Abstract: AbstractThe time-integrated normalized difference vegetation index (iNDVI) provides key remote-sensing-derived information on the interactions between vegetation growth, climatic and soil conditions, and land use. Using a time-series of Landsat imagery obtained for Queensland, Australia, it has been demonstrated how robust geostatistics can be used to predict iNDVI. This approach is novel because it explicitly quantifies the uncertainty of prediction and uses Winsorizing, a data-censoring method, to minimize the distorting effects of outliers. Robust prediction of iNDVI, as opposed to non-robust prediction, was justifiable in 79% of the study area, highlighting the need for methods that deal with outliers in time-series analysis of remotely sensed imagery. There was a strong coarse-scale association between Queensland's bioregions and iNDVI, and also between bioregion and the rain-induced difference in iNDVI through time (effects that were significant at p < 0.001 in both cases). At a finer spatial scale, prediction of iNDVI also appeared to be a promising way to distinguish long-term cropping land from adjacent long-term grazing land (effect significant at p < 0.001). The method is tied to a set of assumptions concerning image radiometry, cloud detection, variogram estimation, and variable additivity. The first two are fundamental remote-sensing issues that can be improved with additional labour; the last two can be improved statistically but would greatly increase the processing time per pixel. Robust geostatistical analysis of time-series has immediate relevance to gap-filling of SLC-off Landsat 7 Enhanced Thematic Mapper Plus (ETM+) imagery, and for generating novel covariates for digital soil mapping. AcknowledgementsThis work was partially funded by the Australian Department of Agriculture, Fisheries and Forestry, and stimulated by the research requirements of Dr Ram Dalal, Dr Diane Allen, and Dr Kathryn Page. Thanks to Dr Peter Scarth and Dr Ben Marchant, and Mr Tom Orton for constructive comments on a draft; the latter two also provided invaluable numerical advice. Thanks also to Dr Robert Denham for help with the conditional probabilities in Section 2.3.3.
Publication Year: 2013
Publication Date: 2013-04-02
Language: en
Type: article
Indexed In: ['crossref']
Access and Citation
Cited By Count: 16
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot