MODIS tasselled cap: land cover characteristics expressed through transformed MODIS data

Abstract

The tasselled cap concept is extended to Moderate Resolution Imaging Spectroradiometer (MODIS) Nadir BRDF‐Adjusted Reflectance (NBAR, MOD43) data. The transformation is based on a rigid rotation of principal component axes (PCAs) derived from a global sample spanning one full year of NBAR 16‐day composites. To provide a standard for MODIS tasselled cap axes, we recommend an orientation in MODIS spectral band space as similar as possible to the orientation of the Landsat Thematic Mapper (TM) tasselled cap axes. To achieve this we first transformed our global sample of MODIS NBAR reflectance values to TM tasselled cap values using the existing TM transformation, then used an existing algorithm (Procrustes) to compute the transformation that minimizes the mean square difference between the TM transformed NBAR values and NBAR PCA values. This transformation can then be used as a standard to rotate the MODIS NBAR PCA axes into a new MODIS Kauth–Thomas (KT) orientation. Global land cover patterns in tasselled cap space are demonstrated graphically by linking the global sample with several other products, including the MODIS Land Cover product (MOD12) and the MODIS Vegetation Continuous Fields product (MOD44). Patterns seen at this global scale agree with previous explorations of TM tasselled cap space, but are shown here in greater detail with a globally representative sample. Temporal trends of eight smaller‐scale BigFoot Project (www.fsl.orst.edu/larse/bigfoot) sites were also examined, confirming the spectral shifts in tasselled cap space related to phenology.

Acknowledgements

The authors would like to acknowledge Claire Hay's convincing arguments over 20 years ago concerning the relative value of the tasselled cap transformation amongst its numerous alternatives. We also thank BigFoot (http://www.fsl.orst.edu/larse/bigfoot/) for providing data used in analysis and figures in this paper. Anne Nolin provided guidance in working with MODIS data and assisted in refining the analyses with respect to potential sources of error. Fred Ramsey contributed the idea of using the Procrustes method of rotation and Robert Kennedy provided technical support and guidance throughout the initial analyses. Todd Schroeder performed the atmospheric correction on the Landsat images used. We are also grateful to the US National Aeronautics and Space Administration (NASA) Terrestrial Ecology Program for funding the research described in this paper.