Comparison of methods for estimation of absolute vegetation and soil fractional cover using MODIS normalized BRDF-adjusted reflectance data

Abstract

Green vegetation (GV), nonphotosynthetic vegetation (NPV), and soil are important ground cover components in terrestrial ecosystems worldwide. There are many good methods for observing the dynamics of GV with optical remote sensing, but there are fewer good methods for observing the dynamics of NPV and soil. Given the difficulty of remotely deriving information on NPV and soil, the purpose of this study is to evaluate several methods for the retrieval of information on fractional cover of GV, NPV, and soil using 500-m MODIS nadir BRDF-adjusted reflectance (NBAR) data. In particular, three spectral mixture analysis (SMA) techniques are evaluated: simple SMA, multiple-endmember SMA (MESMA), and relative SMA (RSMA). In situ cover data from agricultural fields in Southern Australia are used as the basis for comparison. RSMA provides an index of fractional cover of GV, NPV, and soil, so a method for converting these to absolute fractional cover estimates is also described and evaluated. All methods displayed statistically significant correlations with in situ data. All methods proved equally capable at predicting the dynamics of GV. MESMA predicted NPV dynamics best. RSMA predicted dynamics of soil best. The method for converting RSMA indices to fractional cover estimates provided estimates that were comparable to those provided by SMA and MESMA. Although it does not always provide the best estimates of ground component dynamics, this study shows that RSMA indices are useful indicators of GV, NPV, and soil cover. However, our results indicate that the choice of unmixing technique and its implementation ought to be application-specific, with particular emphasis on which ground cover retrieval requires the greatest accuracy and how much ancillary data is available to support the analysis.