VIPER Tools

Introduction to MESMA
A spectral mixing technique called Multiple Endmember Spectral Mixture Analysis (MESMA) is at the core of VIPER Tools.  MESMA is an extension of simple Spectral Mixture Analysis (SMA).  In simple SMA, a spectrum is modeled as the sum of “pure spectra” called endmembers, each weighted by the fraction of an endmember required to produce the mixture (Adams et al., 1993; Roberts et al., 1993; Settle and Drake, 1993). SMA has an advantage over many competing approaches in that it provides physically meaningful measures of cover that account for subpixel mixing.  While SMA is a powerful approach, it fails to account for pixel-scale variability in spectral dimensionality, spectral degeneracy between materials and natural variation in the spectra of most materials.  As a result, simple SMA underutilizes the potential of most remote sensing data sets for discriminating materials, while at the same time producing fractional errors due to the incorrect type or number of endmembers used to unmix a specific pixel.

MESMA extends SMA by allowing the number and types of endmembers to vary on a per-pixel basis (Roberts et al., 1998).  MESMA overcomes limitations of SMA by requiring a model to meet minimum fit, fraction and residual constraints while testing multiple models for each image pixel. MESMA is typically implemented by developing a spectral library, then unmixing an image using every possible combination of two, three and four endmembers applied to each pixel. Using this approach, significantly more than four materials can be mapped across an image, while minimizing pixel-scale fraction errors by selecting the best-fit model for each pixel. Applications of MESMA include snow-cover/snow-area mapping (Painter et al., 1998; 2003), plant species mapping in chaparral (Roberts et al, 1998; Dennison and Roberts 2003a/2003b, Roberts et al., 2003), soil mapping in arid lands (Okin et al., 2001), improved estimation of LAI in boreal forests (Sonnetag et al., 2007), landform mapping in North Africa using MODIS (Ballantine et al., 2005), fire temperature mapping (Dennison et al., 2006), urban remote sensing (Rashed et al., 2003; Powell et al., 2007), shape-based temporal compositing (Dennison et al., 2007) and even several planetary applications (Li and Mustard, 2003; Johnson et al., 2006).  Quantitative comparisons of reference fractions to estimates using SMA or MESMA have proven to be highly accurate over a diverse number of surfaces (e.g. Elmore et al., 2000; Painter et al., 2003; Powell et al., 2007).

Endmembers used in SMA or MESMA can be derived from an image (image endmembers), or from a library of known materials (reference endmembers). Reference endmembers can be derived from the field, laboratory, images or even radiative transfer models. Typical endmembers used in SMA include soil, Green vegetation (GV), Non-photosynthetic Vegetation (NPV) and shade. Identifying a high quality set of reference or image endmembers has been identified as a critical stage of mixture modeling (Tompkins et al., 1997). A number of approaches have been developed for identifying endmembers including  the Pixel Purity Index (PPI: Boardman et al., 1995), Manual selection (Bateson and Curtiss, 1996), Endmember Bundles (Bateson et al., 2000), Constrained Reference Endmember Selection (CRES: Roberts et al., 1993; 1998), Count-based Endmember Selection (COB: Roberts et al., 2003), Endmember Average RMSE (EAR: Dennison and Roberts, 2003a) and Minimum Average Spectral Angle (MASA: Dennison et al., 2004). Four techniques used for identifying optimal endmembers are included in VIPER Tools: CoB, EAR, MASA and CRES.

While MESMA is a powerful approach, several factors have limited its use. One of the major limitations has been a lack of easily used software. One of the key goals of VIPER Tools is to address this lack of software. Another limitation has been the challenges in selecting endmembers to be used in MESMA. In its earliest implementations, MESMA was typically employed using hundreds of endmembers, resulting in significant challenges in interpretation and computation. More recently, efforts have been made to fine tune the analysis by selecting optimal sets of endmembers from a spectral library prior to running MESMA. A key goal is to identify those spectra that are most representative of a specific class, yet also least likely to be confused with spectra from a different class. To facilitate the development of optimal spectral libraries, several techniques have been developed including:

1. Count-based Endmember Selection (CoB): Using this approach, endmembers are selected that model the greatest number of endmembers within their class (Roberts et al., 2003). As a variation on this approach, a Count-Based Index can be used to rank endmember selection based on maximizing the models selected within the correct class, while minimizing confusion with other classes (Clark, 2005).
2. Endmember Average RMSE (EAR): Using this approach, endmembers are selected that produce the lowest RMSE within a class (Dennison and Roberts, 2003).
3. Minimum Average Spectral Angle (MASA): Using this approach, endmembers are selected that have the lowest average spectral angle (Dennison et al., 2004).

A promising alternative to MESMA, called Automatic Monte Carlo Unmixing is described by Asner and Lobell (2000) and Asner et al., (2003). Using this approach, rather than selecting a single set of endmembers or predefined models for a pixel, each pixel is unmixed using sets of endmembers randomly selected from “bundles” of similar materials to generate a mean and standard deviation for each fraction. Automatic Monte Carlo Unmixing, while similar to MESMA in principle, has an advantage in that it does not depend on the selection of a single set of endmembers and is readily automated. However, the fractions derived using this approach will be highly dependent on how representative library spectra are of a region and cannot account for species-specific spectral differences as MESMA does.