The primary goal of this study was to evaluate and improve the quality of Moderate Resolution Imaging Spectroradiometer (MODIS) 8-daily Normalized Difference Vegetation index (NDVI) time-series to monitor changes in forest canopy of Pinus radiata plantations located at Green Hills State Forest in the Southern Highlands of New South Wales, Australia. The utility of MODIS NDVI time-series are limited by the availability of high-quality data (e.g. cloud and atmospheric scatter free). The availability of high-quality data is a critical factor in assessing forest dynamics or phenological parameters such as the onset of crown flushing. The MODIS 250 m spatial resolution NDVI 8-daily (MOD09Q1) and 16-daily time-series (MOD13Q1) from 2000 to 2006 were employed in this study. Firstly, it was shown that the vegetation usefulness index provided with the MOD13Q1 data is dominated by a sun zenith angle threshold set at 60°. The flagging of data errors was optimized by using information on aerosol quantity, atmosphere adjacency correction, mixed cloud, shadow, and satellite viewing angle. Secondly, major data errors were detected in the 8-daily time-series by using the 16-daily time-series. This approach increases the usefulness of the 8- daily time-series which offers a higher temporal resolution and a smaller influence by the compositing technique than the MOD13Q1 product. Thirdly, the Fast Fourier Transform and to a lesser extent the 'running median' approach were influenced by these major data errors. This study has shown that it is crucial to evaluate properties of remote sensing time-series before studying temporal forest growth and health dynamics.
Funding
Australian Research Council
Forestry Tasmania
Forests and Forest Industry Council of Tasmania
Gunns Limited
WA Plantation Resources
History
Publication title
Proceedings of Forest Sat 2007 Conference & Scientific Workshop