A remote sensing exploration of land surface phenology in the Australian Alps

Download files
Access & Terms of Use
open access
Copyright: Thompson, Jeffery
Altmetric
Abstract
Seasonal snow-cover is a significant environmental factor influencing the vegetative phenology of alpine areas. As such, climate induced changes in snow country and trends towards fewer days with persistent snow-cover are likely to significantly alter alpine landscapes both globally and in Australia. In comparison with other mountainous environments, Australia’s alpine bioregion is somewhat diminutive in area and elevation range. Because of this constraint, there is little room for highland vegetation to expand as the climate warms. As a consequence, it has been predicted that Australia’s alpine bioregion will be significantly impacted by climatic change. In recent years, a number of studies have begun using remotely sensed time-series imagery to examine the land surface phenology of alpine areas. Early studies tended to employ time-series derived from AVHRR data, but increasing studies of alpine land surface phenology have utilized MODIS derived time-series. Broadly, most of the studies of alpine land surface phenology have noted the influence of elevation gradients, whereby higher elevations tend to have later starts to the growing period and have earlier ends to the growing period. Aspect has been noted as having a minor, but detectable influence on the phenology of alpine land surfaces. In terms of temporal considerations, there is little agreement within the literature on the direction of trends between regions, and virtually none of the studies have examined the influence of landscape disturbances on phenological changes. To date, there have been no studies that have explored the links between snow-cover and land surface phenology in Australia. In order to help fill the gap, this study used remotely sensed time-series imagery to explore links between changes in snow seasonality and the phenological responses of the land surface. This was accomplished using a number of 12-year time-series images (2000 – 2012) generated using the MODIS Daily Surface Reflectance Product (MOD09). The generation of seasonality descriptor time-series required the development of an improved cloud-mask as well as new methods for deriving descriptors of snow seasonality and vegetative phenology. The improved cloud-mask built upon previous methods and utilizing the existing MODIS cloud-mask results in conjunction with new thresholds to improved the discrimination of snow-cover from clouds. The cloud-masking algorithm and the new methods for deriving descriptors of snow seasonality and vegetative phenology represent important methodological contributions of this thesis. From the remotely sensed observations, three descriptor time-series of snow seasonality were generated and analyzed. These descriptors corresponded with the start, end and duration of the snow-covered period. Over the course of the time-series a slight trend towards a later start of the snow-covered period and a stronger trend towards an earlier end to the snow-covered period was detected across the alpine bioregion. Overall, there was a downward trend in the duration of the snow-covered period for the study area. For all three snow-seasonality descriptors, there was substantial spatial variability in these trends. Patterns from the remotely sensed image time-series for the start, end and duration of the vegetative growing period were also derived. A trend towards an earlier start of the growing period was observed across the alpine bioregion. This was more pronounced at lower elevations. A slight trend in the end of the growing period indicated that it was ending later over the course of the time-series. Corresponding with the changes in both the start and end of the growing period, a trend towards a longer growing period was detected in the time-series. Significant spatial variability was observed in the trends for the phenological descriptors. The relationships between the descriptors of snow seasonality and the land surface phenological descriptors were analyzed. To date, there have been no studies reported in the literature that have explicitly examined the relationships between snow seasonality and alpine land surface phenology using remotely sensed image time-series. Whist it was found that the end of the snow-covered period was often a reasonable predictor for the start of the growing period for some areas, it was not observed in all parts of the study area across the time-series. In contrast, the links between the start of the persistent snow-covered period and the end of the growing period were (almost) universally poor across the time-series for most of the study area. It is thought that these results are entirely new. Regarding disturbance, the 2003 alpine fires had a significant impact on the vegetation within the bioregion. An analysis of the fire severity in the northern section of the bioregion was undertaken. The analysis indicated that there were greater proportions of fire-affected pixels with negative trends in the duration of the snow-covered period. This suggests that some of the negative trends in the snow-covered period and associated trends in the vegetative growing period may be associated with the impacts of the alpine fires. This thesis makes several contributions that are both methodological and empirical in nature. Methodologically, an improved liberal cloud-masking algorithm was developed, which allowed for the enhanced detection of snow-cover in an alpine environment. The method for deriving seasonally persistent snow-cover descriptors is another methodological contribution of the work. Similarly, the development of a new phenology algorithm that was relatively free of the influence of snowmelt is another contribution. These new methods are not region specific, and can thus be readily applied to other snow-covered environments. Empirically, this thesis also demonstrates the utility of remotely sensed time-series imagery for detecting patterns in Australia’s alpine bioregion. It suggests that MODIS time-series imagery can be used to monitor this region, as well as others, for climate related impacts. The research also highlighted a link between observed trends in snow-cover duration and fire disturbance, which does not appear to have been discussed previously. This link has potential implications for alpine regions in a changing climate.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Author(s)
Thompson, Jeffery
Supervisor(s)
Paull, David
Lees, Brian
Creator(s)
Editor(s)
Translator(s)
Curator(s)
Designer(s)
Arranger(s)
Composer(s)
Recordist(s)
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
2013
Resource Type
Thesis
Degree Type
PhD Doctorate
UNSW Faculty
Files
download whole.pdf 76.02 MB Adobe Portable Document Format
Related dataset(s)