Imaging spectroscopy successfully predicts vegetation quantity and quality in heterogeneous alpine grasslands

Background/Question/Methods

Alpine areas are characterized by their remoteness as well as their topographic variability and complexity. The topographic complexity causes small-scale variability in environmental parameters (air/soil temperature, radiation, soil moisture, soil nutrients) which in turn shape vegetation properties such as community composition, productivity (above- and belowground), or shoot- and root nutrient content. Both the general problem of accessibility of alpine areas, as well as the highly variable vegetation patterns makes it difficult to assess vegetation properties with systematic in situ sampling. We investigated possibilities of modeling plant biomass, nitrogen (N) and fiber content in highly heterogeneous alpine grasslands in the Swiss National Park using imaging spectroscopy. Although imaging spectroscopy has become an important tool for the spatially inclusive acquisition of environmental data, vegetation and terrain complexity pose major challenges to this technique. In our study, we combined data from the airborne imaging spectrometer APEX collected between 2010 and 2013 with ground references from up to 100 field plots per year. We calculated simple ratios from all spectral bands and modeled plant biomass, plant N concentration and fiber content using linear, exponential and second order polynomial functions.

Results/Conclusions

Our models predicted all three vegetation properties with high accuracies (biomass: R² = 0.70, RMSE = 156 g.m^-2 (fresh weight); plant N concentration: R² = 0.53, RMSE = 0.5 %; plant fiber content: R² = 0.79, RMSE = 2.5 %). Despite the heterogeneity in our study area, imaging spectroscopy thus successfully predicted some key vegetation characteristics. The reasons for this are likely due to the fact that our ground reference plots covered the entire expected range of the vegetation characteristics of interest. More specifically, we covered the entire altitudinal gradient, and also all different levels of productivity and variations in plant species communities. To do so, expert knowledge becomes crucial during the planning process. The products we generated for our study area – plant biomass, nitrogen and fiber maps with a spatial resolution of 2 x 2 m – provide valuable baseline data for landscape protection, wildlife conservation and the park’s management.