Using MODIS Normalized Difference Vegetation Index to monitor seasonal and inter-annual dynamics of wetland vegetation in the Great Artesian Basin: a baseline for assessment of future changes in a unique ecosystem

The Great Artesian Basin mound springs (Australia) are unique wetland ecosystems of great significance. However, these unique ecosystems are endangered by anthropogenic water extraction. Relationships have been established between the vegetated wetland area and the discharge associated with individual springs, providing a potential means of monitoring groundwater flow using measurements of wetland area. Previous studies using this relationship to monitor Great Artesian Basin springs have used aerial photography or high resolution satellite images, giving sporadic temporal information. These “snapshot” studies need to be placed within a longer and more regular context to better assess changes in response to aquifer draw-downs. In this study, the potential of medium resolution MODIS Normalized Difference Vegetation Index data for studying the long-term and high frequency temporal dynamics of wetland vegetation at the Dalhousie Spring Complex of the GAB is tested. Photosynthetic activity within Dalhousie wetlands could be differentiated from surrounding land responses. The study showed good correlation between wetland vegetated area and groundwater flow, but also the important influence of natural species phenologies, rainfall, and human activity on the observed seasonal and inter-annual vegetation dynamic. Declining trends in the extent of wetland areas were observed over the 20002009 period followed by a return of wetland vegetation since 2010. This study underlined the need to continue long-term medium resolution satellite studies of the Great Artesian Basin as these data provide a good understanding of variability within the wetlands, give temporal context for less frequent studies and a strong baseline for assessment of future changes.


INTRODUCTION
The Great Artesian Basin (GAB) of Australia is one of the largest groundwater resources in the world (Ponder, 2002), underlying about one fifth of the Australian continent ( Figure  1a). Groundwater from the GAB supports a series of mound spring which contain a number of rare and relic endemic flora and fauna (Fensham and Fairfax 2003;Gotch et al., 2008;Ponder, 2002). The GAB springs are of great national and international importance for their ecological, scientific and economic values, and are culturally significant to indigenous Australians. In the recent decade, the sustainability of these groundwater-dependent ecosystems has become uncertain and numerous cases of spring extinction have been observed. In the near future, demands for this precious groundwater resource are likely to be considerable, particularly from proposed mining and petroleum activities, (Fensham et al., 2010;Mudd, 1998). It is thus particularly important to develop reliable methods for monitoring the long term artesian flow to inform management of the GAB and better protect this sensitive environment and its ecosystems.
Direct measurements of spring flow are notoriously difficult and expensive if precise estimates are required (Niejalke and Lamb, 2001). Relationships have been established between the wetland area associated with individual springs and their discharge (L.s -1 ), providing a potential means of monitoring groundwater flow using measurements of vegetated wetland area (Fatchen, 2001;Fatchen and Fatchen, 1993;White and Lewis, 2011;Williams and Holmes, 1978). Previous studies using this relationship to monitor Great Artesian Basin springs have used sporadic aerial photography or high resolution satellite images calibrated to Normalized Difference Vegetation Index (NDVI) (Niejalke andLamb, 2001, White andLewis, 2011). These intermittent high resolution studies have only given "snapshots" of vegetated wetland area dynamics at selected springs in the DSC. They need to be placed within a longer and more regular context to better assess changes in response to aquifer draw-downs (White and Lewis, 2011).
Since the launch of the MODIS satellite in 1998, medium resolution (250-m) MODIS NDVI time-series data have been applied successfully to quantify vegetation activity and to monitor vegetation dynamics (e.g. Jacquin et al., 2010;Jönsson et al., 2010) at high temporal frequency (16 days composites). In this study, 8 to 10 years of MODIS NDVI images were used to monitor and quantify vegetation activity and dynamics of the Dalhousie Spring Complex wetlands located on the southwest margin of the Great Artesian Basin (Figure 1a and b). DSC is the most important spring complex in the GAB and discharges 90% of total spring flow in South Australia through 148 separate spring vents (Gotch, 2010). Recent field work (March 2009) increased knowledge about wetland vegetation associated with active Dalhousie springs (White and Lewis, 2010a, b  On the basis established, ab indicative of w pixels tested e was applied to Dalhousie Spr 1c) and exclud  However, the dynamics of green wetland areas are also controlled by rainfall events on the GAB, natural intra-annual variability of species present within the wetlands, and by historical events at particular sites. Thus, multi-annual variations in vegetation extent are clearly related to the total rain measured at the Hamilton Station (Figure 3c). Ephemeral wetland and Phragmites phenologies explain the mean monthly variations observed in the area index calculated (Figure 3d). Anthropogenic actions, like uncontrolled fire, also play a crucial part in temporal dynamics of spring wetland vegetation (Figure 3a).

Tempora
Therefore, short-term variations observed in vegetated wetland extent must be interpreted with care and placed in a longer context if they are to be used as indicators of spring flow modification. Long-term medium resolution satellite studies provide a good understanding of variability and a strong baseline for assessment of changes, and give temporal context for less frequent high resolution studies. For example, White and Lewis (2011)  While the MODIS NDVI spatial resolution is not sufficient to give precise measurements of wetland area; it does give objective, repeatable measurements of relative changes in area, and provides new understanding of intra and interannual variability. Variations in the DSC wetland extent observed over the ten years are the result of the combined phenological, natural and anthropogenic influences at different springs throughout the complex. Nevertheless, the trends documented in Figure 10b are quite alarming as they suggest decreases of a third in the area index since 2002 and a return of wetland vegetation since only 2010.
MODIS photosynthetic activity monitoring in association with higher resolution temporally targeted studies must be extended to give a fuller picture of the long term dynamics and trends for DSC. Temporal analyses such as those implemented in this paper help define the range of natural variation in the spring-fed wetlands, and could form the basis for thresholds of acceptable change for environmental management purposes. Change in wetland area beyond such thresholds could provide timely warning of responses to changing aquifer pressures or climate change. Further research at smaller groups of GAB springs is testing the limits imposed by the MODIS NDVI spatial resolution, and will extend our understanding of the spring wetland dynamics.

CONCLUSIONS
The aim of this study was to test the potential of MODIS NDVI data as long-term tracer of wetland vegetation in the Dalhousie Spring Complex of the Great Artesian Basin. Results obtained proved that photosynthetic activity within DSC wetlands can be discriminated from surrounding land responses in this medium resolution imagery. Up to ten years of wetland vegetation extent were documented. The study reaffirmed the correlation existing between vegetated areas and groundwater flow, but also demonstrated the important influence of rainfall, natural species phenologies, and human activity on the observed seasonal and inter-annual vegetation dynamics. Over the whole Dalhousie Spring discharge area, declining trends in the extent of wetland areas were observed over the 2000-2009 period followed by a return of wetland vegetation since 2010. This study underlined the need to continue long-term medium resolution satellite studies of the GAB to fully understand variability and trends in the springfed wetlands. Complemented by high resolution studies, use of MODIS imagery over several decades could become a powerful tool for monitoring potential impacts of aquifer drawdown and climate changes in the Great Artesian Basin. 6.