The use of Landsat derived vegetation metrics in Generalised Linear Mixed Modelling of River Red Gum (Eucalyptus camaldulensis) canopy condition dynamics in Murray Valley National Park, NSW, between 2008 and 2016
School of Earth & Environmental Sciences
Curtis, Evan, The use of Landsat derived vegetation metrics in Generalised Linear Mixed Modelling of River Red Gum (Eucalyptus camaldulensis) canopy condition dynamics in Murray Valley National Park, NSW, between 2008 and 2016, BSci Hons, School of Earth & Environmental Sciences, University of Wollongong, 2016.
Water is a primary determinant of the condition of wetland and riparian vegetation in semi-arid Australia. The species that inhabit these ecosystems, such as River Red Gum (Eucalyptus camaldulensis) (RRG), are well adapted to variability in water supplies intrinsic throughout the Australia climate cycle. Despite climatic variability, many of inland Australia’s wetlands and riparian ecosystems function under a depleted state as a partial consequence of logging and river regulation. Murray Valley National Park, NSW (MVNP) is one such example. It is thought that high levels of intra-stand competition for water resources in MVNP are a primary driver of RRG Forest canopy condition. RRG is well known to respond to water availability through the generation and reduction of its canopy, however little is known about the role of stand density on RRG Forest canopy condition.
This study investigated canopy condition of RRG between 2008 and 2016, a period characterised by high levels of hydro-climatic variability in south-eastern Australia. This study aims to determine how water demand and availability drive canopy condition dynamics of RRG in Murray Valley National Park, NSW during this period. Stem density and site quality were used respectively as surrogate measures of water demand and availability in conjunction with hydro- climatic variables, Landsat derived Normalised Difference Vegetation Index (NDVI) and Foliage Projective Cover (FPC) were used to create a strictly empirical data series. Due to the high degree of temporal fragmentation inherent in Landsat data, analyses were undertaken using Generalised Linear Mixed Effects Modelling (GLMM) to determine statistically significant drivers of RRG canopy condition, while allowing trend, periodic and random noise components to be accounted for.
Models found site quality to be a statistically significant driver of canopy condition in both anomalously dry and wet hydro-climatic periods. Site quality was found to exhibit increasingly different canopy conditions following recovery from drought. Conversely, surrogate measures of water demand were not found to be statistically significant, suggesting that despite high stand densities, water supplies across the eight year period have been sufficient to maintain homogeneous canopy condition dynamics throughout MVNP. In concurrence with other studies throughout the Murray-Darling Basin, drivers of RRG canopy condition were modelled as being primarily climatic, and in particular the Southern Oscillation Index (SOI) was a primary driver throughout the eight years. While drivers were primarily climatic, river discharge influenced canopy condition during anomalously wet and dry phases. Periodicity modelling showed a dampened response during the drought phase, and became more pronounced following drought recovery. In many cases, a sub-hectare spatio-temporal investigation such as that presented here would rely on data interpolation to model trend, periodic and random noise components of a data series. This study has been able to model the same components while relating them to the drivers of canopy condition dynamics using an entirely empirical dataset. The research presented here provides scientists at OEH with an empirically derived baseline understanding of RRG canopy condition dynamics in MVNP over a highly variable hydro-climatic multi-year period using remotely sensed vegetation metrics. Furthermore, the methods used have the potential enhance ecosystem research globally, by facilitating the investigation of sub-hectare scale phenomena without the need to rely exclusively on time series analyses or sacrificing data integrity.