Spanglet, HJ, Ustin, SL, Rejmankova, E (1998). Spectral reflectance characteristics of California subalpine marsh plant communities. WETLANDS, 18(3), 307-319.
Reflectance spectra (414-948 nm) were recorded at ground level from leaves and canopies of subalpine marsh macrophytes near Lake Tahoe, California in 1994. Canopy architecture of the dominant species suggested that canopy reflectance should differ from one another: yellow pond-lily (Nuphar polysepalum) has a horizontal canopy; hardstem bulrush (Scirpus acutus) has a vertical canopy; beaked sedge (Carex rostrata) has a spherical (randomly distributed) canopy. Data were used to 1) analyze effects of canopy architecture on reflectance properties, 2) examine effects of canopy architecture on use of spectral vegetation indices as predictors of biomass, and 3) develop methods for detecting species distributions. Leaf reflectance spectra were different in shape and magnitude across the spectrum, with canopy architecture and cover having marked effects on albedo and wavelength-dependent variance among samples. Nuphar polysepalum (horizontal canopy) showed the least difference between canopy and leaf reflectance, S. acutus (vertical canopy) reflectance was most affected by canopy architecture, and C. rostrata (spherical canopy) was intermediate. Reflectance in the near infrared (NIR) region was strongly reduced, while reflectance in the visible region was minimally affected, producing small red/NIR differences and low vegetation indices. Seven vegetation indices were tested for correlation with plant biomass. Fresh and dry green biomass of C. rostrata were significantly correlated with all indices (p<0.002), while no correlation was found between any index and fresh or dry green biomass of N. polysepalum (p>0.13) or S. acutus (p>0.21), suggesting that spectral indices as estimators of biomass assume a spherical vegetation canopy. Differences between Thematic Mapper (TM) bands 3 (red) and 2 (green) normalized by the sum of reflectance ((TM3 - TM2)/(TM3 divided by TM2)), was significantly different among species (p<0.05) and was used as the basis of a supervised classification of a large Lake Tahoe marsh. The classification agrees with the data on distributions of species and expected distributions in relation to water depth.