The Impact of Phenological Variation on Texture Measures of Remotely Sensed Imagery

Abstract

Measures of image texture derived from remotely sensed imagery have proven useful in many applications. However, when using multitemporal imagery or multiple images to cover a large study area, it is important to understand how image texture measures are affected by surface phenology. Our goal was to characterize the robustness to phenological variation of common firstand second-order texture measures of satellite imagery. Three North American study sites were chosen to represent different biomes. At each site, a suite of image textures were calculated for three to four dates across the growing season. Texture measures were compared among dates to quantify their stability, and the stability of measures was also compared between biomes. Interseasonal variability of texture measures was high overall (mean interseasonal coefficient of variation = 0.79), indicating that care must be taken when using measures of texture at different phenological stages. Certain texture measures, such as first-order mean and entropy, as well as second-order homogeneity, entropy, and dissimilarity, were more robust to phenological change than other measures.