We present results of a detailed study of drip rate variations at 12 drip discharge sites in Glory Hole Cave, New South Wales, Australia. Our novel time series analysis, using the wavelet synchrosqueezed transform, reveals pronounced oscillations at daily and sub-daily frequencies occurring in 8 out of the 12 monitored sites. These oscillations were not spatially or temporally homogenous, with different drip sites exhibiting such behaviour at different times of year in different parts of the cave. We test several hypotheses for the cause of the oscillations, including variations in pressure gradients between karst and cave due to cave breathing effects or atmospheric and earth tides, variations in hydraulic conductivity due to changes in viscosity of water with daily temperature oscillations, and solar-driven daily cycles of vegetative (phreatophytic) transpiration. We conclude that the only hypothesis consistent with the data and hydrologic theory is that daily oscillations are caused by solar-driven pumping by phreatophytic trees which are abundant at the site. The daily oscillations are not continuous and occur sporadically in short bursts (2–14 days) throughout the year due to non-linear modification of the solar signal via complex karst architecture. This is the first indirect observation leading to the hypothesis of tree water use in cave drip water. It has important implications for karst hydrology in regards to developing a new protocol to determine the relative importance of trends in drip rate, such as diurnal oscillations, and how these trends change over timescales of weeks to years. This information can also be used to infer karst architecture. This study demonstrates the importance of vegetation on recharge dynamics, information that will inform both process-based karst models and empirical estimation approaches. Our findings support a growing body of research exploring the impact of trees on speleothem paleoclimate proxies.
Karst architecture determines the flow and storage of water from the surface to the underlying cave and is a major influence on drip discharge. Karst systems are characterised by three principle flow types. Primary flow occurs where the water travels through the primary porosity of the rock matrix, secondary flow pathways are characterised by water transported along fractures in the bedrock and tertiary flow pathways consist of conduits enlarged by dissolution. The dominance of a particular flow regime changes over time; for example, older limestone tends to have higher secondary porosity (more fractures and enlarged conduits) and a lower primary porosity due to compaction or cementation (Ford and Williams, 1994). The relationship between karst architecture and delivery of water to cave drip discharge sites has been studied to constrain uncertainty in paleoclimate studies (Bradley et al., 2010; Markowska et al., 2015), identify suitable speleothems as climate archives (McDonald and Drysdale, 2007) and, in conjunction with drip water geochemistry, determine water residence times in karst aquifers (Arbel et al., 2010; Fairchild et al., 2000; Lange et al., 2010; Sheffer et al., 2011; Tooth and Fairchild, 2003; Bradley et al., 2013). Recent research examining drip hydrology and fluctuations in drip rate have used hydrological response to characterise flow paths. For example, Markowska et al. (2015) used statistical analysis of drip hydrology data to identify storage flow, in both the epikarst and overlying soil, to develop conceptual models of a karst system.
Over a timescale of months to years, fluctuations in drip discharge are
typically driven by seasonal variation in water availability (Hu
et al., 2008; Sondag et al., 2003) and long-term climate forcings such as
the North Atlantic Oscillation or El Niño-Southern Oscillation
(McDonald, 2004; Proctor et al., 2000). On a daily
to weekly timescale, drip rate responds to individual rainfall events
(Baldini et al., 2012) and barometric changes (Genty
and Deflandre, 1998; Jex et al., 2012; Tremaine and Froelich, 2013).
Tremaine and Froelich (2013) found weekly and daily fluctuations at one drip
site where an increase in barometric pressure decreased volumetric drip
rate. This was attributed to atmospheric tides, the heating and cooling of
the atmosphere, as the diurnal cycles occurred at 2 h before the solar
noon (S1) and solar midnight (S2) each day. The cave was situated in poorly
to moderately indurated Oligocene limestone with a high likelihood of
primary porosity (Scott, 2001). Jex et al. (2012) observed a negative
correlation between weekly barometric pressure changes and drip rate at 2
out of 40 drip sites monitored at the base of a paleokarst feature in the
marmorised and fractured Devonian limestone at Cathedral Cave, NSW. One drip
discharge site had a relatively strong anti-correlation (
Non-linear and chaotic behaviour of drip discharge has been observed over very short (second to minutes) timescales. Chaotic drip regimes were first noted by Genty and Deflandre (1998) in the Devonian limestone of southern Belgium (Genty and Deflandre, 1998). Chaotic and non-linear drip responses were also observed at an event scale in the fractured-rock limestone of Cathedral Cave, NSW (Mariethoz et al., 2012). These were attributed to the filling and draining of subsurface karst stores within a recharge event, with increasing homogenisation of flow with the filling of the stores. Baker and Brunsdon (2003) observed non-linear responses to rainfall in multi-year drip time series from a fractured rock (Carboniferous limestone) in Yorkshire, UK. With the exception of Tremaine and Froeclich (2013), daily fluctuations have not been observed in cave drip water hydrology.
In this paper we aim to increase our understanding of karst architecture by using a novel approach, the wavelet synchrosqueezed transform, to analyse drip discharge time series from 12 drip discharge sites in Glory Hole Cave, SE Australia. This analysis allows us to characterise daily and sub-daily fluctuations in drip rate and identify the processes driving these oscillations. This study has important implications for understanding karst unsaturated flow processes and karstic groundwater recharge. Currently, most karst models use very simplistic representations of unsaturated flow, if it is considered at all (Hartmann et al., 2014a). This study highlights the importance of vegetation dynamics on vadose flow and recharge, making it significant to karst modelling research and speleothem-based paleoclimate studies which focus on the impact of vegetation dynamics on proxy records (Treble et al., 2015, 2016a).
Glory Hole Cave is part of the Yarrangobilly Caves National Park located in
the Snowy Mountains, New South Wales, Australia (35
Location of Yarrangobilly Caves in New South Wales, Australia, with
photos of surface vegetation
Glory Hole Cave is formed of two main sections connected by a narrow
constriction
The vegetation is classified as sub-alpine open snow gum (
Drip discharge rate was recorded at 12 drip sites in three locations (Fig. 1
and Table 1) within Glory Hole Cave using Stalagmate© drip
loggers between December 2012 and September 2015, and monitoring is ongoing.
The drip sites were chosen using a stratified sampling method. A transect of
the cave was used to select three locations (G, M and LR) that satisfied the
following criteria: (1) there were actively dripping speleothems, (2) spatially
distant from the other locations and (3) different depths within the cave.
Individual drips were sampled randomly at each location, with selection
guided by practical constraints such as stalagmite surface being suitable
for placement of logger and the drip falling from high enough to activate
pressure sensor on the logger. Drip loggers recorded the frequency of drips
falling onto the surface of the sealed box containing an acoustic sensor in
15 min intervals. The number of drips were converted to mL min
Summary of drip sites and location within cave as indicated in Fig. 1, the monthly mean and standard deviation (std) of total flow volume and maximum and minimum drip rate in summer (December–February) and winter (June–August).
Barometric pressure and air temperature were recorded at two locations
within the cave (Fig. 1) using Solinst level loggers at 15 min intervals
from January to September 2015. Precipitation (accuracy
Daily potential evapotranspiration was estimated using “ETo Calculator”
software developed by the Land and Water Division of the Food and
Agriculture Organisation of the United Nations (
A new advance in signal processing was used to analyse the time–frequency
content of measured cave drip discharge rate, temperature and barometric
pressure. Here, the frequencies of interest are 1 cycle per day (cpd) and
faster, i.e. diurnal to sub-diurnal. Daubechies et al. (2011) first presented the
wavelet synchrosqueezed transform (WSST) as an empirical mode decomposition
like tool for disentangling an amplitude and phase modulated signal into
approximately harmonic components. Thakur et
al. (2013) adapted the WSST to discretised data (rather than continuous
functions) and developed a MATLAB® Synchrosqueezing Toolbox
(available for download:
The drip discharge rate time series, barometric pressure and air temperature
(potential weather-related drivers of drip discharge oscillations) were
analysed for time–frequency content in the following way:
The WSST functions in MATLAB® (version R2016a
or later) or the Synchrosqueezing Toolbox (Thakur et al., 2013) are applied to compute
the signal's frequency content over time. The output is a 2-D matrix
containing the complex frequency domain response The component amplitudes according to the standard signal
processing procedure are calculated using The component amplitudes are normalised using In order to highlight the main frequency components of interest (1 and
2 cpd) we chose The normalised amplitude matrices are visualised in pseudo-colour plots.
Distinct frequency components (signals with contrasting amplitudes whose
frequency does not significantly change over time) can easily be
distinguished from chaos (i.e. lack of regular oscillations identified as
signals with varying amplitude and frequency over time). Stronger periodic
components would yield larger amplitudes and therefore also a value that is
closer to 1 in the respective WSST plots. While this analysis is conducted
manually, it could be automated using criteria for the strength, continuity
and stability of any frequency component of interest.
An example of the time–frequency mapping conducted according to the above described method is illustrated in Fig. 2. The results obtained by applying the WSST (Fig. 2b) can be compared to the results from a continuous wavelet transform (CWT) with a Morlet mother wavelet (Fig. 2c) (Torrence and Compo, 1998). From this example it is clear that WSST features significantly less time–frequency smearing and therefore allows improved identification and delineation of close-by frequency components such as those at 1 or 2 cpd (compare Fig. 2b and c). Therefore, WSST presents a significant advantage over traditional signal processing methods such as the continuous wavelet transform when identifying the timing and duration of multiple frequency components embedded in measurements.
Using this methodology, a periodic drip discharge rate could be defined as consisting of continuous periods of (a) stable 1 cpd frequency, (b) stable 1 and 2 cpd frequency and (c) chaos (components with randomly varying frequency and amplitude). We used (a) and (b) as spectral “fingerprints” to identify and mark periods of continuous occurrence of daily and sub-daily oscillations in the drip discharge rate dataset.
The drip discharge time series are presented in Fig. 3. The drip discharge sites are spatially clustered in three groups within the cave (Fig. 1 and Table 1). Sites with the G prefix are located near the main entrance of the cave on the western side. The location is highly decorated with speleothems. M sites are located in the middle section of the cave in a large chamber with a high ceiling populated by soda straw formations. Location LR1 is situated near the cave exit at the top of a flow stone.
Comparing the time–frequency content of the drip discharge rate
Drip discharge rate time series for all drip sites in Glory Hole Cave with periods where daily fluctuations occur highlighted in light grey (1 cpd) and dark grey (1 and 2 cpd). The time periods examined in more detail in Figs. 4–6 are indicated by bolder outline. Daily evapotranspiration (19 December 2012–3 July 2014), rainfall, barometric and air temperature are also shown.
The raw drip rate, evapotranspiration and surface temperature data
with the corresponding drip rate WSST plot for time periods where a 1 cpd
signal is present for sites
The drip discharge rate at G1 and G3 varies seasonally, with higher drip
rates in winter, total flow volume of 133.37 and 109.52 L, respectively,
than summer (64.56 and 14.1 L, respectively). Drip rate increases in
response to rainfall events during the wet season and gradually decreases
through the drier part of the year. Drip rate is lowest during April and May
and highest during June and July. Similarly, G6 exhibits seasonal variation
with a higher volume of discharge during the winter than summer. The drip
rate at G10 increases sharply from 0.14 mL min
Daily fluctuations in drip discharge rate were identified in 8 out of 12 sites using WSST. There was no connection between the sites that did not exhibit the fluctuations with respect to spatial location, flow volume or flow regime type. The temporal and spatial pattern of daily oscillations are shown by the grey shaded areas in Fig. 3. The length of time the signal is present varied temporally for each drip site. For example, there was a strong 1 cpd signal in the drip water at G1 for 10 days in February 2013 whereas in January 2014 1 cpd fluctuations only lasted 5 days (Fig. 4). The timing of when the signal occurs on an annual scale varied within and between drip sites. For example, a 1 cpd signal only occurred during the first 3 months of the year for G1, whereas a 1 cpd signal occurred sporadically at G3 throughout the calendar year (December 2012, February and March 2013, January 2014, September 2014, January 2015).
The time lag calculated using cross correlation analysis between air
temperature and daily drip rate for each period of drip rate oscillation; the
timing of when minimum and maximum drip rate occurred within the time periods
The daily timing of minimum and maximum drip rate varied within and between
individual drip sites. At G1 the 1 cpd maximum and minimum drip rate
generally around 06:00–00:00 and 00:00–21:00 LT, respectively. Daily oscillations were
only observed once at G8 between 14 and 21 May 2014 with minimum drip rate at
03:00–09:00 LT and maximum drip rate around 00:00–21:00 LT.
Both 1 and 2 cpd signals were observed at M10 for all the periods of drip rate oscillation with the larger
peak occurring in the afternoon around 15:00–18:00 LT, minimum drip rate appeared
consistently between 06:00 and 09:00 LT. Time lag between air temperature and drip rate
was quantified by performing a cross correlation analysis with a shift
interval of 15 min up to
1 and 2 cpd signals can occur concurrently, for example, at M4 between 1 and 9 September 2013 (Fig. 5). This trend, in which the 2 cpd is weaker than the 1 cpd, is consistent across all sites where the two signals coincide. The 2 cpd signal can be visually determined in the raw drip rate data by a second smaller peak. Examples of characteristic WSST plots alongside the corresponding raw drip rate and surface temperature data will be discussed in greater detail below. All WSST analyses have been plotted in the Supplement.
WSST identified a 1 cpd oscillation in drip rate between 8 and 21 February 2013
at G1 and G3 (Fig. 4a and b). At G1 (Fig. 4a), the signal was
initially chaotic, but from 8 to 21 February 2013 the drip rate oscillates
sharply at 1 cpd. The maximum drip rate ranging from 4.03 to 3.75 mL min
The drip rate at G3 (Fig. 4b) oscillated at 1 cpd for 8 days from
12 to 20 February 2013. In contrast to G1, the maximum drip rate appeared in the
evening and the minimum drip rate occurred in the morning. The maximum drip
rate ranging from 1.63 to 2.01 mL min
From 1 to 27 February 2013, daily barometric pressure peaked between 08:30 and 09:00 LT with a magnitude of 0.1 to 0.5 kPa with a smaller second peak between 20:00 and 22:00 LT with a magnitude of 0.1–0.3 kPa (Fig. 4c). There were larger changes in air pressure on a mesoscale with peaks in air pressure on 16, 22, 26 February 2013 and minimum air pressure on 19, 24 and 28 February 2013. The air pressure changes in these cycles were as much as 1.5–2 kPa. The drip rate at G1 and G3 did not appear to be affected by the daily or weekly changes in air pressure. For example, when air pressure decreased dramatically on 27 February 2013 (Fig. 4c) there was no substantial change in drip rate at either G1 or G3.
Insolation drives daily cycles in surface air temperature with maximum
temperatures recorded between 11:30 and 16:00 LT and minimum temperatures recorded
between 04:00 and 08:00 LT (Fig. 4d). The difference in daily minimum and maximum air
temperature varied greatly. For example, between 12 and 20 February 2013 the
difference was 17.05–22.2
The raw drip discharge data, evapotranspiration, surface temperature and wavelet synchrosqueezed transform (WSST) plot of the drip discharge for site M4 from 1 to 11 September 2013.
The complexity of the Glory Hole Cave karst system is evident in the variety of drip regimes. For example, the drip rate at G1, G6 and G3 is seasonally driven with high discharge rates during the wettest period of the year. In contrast, drip discharge at G10 and M10 is likely driven by a storage component which discharges via a less permeable pathway which limits the store at a particular level during wet periods. The drip site is fed via the main water store rather than the overflow pathway itself (Baker et al., 2012; Bradley et al., 2010). Sites LR1, M4, M13 and M2 behave similarly in that they are all very responsive to rainfall events and have low base flows during periods of low rainfall. The response to rainfall events occur within 24 h across these sites. Calculated flow volumes indicate the storage capacity of the stores feeding the discharge sites. For example, there was an infiltration event on 1 June 2013 which caused a dramatic increase in drip rate for sites LR1, M2, M4 and M13. The flow volumes for each site from the start of the event to the point where the discharge returns to a constant rate are as follows: LR1 (1.60 L), M4 (2.99 L), M13 (8.09 L) and M2 (11.30 L). The length of the recession limb, calculated from the peak of the hydrograph until the drip rate returns to base rate, is indicative of the speed at which the store drains. For example, the decay in drip rate is 12 days for site M2 compared to 4 days for M13. The time it takes for the store to drain is not dependent on flow volume, as M13 has a flow volume of more than 5 times that of site LR1, but they both have drainage periods of 5 days. The discrepancy in drainage time could indicate variation in flow pathway length between sites. G8 is the only site with a relatively lower total flow volume during winter than summer. M1 has a low drip rate that shows a small seasonal fluctuation but does not visibly respond to individual events. This site is likely being fed by a store that is large enough to assimilate short-term inputs from the surface without impacting drip rate. This type of store has been described as a karst hydrological model component in a number of studies (Arbel et al., 2010; Hartmann et al., 2014b; Markowska et al., 2015).
Constant frequency oscillations in drip discharge (1 and 2 cpd) occur sporadically throughout the monitoring period December 2012–April 2015 at 8 out of 12 monitored drip sites. This phenomenon could be explained by a number of daily drivers including variations in pressure gradients between karst and cave due to cave ventilation effects, atmospheric and earth tides or variations in hydraulic conductivity (due to changes in viscosity of water with daily temperature oscillations), and solar-driven daily cycles of vegetative (phreatophytic) transpiration. These drivers are now considered in turn.
Surface air pressure and cave air pressure were significantly correlated
(
Atmospheric tides are caused by changes in air pressure due to the heating and cooling of air masses during the day and night. Correlations between atmospheric tides and drip rates can occur since increases (decreases) in atmospheric pressure at the ground surface are partitioned into stress increase (decrease) in the soil/rock mass and pore pressure increase (decrease) within the formation (Acworth et al., 2015). Drip rates could be affected if this changes the pressure gradient between the groundwater in karst stores and the cave (Tremaine and Froelich, 2013). Such a pressure imbalance is dependent on the hydromechanical properties and karst architecture as well as the degree of pneumatic connection between both the surface and the water table, and the surface and the cave at the location of the drip. Maximum and minimum atmospheric pressure occur at the same time each day (Fig. 4d).
Atmospheric tides were eliminated as a process to explain the daily oscillation phenomenon for several reasons. Firstly, there was no relationship between drip discharge rate and the longer-term barometric changes caused by synoptic weather patterns (Fig. 4). The mesoscale fluctuations in pressure caused by synoptic weather patterns are an order of magnitude higher than those caused by daily atmospheric tides. Since the drip rate did not respond to pressure changes of this size, they will not respond significantly to changes of a smaller magnitude at a higher frequency because higher frequency signals will be more highly damped and lagged. Secondly, the timing of the daily maximum and minimum drip rates in Glory Hole Cave varied within each drip site over time. For example, the peak discharge time for site G6 varied between 13:24 and 19:48 LT for the period 11 August 2013–25 August 2015. This finding contrasts with previous studies where drip rate is negatively correlated with barometric pressure and responds to daily pressure changes linearly (Tremaine and Froelich, 2013). However, this could indicate that the daily drip water variations in Glory Hole Cave are being driven by a non-linear process and this is discussed further below. Thirdly, the karst architecture of Glory Hole Cave is well-developed, has little to no primary porosity and is unconfined. Hence, it is unlikely to exhibit barometric responses such as seen in confined systems (Merritt, 2004), whereby pore pressure changes due to barometric loading are substantially lower than the change of cave air pressure.
Earth tides are solid deformations of the Earth's surface caused by the gravitational pull of the moon and sun (Merritt, 2004). It has been previously shown that earth tides can cause regular oscillations in groundwater level if the aquifer is sufficiently confined (Acworth et al., 2015). However, at Glory Hole Cave this process can be ruled out due to the unconfined conditions and the fact that the compressibility of limestone is smaller than that of water and because fluctuations in pressure caused by earth tides are so small.
The study site has large surface temperature variations, particularly in
summer where daytime and nighttime temperatures can vary up to
31.1
The timing of the daily drip rate signal appears to be associated with the
difference in maximum and minimum surface temperature. In the examples
examined in more depth in Fig. 4a and b, when the difference between the maximum
and minimum temperature was high (17–22
During periods when there are 1 cpd oscillations in drip rate, there was a
relationship between drip rate and surface temperature on a weekly
timescale. The best example is in Fig. 6, where
Surface air temperature, evapotranspiration and drip discharge rate with the corresponding daily moving average for site G1 1–19 February 2014.
The area above the cave and in the small uphill catchment is dominated by
Tree water use from deep roots occurs when the upper layers are too dry and have a lower water potential than the soil water at deeper levels (Dawson and Pate, 1996; Zapater et al., 2011). Maximum tree water use by the roots is therefore expected in the afternoon during the period of maximum solar radiation, possibly lagged due to the time taken to hydraulically lift the water. Conversely, minimum tree water use is expected at the end of night around 06:00 LT. Burgess et al. (2001) measured sap flow in Eucalypt tap roots, finding tap root sap flow peaked around 1 pm and negative sap flow values indicated reverse (acropetal) flow between 19:00 and 07:00 LT. In consideration of this, drip water that comes from fractures and stores which contain tree roots would be expected to have a minimum drip discharge in the afternoon and maximum around sunrise. In reality, we observe more complex daily drip oscillations, with peak drip rate occurring at different times of the day and different times of the year. This is to be expected from a karstified system with flow routed through a varied and complex fractured network. Different scenarios for daily oscillations in a karst system will be discussed in detail below.
A conceptual representation of tree water use from karst stores
under different circumstances.
The depth of a store could affect the timing of daily drip rate oscillations
due to the delay in tree water transport. For example, consider the
hypothetical, identical trees with roots intercepting identical karst stores
or fractures at
The size of the karst store, or volume of water within the store, could
determine whether the daily oscillation is observable or not. Consider the
conceptual Fig. 7c and d, where identical trees have roots intercepting
different karst stores at the
Tree water use responds to annual variation in insolation. Consider Fig. 7e and f, where one tree root intercepts the same karst store over the course of a year. During winter (Fig. 7e), there is less insolation than the summer (Fig. 7f); therefore the rate of evapotranspiration is lower. This means that in winter the hydraulic lift is low or negative and daily oscillations in drip discharge could be dampened or absent. Our analysis reveals that only 2 out of 41 periods of 1 cpd oscillation occur during winter months June–August (G6 between 14 and 24 August 2013 and M2 between 8 and 13 July 2013). However, our analysis also revealed that season did not explain a significant amount of variance in lag time, thus suggesting that more variables, such as karst architecture, are affecting the timing of drip rate oscillations.
In reality, there are multiple trees of different ages above the cave,
further complicating the flow variability. Figure 7g–i present a conceptual
representation of tree tap root length increasing (
Karst architecture controls flow regimes and drip discharge rates of water
infiltrating into caves (e.g. Markowska et al., 2015). Flow rate influences
speleothem climate proxies, such as the
This is the first indirect volumetric observation of tree water use in cave
drip water. This supports a growing number of studies examining the impact
of trees on karst processes and paleoclimate proxies. For example, tree root
respiration provides a source of CO
We demonstrated a novel method of analysing recurring patterns in cave water drip rate using the WSST. Our analysis revealed daily and sub-daily oscillations with variable temporal and spatial signatures. We tested competing hypotheses for causes of daily oscillations using drip rate, barometric and temperature data. The only hypothesis which all the data and hydrologic theory were consistent with was that daily fluctuations in drip rate were driven by tree water use. We proposed that the complexity of flow pathways in the karst system accounted for the spatial and temporal variation in the daily fluctuations of drip rates. This was explored in detail using conceptual models. The results have wider implications for karst research, including providing a new protocol for inferring karst architecture, informing selection of speleothem specimens for tree water use paleoclimate studies and highlighting the importance of vegetation dynamics on karst recharge.
Data are available on figshare under
Katie Coleborn, Mark O. Cuthbert, Gabriel C. Rau and Andy Baker wrote the manuscript, discussed the results and implications and commented on the manuscript at all stages. Katie Coleborn, Andy Baker and Owen Navarre collected data. Gabriel C. Rau performed the WSST analysis and generated the WSST and CWT figures. Gabriel C. Rau and Owen Navarre created the location map. Katie Coleborn and Gabriel C. Rau generated other graphs and conceptual figures.
We acknowledge that Katie Coleborn was supported the Australian Research Council (LP130100177). Mark Cuthbert was supported by Marie Curie Research Fellowship funding from the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 299091. We would also like to thank Stuart Hankin for allowing us access to the weather station data and the National Parks and Wildlife Service staff at Yarrangobilly Caves. Solar exposure data derived from satellite imagery processed by the Bureau of Meteorology from the Geostationary Meteorological Satellite and MTSAT series operated by Japan Meteorological Agency and from GOES-9 operated by the National Oceanographic & Atmospheric Administration (NOAA) for the Japan Meteorological Agency. We would also like to acknowledge the use of equipment funded by the Australian Government National Collaborative Research Infrastructure Strategy (NCRIS). Edited by: T. Blume Reviewed by: two anonymous referees