Linking rangeland productivity to species composition and diversity through airborne remote sensing

Elisa Van Cleemput and Katharine N. Suding, Institute of Arctic and Alpine Research, University of Colorado, Boulder, Boulder, CO, Elisa Van Cleemput and Katharine N. Suding, Department of Ecology and Evolutionary Biology, University of Colorado, Boulder, Boulder, CO

Background/Question/Methods
Due to climate change, ranchers are facing uncertainties about forage production in grasslands. Although it is generally assumed that species composition and biodiversity influence productivity, it is often unclear how they should be incorporated in rangeland management. Here we address this issue using airborne and satellite imagery as data sources of biological information in various rangeland systems across the USA. We ask 1) Do spectral measurements hold clues on species composition?; 2) Are patterns of optically estimated species composition related to productivity and productivity stability, at a regional scale?; 3) How variable are these relations across systems? We conduct our work on grassland sites included in the National Observatory Ecological Network (NEON), utilizing NEON airborne and plot datasets and aboveground herbaceous biomass provided by the Rangeland Analysis Platform (RAP) for 2017. The dimensionality of vegetation composition (in 1 m × 1 m NEON plots) at each site was reduced via Principal Coordinate Analysis. We examined the link between hyperspectral signatures and the first ordination composition axis through partial least squares regressions. These models were subsequently upscaled to the site level. We then assessed potential site-specific relations between species composition, biomass and temporal biomass stability through Pearson’s correlation.

Results/Conclusions
The airborne hyperspectral signatures represented the gradient of species composition (first ordination axis) well, with R²cv ranging between 0.56 and 0.59 depending on the site. Across sites, band importance in the models highly varied, indicating that these models are not transferable and hyperspectral signatures may represent different biological aspects in various systems. Relations between site-level species composition, biomass and biomass stability also varied across sites. At some sites (e.g. Northern Great Plains Research Laboratory), community compositions that promoted biomass were also related to higher biomass stability (r = -0.38 and 0.20 respectively, p < 0.001). At other sites (e.g. Oklahoma Agricultural Experiment Station), we observed a trade-off between biomass production and biomass stability (r = 0.40 and 0.33 respectively, p < 0.001). In other words, increasing site-level species diversity on the latter sites may be a good strategy to ensure a stable forage production. We conclude that (i) hyperspectral remote sensing technology has potential to assess diversity patterns in rangelands, but that the ecological information entailed in these measurements is site-specific; and (ii) nature-based solutions that ensure biomass stability do not necessarily lead to higher annual biomass.