Hyperspectral remote sensing of tropical grass quality and quantity

Resource distribution is a fundamental factor governing the movement and distribution of herbivores. Specifically, the quality (foliar concentration of nitrogen, phosphorous, calcium, magnesium, potassium and sodium) and quantity (biomass) of vegetation are important factors. In this regard, the development of techniques that can model the distribution of vegetation quality and quantity are critical for an improved understanding of wildlife distribution as well as facilitating an optimal management of wildlife resources. The advent of hyperspectral remote sensing has offered unprecedented opportunities to accomplish this task.

This study aimed to investigate the potential of hyperspectral remote sensing in estimating biomass of tropical grass at full canopy cover (a task that could not be achieved using broad band satellite images) and to predict and map the quality of tropical grasses at canopy level. Our approach was to investigate the potential of hyperspectral remote sensing at three levels of investigation - laboratory level, field level and airborne platform level.

Our results showed that, at full canopy cover, tropical grass biomass is more accurately estimated by vegetation indices based on narrow wavelengths located in the red edge than the standard NDVI. At laboratory level, we could discriminate between different foliar nitrogen treatments using high-resolution spectra measured at canopy level. We also showed that there was a shift of the red edge position to longer wavelengths with an increase in nitrogen concentration. The laboratory experiment permitted the extension of the developed techniques to the field level. Using continuum-removed absorption features calculated from field spectra, we could reliably predict the quality (N, K, P, Ca, Mg, Na) of in situ grass measured in the Kruger National Park, South Africa. We also showed a strong interaction between species type and biochemical concentration in effecting spectral reflectance. This provided a basis for the algorithms to use in mapping foliar biochemicals in a mixed species environment using airborne hyperspectral image. Therefore the techniques developed for accomplishing the final stage (airborne platform level) were largely built upon the laboratory and field observations. The new integrated approach, involving the red edge position, continuum-removed absorption features as well as a neural network was applied to map foliar nitrogen concentration in the Kruger National Park, South Africa.

Overall, the study has shown the potential of hyperspectral remote sensing to predict the quality as well as the quantity of tropical grasses. The result is important for wildlife habitat modelling.