A standardised Landsat time series (1973–2016) of forest leaf area index using pseudoinvariant features and spectral vegetation index isolines and a catchment hydrology application

Remotely-sensed imagery from Landsat dating back to 1972 is now available from the United States Geological Survey free-of-charge as a level 1 terrain corrected product. However, to take full advantage of the time series requires that the earlier multispectral scanner (MSS) imagery be integrated with the later thematic mapper (TM), enhanced thematic mapper (ETM+) and operational land imager (OLI) imagery. Here we describe a simple, generic approach to processing Landsat scenes to develop a standardised time series of leaf area index, a key biophysical attribute of forests, for a study region in the northern jarrah forest of south-western Australia. The five-step approach utilised the main features in near infra-red (NIR)-red space of a dark point, soil line and a radiating set of ratio index isolines. Firstly, digital numbers were converted to top of atmosphere reflectance. Secondly, the red and NIR bands of all scenes were atmospheric corrected using long-established deep water supply reservoirs within the study region as invariant ‘dark objects’. Thirdly, adjustment of the NIR band such that the soil line was consistently located on the 1:1 line in NIR-red space across scenes both from the same sensor and from different sensors, taking advantage of a >40 year record of bauxite mining operations to identify pseudo-invariant bare-ground targets. Fourthly, an inter-sensor calibration of spectral vegetation indices (SVIs) to ensure that SVI isolines from different sensors were consistent. Finally, application of a relationship between SVIs and an extensive multi-year dataset of ground-based LAI measurements to generate a time-series of leaf area index (LAI). We demonstrate a simple application of the LAI time-series by investigating the relationship between rainfall, forest LAI as affected by different land use activities, and streamflow in the study region in the south west of Australia. The time-series could be used to improve prediction and modeling of hydrological changes in the region resulting from changes in catchment forest cover.