The influence of canopy strata on remotely sensed observations of savanna-woodlands

Upwelling radiance from savanna woodlands may originate from two separate layers: (1) the held layer, which is a mixture of soil, litter and herbs, and (2) the tree layer composed of woody parts and leaves. Unless detailed field data are available for a particular savanna location, it is unknown how the individual layers may influence the red and near-infrared signals and whether radiative interactions between layers are important. We employed an existing radiative transfer model (SAIL) in conjunction with a simple, single-scattering model to analyse the variation in Advanced Very High Resolution Radiometer (AVHRR) channel 1 and 2 response as well as NDVI for savanna-woodland vegetation in eastern Zambia. Linear fits between predicted and observed values of reflectance and NDVI were significant (p < 0.05) in the red and in NDVI, however, the model failed to explain a high proportion of the variation in near-infrared. Red and NDVI in sites where tree cover was high were also poorly modelled, which suggests that multiple interactions between canopy layers make a single-scattering model unreliable, particularly in the near-infrared. Modelled results were also compared to aircraft radiometer measurements provided by the integrated camera and radiometer instrument (ICAR). Simulations parameterized with field data suggest that the model may be used to infer tree and field layer influences at different points during the seasonal cycle. Results also suggest that the field layer dominated the signal in our savanna woodland sites throughout most points of the seasonal cycle, which is consistent with other canopy radiative-transfer studies. Simulations indicated that the tree layer was a relatively more important component of NDVI during the dry season when the field layer was largely senescent, accounting for 20-40 per cent of the satellite signal.