Mapping flood events using remote sensing in an arid zone river, central Australia

Abstract

Rivers in arid and semi-arid zones are distinguished by high variations in flow patterns both in spatial and temporal scale. The variable flows, often ephemeral, have a strong impact on the local hydrological, ecological and landscape characteristics. In Australia, rivers in the arid zone are often ungauged and therefore little information is available for further hydrological and ecological analysis. An assessment of the water resources and ecological importance of these rivers requires the fundamental information on the occurrence and extent of flow events. Remote sensing technologies can be applied where ground field data are few, or completely absent, but have rarely been operationally applied as the spatial and temporal resolutions of the sensors have traditionally been considered to be inadequate for mapping short-term flood events or those obscured by vegetation. However, the relationship between flow events and the following vegetation response can provide a useful indirect measurement of flow occurrences. Arid regions typically have low vegetation content and greenness for most of the year but the provision of water via a flow pulse can cause abrupt increases in vegetation density and greenness for a period following the flow event. This paper explores a method using daily observations of the Normalized Difference Vegetation Index (NDVI) collected by the Moderate Resolution Imaging Spectrometer (MODIS) onboard the Terra satellite to map flow events in the Neales River catchment of the Lake Eyre Basin, Australia. MODIS bands 1 and 2 of 250-m spatial resolution surface reflectance are used to quantify NDVI. Larger flow events result in significant areas of open water, which are directly observed by negative NDVI values, while smaller events are implied by an increase in NDVI along the river channels a certain period after a flow pulse (Figure 1).(graph presented) Flow size, floodplain width, vegetation type and vegetation density are considered as the major factors which influence the NDVI response during the flow periods. It is inferred that errors in timing the flow occurrences can be moderated by refined site selection, while characterising flow size and duration is more complicated.