Interpretation of the relationship between skin temperature and vegetation fraction: Effect of subpixel soil temperature variability

Both land surface/skin temperature and vegetation indices data provided routinely and globally by NASA MODIS sensors at 1-km grid resolution represent an important piece of information assimilated into various environmental applications/models. Previous studies based on these and similar remotely data sets and on two-component pixel representation (accounting for pixel-aggregated vegetation and bare soil temperatures only) have shown a rather strong linear relationship between the pixel's skin temperature and the vegetation index/fraction. Deviations (Δ₀) from this relationship are frequently used for soil moisture content estimates at a pixel scale. As the two-component pixel model does not account for subpixel heterogeneity (associated, for example, with bare soil temperature variability within the pixel), its role in controlling a magnitude of Δ₀ has been examined. A simple tri-component pixel model describing vegetation and wet and dry bare soil temperatures was suggested to analyse an impact of this heterogeneity on Δ₀ estimates. This model was considered to provide a 'true' estimate of Δ₀ as compared with Δ₀ evaluated from the two-component pixel model. A comparison between the models shows that a substantial underestimation of Δ₀ was likely to occur at a level of individual pixels when the two-component approach was applied for interpretation of the observed relationship between the skin temperature and the vegetation index. Depending on the fraction of pixel occupied by the dry soil, this underestimation might be as much as 100%.