the canegrub, it is essential that the location of the affected areas is identified. However, identifying canegrub damage in the field is difficult due to the often impenetrable nature of sugarcane. Satellite imagery offers a feasible means for achieving this by using the visual characteristics of sprawling, changed leaf colour and exposure of soil in damaged areas. The objective of this research was to use object-based image analysis (OBIA) and high spatial resolution satellite imagery to map canegrub damage. The OBIA mapping approach used in this research was based on four key steps for three selected study sites in Queensland, each covering 50 – 100 km2 around Mackay, Home Hill and Gordonvale. The steps were: (1) initial segmentation of sugarcane block boundaries based on existing GIS layers provided by the respective Mills and further segmentation of each block into smaller homogenous objects; (2) classification and subsequent omission of fallow/harvested fields, tracks and other non-sugarcane features within the block boundaries; (3) identification of ‘potentially’ grub-damaged areas within each block with the lowest amounts of green leaves (low Normalised Difference Vegetation Index (NDVI) values) and highest level of image texture; and (4) the further refining of ‘potentially’ grub damaged areas to ‘likely’ affected areas based on the absolute difference in the amount of green leaves (NDVI values) and texture between the ‘potentially’ grub damaged areas and the remaining parts of each block. The initial validation based on field observations of greyback canegrub damage at the time of the satellite image capture in June 2013 yielded overall accuracies between 53–80%. However, this included a number of false positives resulting from sprawling, drainage issues, weed and pig damage. Further research will focus on reducing these false positives as well as investigating the inclusion of additional data layers to increase the predictive accuracies. Such data layers may include distance from damage to tree corridors, distance to neighbouring grub damage and, potentially, soil type, cane variety and treatment information. Analysis of archived imagery may also provide some insight into the historic location and distribution of grub damage, thus assisting with improved understanding of potential risk for the subsequent year. The results of this research will help cane growers to manage and reduce damage caused by canegrubs and increase future yields.